Citation
In vitro morphogenesis and inheritance of in vitro traits in desmodium

Material Information

Title:
In vitro morphogenesis and inheritance of in vitro traits in desmodium
Creator:
Krottje, Peter A., 1951-
Publication Date:
Language:
English
Physical Description:
vii, 107 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Agronomy thesis, Ph. D
Dissertations, Academic -- Agronomy -- UF
Callus ( jstor )
Statistical discrepancies ( jstor )
In vitro fertilization ( jstor )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1995.
Bibliography:
Includes bibliographical references (leaves 83-90).
General Note:
Typescript.
General Note:
Vita.
Statement of Responsibility:
by Peter A. Krottje.

Record Information

Source Institution:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
021933836 ( ALEPH )
33662107 ( OCLC )

Downloads

This item has the following downloads:

AA00029821_00001 ( .pdf )

AA00029821_00001_0047.txt

AA00029821_00001_0056.txt

AA00029821_00001_0092.txt

invitromorphogen00krot_0080.txt

invitromorphogen00krot_0015.txt

AA00029821_00001_0109.txt

invitromorphogen00krot_0059.txt

invitromorphogen00krot_0083.txt

invitromorphogen00krot_0011.txt

invitromorphogen00krot_0064.txt

AA00029821_00001_0076.txt

AA00029821_00001_0081.txt

invitromorphogen00krot_0092.txt

invitromorphogen00krot_0035.txt

AA00029821_00001_0008.txt

AA00029821_00001_0075.txt

AA00029821_00001_0002.txt

invitromorphogen00krot_0077.txt

invitromorphogen00krot_0029.txt

invitromorphogen00krot_0109.txt

invitromorphogen00krot_0040.txt

invitromorphogen00krot_0037.txt

invitromorphogen00krot_0108.txt

invitromorphogen00krot_0104.txt

AA00029821_00001_0078.txt

AA00029821_00001_0115.txt

invitromorphogen00krot_0019.txt

invitromorphogen00krot_0030.txt

AA00029821_00001_0061.txt

AA00029821_00001_0098.txt

invitromorphogen00krot_0004.txt

AA00029821_00001_0101.txt

AA00029821_00001_0095.txt

AA00029821_00001_0035.txt

invitromorphogen00krot_0081.txt

AA00029821_00001_0062.txt

invitromorphogen00krot_0079.txt

AA00029821_00001_0040.txt

invitromorphogen00krot_0016.txt

AA00029821_00001_0034.txt

AA00029821_00001_0025.txt

AA00029821_00001_0038.txt

invitromorphogen00krot_0010.txt

invitromorphogen00krot_0028.txt

AA00029821_00001_0030.txt

AA00029821_00001_0111.txt

invitromorphogen00krot_0027.txt

invitromorphogen00krot_0103.txt

invitromorphogen00krot_0007.txt

AA00029821_00001_0063.txt

invitromorphogen00krot_0112.txt

invitromorphogen00krot_0068.txt

invitromorphogen00krot_0005.txt

AA00029821_00001_0100.txt

invitromorphogen00krot_0003.txt

AA00029821_00001_0032.txt

invitromorphogen00krot_0058.txt

AA00029821_00001_0019.txt

invitromorphogen00krot_0098.txt

invitromorphogen00krot_0096.txt

AA00029821_00001_0069.txt

AA00029821_00001_0077.txt

AA00029821_00001_0049.txt

AA00029821_00001_0014.txt

invitromorphogen00krot_0113.txt

AA00029821_00001_0070.txt

invitromorphogen00krot_0012.txt

invitromorphogen00krot_0039.txt

AA00029821_00001_0080.txt

AA00029821_00001_0108.txt

AA00029821_00001_0010.txt

AA00029821_00001_0007.txt

AA00029821_00001_0103.txt

invitromorphogen00krot_0054.txt

AA00029821_00001_0011.txt

AA00029821_00001_0005.txt

AA00029821_00001_0054.txt

invitromorphogen00krot_0110.txt

invitromorphogen00krot_0047.txt

invitromorphogen00krot_0088.txt

invitromorphogen00krot_0100.txt

invitromorphogen00krot_0014.txt

invitromorphogen00krot_0105.txt

AA00029821_00001_0028.txt

AA00029821_00001_0079.txt

AA00029821_00001_0004.txt

AA00029821_00001_0052.txt

invitromorphogen00krot_0070.txt

AA00029821_00001_0105.txt

AA00029821_00001_0097.txt

invitromorphogen00krot_0097.txt

AA00029821_00001_pdf.txt

AA00029821_00001_0018.txt

AA00029821_00001_0031.txt

AA00029821_00001_0016.txt

invitromorphogen00krot_0093.txt

AA00029821_00001_0020.txt

invitromorphogen00krot_0048.txt

AA00029821_00001_0029.txt

AA00029821_00001_0057.txt

invitromorphogen00krot_0085.txt

AA00029821_00001_0036.txt

AA00029821_00001_0058.txt

invitromorphogen00krot_0082.txt

invitromorphogen00krot_0008.txt

invitromorphogen00krot_0026.txt

invitromorphogen00krot_0042.txt

AA00029821_00001_0114.txt

AA00029821_00001_0068.txt

AA00029821_00001_0041.txt

invitromorphogen00krot_0075.txt

AA00029821_00001_0094.txt

AA00029821_00001_0066.txt

invitromorphogen00krot_0115.txt

AA00029821_00001_0033.txt

AA00029821_00001_0015.txt

invitromorphogen00krot_0023.txt

invitromorphogen00krot_0057.txt

invitromorphogen00krot_0021.txt

AA00029821_00001_0037.txt

AA00029821_00001_0116.txt

invitromorphogen00krot_0034.txt

AA00029821_00001_0003.txt

AA00029821_00001_0084.txt

invitromorphogen00krot_0060.txt

AA00029821_00001_0026.txt

AA00029821_00001_0102.txt

invitromorphogen00krot_0071.txt

AA00029821_00001_0017.txt

AA00029821_00001_0067.txt

invitromorphogen00krot_0025.txt

AA00029821_00001_0096.txt

invitromorphogen00krot_0086.txt

AA00029821_00001_0053.txt

invitromorphogen00krot_0002.txt

AA00029821_00001_0043.txt

AA00029821_00001_0013.txt

AA00029821_00001_0089.txt

invitromorphogen00krot_0076.txt

AA00029821_00001_0106.txt

AA00029821_00001_0085.txt

AA00029821_00001_0112.txt

AA00029821_00001_0099.txt

invitromorphogen00krot_0055.txt

AA00029821_00001_0039.txt

invitromorphogen00krot_0074.txt

AA00029821_00001_0012.txt

AA00029821_00001_0042.txt

invitromorphogen00krot_0052.txt

invitromorphogen00krot_0013.txt

invitromorphogen00krot_0024.txt

AA00029821_00001_0065.txt

invitromorphogen00krot_0063.txt

AA00029821_00001_0093.txt

AA00029821_00001_0006.txt

AA00029821_00001_0083.txt

invitromorphogen00krot_0043.txt

invitromorphogen00krot_0001.txt

invitromorphogen00krot_0090.txt

AA00029821_00001_0086.txt

invitromorphogen00krot_0000.txt

invitromorphogen00krot_0009.txt

AA00029821_00001_0045.txt

invitromorphogen00krot_0033.txt

E604ZTRQX_PA27DJ_xml.txt

invitromorphogen00krot_0072.txt

invitromorphogen00krot_0051.txt

AA00029821_00001_0022.txt

AA00029821_00001_0051.txt

AA00029821_00001_0048.txt

invitromorphogen00krot_0099.txt

invitromorphogen00krot_0036.txt

AA00029821_00001_0074.txt

AA00029821_00001_0046.txt

invitromorphogen00krot_0091.txt

invitromorphogen00krot_0107.txt

AA00029821_00001_0104.txt

invitromorphogen00krot_0067.txt

invitromorphogen00krot_0066.txt

AA00029821_00001_0082.txt

AA00029821_00001_0024.txt

invitromorphogen00krot_0031.txt

invitromorphogen00krot_0078.txt

invitromorphogen00krot_0038.txt

invitromorphogen00krot_0045.txt

AA00029821_00001_0023.txt

invitromorphogen00krot_0101.txt

invitromorphogen00krot_0018.txt

invitromorphogen00krot_0053.txt

invitromorphogen00krot_0032.txt

invitromorphogen00krot_0022.txt

AA00029821_00001_0055.txt

invitromorphogen00krot_0046.txt

AA00029821_00001_0071.txt

AA00029821_00001_0009.txt

AA00029821_00001_0073.txt

AA00029821_00001_0091.txt

invitromorphogen00krot_0106.txt

invitromorphogen00krot_0006.txt

invitromorphogen00krot_0062.txt

invitromorphogen00krot_0094.txt

invitromorphogen00krot_0111.txt

invitromorphogen00krot_0084.txt

AA00029821_00001_0107.txt

AA00029821_00001_0027.txt

invitromorphogen00krot_0020.txt

AA00029821_00001_0060.txt

invitromorphogen00krot_0069.txt

invitromorphogen00krot_0044.txt

AA00029821_00001_0044.txt

invitromorphogen00krot_0061.txt

invitromorphogen00krot_0089.txt

invitromorphogen00krot_0102.txt

invitromorphogen00krot_0056.txt

AA00029821_00001_0064.txt

AA00029821_00001_0090.txt

invitromorphogen00krot_0041.txt

AA00029821_00001_0059.txt

AA00029821_00001_0050.txt

invitromorphogen00krot_0065.txt

invitromorphogen00krot_0017.txt

AA00029821_00001_0113.txt

invitromorphogen00krot_0049.txt

invitromorphogen00krot_0114.txt

invitromorphogen00krot_0050.txt

AA00029821_00001_0087.txt

AA00029821_00001_0072.txt

AA00029821_00001_0110.txt

invitromorphogen00krot_0095.txt

AA00029821_00001_0088.txt

AA00029821_00001_0021.txt

invitromorphogen00krot_0087.txt

AA00029821_00001_0001.txt

invitromorphogen00krot_0073.txt


Full Text







IN VITRO MORPHOGENESIS AND INHERITANCE OF IN VITRO
TRAITS IN DESMODIUM












By

PETER A. KROTTJE


















A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1995










ACKNOWLEDGEMENTS


I gratefully acknowledge Dr. D.S. Wofford for his sustained support,
guidance, and encouragement throughout my course of study. Thanks are also extended to Dr. K.H. Quesenberry, who has generously provided insight and assistance in conducting this project. I would like to thank the other members of my committee, Dr. R.L. Smith, Dr. P.M. Lyrene, and Dr. G.A. Moore for their interest in my program and their readiness to assist and to share their expertise.
































11








TABLE OF CONTENTS


ACKNOWLEDGMENTS ..................... ii

LIST OF TABLES ................. .. ..... iv

LIST OF FIGURES ................... ..... v

ABSTRACT . . . . . . . . . . . . vi

CHAPTERS

1 INTRODUCTION .............. 1

2 OPTIMIZATION OF IN VITRO REGENERATION PROTOCOL 6

Materials and Methods .............. 11
Results and Discussion .............. 15

3 INHERITANCE OF IN VITRO REGENERATION AND
ASSOCIATED CHARACTERS ........... 30

Materials and Methods .............. 35
Results and Discussion ............. 37

4 SUMMARY AND CONCLUSIONS .... ... . 78 REFERENCES . . . . . . . . . . . . 83

APPENDICES

A DERIVATION OF BETWEEN- AND WITHIN-FAMILY
VARIANCES FOR CALLUS GROWTH ........ 91

B DERIVATION OF VARIANCE PARTITIONING MATRICES
FOR CALLUS GROWTH .............. . 92

C RESULTS OF LATER ITERATIONS OF CALLUS GROWTH
VARIANCE PARTITIONING .......... . 94

D RAW CALLUS GROWTH AND REGENERATION DATA 97 BIOGRAPHICAL SKETCH .................... 107






iii








LIST OF TABLES

Table p.ge

2-1 Response of genotype IRFL 6123 to 2,4-D and kinetin levels in the
shoot bud induction medium. . . . ........ 22

2-2 Response of genotype IRFL 6123 to duration of shoot bud induction
treatment. ........................ 23

2-3 Response of genotype IRFL 6123 to benzyladenine and gibberellic
acid levels in the shoot elongation medium. ......... 24

2-4 Response of genotype IRFL 6123 to picloram and benzyladenine
levels in the shoot elongation medium. ........ . 25

3-1 Number of F2 plants, F3 plants, and F3 families from which in vitro
data were collected .. .. . . . . . . . . 59

3-2 Univariate statistics for parental, F2, and F3 populations in linear,
logarithmic, and square root scales. ...... ........ 60

3-3 Model coefficients used to construct joint scaling tests ..... 62 3-4 Estimates of m, [d], and [h] obtained from joint scaling tests for callus
growth. . . . . . .. .. . . . .. .. 63

3-5 Coefficients used for variance partitioning for callus growth. 64 3-6 Observational variance components and initial weights used for
variance partitioning for callus growth ............. 65

3-7 Estimates of causal variance components and associated
probabilities obtained from variance partitioning for callus growth. 67 3-8 Regeneration score means and variances derived from raw and
uncensored data sets ............... . . 68







iv








LIST OF FIGURES


Figure P-M 2-1 Response in genotype IRFL 6123 to 2,4-D concentration in shoot
induction medium. ..................... 26

2-2 Unifoliate shoots produced by genotype IRFL 6123. .. ..... 27 2-3 Multifoliate shoots produced by genotype IRFL 6123 ..... 28 2-4 Spatulate shoots produced by genotype IRFL 6123 in the presence
of gibberellic acid (GA3). ................ 29

3-1 Relationship of F3 family means and variances for callus growth
under (a) linear, (b) logarithmic, and (c) square root scales .. 70 3-2 Regression of F3 family mean callus weight on F2 parental callus
weight under linear, logarithmic, and square root scales. . 71 3-3 Two hypothetical distributions with three response classes and two
thresholds . . . . . . . . . .. .. 72

3-4 Normal quantile plots of a hypothetical (a) intact and (b) censored
population .. . . . . . . . . ... . 73

3-5 Regeneration score histogram for genotype IRFL 6123 . . 74 3-6 Regression of F3 family mean regeneration score on F2 parent
regeneration score .................... 75

3-7 Profuse regeneration exhibited by members of two F3 families
derived from cross 507 . ... ............ .76

3-8 Regression of callus weight on regeneration score for combined F3
populations .......................... 77







v










Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

IN VITRO MORPHOGENESIS AND INHERITANCE OF IN VITRO TRAITS IN DESMODIUM

By

Peter A. Krottje

August, 1995

Chairman: David S. Wofford
Major Department: Agronomy

Desmodium, Desmodium sp., is a forage legume widely cultivated in the
tropics and of growing importance in the southern United States. Previous work aimed at application of biotechnological methods to this crop had obtained limited in vitro regeneration from seedling hypocotyl explants in a single genotype. The objective of the present work was to examine the potential for improving regeneration response in desmodium through optimization of culture protocols and through genetic improvement.
Efforts at identifying alternative explant sources focused on seedling
cotyledons, mature leaf disks, and mature petioles. Regeneration could not be induced from any of these explant types. Several combinations of 2,4-D and kinetin concentrations in the shoot bud induction medium were examined in an unsuccessful effort to obtain regeneration from hypocotyl explants of recalcitrant genotypes. Direct regeneration from hypocotyl explants without an intervening callus growth step was also unsuccessful. Several treatments were investigated with the aim of enhancing regeneration response in the previously identified regenerating genotype. The best protocol identified consisted of a

vi








28-day callus induction step with 0.06 mg L-1 picloram (4-amino-3,5,6trichloropicolinic acid) and 0.1 mg L-1 BA (6-benzylaminopurine), a 14-day shoot bud induction step with 1.0 mg L-1 2,4-D and 2.0 mg L-1 adenine, and a 28-day shoot elongation step with 0.012 mg L-1 picloram and 0.2 mg L-1 BA. Under this protocol, shoot production occurred in 63% of explants.
Extensive efforts at hybridizing regenerating and nonregenerating
genotypes yielded two crosses. Callus growth and regeneration capacity were evaluated in the parental lines, the F2 and F3 generations. Based on joint scaling tests and variance partitioning, neither trait was found to fit a simple additive-dominant genetic model. Both traits were moderately to highly heritable, as determined by parent-offspring regression. Heritability of callus growth ranged from 0.52 to 0.77, depending on the cross and on the numerical scale employed. Heritability of the regeneration trait ranged from 0.14 to 0.46. Members of two F3 families exhibited much more vigorous and prolific regeneration than the regenerating parental genotype.






















vii












CHAPTER 1
INTRODUCTION


The last two decades have seen a surge of interest in biodiversity within the agricultural research community. In order to exploit diverse agricultural environments in an efficient, environmentally benign manner, the full range of available genetic resources must be examined and utilized. Extensive research effort is aimed at developing new crops and identifying new regions of application for existing crops. Desmodium (Desmodium spp.) is one of several exotic forage legumes that show potential for use in the pastures of Florida and the adjacent southeastern United States.
Desmodium is a large genus consisting primarily of perennial herbs and shrubs. The genus is distributed in tropical and subtropical regions worldwide, with a probable center of origin in Southeast Asia and a secondary center of diversity in Mexico (Ohashi, 1973; Schubert, 1963). Classification within this genus is difficult due to continuity of taxonomic characters among species, and estimates of the number of species range from 200 (Takahashi, 1952) to 500 (Younge et al., 1964). Most of the species studied have chromosome numbers of 2n=2x=22, and are predominantly self-pollinated (Chow and Crowder, 1972, 1973; Rotar and Uruta, 1967).
The agronomic role of the genus Desmodium has been reviewed by Imrie et al. (1983). Several desmodium species are grown in significant acreages in tropical or subtropical pastures. D. intortum (Mill.) Urb. (greenleaf desmodium) and D. uncinatum (Jacq.) DC. (silverleaf desmodium) are widely used in humid subtropical regions of Australia, Africa, and South America.


1






2

D. heterocarpon (L.) DC. (carpon desmodium) has been released as the cultivar 'Florida' for pastures in the southeastern United States (Kretchmer et al., 1976). This species flowers earlier in the fall than either greenleaf or silverleaf, and is thus adapted to areas where occasional early frosts threaten seed production. D. ovalifolium Guill and Perr. is grown in Southeast Asia and has attracted attention elsewhere for its tolerance to acid, low fertility soils and shaded conditions (Schultz-Kraft and Pattanavibul, 1985). Other desmodium species with agronomic potential include D. heterophyllum (Willd.) DC., D. sandwicense E. May., and D. barbatum Benth.
Most desmodium cultivars are simply selections from germplasm collections, and genetic improvement of this crop has received minimal attention until the last decade. A desmodium breeding program was initiated at the University of Florida in the late 1980's, with the primary breeding objectives of improved forage quality, rapid establishment, increased seed production, and resistance to root knot nematodes. The University of Florida program utilizes germplasm from Desmodium heterocarpon ssp. heterocarpon, D. heterocarpon ssp. angustifolium (Craig) Ohashi, and D. ovalifolium. The species heterocarpon and ovalifolium are closely related and yield fertile progeny upon intercrossing (Quesenberry et al., 1989). Some authors include the species ovalifolium within species heterocarpon (Ohashi, 1983), but the two are morphologically distinct and are generally regarded as separate entities by agronomic researchers (Imrie et al., 1983; Schultze-Kraft and Benavides, 1988). D. heterocarpon produces elongated inflorescences and glabrous to slightly pubescent seed pods, while D. ovalifolium has compact inflorescences and bears heavily pubescent pods. Leaves are opaque in ovalifolium, glabrous to slightly pubescent in heterocarpon ssp. heterocarpon, and coriacious in heterocarpon ssp. angustifolium. Subspecies angustifolium is further






3

distinguished by a leaf length/width ratio of 3.0 or greater, as opposed to 1.0 to
2.5 in ssp. heterocarpon and D. ovalifolium (Quesenberry et al., 1989), and by a more erect growth habit.
Recent advances in biotechnology have extended the potential for genetic improvement of both new and established crop species. Plant breeding has traditionally relied upon genetic variation existing within the species of interest or in closely related, sexually compatible species. Current genetic transformation methods allow desirable genes to be moved between distantly related or nonrelated species, thereby greatly expanding the potential gene pool for any specific crop (Fraley et al., 1986). Among forage legumes, Agrobacterium-mediated genetic transformation has been used to insert reporter genes into alfalfa (Deak et al., 1986), white clover (White and Greenwood, 1987), and red clover (Quesenberry et al., 1992), and Stylosanthes (Sarria et al. 1994). Future application of transformation techniques to forage legumes may involve introduction of specific genes for pathogen resistance, herbicide resistance, or forage quality.
Because genetic transformation occurs on a single cell basis, obtaining
transformed plants requires that whole plants be regenerated from single cells, or from small cell aggregates. A protocol for efficient in vitro plantlet regeneration is therefore highly desirable for transformation work. Besides serving as a tool for genetic transformation, cell and tissue culture methodologies can be useful for in vitro screening of germplasm for resistance to pathogens, herbicides, and environmental stresses (Hughes, 1983).
Published work on the in vitro culture of desmodium is sparse. Angeloni et al. (1988) regenerated D. affine and D. incanum from shoot tip cultures using a modified MS medium (Murashige and Skoog, 1962), but failed to obtain regeneration from leaf or anther explants. Using seedling hypocotyl explants






4

from six diverse desmodium genotypes, Wofford et al. (1992b) evaluated two tissue culture protocols that had been previously demonstrated effective for legumes. Regeneration was only observed in a single genotype of D. heterocarpon ssp. angustifolium, using an L2 based protocol originally developed for use with Trifolium (Phillips and Collins, 1980).
If in vitro techniques are to be most effectively employed for the genetic improvement of desmodium, it will first be necessary to broaden the range of regenerating genotypes. This may be accomplished either by modification of the culture protocol or by breeding for regeneration ability. Nonregenerating genotypes can often be induced to regenerate by manipulating growth regulator levels or any of several other culture parameters. In many species, however, the majority of genotypes have resisted exhaustive efforts to induce regeneration (Flick et al., 1983; Ammirato, 1983). Breeding for regeneration ability has been effective for other forage crop species, including alfalfa (Reisch and Bingham, 1980) and red clover (Quesenberry et al. 1992). The efficiency of such a breeding effort is dependent on the genetic basis of the regeneration trait. Investigations of the inheritance of this trait in several higher plant species has shown that the trait may be either qualitatively or quantitatively controlled, and can be subject to cytoplasmic effects (Keyes et al., 1980; Kumar et al., 1985; Wan et al., 1988).
The remainder of this dissertation addresses both of these approaches to broadening the range of regenerable desmodium lines. Chapter 2 examines a variety of culture protocols with the objectives of inducing regeneration in a wider range of genotypes and optimizing the response of established regenerator genotypes. In Chapter 3 the genetic basis of callus growth and regeneration is examined through the production of hybrids between regenerating and nonregenerating lines and evaluation of the resulting F2 and






5

F3 generations. Chapter 4 presents a summary and conclusions, and discusses the implications of this work within the broad context of genotype by environment interaction under in vitro conditions.












CHAPTER 2
OPTIMIZATION OF IN VITRO REGENERATION PROTOCOL


Although numerous members of the family Leguminoseae have been
regenerated in vitro, the family is regarded to be difficult with respect to in vitro regeneration (Flick et al., 1983). Regeneration frequency is low for many species, and specific culture requirements can vary widely both among and within species (Phillips and Collins, 1983).
Regeneration in angiosperms can be accomplished via either of two
conceptually distinct pathways-organogenesis or somatic embryogenesis. Both types of regeneration can be induced either directly from the initial explant source, or from callus or suspension cells generated in culture. Organogenesis usually involves production of a shoot meristem followed by shoot elongation and rooting. Shoot and root regeneration are discrete processes occurring in response to specific culture conditions, particularly the types and concentrations of plant growth regulators in the medium. In somatic embryogenesis, shoot and root meristems are produced simultaneously in a process similar to zygotic embryo development. Maturation and germination of somatic embryos can be induced by manipulating plant growth regulator levels, or may occur spontaneously under constant culture conditions (Hazra et al., 1989).
In practice, the distinction between organogenesis and embryogenesis is not always so clear (Ammirato, 1983). Ideally, somatic embryos should closely resemble sexual embryos and possess clearly identifiable shoot, root, and cotyledonary primordia that subsequently develop into their respective organs. Deviations from this ideal situation include abnormal or absent cotyledons and


6






7

failure of the radicle to produce a root (Tisserat et al., 1978; Sellars et al., 1990). The situation can be further complicated by complete failure of the somatic embryo to germinate, sometimes followed by adventitious bud production from embryo tissue (Saunders et al., 1987; Vasil, 1987). Due to the frequent occurrence of these and other developmental abnormalities during somatic embryogenesis, histological examination is necessary to conclusively distinguish between embryogenesis and organogenesis. Somatic embryos are characterized by two critical features: a bipolar morphology with discrete coleoptile and coleorhiza, and a lack of vascular connections to the surrounding tissue (Haccuis, 1978).
A successful regeneration protocol depends on choosing the appropriate type of tissue to initiate the in vitro culture. In easily cultured nonlegumes such as carrot, almost any part of the plant taken at any developmental stage can serve as an explant source (Ammirato, 1983). In most legumes, however, the choice is quite limited. As a general rule, regenerable cultures are most easily obtained from immature tissue. Somatic embryogenesis from immature embryos or portions of embryos has been reported for white clover, peanut, and soybean (Maheshwaran and Williams, 1984; Sellars et al., 1990). Mature embryos have been used as explant sources for mung bean, pigeon pea, peanut, and soybean (Mathews and Rao, 1984; Mehta and Mohan Ram, 1980; McKently et al., 1990).
Various types of seedling tissue have been used as explant sources for a wide range of legume species. Seedling hypocotyls are perhaps the most widely utilized explant source for legumes, having been used for alfalfa, red clover, pea, pigeon pea, and several other species (Bingham et al., 1975; Phillips and Collins, 1979; Kumar et al., 1984; Oelck and Schneider, 1983). Epicotyls or cotyledons have been used for pea, red clover, alfalfa, and pigeon






8

pea (Malmberg, 1979; Phillips and Collins, 1980; Lupotto, 1983; Kumar et al., 1984).
Among mature tissue types, regeneration has been obtained from leaf explants in alfalfa and peanut (Oelck and Scheider, 1983; McKently et al., 1992). Mature petioles have served as explant sources in various clover species (Choo, 1988), and mature stems in birdsfoot trefoil and Stylosanthes (Swanson and Tomes, 1980; Meijer, 1981).
Several different basal medium formulations have been used for tissue culture of legume species. The majority of work has utilized Murashige and Skoog's (1962) MS medium (Malmberg, 1979; Kao and Michayluk, 1981; Hazra et al., 1989; McKently et al., 1990). Gamborg et al. (1968) developed the B5 medium for tissue culture of soybean, and this medium has proven useful for several other legume species, including red clover (Quesenberry et al., 1992) and birdsfoot trefoil (Swanson and Tomes, 1980). The B5 medium differs from MS primarily by its greatly reduced ammonium nitrogen level and a lower calcium concentration. Schenk and Hildebrandt's (1972) SH medium is similar to B5, but contains much higher levels of inositol. This medium has been used for Stylosanthes (Scowcroft and Adamson, 1976) and crimson clover (Horvath et al., 1979). Phillips and Collins' (1979) L2 medium was developed for red clover culture and has subsequently found application in the culture of peanut and soybean (Sellars et al., 1990), and alyceclover (Wofford et al., 1992a). This medium is somewhat higher in calcium than MS, B5, or SH, and also differs from these media in that it lacks nicotinic acid. Blaydes' (1966) medium contains less nitrate than the above media, and has been employed for soybean, alfalfa (Bingham et al., 1975), and pigeonpea (Kumar et al., 1984).
It is generally acknowledged that plant growth regulator levels are of critical importance in plant tissue culture. For angiosperms in general,






9

organogenesis can often be induced by "reversal transfer" of callus from a high auxin, low cytokinin to a low auxin, high cytokinin medium (Dodds and Roberts, 1985). Somatic embryos can often be induced on media containing high levels of auxin-particularly 2,4-D-sometimes in combination with low levels of cytokinin (Ammirato, 1983). Reviews of legume work (Allavena, 1983; Ammirato, 1983; Flick et al., 1983; Phillips and Collins, 1983) indicate that the above generalities are applicable. Earlier work favored use of kinetin to induce organogenesis, but more recently, BA (6-benzylaminopurine) generally in the concentration range of 0.5 to 5.0 mg L-2, has been widely used (Webb et al., 1987; Vieira et al., 1990; McKently et al., 1990). Somatic embryogenesis in legumes has been induced with 2,4-D concentrations ranging from
0.001 mg L-1 (Phillips and Collins, 1980) to 80 mg L-1 (Saunders et al., 1987).
In callus based regeneration systems, there is strong evidence that culture conditions for the initial establishment of callus can strongly influence the potential for subsequent organogenesis or embryogenesis (Saunders et al., 1985; Tisserat et al., 1978). This issue has received relatively little attention in legumes, and is generally approached empirically. Callus induction and regeneration can sometimes occur on the same medium (Santos et al., 1983; Bovo et al., 1986), but most work has employed a separate callus induction step (Phillips and Collins, 1979; Walker et al., 1979; Flick et al., 1983).
Little has been published specifically on tissue culture of desmodium.
Angeloni et al (1988) produced multiple shoots from from shoot tip cultures of D. incanum on MS medium supplemented with 1.0 mg L-1 NAA (naphthalenacetic acid), 0.1 mg L-1 BA, and 1.0 mg L-1 GA (gibberellic acid). Shoot tip culture is regarded as meristem cloning rather than true regeneration, however. Wofford et al. (1992b) evaluated regeneration from seedling hypocotyl explants of six genotypes of D. heterocarpon and D. ovalifolium






10

under two protocols-an MS-based procedure intended to induce organogenesis and an L2-based procedure originally used by Collins and Phillips (1982) to induce embryogenesis in red clover. The MS procedure consisted of a callus induction step with 2.0 mg L-1 IAA (indole-3-acetic acid) and 1.0 mg L-1 kinetin, shoot induction with 0.1 mg L-1 IAA and 4.0 mg L-1 BA, and rooting on medium lacking plant growth regulators. The L2 protocol consisted of callus induction with 0.06 mg L-1 picloram (4-amino-3,5,6trichloropicolinic acid) and 0.1 mg L-1 BA, embryo induction with 0.01 mg L-1 2,4-D and 2.0 mg L-1 adenine, and embryo germination with 0.002 mg L-1 picloram and 0.2 mg L-1 BA. The MS protocol resulted in production of shoot meristems in five of six genotypes, but in all cases shoots failed to elongate. The L2 protocol yielded shoot meristems in two genotypes and whole plants were obtained in one genotype of D. heterocarpon ssp. angustifolium. The response on L2 superficially resembled organogenesis, but histological examination was not performed.
This chapter will evaluate a number of variations on the L2-based procedure utilized by Wofford et al. (1992b). Primary objectives were to determine if regeneration could be obtained in a wider range of desmodium genotypes and to optimize the response of previously identified regenerating genotypes. A bewildering range of culture variables can potentially influence regeneration response. The work presented here attempts to examine some of these variables in an orderly, stepwise manner in which the treatments in each experiment are based on the results of the previous experiment. The work begins with an examination of explant sources, proceeds to an investigation of various shoot bud induction treatments, and then examines several shoot elongation treatments. The chapter concludes with histological data presented with the objective of clarifying the mode of regeneration in this system.










Materials and Methods


Germplasm. Four genotypes were selected to represent a wide range of in vitro responses. In previous work (Wofford et al., 1992b) IRFL 6123 had been identified as a strong regenerator, CIAT 13083 as a possible regenerator, and UF 20 and UF 144 as nonregenerators. IRFL 6123 is classified as D. heterocarpon ssp. angustifolium, UF 20 represents D. heterocarpon ssp. heterocarpon, UF 144 represents D. ovalifolium, and CIAT 13083 possesses characteristics intermediate between D. heterocarpon ssp. heterocarpon and D. ovalifolium.

General protocol. Young, fully expanded leaves from greenhouse-grown plants were used to obtain leaf disc and petiole explants. Leaves and petioles were sterilized by immersion in 70% ethanol for 30 seconds followed by immersion in a 15% (v:v) Chlorox solution for one minute and several rinses with sterile deionized water. A stainless steel cork borer was used to cut 5 mm leaf discs, each containing a portion of midvein. Leaf discs were cultured with abaxial surface upward. Petioles were cut into sections approximately 10 mm in length.
For hypocotyl and cotyledon explants, seeds were scarified and sterilized by a 12-minute immersion in concentrated sulfuric acid followed by several rinses in sterile deionized water. Seeds were then placed in petri dishes containing SGL medium (Collins and Phillips, 1982) and incubated at 270 C. Hypocotyls and cotyledons were excised and placed onto callus induction medium after the hypocotyls reached a length of 7 to 10 mm. This occurred between 7 and 10 days after scarification. Hypocotyl explants were excised to a length of approximately 5 mm, with care taken to avoid radicle tissue and






12


cotyledon node tissue. Cotyledon explants consisted of the distal two-thirds of the cotyledon, placed abaxial surface upward.

All culture protocols used L2 basal medium (Phillips and Collins, 1980) solidified with 0.8 % (w:v) Phytagar (Gibco, Inc.; Island City, N.Y.). Growth regulators were coautoclaved with the basal medium for 20 minutes at 1210 C. Cultures were maintained at 270 C with a 16/8 h light/dark cycle at an illumination of approximately 100 gE m-2 sec-1. Callus induction and shoot bud induction was carried out in 60 mm disposable petri dishes with two or three explants per dish. For shoot elongation experiments, six calli were placed in each 100 mm petri dish.

All experiments utilized completely randomized designs. For number of shoots per explant, each explant comprised an observation. For percentage responding explants, each petri dish comprised one observation. Prior to statistical analysis, data on number of buds per explant were transformed to the square root of the quantity, observed value plus one half, in order to render variance independent of mean. Percentage data were transformed to arcsine of the square root of the observed percentage.

Evaluation of explant sources. Twenty-one leaf discs, petioles, hypocotyls, and cotyledons from each of genotypes UF 20, UF 144, IRFL 6123, and CIAT 13083 were cultured in a three-step protocol based on that described by Phillips and Collins (1980). Explants were placed on L2-based callus induction medium containing 0.06 mg L-1 picloram (4-amino-3,5,6-trichloropicolinic acid) and 0.1 mg L-1 BA (6-benzylaminopurine) for 28 days. Calli were then weighed and transferred to shoot induction medium containing 1.0 mg L-1 2,4-D and
2.0 mg L-1 adenine, again for 28 days. Following the shoot induction treatment, calli with or without visible shoot buds were transferred to shoot elongation






13

medium with 0.002 mg L-1 picloram and 0.2 mg L-1 BA for another 28 days. Shoot bud production and callus appearance were evaluated regularly throughout the culture period.
Effect of 2.4-D level and kinetin level on shoot bud induction. Seedling hypocotyls from genotypes UF 20, UF 144, and IRFL 6123 were cultured for twenty eight days on callus induction medium as described above. Calli were then transferred to the following L2-based shoot bud induction media:


(1) 0.1 mg L-1 2,4-D, 2.0 mg L-1 adenine (2) 0.3 mg L-1 2,4-D, 2.0 mg L-1 adenine (3) 1.0 mg L-1 2,4-D, 2.0 mg L-1 adenine (4) 3.0 mg L-1 2,4-D, 2.0 mg L-1 adenine
(5) 0.1 mg L-1 2,4-D, 0.15 mg L-1 kinetin, 2.0 mg L-1 adenine (6) 1.0 mg L-1 2,4-D, 0.15 mg L-1 kinetin, 2.0 mg L-1 adenine


The 28-day shoot induction treatments were followed by a 28-day shoot elongation treatment as described for the previous experiment. Ten explants (two per dish) were used per genotype-treatment combination.
Induction of direct regeneration from hypocotyl explants without an
intervening callus induction step. This experiment was carried out identically to the previous experiment with the exception that explants were placed directly onto the six shoot induction treatments rather than onto callus induction medium. Explants were rated after 28 days of shoot bud induction treatments and after 28 days on elongation medium.
Effect of duration of shoot bud induction treatment. Twenty-eight-day
hypocotyl-derived calli from IRFL 6123 were transferred to shoot bud induction medium containing 2.0 mg L-1 adenine and either 1.0 or 0.1 mg L-1 2,4-D for 0,





14

3, 7, 14, or 28 days. Following the induction treatment, calli were transferred to the shoot elongation medium described above. The 28-day induction treatment was followed by 28 days on elongation medium and the 14-, 7-, 3-, and 0-day induction treatments were followed by 42, 49, 53, or 56 days on elongation medium, respectively, to yield a total of 84 days in culture for each treatment. Calli were evaluated for bud formation at frequent intervals throughout the study. Twenty-one explants (three per dish) were used for each of the nine treatments.
Effect of BA and GA3 (aibberellic acid) levels on shoot elongation.
Hypocotyls from genotype IRFL 6123 were cultured on the callus growth medium described above for 28 days followed by 28 days on bud induction medium containing 2.0 mg L-1 adenine and 1.0 mg L-1 2,4-D. Twenty-four calli, each containing at least one well-formed shoot bud or small bud cluster, were transferred to each of the following elongation media:


(1) 0.002 mg L-1 picloram, 0.2 mg L-1 BA, 0 mg L-1 GA3
(2) 0.002 mg L-1 picloram, 0.2 mg L-1 BA, 0.2 mg L-1 GA3
(3) 0.002 mg L-1 picloram, 0.6 mg L-1 BA, 0 mg L-1 GA3
(4) 0.002 mg L-1 picloram, 0.6 mg L-1 BA, 0.2 mg L-1 GA3
(5) 0.002 mg L-1 picloram, 2.0 mg L-1 BA, 0 mg L-1 GA3
(6) 0.002 mg L-1 picloram, 2.0 mg L-1 BA, 0.2 mg L-1 GA3


Shoot elongation was visually evaluated after 28 days of elongation treatment.






15


Effect of picloram/BA ratio on shoot bud elongation. This experiment was conducted in the same manner as the previous experiment, but with the following elongation treatments:


(1) 0.012 mg L-1 picloram, 0.2 mg L-1 BA (2) 0.012 mg L-1 picloram, 0.6 mg L-1 BA
(3) 0.04 mg L-1 picloram, 0.2 mg L-1 BA (4) 0.04 mg L-1 picloram, 0.6 mg L-1 BA


An additional elongation treatment of 0.002 mg L-1 picloram and
2.0 mg L-1 BA was included as a check.
Histological examination. After twenty-eight days of callus growth medium, calli of IRFL 6123 were transferred to shoot induction medium containing
1.0 mg L-1 2,4-D and 2.0 mg L-1 adenine. Calli were removed for sectioning after 1, 2, and 3 weeks on induction medium. Specimens were embedded in Tissue-Tek O.T.C. Compound (Miles, Inc., Elkhart, IN) and sectioned on a CTF Microtome-Cryostat (International Equipment Co., Needham, MA). Sections eight microns in thickness were observed under the light microscope without staining.


Results and Discussion


Evaluation of explant sources. Wofford et al. (1992b) showed that seedling hypocotyls are satisfactory explant sources for in vitro regeneration in desmodium. The use of hypocotyls presents certain practical problems, however. Since the seedling must be dissected in order to place the hypocotyl






16

into culture, it is difficult or impractical to obtain both in vitro and in planta data from a single individual. Because an individual seedling possesses only a single hypocotyl, it is impossible to obtain replicated data on in vitro response unless quantities of genetically uniform seed are available. In the hope of eliminating these difficulties, an experiment was conducted to determine if regenerable callus could be produced using cotyledon, leaf disc, or petiole explants from genotypes IRFL 6123, CIAT 13083, or UF 20, with hypocotyl explants included as a check.
Results were discouraging. Cotyledons from IRFL 6123 and CIAT 13083 produced small amounts of necrotic callus while UF 20 cotyledons produced fairly abundant, deep green callus that showed no signs of regeneration. Leaf disc explants followed a similar pattern, with IRFL 6123 and CIAT 13083 producing meager quantities of nonregenerating callus primarily from major veins, and UF 20 producing larger quantities of nonregenerating callus. Petioles yielded somewhat greater callus mass than either cotyledons or leaf discs, but still failed to regenerate.
Hypocotyls yielded the greatest mass of callus from all genotypes. Ten of 21 hypocotyl-derived calli from IRFL 6123 produced shoot buds and elongated shoots were obtained from three of these. Shoot regeneration was observed from one UF 144 hypocotyl and one CIAT 13083 hypocotyl. In both cases, a single shoot bud appeared to form directly from explant tissue. This response could not be repeated in later experiments. It is possible that in both cases regeneration was the result of a small portion of meristematic tissue from the cotyledonary node being inadvertently included in the excised hypocotyl explants. Thus, shoots may have resulted from meristem cloning rather than true regeneration.






17

This work was rather limited in that the various explants were only tested under a single culture protocol. While it is possible that a different protocol might be more effective in inducing regenerable callus from leaf, petiole, or cotyledon explants, it is also possible that these explants would prove unresponsive to a wide range of protocols. In view of this uncertainty, further investigation of alternative explant sources was abandoned, and the remainder of this dissertation deals only with hypocotyl explants.
Effect of 2.4-D level and kinetin on shoot induction. The selection of
treatments for this experiment were based on reports that regeneration in other legume species can be sensitive to 2,4-D concentration (Phillips and Collins, 1980; Saunders et al., 1987) and that low concentrations of kinetin in combination with 2,4-D can stimulate regeneration (Sellars et al., 1990).
None of the 2,4-D treatments, either with or without kinetin, resulted in shoot production in genotypes UF 20 or UF 144. Response of IRFL 6123 to shoot induction treatments is presented in Table 2-1. The presence of kinetin in the induction medium resulted in a slight browning of the callus and complete inhibition of shoot formation. In the treatments lacking kinetin, a clear response to 2,4-D concentration was observed. Analysis of variance (ANOVA) indicates that with respect to mean number of shoot buds per explant, 1.0 mg L-1 2,4-D was superior to 0.1 mg L-1, but not different from either 0.3 or 3.0 mg L-1. Linear regression analysis of shoot buds per explant on 2,4-D level yields a significant quadratic relationship with a maximum between 1.0 and 3.0 mg L-1 (Figure 2-1). Thus, although ANOVA fails to indicate clear superiority of the 1.0 mg L-1 treatment, regression suggests that this treatment is close to the optimum 2,4-D level, and this is the concentration adopted for later work.
Consistency of regeneration response, as indicated by percent responding explants, is presented in the last column of Table 2-1. Although values in this





18

column span a wide range, no significant differences were found. Failure to detect statistically significant differences may be due to the low number of observations (10 per treatment) and the large number of nonresponding explants.
Direct regeneration. Direct regeneration from embryonic and seedling

explants without an intervening callus phase has been demonstrated in several legume species (Ammirato., 1983). In an effort to achieve this in desmodium, the induction treatments applied to callus in the previous experiment were applied to newly excised seedling hypocotyls of three genotypes. Results were disappointing. Genotype UF 20 produced small quantities of callus on all treatments, but showed no sign of regeneration. No growth whatsoever was seen in UF 144. At 2,4-D concentrations of 0.1 and 0.3 mg L-1, IRFL 6123 hypocotyls showed no growth. At the two higher 2,4-D concentrations IRFL 6123 produced roots from the radicle end of nearly every explant. When these were transferred to shoot elongation medium, tiny clusters of budlike structures less than 1 mm long formed at the end of the hypocotyl opposite the roots, but failed to develop further.
Effect of duration of shoot bud induction treatment. Since IRFL 6123 was the single responding genotype in the experiments described above, this experiment and the remainder of the chapter focus on optimizing protocols for this genotype. This experiment was designed not to enhance regeneration response, but rather to determine to what extent shoot bud induction period can be reduced without a corresponding reduction in regeneration.
Results presented in Table 2-2 show an increase in the mean number of shoot buds per explant through 14 days of induction and a leveling off thereafter. Despite the higher number of observations per treatment (21) than in the previous experiments, no significant differences were seen among the 3, 7,






19

14, and 28 day treatments with respect to percent responding explants. The zero day treatment yielded significantly fewer responding explants than all except the 7 day treatment. The zero day treatment is essentially a reversal transfer from high auxin, low cytokinin to low auxin, high cytokinin medium. Reversal transfer is a widely utilized method for inducing organogenesis. The rather poor response to this treatment relative to the 2,4-D treatments suggests that regeneration may be occurring via somatic embryogenesis rather than organogenesis. Histological evidence on this question will be presented later in this chapter.
The fact that a percentage of IRFL 6123 calli produce shoot buds under all treatments is evidence of this genotype's strong predisposition to regenerate. The failure of many of the explants to produce shoot buds under any of the treatments and the lack of significant differences among many of the treatments underscores the high level of variability often encountered in tissue culture work, and may also be an indication that we have not yet found-and may never find-the optimal regeneration protocol for this recalcitrant species.
Effect of BA and GA3 levels on shoot elongation. Both shoot meristem formation and shoot elongation are difficult to induce in desmodium. In previous work (Wofford et al., 1992b) as well as in the experiments described above, a large percentage of shoot buds failed to develop into plants, either due to poor elongation or failure to root. Three levels of BA were evaluated in the first of two experiments intended to improve elongation response. Gibberellic acid has been reported to stimulate development and elongation of shoot meristems in cultured soybean and pigeonpea (Ghazi et al., 1986; Kumar et al., 1983). To determine if this compound might be useful for desmodium, each BA level was tested both with and without the addition of 0.2 mg L-1 GA3.






20

Results are presented in Table 2-3. The two right-hand columns of this table represent two distinct types of elongation response. In the majority of cases shoot development ceased at the single leaf stage and at a length of one cm or less (Figure 2-2). These unifoliate shoots exhibited low vigor and could be rooted only with difficulty by transfer to basal medium supplemented with
1.0 mg L-1 indoleacetic acid (IAA). A smaller proportion of buds developed into multifoliate shoots (Figure 2-3). The more vigorous multifoliate shoots often spontaneously formed roots after two to three weeks in elongation medium, or could easily be induced to root upon transfer to basal medium lacking plant growth regulators.
Percentage of buds showing the unifoliate elongation response decreased with increasing BA level. In addition, a qualitative difference could be seen between the treatments with GA3 and without. In the presence of GA3, unifoliate shoots tended to bear abnormal, spatulate leaves (Figure 2-4). Because the degree of abnormality was variable and graded continuously into normal leaf development, abnormal and normal unifoliate elongation are not differentiated in Table 2-3.
Significantly more multifoliate shoots were produced on the medium
containing 2.0 mg L-1 BA and lacking GA3 than on the other treatments. The 21% multifoliate shoot production obtained with this treatment was nonetheless rather disappointing, and the following experiment was designed in an effort to improve on this figure.
Effect of picloram/BA ratio on shoot elongation. All of the treatments in the above experiment followed the precedent of Phillips and Collins (1980) and Wofford et al. (1992b) in utilizing the relatively low picloram concentration of
0.002 mg L-1. Since neither increasing BA level nor addition of GA3 produced dramatic improvement of elongation response, a second experiment was






21

performed to examine the effect of increasing picloram concentration and manipulating picloram/BA ratios. The most effective treatment from the previous experiment was included as a check. Results of this experiment are presented in Table 2-4. Just as in the previous experiment, unifoliate shoots substantially outnumbered the more vigorous multifoliate shoots. In contrast to the previous experiment, all of the leaves produced were relatively normal in morphology. No significant differences among treatments were observed for unifoliate shoot production, but the 0.012 mg L-1 picloram, 0.2 mg L-1 BA treatment yielded significantly more multifoliate shoots (38%) than the other treatments, including the best treatment from the previous experiment. It appears from these two experiments that when auxin level is insufficient, high cytokinin levels can serve a partial compensatory function in stimulating elongation. Superior elongation response can be obtained with a somewhat higher auxin level than that used in the original experiment in combination with fairly low level of cytokinin.

Histological examination of bud formation. As noted above, the method employed to induce regeneration in this study is consistent with an embryogenic pathway. The regeneration response appeared on a gross morphologic level to be organogenic. Histological examination revealed a mixture of these two regeneration mechanisms. At one and two weeks after transfer to induction medium, a low frequency of somatic embryos with clearly bipolar morphology and no vascular connections to the surrounding callus could be seen. These were mixed with a much higher frequency of clearly organogenic structures possessing a budlike morphology and definite vascular connections to surrounding callus. At three weeks and later stages embryos could no longer be found, while shoot buds at various stages of development were easily observed. Thus, it appears that regeneration in this system is predominantly via organogenesis.






22

Table 2-1. Response of genotype IRFL 6123 to 2,4-D and kinetin levels in the shoot bud induction medium.

2,4-D Kinetin Mean number Mean percent
conc. conc. of buds per responding
-------------mg L1 ---------- explant t explants, %t

0.1 0 0.5 b 30 a 0.3 0 2.2 ab 70 a 1.0 0 3.3 a 80 a 3.0 0 1.6 ab 50 a 0.1 0.15 0 c 0 b 1.0 0.15 0 c 0 b



t Mean separation by Tukey's honestly significant difference (HSD a=0.05).






23

Table 2-2. Response of genotype IRFL 6123 to duration of shoot bud induction
treatment.


Length of induction Mean number Mean percent
period, days of buds per responding explant t explants, % t

0 0.43 c 28 b 3 1.04 b 67 a
7 1.14 b 38 ab
14 2.52 a 57 a 28 2.50 a 61 a


t Mean separation by Tukey's HSD (a=0.05).






24

Table 2-3. Response of genotype IRFL 6123 to benzyladenine and
gibberellic acid levels in the shoot elongation medium.


BA GA3 Mean frequency responding explants t conc. conc. Unifoliate Multifoliate
----------mg L-1......................... ----------------------%----------------0.2 0 50 a 0 a 0.2 0.2 54 a 0 a 0.6 0 33 a 4 a 0.6 0.2 38 a 0 a 2.0 0 13 b 21 b 2.0 0.2 13 b 0 a



t Mean separation by Tukey's HSD (a=0.05).






25

Table 2-4. Response of genotype IRFL 6123 to picloram and benzyladenine
levels in the shoot elongation medium.


Picloram BA Mean frequency responding explants t conc. conc. Unifoliate Multifoliate
----------mg L-1 ____ ---------------------- % ----------------0.012 0.2 25 a 38 a 0.012 0.6 38 a 8 b 0.04 0.2 33 a 4 b 0.04 0.6 29 a 4 b 0.002 2.0 17 a 17 b



t Mean separation by Tukey's HSD (a=0.05).







26














10

y = -1.697x 2 + 5.405x + 0.333 r 2 0.302 8




0
*


4- *




Is IlI






0 1 2 3 4 2,4-D Concentration, mg/L















Figure 2-1. Response of genotype IRFL 6123 to 2,4-D concentration in shoot
induction medium.






27












































Figure 2-2. Unifoliate shoots produced by genotype IRFL 6123.





28













































Figure 2-3. Multifoliate shoots produced by genotype IRFL 6123.





29












































Figure 2-4. Spatulate shoots produced by genotype IRFL 6123 in the presence
of gibberellic acid (GA3).











CHAPTER 3
INHERITANCE OF IN VITRO REGENERATION AND ASSOCIATED CHARACTERS


The high genotype-specificity of in vitro regeneration in desmodium has been clearly demonstrated in the previous chapter. An understanding of the genetic control of regeneration is desirable for the efficient application of biotechnological methods to this crop. The mode of inheritance of regeneration determines whether it is feasible to transfer the trait to agronomically desirable nonregenerating lines and, if so, the appropriate method of accomplishing this.
The genetic basis of regeneration and associated in vitro traits has been studied in few higher plant species. In most studies regeneration was accomplished through organogenesis, and organogenesis is assumed in the current discussion unless otherwise stated. Regardless of whether organogenic or embryogenic response was examined, regeneration in most species was treated as a quantitatively inherited trait, although evidence supporting this assumption was often not presented. Several studies have been directed at partitioning genetic variance components and determining heritabilities for the regeneration trait in different crops. Buiatti et al. (1974) performed a diallel analysis, without reciprocals, of callus growth and shoot regeneration from flower petal explants in cauliflower. Additive gene effects were high for both traits, and narrow sense heritability estimates were 0.81 for callus growth and 0.09 for percent explants forming shoots. The low heritability for the latter trait resulted from extremely high epistatic and error effects. Based on North Carolina Design II matings among twenty four red clover genotypes, Keyes et al. (1980) determined that embryogenic regeneration frequency from

30






31

seedling hypocotyl-derived callus was subject to primarily additive genetic effects. Narrow sense heritabilities ranged from 0.25 to 0.54, depending on culture medium. Analysis of a complete diallel in tomato (Frankenberger et al., 1981) revealed that number of shoots produced per leaf disc explant was controlled predominantly by additive genetic effects, and the narrow sense heritability for this trait was estimated to be 0.98. A similar study in pigeonpea (Kumar et al., 1985) showed that callus growth was controlled largely by additive gene effects, while the number of shoots regenerated per cotyledon explant was subject to primarily nonadditive effects. Komatsuda et al. (1990) performed a detailed diallel analysis of callus proliferation and shoot regeneration from immature embryo explants in barley. Narrow sense heritabilities were approximately 0.7 for both traits. Significant dominant effects were detected for both traits, and significant epistatic effects were observed for shoot regeneration.
Evidence for qualitative inheritance of regeneration capacity has been reported in tomato, alfalfa and petunia. Koornneef et al. (1987) examined regeneration from leaf disc-derived callus in the progeny of a cross between tomato (Lycopersicon esculentum) and L. peruvianum. Segregation ratios in F2, F3, and backcross generations indicated that the trait was controlled by two dominant, complementary loci contributed by L. peruvianum.
Bingham et al. (1975) found that the frequency of regeneration from
hypocotyl-derived callus in alfalfa could be increased from 12% to 67% in two cycles of recurrent phenotypic selection. Reisch and Bingham (1980) examined several F1, F2, and backcross populations produced from crosses between regenerating and nonregenerating clones of diploid alfalfa. Segregation ratios were consistent with a two gene model in which both genes were dominant and the presence of the dominant allele at either locus resulted in low frequency






32


regeneration, while the presence of both dominant alleles yielded a high regeneration frequency. Crosses between two different sets of nonregenerating clones yielded no regenerating progeny, suggesting that inheritance was not quantitative.
Wan et al. (1988) examined the genetics of regeneration from petiole
explants in tetraploid alfalfa, employing a different culture protocol from that of Reisch and Bingham. Based on data from several F1 and S1 populations, it was concluded that regeneration was controlled by a two gene system distinct from that described by Reisch and Bingham. Duplicate recessive epistasis was postulated; that is, the presence of dominant alleles at both loci was necessary for any regeneration to occur. Variation in the frequency of regeneration among regenerating clones was attributed to either dosage effects or the presence of modifier genes.
Dulieu (1991) obtained rather tentative evidence of single gene control of regeneration from hypocotyl explants in petunia. Two backcross populations yielded approximately 1:3 ratios of "high regenerating" to combined "low regenerating" and "nonregenerating" plants, but the distinction between high and low level regeneration was somewhat arbitrary, and quantitative inheritance could not be ruled out.
All of the studies discussed above have dealt with regeneration from
sporophytic (2n) tissue. The process of androgenesis, or embryogenesis from tissue derived from gametophytic (n) pollen mother cells, is markedly distinct from regeneration from sporophytic tissue. However, the two processes are sufficiently similar that work conducted on the genetics of androgenesis may provide insight into the genetics of regeneration from 2n tissue. Several papers have been published on the genetic control of androgenesis in cereals, and all






33

have treated both callus growth and regeneration frequency as quantitatively inherited traits. Diallel analysis of androgenesis in both wheat (Lazar et al., 1984) and triticale (Charmet and Bernard, 1984) revealed highly significant GCA, SCA, and reciprocal effects. In both cases, the GCA effect was predominant, and the wheat study reported a narrow sense heritability of 0.6 for plantlet formation. Earlier work involving a different set of wheat lines (Bullock et al., 1982) found no significant reciprocal effects, underscoring the fact that the results of any genetic study are highly specific to the germplasm under consideration. Two studies conducted on the inheritance of androgenic ability in barley yielded similarly disparate results. A study involving inbred lines and reciprocal F1 hybrids (Foroughti-Wehr et al., 1982) found significant differences among reciprocal crosses, while a similar study (Dunwell et al., 1987) involving some of the same germplasm found no reciprocal differences. The latter work included F2 and backcross generations and concluded that plantlet production from anther culture was highly complex genetically, with large epistatic effects. The work of Dunwell et al. (1987) also points out the importance of clearly defining the response variable. When response was defined as number of green plants produced per anther (the most common practice), the dominance effect was large and positive, with no significant additive effect detected. When response was expressed as percent responding anthers, however, the dominance effect was negative and a large, positive additive effect was seen. This study also reported that variation among different spikes on the same plant exceeded variation due to genetic factors. This suggests that physiological status of the anther can be a major determinant of capacity for androgenesis. A diallel analysis in rice (Quimio and Zapata, 1990) found that callus production and number of regenerated plants per anther were both influenced primarily by additive genetic effects, with no significant reciprocal effects. Variance (Vr) and






34


covariance (Wr) analysis indicated that regeneration was largely controlled by partially recessive genes.
The limited number of studies that have addressed phenotypic or genetic correlations between in vitro regeneration and various in vitro and in planta characters have yielded both expected and unexpected results. Oelck and Scheider (1983) found that in Melilotus officinalis, Trifolium pratense, and T. resupinatum, the ability to form shoots from callus was correlated with the production of side shoots from in vitro shoot tip cultures. Kern et al. (1986) observed a similar correlation between capacity for somatic embryogenesis and in vitro axillary shoot development in soybean. These correlations may reflect a generalized tendency among responding genotypes to form shoots, or may simply indicate a tolerance to the specific culture conditions employed in the study. It has been observed in alfalfa (Bingham et al., 1975; Brown and Atanassov, 1985) and alyceclover (Wofford et al., 1992a) that regenerating genotypes often have a creeping growth habit and readily produce adventitious shoots. Thus it appears that in at least some cases the ability to regenerate in vitro reflects a general proclivity for shoot production both in vitro and in planta. Oelck and Scheider (1983) suggested that a general tendency to produce adventitious shoots may be a useful indicator in preliminary screenings of germplasm for regenerating lines.
Perhaps surprisingly, there appears to be little correlation between callus growth rate and regeneration frequency ( Baroncelli et al., 1974; Kumar et al., 1985; Lazar et al., 1984). However, shoot regeneration from callus can be restricted in genotypes with very poor callus growth (Bingham et al., 1975). It is well established that callus appearance frequently bears a strong relationship to regeneration potential (Bingham et al., 1975; Ketchum et al., 1987; Delieu






35

1991), but due to the difficulty of quantifying callus appearance the relationship has not been subjected to statistical analysis.
The primary objective of the work described in this chapter was to

determine the mode of genetic control of in vitro regeneration in desmodium. A secondary objective was investigation of the mode of inheritance of callus growth and examination of the relationship between callus growth and regeneration.


Materials and Methods


Production of hybrids. Two regenerating and three nonregenerating lines were selected for crossing. The regenerators, IRFL 6123 and IRFL 6128 are classified as D. heterocarpon ssp. angustifolium, and are morphologically distinguishable from the nonregenerating genotypes by their lanceolate, coriacious leaves, upright growth habit, elongated racemes, and glabrous seed pods. Lines 6123 and 6128 are very similar except for a distinct leaf mark on 6123 that is absent on 6128. All of the selected nonregenerators-UF 20, UF 144, and CIAT 13083-have ovate or obovate, noncoriacious leaves, spreading growth habits, compact racemes, and pubescent pods. Genotype UF 20 ('Florida' carpon) represents D. heterocarpon ssp. heterocarpon and is distinguished by its thin, slightly pubescent leaves and prominent leaf marks. Genotype UF 144 is classified as D. ovalifolium, and possesses thicker, glabrous leaves. Genotype CIAT 13083 is intermediate between D. heterocarpon and D. ovalifolium.
Parent plants were moved from the field to the greenhouse in April, 1989. To induce flowering, a ten hour day length was simulated by covering the plants with a tarpaulin from approximately 7:00 P.M. until 9:00 A.M. Pollinations were






36

done in June and July. Donor pollen was obtained by tripping flowers using a toothpick with a small piece of fine sandpaper glued to one end in such a way that the anthers and stigma would strike the sandpaper and pollen would adhere. Pollen was then transferred to recipient flowers by tripping in a similar manner. Recipient flowers were not emasculated because the flowers were extremely sensitive to handling, and emasculation tended to result in flower abscission. Pollinations were carried out at a time of day when the flowers had fully opened, but self-tripping had not yet occurred. The specific time varied according to night temperature and daytime cloud cover, but was generally between 9:00 A.M. and noon. At least one hundred flowers for each possible combination of regenerator and nonregenerator parent, including reciprocals, were pollinated in this manner.
Seeds were harvested in August,1989 and germinated and planted in the greenhouse the following winter. Hybrids were clearly identifiable based on the morphological characteristics described above. Flowering of hybrids was induced in the summer of 1990. Plants were allowed to self-pollinate, but flowers were hand tripped to increase seed set.
Evaluation of callus growth and regeneration in the parental, F2 and F
generations. The F2 seed harvested from the original hybrids was scarified and germinated on SGL medium (Collins and Phillips, 1982) as described in Chapter 2. Hypocotyls were excised and placed onto callus induction medium and epicotyls were returned to SGL medium for rooting. Rooted epicotyls were transferred to potting soil after ten days and grown to maturity in the greenhouse where they were allowed to self-pollinate to produce the F3 generation.
Hypocotyls from both F2 and F3 populations were cultured according to the optimal protocol established in Chapter 2. Initial callus production was on L2 medium supplemented with 0.06 mg L-1 picloram and 0.2 mg L-1 BA for






37

28 days. Shoot buds were induced on L2 with 1.0 mg L-1 2,4-D and 2.0 mg L-1 adenine for 14 days, and shoot elongation was induced on L2 with
0.012 mg L-1 picloram and 0.2 mg L-1 BA for 28 days. Shoots were rooted on L2 lacking plant growth regulators. Hypocotyls from parental lines were cultured simultaneously with the F2 and F3 populations to serve as checks.
Callus fresh weights were measured at the time of transfer from callus medium to bud induction medium. Each callus was visually scored for shoot bud formation and elongation at the end of both the bud induction and bud elongation steps. The scoring system and statistical transformations used will be discussed in the results.

Statistical analysis. General statistical analyses were conducted using JMP Version 3.0, the SAS Institute general statistics product for Macintosh computers (SAS Institute, 1994). Means and variances of truncated data sets were estimated using UNCENSOR Version 3.0 (Newman and Dixon, 1989).


Results and Discussion


Results of crossing efforts were disappointing. The approximately 2200
flowers that were cross-pollinated produced only 590 seeds. Poor seedset was likely due in part to inadequate greenhouse ventilation, which resulted in occasional daytime temperatures sometimes exceeding 380 C. Hybrid yield was further reduced by the low ratio of hybrids to selfs in the seeds that were produced. Approximately 550 of the 590 seeds germinated, but only eight of the resulting plants were identified as hybrids. Hybrids could be readily identified by leaf shape and growth habit, both of which were intermediate between the parental types. Two hybrid plants were progeny of the cross IRFL 6123 x CIAT 13083 (hereafter referred to as cross 501), three were from cross





38

IRFL 6123 x UF 144 (cross 507), and three were from CIAT 13083 x IRFL 6128 (cross 510). Since genotypes 6123 and 6128 are nearly identical, both in gross morphology and in vitro performance, crosses 501 and 510 are essentially reciprocals.
Hybrids were selfed through the F3 generation. Table 3-1 summarizes numbers of individuals and number of families from which callus growth and regeneration data were collected. In the case of cross 510, data collection was terminated at the F2 generation. Severe morphological abnormalities appeared in the F1 of this cross, and persisted through F3. The three F1 seedlings appeared normal through approximately 6 weeks of age, after which new growth showed severely stunted leaves and internodes. Stunting was evident from the early seedling stage in many F2 and F3 individuals, and appeared at late seedling stages in all individuals. Stunting was accompanied by extremely poor flower production and seedset. The F2 of cross 510 also exhibited a depressed callus growth rate relative to the two other crosses and to the parental lines with a high incidence of callus necrosis and death. Since cross 510 is essentially the reciprocal of cross 501, the abnormalities may be the result of interactions between the cytoplasm of CIAT 13083 and nuclear genes of IRFL 6127. (Evidence of maternal effects in the other crosses will be discussed later in this chapter.) Whatever the cause, the observed abnormalities have the potential effect of masking expression of the in vitro characters of interest, as well as exerting confounding selective pressures in the F2 and F3 generations. Therefore, genetic analysis was not attempted on cross 510, and the remainder of this chapter deals exclusively with crosses 501 and 507.
Scaling of the callus growth trait. The logical first step in genetic analysis of any trait is to identify an appropriate numerical scale; that is, to determine if






39

mathematical scale transformation is necessary. This task can be approached in at least two general ways. A purely statistical approach, described in any number of statistics texts (Little and Hills, 1978; Zar, 1984), seeks a scale that yields a data structure satisfying the general assumptions of parametric statistics-in this case, normally distributed error terms with error variances independent of means. A more genetically oriented (and somewhat more complex) approach is not directly concerned with statistical assumptions, but stresses instead whether a scale will facilitate partitioning of genetic variation into underlying genetic causes, or components; e.g., additive, dominant, and epistatic gene action (Mather and Jinks, 1971). Implicit in this second approach is the statistical assumption that factor levels act in an additive fashion. Both approaches will be presented below.
Three scales were examined in the present study. The simplest is the untransformed linear scale, which in this case is callus weight in grams. This scale possesses the virtue of allowing various statistics and parameters to be reported in grams, permitting easy interpretation. A logarithmic scale was also investigated. While this scale results in somewhat non-intuitive statistical outputs, the scale is theoretically more applicable to analysis of growth variables than is a linear scale. The logarithmic scale asks the question, how many times has the cell mass doubled, while the linear scale asks how many mass units have been added to the starting mass. The logarithmic transformation is frequently used to stabilize variances in cases where standard deviation is proportional to mean in the untransformed scale. The third scale tested was the square root transformation. This scale is intermediate between the linear and logarithmic scales in terms of skewing of distributions, and is useful when variance is proportional to mean.





40


To examine the effect of the three scales on normality, the scales were applied to populations of each of the three parental lines and to F2 and F3 populations. The parental populations are theoretically normally distributed, while the F2 and F3 populations are expected to be normally distributed if little or no unidirectional dominant or epistatic gene action is present. The later generation populations are included here because, as a result of a temporarily limited seed supply, the size of the tested parental populations was quite small (20 5 n 36). Univariate statistics for the various populations are presented in Table 3-2. The linear scale tends to produce rightward (positive) skewing, while the log transformed populations are moderately left-skewed (negative), and the square root transformed populations are slightly left-skewed. The Shapiro-Wilk W statistic is used to test for normality in populations as small as ten observations; a low probability associated with a calculated value of W indicates a significant deviation from normality (SAS Institute, 1989; Gilbert, 1987). None of the W values for the parental populations indicated deviation from normality.
Results for F2 and F3 populations are similar to those for the parental populations, but are somewhat clearer. The linear scaling results in gross leftward skewing and yields highly significant W statistics. This may be attributable to genetic effects if there is a strong preponderance of negatively acting dominant genes in these populations. Strong, unidirectional epistatic effects, which could result in a relatively small proportion of the populations possessing the necessary combination of alleles for rapid callus growth, is another possible genetic cause for the observed skewing. However, the fact that the leftward skewing occurs in all populations can be interpreted to indicate that the linear scale is inadequate and that with proper scaling the F2 and F3






41

will in fact be approximately normally distributed. Further evidence regarding this hypothesis will be presented later in this chapter.
Skewness and W statistics for the F2 and F3 populations fail to clearly
indicate whether the logarithmic or the square root transformation yields a more normal distribution. As in the case of the parental populations, the log transformed populations are skewed to the right, while the square root transformed populations are skewed to the left. In the case of F3 populations, the magnitude of skewing is substantially less for the logarithmic than for the square root scales. The only non-significant W statistic was obtained from the square root transformed cross 507 F2 population.
If a variable is properly scaled, populations with high means should, in
general, have variances no greater than those of populations with lower means. This can be tested by regressing F3 family means versus variances, as presented in Figure 3-1. The linear scale exhibits a strong, positive relationship between mean and variance. Logarithmic transformation results in a weak, negative relationship, and the square root transformation produces a weak, positive relationship. It is feasible that the expectation of independence of means and variances can be confounded by genetic effects, since different F3 families may possess differing amounts of genetic variance. For example, in the not unlikely situation that a large, positive dominance effect exists, families with high means will be those whose F2 parents had high levels of heterozygosity for genes controlling callus growth. These families will also exhibit elevated variances due to the presence of segregating genes. This genetic effect can only be large, however, if total genetic variance is much greater than environmental variance. It will be shown later in this chapter that this is not the case. Therefore, the strong relationship between means and variances for F3 families provides further evidence of the inadequacy of the linear scale.






42

Because only three parental populations were examined, meaningful regression of mean versus variance for parental populations is not possible. However, simple examination of parental means and variances (Table 3-2) can shed further light on scale effects. If variance is independent of mean, the variances of the three parental populations should be similar in spite of the differences in means. The effects are similar to those seen in the F3 populations, but the square root transformation stands out in that it yields a smaller relative range in variances (maximum variance 48% greater than minimum) than either the linear (177%) or logarithmic (78%) scales.
In summary, a general statistical examination of the callus growth trait
reveals the clear inadequacies of linear scaling. Logarithmic and square root scales are similar in their conformity to the statistical requirements of normality and independent variance.
Joint scaling test for callus growth. A more genetically oriented approach to scaling was developed by Cavalli (1952) and elaborated upon by Mather and Jinks (1971). Known as the joint scaling test, in its simplest form this technique involves estimating the effects of additive and dominant gene action on generation means. This is done by establishing a series of simultaneous equations, one for each generation for which data are available. The left side of each equation is the observed generation mean. The right side consists of three terms, each of which is composed of an unknown and an associated coefficient. The unknowns are model parameters: the grand mean (essentially the genetic-neutral component of the various generation means, designated as m), additive genetic effect (designated [d]), and dominant genetic effect (designated [h]). Coefficients correspond to the theoretical causal contributions made by these parameters to each generation mean. Since there are three unknowns, data from at least four generations (each parent represents one





43


generation) are needed to obtain a non-singular solution and a measurement of statistical significance of the solution. Each equation is weighted in proportion to the level of certainty associated with each generation mean, which is equivalent to the reciprocal of the variance of the mean. As originally described, a solution is obtained through cumbersome matrix manipulations. This is computationally identical to performing weighted least-squares regression using the observed means as the dependent variable and the model coefficients as independent variables (Rowe and Alexander, 1980). The partial regression coefficients obtained in this way are equivalent to the unknowns in the above-described equations. The adequacy of solutions obtained by this method is assessed in two ways. First, each predicted model parameter has an associated F test for the hypothesis that the parameter differs from zero. Secondly, the whole model can be tested by examining the error sum of squares (SSE). In this particular form of analysis, SSE exhibits a X2 distribution, and the reported SSE can be compared to tabular x2 values to determine goodness of fit (Rowe and Alexander, 1980).
Six joint scaling tests were conducted in order to test each of the three scales using data from each cross. Model coefficients used to construct the tests are given in Table 3-3. Observed means and weights used in the tests are are based on the means and variances presented in Table 3-2. Predicted values of m, [d], and [h], and the results of goodness of fit testing are given in Table 3-4. Adequate goodness of fit is obtained only for the linear and square root scales and only for cross 501. In both cases, a large, negative dominance effect is indicated.
The joint scaling test suffers from limitations that necessitate cautious
interpretation of results. One serious limitation is that the test assumes that one parent possesses most of the positively acting alleles and the other possesses






44

the majority of negatively acting alleles, and that dominance acts primarily in a single direction. Judging from the large difference in callus growth rates among the parents in the current work, the first assumption is probably at least partially valid. The validity of the second assumption is difficult to assess. A second problem with the joint scaling test is that two separate hypotheses are simultaneously tested: that the scale is adequate, and that the genetics of the trait in question can be adequately described by an additive-dominant genetic model. When an adequate goodness of fit is not obtained, the joint scaling test cannot by itself distinguish which of these hypotheses is incorrect. It is possible to partially solve this problem by adapting the test to more complex genetic models involving epistatic and perhaps other types of genetic effects. Unfortunately, more complex models require data from more generations and are dependant on an increasing number of rather dubious assumptions about gene interactions (Mather and Jinks, 1971). As a result of these limitations, spurious positive results are quite possible, particularly when the test involves only a single degree of freedom. Considering the statistical evidence for the inadequacy of the linear scale, it is likely that the good fit obtained when this scale was applied to cross 501 is in fact spurious.
Examination of the data presented in Table 3-2 suggests a factor that may confound the joint scaling test in the present case. In both crosses, and for all scales, the F2 and F3 means are significantly lower than the mid-parent mean, and are fairly close to one another. In the case of cross 501, the F3 mean is higher than the F2 mean, while in cross 507 the F3 mean is lower. If a simple additive-dominant model is assumed, the deviation of the F2 and F3 from the mid-parent mean suggests a large, negative dominance effect, just as indicated by the joint scaling test. However, the relationship between F2 and F3 means suggests a small dominance effect-a negative effect for cross 501, and a






45

positive effect for cross 507. The discrepancy between the dominance effect predicted by the relationship of the progeny means to the mid-parent mean and that predicted by the relationship between F2 and F3 means accounts for the generally poor joint scaling test results. These observations suggest that some factor not considered in the model may be acting to depress both F2 and F3 means relative to parental means. This factor could be a maternal or cytoplasmic effect-nuclear genes from parents 144 and 13083 may interact in a negative manner with cytoplasmic genes from 6123 (the female parent in both crosses). The depressed callus growth observed in cross 510 is consistent with this explanation. An alternative cause would be negative epistatic interactions between nuclear genes originating from the different parents. If either explanation is correct, then the results of the additive-dominant joint scaling tests are invalid. Unfortunately, there are insufficient generations available to allow construction of an additive-dominant-maternal or additive-dominantepistatic joint scaling test.
Variance partitioning for the callus growth trait. Variance partitioning methods are not subject to all of the limitations described for the joint scaling test. Specifically, variance partitioning is not affected by matemal factors that affect means, by distribution of positive-acting alleles between parents, nor by dominance acting in different directions at different loci. A sophisticated method of deriving genetic variance parameters from observed variances and covariances in segregating generations derived from crosses of inbred lines is presented by Mather and Jinks (1971). The technique is similar in many ways to the joint scaling test, but observed variances and theoretical causal variance components are utilized to construct the model. As with the joint scaling test, fairly complex genetic models are possible, but, due to data limitations, the present discussion will be limited to a simple additive-dominant model. In the






46

present case, five observational variances or covariances are available: pooled parental variance (E1), F2 variance (V1 F2), among-family F3 variance (V1F3), within-family F3 variance (V2F3), and F2-F3 covariance (W1 F23). Among- and within-family F3 variances are obtained by analysis of variance using F2 parent as the grouping variable, followed by partitioning of variances based on expected mean squares (Appendix A). The F2-F3 covariance is the covariance between F2 parents and F3 family means.
The causal variance components described by Mather and Jinks are
somewhat complex. In addition to additive and dominant genetic components (designated D and H, respectively), two types of environmental variance and a so-called sampling variance are described. The first environmental variance
(Ew) is simply the error variance among individuals. The second (EB) is the error among plots or rearing environments and is applicable only to V1 F3 in the present study. Since explants from the different F3 families were randomly distributed among petri dishes, which were randomly distributed within a single incubator, there is no intrinsic "plot" error. Instead, EB is equivalent to EW divided by the harmonic mean of the number of individuals per F3 family (Mather and Jinks; 1971). Since EB is a function of EW, it can be combined into EW, thereby eliminating a column from the model and gaining a degree of freedom. The sampling variance applies only to among-group variances. It is similar in concept to EB, but is a function of within-group genetic variances rather than error variances. Therefore, in the case of V1 F3 the sampling variance is equivalent to V2F3 divided by the harmonic mean of the number of individuals per F3 family. Sampling variance can be algebraically included into the model row for V1 F3 and does not have an associated model column. Derivation of model matrices is described in detail in Appendix B.






47

The method of Mather and Jinks employs an iterative approach for
determining model weights. Weights for each of the matrix rows are initially established as the reciprocals of the theoretical variances of each observed variance or covariance. The variance of a variance is equal to 2(V2)/df, where V is the observed variance and df is the number of degrees of freedom on which the observed variance is based. The variance of a covariance is equal to (W2 + V1V2)/df, where W is the observed covariance, V1 and V2 are the variances of the populations from which the covariance was obtained, and df is the degrees of freedom upon which the covariance is based. The model is then solved, and predicted (Mather and Jinks use the term expected) observational variances are obtained. These are used to derive new weigh estimates, and the model is solved again. This is continued until no improvement in model fit is obtained.

Coefficients used in variance partitioning for the callus growth trait are

presented in Table 3-5. Observational variance components and initial weights are presented in Table 3-6. Results of the first iterations are given in Table 3-7. In all cases, second and third iterations failed to improve model fits. (Results of later iterations are presented in Appendix C.) The linear and logarithmic scales failed to produce significant whole-model F tests for either cross. The square root scale produced significant whole-model tests for both crosses. For cross 501, estimated additive genetic variance was large, and significantly different from zero (assuming a rather generous a of 0.1), while dominant variance was smaller and not significantly different from zero. Cross 507 yielded no significant genetic variance estimates.
The analysis presented thus far gives a rather unclear and somewhat
contradictory picture of the genetic mechanism controlling the callus growth trait in the two crosses. Both the joint scaling and variance partitioning approaches






48


indicate that cross 507 cannot be adequately described by an additivedominant genetic model. Based on the joint scaling test, the genetics of cross 501 may be characterized by a large, negative dominance effect and a moderate additive effect. An altemate interpretation is that large, negative matemal or epistatic effects exist in both crosses. In contrast, variance partitioning indicates a large additive genetic variance and a much smaller dominant variance in cross 501.
A general shortcoming of the types of analysis presented to this point is the large number of simplifying genetic assumptions required. The results derived from variance partitioning may be more reliable than those of the joint scaling test, due to the fewer genetic assumptions required by this method. Many assumptions, such as absence of linkage and equal magnitude of effect from each contributing locus, are common to both techniques. The generally poor model fits obtained by both methods may be due to failure of the experimental material to conform to these assumptions, or to the presence of significant nonadditive, nondomininant genetic effects. It is likely that experimental error also contributes to poor model fit.
Parent-offspring regression is a simpler, more empirical method of analysis, relatively free of genetic assumptions. Regressions of F3 family means on F2 parents are presented in Figure 3-2. The slopes (regresion coefficients) represent, by definition, heritabilities. It should be noted, however, that in the case of selfing, parent-offspring regression is in part a function of dominant and epistatic genetic effects, but these effects play a much smaller role in the regression than does additive genetic variance. Regression coefficients are remarkably high, regardless of cross or scale, ranging from 0.524 to 0.769. The practical implication of this is that substantial progress





49

should be possible in selecting for callus growth among segregating generations of this material.
The nature of the regeneration variable. The callus growth trait was amenable to quantitative analysis due to the fact that this trait could be expressed in terms of a continuous, metric variable for each individual plant. This was not the case for the regeneration trait. In an ideal study of the genetics of regeneration, each plant can be evaluated in terms of a metric variable such as percent responding explants or number of shoots per explant. Because hypocotyl explants had to be used in the present study, only one explant existed per plant, so the data could not be expressed as percent responding explants. Expression of regeneration as number of shoots per explant was impossible because the majority of explants produced no buds or shoots. Instead, the trait was expressed on a visual rating scale of one through seven. A rating of one represented no evidence of regeneration. Two represented a slight indication of regeneration in the form of localized deep green coloration. Three represented formation of a single, well-defined but nonelongated bud. Four indicated multiple bud formation, usually with some elongation. A rating of five indicated one or two well elongated shoots. Six indicated between three and five elongated shoots, and a rating of seven indicated more than five elongated shoots.

A significant number of individuals in all populations showed no sign of
regeneration and received scores of one. All individuals in the nonregenerating parent populations (genotypes 144 and 13083) received scores of one. This type of data structure can be described as "threshold" or "censored" data. Falconer (1981) discussed threshold traits at length. He assumed that there exists an underlying, continuous scale of proclivity (or as Falconer describes it, "liability") towards a certain condition-in the case of the present work, tendency





50


to regenerate-that can only be detected when the level of proclivity rises above a given threshold.
Distributions of this type cannot be rendered normal by any mathematical transformation, since we lack any information about differences among those individuals that fall below the observable threshold, and no transformation can restore this missing information. For example, no matter what transformation is used, the two nonregenerating parental populations will always have equal means, and variances of zero. Fortunately, useful analysis can still be performed on this type of data, provided it can be assumed that the populations under consideration are normally distributed.
In Falconer's treatment, threshold traits can be manifested at either two or three discrete levels, or classes. This can be done with the present data by collapsing the rating scale into two (nonregeneration versus regeneration), or three (nonregeneration versus partial regeneration versus full regeneration) classes. While little useful analysis can be done when only two classes are present, meaningful analysis is possible in situations with three classes, contingent on the assumption that the underlying variable is normally distributed. For any population, if the percentage of individuals falling into each of three threshold classes is known, then the mean and variance of the population, expressed in threshold units, can be deduced. The principle is illustrated in Figure 3-3, adapted from Falconer (1981), but based on actual F3 data. The classes are separated by the two vertical lines on each graph, nonregenerators lying to the left of both lines, partial regenerators lying between the lines, and full regenerators lying to the right. While both populations contain 68% full regenerators, population (a) contains more partial regenerators and fewer nonregenerators than population (b). As a result,






51

population (a) has a lower standard deviation and lower mean than population (b).
An attractive advantage of analyzing data based on three ordinal classes with two thresholds is that scaling problems are avoided. This is because only one real unit exists on the x axis-the unit separating the two thresholds. Values either to the left or to the right of both thresholds can never be experimentally measured, and are simply extrapolations of the central threshold unit. A serious disadvantage of the three class approach is that populations that do not contain at least one observation in each of the three classes cannot be analyzed. Many of the F3 families in the present study contain no fully regenerating individuals, and therefore cannot be analyzed using this approach.
While the methods described by Falconer for analysis of threshold traits
cannot be directly applied to data with more than three classes of observations, it is possible to analyze data with an unlimited number of classes by applying techniques developed for censored data sets. Censored data are similar to threshold data in that both data types are subject to a threshold below which no information is known (Schneider, 1986). The difference is that in censored data there is only one threshold, and data above this threshold are of a continuous, or at least semicontinuous nature. In both cases, data lying above the detection threshold are used to make inferences about data below the threshold; therefore, assumptions of normality are necessary in both cases. By treating the seven-category regeneration rating scale as a quasi-continuous variable, the data collected in the present study are amenable to analysis as censored data.
Several statistical techniques have been developed for "uncensoring"
censored data sets, and, as in many fields of statistics, the relative strengths and






52

weaknesses of each are subject to considerable debate among statisticians. The most intuitively appealing method is termed regression order statistics. This method is graphically illustrated in Figure 3-4. Figure 3-4 (a) is a normal probability plot of a hypothetical, approximately normally distributed population. The Y-axis is observed value of the variable of interest, and the X-axis is normal quantile score (a transformed Z-score scale). The points represent the individual observations, and the line represents the simple linear regression line obtained from the individual data points. The slope and intercept of this line can be used to approximate the variance and mean of the population. Figure 3-4 (b) is a similar plot of the same population with a censoring threshold applied such that approximately 30% of the observations yield no observable response. These observations, of course, all have the same values for both axes, and therefore all lie atop one another at the lowest left data point on the graph. The result is a truncated plot. The regression line is derived only from those observations lying above the censoring threshold, but is similar to the line obtained from the uncensored population, and would yield variance and mean estimates close to the actual parameters of the original population. The regression order statistics method is a numeric analog of Figure 3-4 (b). That is, it performs a normal quantile regression on a truncated population in order to approximate the mean and variance of the intact population.
The accuracy of population parameters estimated by uncensoring through regression order analysis is limited by the number of observations lying above the censoring threshold. However, reasonably accurate results can be obtained even when the majority of observations are below the threshold, as long as several observations lie above the threshold. Regression order analysis was used to estimate means and variances of all populations in the current study that had at least four individuals with regeneration ratings higher






53

than one. Although this is a less than ideal approach, there appears to be no acceptable alternative. The option of simply analyzing the raw regeneration scores-either transformed or otherwise-is undesirable because this entails accepting the clearly incorrect assumption that all individuals receiving regeneration scores of one have the same proclivity for regeneration. The contrasting options, then, are to ignore the left-hand tails of many of the population distributions, or to use an acknowledgedly problematic method of reconstructing these tails. Means and variances determined by uncensoring, as well as means and variances of the raw data are presented in Table 3-8. In general, means derived from regression order analysis are lower than those obtained from the raw data, and variances are higher. This is not unexpected, since truncation of the low end of a data set-as occurs in the raw data-will artificially raise the mean and lower the variance. For the raw data, a significant (a=0.05), positive relationship exists between F3 family variance and family mean, due at least in part to the fact that the lower the mean, the more severely truncated the distribution. This relationship is absent from the uncensored data because the truncated lower ends of the distributions have been reconstructed. The unusually high variances observed for some F3 families may be due to segregation of major genes within these families, resulting in a somewhat bimodal distribution and exaggerated variance. This effect is most pronounced in the uncensored data. Unfortunately, population sizes for the F3 families are too small to clearly distinguish between spurious bimodality and bimodality resulting from genetic causes. Bimodality was not observed in the F2 populations, nor in the complete F3 populations.
The accuracy of the variance estimates presented in Table 3-8 can be compromised by non-normality resulting from inadequate scaling. Major scaling problems can be detected by examining the distribution of a population






54

containing no genetic variance (since genetic factors can also cause deviations from normality) and having a mean significantly higher than the censoring threshold. The regenerating parental population (genotype 6123) is such a population, and a histogram of this population is presented in Figure 3-5. The distribution is not seriously skewed, nor is it bimodal, suggesting that there is little scale-induced non-normality.
Genetic analysis of the regeneration trait. Due to the difficulties presented above, the types of analysis that can be performed on the regeneration trait are severely limited. The joint scaling test cannot be used because this test requires an estimate of the mean for each parent. Neither of the nonregenerating parent populations (genotypes 144 and 13083) gave any evidence of regeneration, so meaningful estimates of mean regeneration score could not be obtained for these parents. This reiterates the case for analyzing the uncensored rather than raw regeneration data. All calli for each nonregenerating parental genotype received regeneration scores of one, so based on the raw data the two genotypes have equal mean regeneration scores. There is no reason to believe that both of the nonregenerating parents have the same genetic proclivity for regeneration. The scores merely indicate that neither genotype has enough proclivity to produce any visible evidence of regeneration under the culture protocol used.
The iterative variance partitioning method used for the callus growth trait could not be applied to the regeneration trait. This method requires good estimates of the uncertainty associated with variance estimates and it was felt that these were lacking for the regeneration trait. Mather and Jinks (1971) described a simpler, nonweighted, noniterative variance partitioning method. Preliminary efforts were made to apply this method to the regeneration trait. Both raw and uncensored data were examined in this way, despite the






55

truncation-related biases in the raw data. Preliminary variance partitioning yielded no significant (or even nearly significant) estimates for any parameters. In the case of cross 501, this appeared to be due to an excessively high withinfamily variance in the F3 and extremely low among-family F3 variance and F2F3 covariance. For cross 507, the primary confounding factor appeared to be an overly high F2 variance. In both cases these factors made it impossible to obtain solutions to the simultaneous linear equations established in the variance partitioning matrix.
The remaining option for genetic analysis of the regeneration trait is
parent-offspring regression. Regressions of F3 family mean on F2 parent for the two crosses are presented in Figure 3-6. Cross 507 shows a strong parentoffspring relationship, with a moderately regression coefficient, or heritability. The coefficient of determination (r2) is higher for the raw data than for the uncensored data, but the slope is greater for the uncensored data. In the case of cross 501, the regression is not significant. One reason for the poor regression observed for cross 501 is the very small sample size, especially the low number of parents with regeneration scores higher than one. Another reason is the broad range of mean offspring regeneration scores for parents with regeneration scores of one. It is likely that the "true" regeneration scores for some of the parents (particularly the parent at the far lower left of the plot) lie somewhat below the censoring threshold. If it were possible to measure parental regeneration scores of less than one, the observations at the lower left corner of the plot might be distributed farther to the left, resulting in a stronger regression.
The previous discussion of both the callus growth and regeneration traits contained the implicit assumption that the traits were quantitatively inherited. For any genetic system there exists a continuum of possibilities, ranging from






56


absolute control by a single gene, through control by one or a few major genes in conjunction with a number of minor, or "modifier" genes, to classical quantitative inheritance involving control by many genes with approximately equal effects. The large error variances and non-ideal data structure of this study make it impossible to formally distinguish among these possibilities. However, several general observations and considerations combine to suggest that both of the traits examined are controlled at least to some degree by multiple genes.
First, both traits exhibited transgressive segregation. Callus weights of many F3 individuals in both crosses greatly exceeded that of any parental individual, and four F3 individuals from cross 507 received regeneration scores of seven-a score never achieved by the regenerating parent (Figure 3-7; Appendix D). Transgressive segregation in crosses between inbred lines can be attributable to either overdominance or the presence of positive-acting alleles at different loci in each parent. True overdominance is so rare that many authorities doubt its existence (Simmonds, 1979). If single-gene overdominance existed for the regeneration trait, approximately one quarter of F2 individuals would carry the overdominant genotype, and some proportion of these would likely exhibit the overdominant (transgressive) phenotype. The transgressive phenotype was not observed in the F2. Thus, there appear to be at least two loci controlling the regeneration trait. In view of the moderately large size of the F2 populations (57 for cross 501 and 80 for cross 507), the absence of the transgressive phenotype in this generation suggests that more than two loci may be involved.
A second indication that at least two loci control the regeneration trait is the large difference in regeneration scores between Crosses 501 and 507, despite






57

the fact that the crosses share a regenerating female parent. If ability to regenerate were conferred entirely by alleles originating from one or two loci in the regenerating parent, it would be expected that the best regenerators among the F2 and F3 would be those bearing the fixed genotype of the regenerating parent. The best regenerators from each cross would then be expected to perform approximately equally, regardless of the identity of the nonregenerating parent. This was not observed. Instead, the best regenerators from cross 507 consistently outperformed those of cross 501 in both the F2 and F3. The transgressive phenotype occurred only in cross 507, suggesting that the nonregenerating parent in this cross contributes at least one positive-acting allele. The issue of quantitative versus qualitative inheritance could be clarified by culturing of a large number of progeny from several high-regenerating F3 plants.
Some final insights can be obtained by examining the relationship
between regeneration and callus growth. Figure 3-8 presents a scattergram of regeneration score versus logarithm of callus weight for the combined F3 population. A very weak, but significant correlation exists, and perhaps as meaningful is the observation that the high regeneration scores are generally associated with moderate callus growth. This apparent clustering is in part an illusion stemming from the smaller population sizes for the higher regeneration scores. However, there is a significant (a=0.05) decrease in population variances for the callus growth trait as regeneration score increases, indicating that the clustering is real. The association can be either causal, accidental, or both. In both crosses, callus growth was greater in the nonregenerating parent. The absence of regeneration in the high growth calli may therefore be the result of genetic linkage between the two traits-an accident of the distribution of genetic material among the parents. The lack of regeneration among low






58

growth calli cannot be attributed to linkage. A likely explanation is that a certain degree of in vitro vigor, as manifested by at least a moderate callus growth rate, promotes regeneration. This would constitute pleiotropy in a broad sense. An alternative possibility is that there exist genes that act in a physiologically pleiotropic manner; that is, there may be genes that either promote or inhibit specific metabolic pathways that are necessary for both callus growth and regeneration. To experimentally distinguish between these subtly different hypotheses is far beyond the scope of the present study, or of any study undertaken to date on the genetics of in vitro traits.






59

Table 3-1. Number of F2 plants, F3 plants, and F3 families from which in vitro
data were collected.


Cross F2 plants F3 plants F3 families
501 65 181 11 507 91 291 18 510 52 t t



t Cross 510 was discontinued after F2.








Table 3-2. Univariate statistics for callus weight for parental, F2, and F3 populations in linear, logarithmic and
square root scales.

Population Scalet Mean Variance Skewness W Prob
144 (Parental) li 0.7442 0.0648 0.247 0.953 0.288 26 lo 2.844 0.0268 -0.702 0.934 0.102 26 sq 26.87 22.89 -0.196 0.985 0.372 26

6123 (Parental) li 0.2897 0.0172 0.38 0.947 0.119 35 lo 2.413 0.0477 -0.528 0.95 0.15 35 sq 16.57 15.44 -0.021 0.964 0.369 35

13083 (Parental) li 0.7215 0.0541 -0.114 0.958 0.514 20 lo 2.833 0.0261 -0.93 0.906 0.055 20 sq 26.49 20.74 -0.503 0.943 0.293 20

501 (F2) li 0.301 0.071 2.111 0.794 <0.001 65 Io 2.319 0.162 -0.432 0.954 0.038 65 sq 15.93 47.88 0.766 0.942 0.008 65

507 (F2) li 0.366 0.069 1.506 0.877 <0.001 92 lo 2.445 0.123 -0.65 0.9953 0.009 92 sq 17.97 43.58 0.431 0.969 0.15 92

t li = linear (grams); lo = log10(milligrams); sq = square root (milligrams)








Table 3-2 continued.


Population Scalet Mean Variance Skewness W Prob

501 (F3) Ii 0.380 0.1138 1.729 0.830 <0.001 181
lo 2.412 0.1643 -0.242 0.9612 0.001 181 sq 17.81 63.47 0.691 0.940 <0.001 181 507 (F3) li 0.254 0.0526 1.509 0.817 <0.001 291
lo 2.231 0.1647 -0.161 0.963 <0.001 291 sq 14.49 43.72 0.737 0.924 <0.001 291


t li = linear (grams); lo = log10(milligrams); sq = square root (milligrams)










>)






62

Table 3-3. Model coefficients used to construct joint scaling tests.

Population Coefficient
m [d] [h]


Parent 1 1 -1 0 Parent 2 1 1 0

F2 1 0 0.5 F3 1 0 0.25









Table 3-4. Parameter estimates and associated probabilities obtained from joint scaling tests for callus growth.


Parameter

m [d] [h] Whole Model Cross Scale
estimate prob>F estimate prob>F estimate prob>F X2 prob>X2 501 Linear 0.497 0.023 0.210 0.057 -0.417 0.091 0.481 0.45 507 Linear 0.401 0.247 0.149 0.567 -0.364 0.637 53.97 <0.001 501 Logarithmic 2.611 0.009 0.210 0.112 -0.672 0.139 2.086 0.24 507 Logarithmic 2.542 0.053 0.228 0.504 -0.582 0.601 89.17 <0.001 501 Square root 21.26 0.020 4.849 0.091 -11.73 0.128 1.360 0.36 507 Square root 19.20 0.148 4.313 0.555 -9.679 0.655 81.66 <0.001






64

Table 3-5. Coefficients used for variance partitioning for callus growth.


Causal Component
Observational df Cross Component D H Ew

501 V1 F2 0.5 0.25 1 64
V1F3 0.5164 0.07076 0.1314 10 V2F3 0.25 0.125 1 170 W1 F23 0.5 0.125 0 10 E1 0 0 1 53

507 V1 F2 0.5 0.25 1 91
V1F3 0.5196 0.07233 0.1571 17 V2F3 0.25 0.125 1 273 W1 F23 0.5 0.125 0 17 E1 0 0 1 59






65

Table 3-6. Observational variance components and initial weighs
used for variance partitioning for callus growth.


Cross Scale Component Variance Weight df

501 Linear V1F2 0.07076 12780 64 V1F3 0.06100 2688 10 V2F3 0.06232 43780 170 W2F23 0.08426 582.6 10 El 0.03040 57370 53

501 Logarithmic V1F2 0.1618 2445 64 V1F3 0.09271 1163 10 V2F3 0.08591 23030 170 W2F23 0.05925 813.8 10 E1 0.03994 33210 53

501 Square Root V1F2 4788 0.02791 64 V1F3 36.46 0.007521 10 V2F3 32.86 0.1575 170 W2F23 36.93 0.002687 10 E1 17.37 0.1757 53






66

Table 3-6. Continued.


Cross Scale Component Variance Weight df

507 Linear V1F2 0.06941 19100 91 V1F3 0.02204 3500 17 V2F3 0.03618 208600 273 W2F23 0.02148 13890 17 E1 0.03746 42040 59

507 Logarithmic V1F2 0.1227 6044 91 V1F3 0.06896 3574 17 V2F3 0.1132 21300 273 W2F23 0.04864 2612 17 E1 0.03879 39210 59

507 Square Root V1F2 43.72 0.04760 91 V1F3 19.19 0.04642 17 V2F3 29.39 0.3160 273 W2F23 15.64 0.02542 17 E1 18.23 0.1701 59








Table 3-7. Estimates of causal variance components and associated probabilities obtained from variance partitioning
for callus growth.


Component

Cross Scale D H Ew Whole Model Estimate Prob>F Estimate Prob>F Estimate Prob>F F Prob>F 501 Linear 0.1258 0.4033 -0.0578 0.8239 0.0324 0.1428 28.57 0.1365 507 Linear 0.0313 0.6534 0.0271 0.9055 0.0284 0.2982 11.50 0.2127 501 Logarithmic 0.1273 0.4244 0.1535 0.6266 0.3881 0.1241 48.08 0.1055 507 Logarithmic 0.0594 0.7487 0.3036 0.5601 0.0438 0.2364 11.64 0.2114 501 Square Root 70.39 0.0592 -17.50 0.4434 17.38 0.0189 1944 0.0189 507 Square Root 23.61 0.2057 46.27 0.2916 18.24 0.0512 332.2 0.0403




4'






68

Table 3-8. Univariate statistics for regeneration score utilizing raw and
uncensored data.


Population Raw Data Uncensored Data Mean Variance Mean Variance N N>1 6123 4.053 1.240 4.075 1.080 38 37 501 F2 (all) 1.868 1.076 1.679 1.774 57 29 507 F2 (all) 1.994 1.952 1.278 4.351 80 43 501 F2 (parents only) 1.636 0.8552 2.078 0.4886 11 4 507 F2 (parents only) 2.267 2.924 1.553 6.529 17 8 501 F3 1.594 0.8158 1.185 1.882 165 67 507 F3 2.058 1.807 1.554 3.733 260 136 501 F3 families:

501-1 2.000 1.166 2.116 1.143 13 7 501-2 1.462 0.4359 1.520 0.533 13 6 501-3 1.333 0.4334 1.391 0.7310 21 5 501-4 2.083 1.356 1.982 1.948 12 7 501-5 2.417 2.629 2.045 3.069 12 8 501-6 1.440 0.5067 1.391 0.832 25 8 501-7 1.600 0.7115 1.612 0.986 10 5 501-8 1.214 0.1813 - 14 3 501-9 1.923 0.4102 2.032 0.3152 13 11 501-10 1.2143 0.1813 - 14 3 501-11 1.389 0.7222 -0.0665 4.920 18 4






69

Table 3-8. Continued.


Population Raw Data Uncensored Data


Mean Variance Mean Variance N N>1 507 F3 Families:
507-1 1.571 0.8791 0.9004 2.974 14 6 507-2 1.769 2.190 - 13 3 507-3 1.500 0.3330 - 4 2 507-4 3.850 2.976 3.700 2.036 20 20 507-5 3.706 2.220 3.667 1.672 17 16 507-6 1.667 1.466 - 6 2 507-7 1.545 0.6406 1.635 0.7582 22 8 507-8 2.000 1.999 1.319 5.197 13 6 507-9 1.500 0.5770 1.438 0.953 14 5 507-10 1.273 0.4182 - 11 2 507-11 2.733 2.065 2.641 2.577 15 12 507-12 1.650 0.5553 1.753 0.5090 20 10 507-13 2.625 0.8383 2.625 0.6933 8 7 507-14 1.714 1.214 1.220 2.965 21 8 507-15 2.000 0.9091 2.356 1.188 12 7 507-16 1.611 0.6050 1.680 0.668 18 8 507-17 1.789 1.397 1.056 4.177 19 8 507-18 1.615 0.7564 1.386 1.365 13 6







70




(a) 0.40 2
y = 0.235x 0.023 r = 0.662


0.30


0.20


0.10


0.00 .
0.00 0.25 0.50 0.75 1.00 Mean


(b) 0.40
y = -0.094x + 0.323 r 2 = 0.144

0.30


0.20

> *
0.10 -*

4* *
0.00 I I
1.50 2.00 2.50 3.00
Mean



(c)y = 1.949x + 1.123 r 2 = 0.245

75


C 50


25 *


0I I I
5 10 15 20 25 30 Mean

Figure 3-1. Relationship of F3 family means and variances for callus growth
under (a) linear, (b) logarithmic, and (c) square root scales.







71



Cross 501 Linear Cross 507 Linear
1.00 0.60


0.75
0.40

0.50 U0.20
0.25- *
0.25y = 0.588x + 0.149y = 0.524x + 0.099
Sr2 = 0.705 r 2 = 0.605
0.00' 1 0.00
0.00 0.50 1.00 1.50 0.00 0.25 0.50 0.75 1.00 F2 F2


Cross 501 Logarithmic Cross 507 Logarithmic
3.00 2.75

2.75 2.502.50.50

(v) LL 2.252.25

2.00- y = 0.653x + 0.784 2.00- y = 0.769x + 0.387 r =0.400 r =0.602
1.75 I I I I 1.75
2.00 2.25 2.50 2.75 3.00 3.25 1.75 2.00 2.25 2.50 2.75 3.00
F2 F2


Cross 501 Square Root Cross 507 Square Root
30- 25

25
20
20- 1

15-1

y = 0.624x + 6.169 10
10 2 0y=0.631x+4.137 r =0.579 2 r =0.633
5 I I I I 5
10 15 20 25 30 35 40 5 10 15 20 25 30 35 F2 F2

Figure 3-2. Regression of F3 family mean callus weight on F2 parental callus
weight under linear, logarithmic, and square root scales.






72





(a)


Partial > Nonregenerators Full Regenerators
C U
25% 68%





Regeneration Response





(b)


Partial Nonregenerators Full Regenerators



68% 20% 12%



Regeneration Response









Figure 3-3. Two hypothetical distributions with three response classes and two
thresholds.







73



(a)

y = 0.227x + 2.442




0












-3 -2 -1 0 1 2 3 Normal Quantile


(b)


y = 0.231x + 2.441



.0




C Cr






-3 -2 -1 0 1 2 3 Normal Quantile


Figure 3-4. Normal quantile plots of a hypothetical (a) intact and (b) censored
population.






74

















20

n = 38 15 10
C"



5




1 2 3 4 56 Regeneration Score
















Figure 3-5. Regeneration score histogram for genotype IRFL 6123.






75









Cross 501 Raw Data Cross 501 Uncensored
5.00 5.00
y = 0.189x + 1.283 y = 0.297x + 0.932
4.00- r2 = 0.287 p = 0.287 4.00- r2 = 0.263 p = 0.158

3.00 3.00

LLM 60- 2.00
2.00

01.00 0.00
0.00 Ii I I I I
0.00 1.00 2.00 3.00 4.00 5.00 0.00 1.00 2.00 3.00 4.00 5.00

F2 F2




Cross 507
Cross 507 Raw Data Uncensored

6.0 y = 0.418x + 1.077 6.0- y = 0.460x + 0.904

5.0 r2= 0.803 p < 0.001 5.0 -r2 = 0.806 p < 0.001

4.0 4.0L. 3.0 L 3.0 2.0 2.0 1.0 1.0

0.0 I I I i 0.0 I I
0.0 1.0 2.0 3.0 4.0 5.0 6.0 0.0 1.0 2.0 3.0 4.0 5.0 6.0

F2 F2







Figure 3-6. Regression of F3 family mean regeneration score on F2 parent
regeneration score.





76












































Figure 3-7. Profuse regeneration exhibited by members of two F3
families derived from cross 507







77















3.50
y =0.045x + 2.279 r 2 = 0.021 p =0.003
8 o
0
0
3.00 o




S2.50 o 0 0
Co o
:o 0


g 2.00 o o o O 0 0
o 0


1.50 o o

0


1.00
0.00 2.00 4.00 6.00 8.00


Regeneration Score











Figure 3-8. Regression of callus weight on regeneration score for combined F3
populations.





78





CHAPTER 4
SUMMARY AND CONCLUSIONS


The point of departure for this dissertation was a study conducted by
Wofford et al. (1992b). Wofford's study identified a three stage (callus induction, shoot induction, and shoot elongation) protocol utilizing the L2-based media of Phillips and Collins (1979) as suitable for tissue culture of desmodium hypocotyls, and obtained limited regeneration from a single genotype--IRFL 6123. The objective of the current work has been to examine the potential for enhancing in vitro response in desmodium through improvements in culture protocol and through breeding.
Chapter 2 focused on optimization of culture protocols. The work was not intended to be an exhaustive investigation of potential methods for enhancing in vitro response in desmodium. Given the wide range of variables that can determine the effectiveness of tissue culture protocols and the limited prior work with this crop, such an investigation would be beyond the scope of any single study. Instead, the intent was to cast a broad net-to determine if the poor, highly genotype-specific regeneration observed in desmodium could be significantly improved by applying a wide range of techniques similar to those that had proven successful in other legumes. Accordingly, within each general culture strategy (e.g., alternative explant sources, manipulation of auxincytokinin ratio, substitution of different auxin sources) only a small number of specific culture protocols were examined. For the same reason, number of replications was often fairly low. Thus, failure, for example, to obtain regeneration from leaf disks or petioles does not indicate that regeneration from


78






79

these explant types can not be achieved--only that none of the methods tested yielded strongly encouraging results.
The first general culture strategy investigated was the use of alternative explant sources. In addition to the previously proven hypocotyl explants, leaf disks, petioles, and seedling cotyledons were tested. One previously established regenerating genotype (IRFL 6123) and two nonregenerating genotypes ( CIAT 13083 and UF 20) were tested. Regeneration was not obtained from any of the alternative explant sources under the single culture protocol tested. The only repeatable regeneration response was from hypocotyls of IRFL 6123, as had been previously reported (Wofford et al., 1992b).
Two more strategies for broadening the range of regenerating genotypes were examined. The first was manipulation of 2,4-D concentration, with and without added kinetin, in the shoot induction medium. This experiment failed to induce regeneration in recalcitrant genotypes, but succeeded in identifying an optimal 2,4-D concentration for regeneration in IRFL 6123. In addition, kinetin was found to inhibit regeneration in this genotype. The last attempt at inducing regeneration in recalcitrant genotypes involved direct regeneration from hypocotyl explants. This approach failed to yield regeneration even from IRFL 6123.
The remainder of Chapter 2 dealt with enhancing the regeneration
response in IRFL 6123. The first experiment demonstrated that reducing the shoot induction period from the original 28 days to 14 days had no effect on regeneration. Two additional experiments investigated the effects of growth regulators in the shoot elongation step. An optimum ratio of 0.012 mg L-1 picloram to 0.2 mg L BA was identified.






80

Histological examination of regenerating cultures revealed vascular
connections between shoots and the surrounding callus tissue, indicating that regeneration was of an organogenic nature. However, early in the shoot induction period, small structures resembling somatic embryos could occasionally be observed. The combination of growth regulators used in the shoot induction step is similar to that used for somatic embryogenesis in other species, so it is possible that the regeneration observed in this study was some aberrant form of somatic embryogenesis.
It was clearly established in Chapter 2 that regeneration in desmodium is highly dependent on genotype and that broadening the range of regenerating genotypes through modification of culture protocols is difficult or impossible. Chapter 3 focused on the genetics of callus growth and regeneration. An extensive effort at crossing regenerating with nonregenerating genotypes yielded only three crosses-501, 507, and 510. Cross 510 proved useless for the analysis at hand due to severe internodal stunting and other morphological abnormalities in the F1 and F2 generations. This may have been the result of cytoplasmic effects, since cross 510 was essentially the reciprocal of cross 507, which showed no abnormalities. Examination of the genetics and physiology of this phenomenon might produce very interesting results.
The remaining two crosses were selfed through the F3 generation.
Analysis of callus growth and and regeneration in these crosses produced many ambiguous results, but also yielded useful insights. A large amount of effort was devoted to scaling of the callus growth trait. General statistical considerations indicated that a simple linear scale was inadequate for describing this trait. Logarithmic and square root transformation were both more satisfactory than the linear scale, but it was not clear which of these transformations was superior.






81

Formal genetic analysis of callus growth was approached via the joint scaling test and variance partitioning. Neither method yielded conclusive results. This may have been due to violation of the genetic assumptions inherent in both methods, or may have been because both methods inappropriately attempted to describe the trait with a simple additive-dominant genetic model. Parent offspring regression yielded heritabilities ranging from
0.52 to 0.77, depending on cross and scale. It appears, then, that callus growth in this germplasm is controlled to a significant extent by additive genetic effects, but that there may in addition be nonadditive, nondominant genetic effects that act to confound the joint scaling test and variance partitioning.
The regeneration trait presented unusual analytical difficulties. Many calli showed no evidence of regeneration. As a result the majority of population distributions were severely truncated. Truncated (or "censored") distributions cannot be rendered normal by any mathematical transformation. A procedure was described by which truncated population distributions can be reconstructed, or "uncensored." While this procedure has apparently not been previously applied to genetic analysis, it is conceptually similar to the threshold trait approach described by Falconer (1981). The method is less than ideal, but is preferable to attempting to analyze raw, severely censored data. Constraints imposed by the data structure and by the uncensoring technique made it impossible to conduct meaningful joint scaling tests or variance partitioning analyses. Parent-offspring regression yielded relatively high heritability estimates for cross 507--0.416 for raw data; 0.460 for uncensored data. No significant regression was obtained for cross 501. It appears likely that this failure is due to shortcomings of the data structure, and that regeneration is in fact weakly to moderately heritable in this cross.






82

Vasil (1987) has suggested that the genetics of regeneration is irrelevant, and that with sufficient insight into the in vitro physiology of a species regeneration can be induced in even the most difficult genotype. While this may be true in theory, this work suggests that such an assumption can be impractical or even counterproductive. Approximately equal effort was directed at culture protocol and genetic approaches and the latter was found to be by far the more productive path. It is noteworthy that the original regenerating parent regenerated under a variety of culture conditions, suggesting a general genetic proclivity to regenerate. A more experienced investigator with greater resources may or may not have obtained better results from culture protocol optimization, but success may well have been very costly in terms of time, effort, and financial resources. The genetic approach has been relatively straightforward and strikingly successful. The magnitude of this success was particularly apparent from visual examination of regenerating calli. Several F3 calli, including that shown in Figure 3-7, grossly outperformed the parental genotype, continuing to produce vigorous shoots through repeated subcultures, and ultimately yielding dozens of shoots.
This study has demonstrated that regeneration in desmodium can be
greatly improved by conventional crossing techniques. Limited light was shed on the genetic basis of regeneration in this crop, and many questions remain unanswered. It is hoped that the F3 material produced in this study may prove useful to any investigator who might desire to perform regeneration-dependent work with this crop.











REFERENCES

Ammirato, P.V. 1983. Basic techniques of plant cell culture: Embryogenesis. In D.A. Evans, W.R. Sharp, P.V. Ammirato, and Y. Yamada (eds.) Handbook of plant cell culture. Vol. 1. Techniques for propagation and breeding. Macmillan Publishing Co., New York.

Angeloni, P.N., H.Y. Rey, and L.A. Mroginski. 1988. Cultivo in vitro de tejidos de Desmodium incanum y D. affine (Leguminosae). Phyton:48:71-76.

Bingham, E.T., L.V. Hurley, D.M. Kaatz, and J.W. Saunders. 1975. Breeding alfalfa which regenerates from callus tissue in culture. Crop Sci. 15:719-721.

Blaydes, D.F. 1966. Interaction of kinetin and various inhibitors in the growth of soybean tissue. Physiol. Plant. 19:748-753.

Bovo, O.A., L.A. Mroginski, and H.Y. Rey. 1986. Regeneration of plants from callus tissue of the pasture legume Lotononis bainessi. Plant Cell Rep. 5:295-297.

Brown, D.C.W., and A. Atanassov. 1985. Role of genetic background in somatic embryogenesis in Medicago. Plant Cell, Tissue, Organ Culture 4:111-122.

Buiatti, M., S. Baroncelli, A. Bennici, M. Pagliai, and R. Tesi. 1974. Genetics of growth and differentiation in vitro of Brassica oleracea var. botrytis. Z. Pflanzenzuchtg. 72:269-274.

Bullock, W.P., and P.S. Baenziger. 1982. Anther culture of wheat (Triticum aestivum L.) Fl's and their reciprocal crosses. Theor. Appl. Genet. 62:155-159.

Cavalli, L. 1952. An analysis of linkage in quantitative inheritance. In E.C.R. Reeve and C.H. Waddington (eds.) Quantitative inheritance. HMSO, London.



83





84

Charmet, G., and S. Bernard. 1984. Diallel analysis of androgenetic plant production in hexaploid Triticale X. triticosecale, Wittmack. Theor. Appl. Genet. 69:55-61.

Choo, T.M. 1988. Plant regeneration in zigzag clover (Trifolium medium L.). Plant Cell Rep. 7:246-248.

Chow, K.H., and L.V. Crowder. 1972. Hybridization of Desmodium canum (Gmel.) Schin. and Thell. and D. uncinatum (Jacq.) DC. Crop Sci. 12:784-785.

Chow, K.H., and L.V. Crowder. 1973. Hybridization of Desmodium species. Euphytica 22:339-404.

Collins, G.B., and G.C. Phillips. 1982. In vitro tissue culture and plant regeneration in Trifolium pratense L. In E.D. Earle and Y Demardy (eds.) Variation in plants regenerated from cells and tissue culture. Praeger Sci. Pub. New York.

Deak, M., G.G. Kiss, C. Koncz, and D. Dudits. 1986. Transformation of Medicago by Agrobacterium mediated gene transfer. Plant Cell Rep. 5:97-100.

Dodds, J.H. 1995. Experiments in plant tissue culture. 3rd Ed. Cambridge University Press, Cambridge.

Dulieu, H. 1991. Inheritance of regeneration capacity in the genus Petunia. Euphytica 53:173-181.

Dunwell, J.M., R.J. Francis, and W. Powell. 1987. Anther culture of Hordeum vulgare L.: a genetic study of microspore callus production and differentiation. Theor. Appl. Genet. 74:60-64.

Falconer, D.S. 1981. Introduction to quantitative genetics, 2nd Ed. Longman Inc., New York.

Flick, C.E., D.A. Evans, and W.R. Sharp. 1983. Basic techniques of plant cell culture: organogenesis. In D.A. Evans, W.R. Sharp, P.V. Ammirato, and Y. Yamada (eds.) Handbook of plant cell culture. Vol. 1. Techniques for propagation and breeding. Macmillan Publishing Co., New York. p. 13-81.





85

Foroughi-Wehr, B., W. Friedt, and G. Wenzel. 1982. On the genetic improvement of androgenetic haploid formation in Hordeum vulgare L. Theor. Appl. Genet. 62:233-239.

Fraley, R.T., S.G. Rogers, and R.B. Horsch. 1986. Genetic transformation in higher plants. CRC Critical Reviews in Plant Sciences 4:1-46.

Frankenberger, E.A., P.M. Hasegawa, and E.C. Tigchelaar. 1981. Diallel analysis of shoot-forming capacity among selected tomato genotypes. Z. Pflanzenphysiol. 102:233-242.

Gamborg, O.L., R.A. Miller, and K. Ojima. 1968. Nutrient requirements of suspension cultures of soybean root cells. Exp. Cell Res. 50:151158.

Ghazi, T.D., H.V. Cheema, and M.W. Nabors. 1986. Somatic embryogenesis and plant regeneration from embryogenic callus of soybean, Glycine max. L. Plant Cell Rep. 5:452-456.

Gilbert, R.O. 1987. Statistical methods for environmental pollution monitoring. Van Nostrand Reinhold Company, Inc., New York.

Haccius, B. 1978. Question of unicellular origin of nonzygotic embryos in callus cultures. Phytomorphology 28:74-81.

Hazra, S., S.S. Sathaye, and A.F. Mascarenhas. 1989. Direct somatic embryogenesis in peanut (Arachis hypogea). Biotechnology 7:949951.

Hills, F.J., and T.M Little. 1978. Agricultural experimentation design and analysis. John Wiley and Sons, New York.

Horvath, K.B., and R.R. Smith. 1979. Plant regeneration from callus culture of red and crimson clover. Plant Sci. Lett. 16:231-237.

Imrie, B.C., R.M. Jones, and P.C. Kerridge. 1983. Desmodium. In R.L.
Burt, P.P. Rotar, J.L. Walker, and M.W. Silvey (eds.) The role of Centrosema, Desmodium, and Stylosanthes in improving tropical pastures. Westview Tropical Agriculture Series No. 6. Westview Press, Boulder, Colorado.






86

Kao, K.N., and M.R. Michayluk. 1981. Embryoid formation in alfalfa cell suspensions from differentiated plants. In Vitro 17:645-648.

Kerns, H.R., U.B. Barwale, M.M. Meyer, Jr., and J.M. Widholm. 1986.
Correlation of cotyledonary node shoot proliferation and somatic embryoid development in suspension cultures of soybean (Glycine max L. Merr.). Plant Cell Rep. 5:140-143.

Ketchum, J.L.F., O.L. Gamborg, G.E. Hamming, and M.W. Nabors. 1987.
Progress report, tissue culture for crops project. Colorado State University Press., Ft. Collins, Colorado.

Keyes, G.J., G.B. Collins, and N.L. Taylor. 1980. Genetic variation in tissue cultures of red clover. Theor. Appl. Genet. 58:265-271.

Komatsuda, T., S. Enomoto, and K. Nakajima. 1989. Genetics of callus proliferation and shoot differentiation in barley. J. Hered. 80:345350.

Koornneef, M., C.J. Hanhart, and L. Martinelli. 1987. A genetic analysis of cell culture traits in tomato. Theor. Appl. Genet. 74:633641.

Kretchmer, A.E., Jr., J.B. Brolmann, G.H. Snyder, and S.W. Coleman. 1976. 'Florida' Carpon desmodium, a perennial tropical legume for use in south Florida. Soil and Crop Sci. Soc. of Fla. Proc. 35:25-30.

Kumar, A.S., T.P. Reddy, and G.M. Reddy. 1983. Plantlet regeneration from different callus cultures of pigeonpea (Cajanus cajan L.). Plant Sci. Lett. 32:271-278.

Kumar, A.S., T.P. Reddy, and G.M. Reddy. 1984. Multiple shoots from cultured explants of pigeonpea and atylosia species. SABRAO J. 16:101-105.

Kumar, A.S., T.P. Reddy, and G.M. Reddy. 1985. Genetic analysis of certain in vitro and in vivo parameters in pigeonpea (Cajanus cajan L.). Theor. Appl. Genet. 70:151-156.

Lazar, M.D., P.S. Baenziger, and G.W. Schaeffer. 1984. Combining abilities and heritability of callus formation and plantlet regeneration in wheat (Triticum aestivum L.) anther cultures. Theor. Appl. Genet. 68:131-134.






87

Lupotto, E. 1983. Propagation of an embryogenic culture of Medicago sativa L. Z. Pflanzenphysiol. 111:95-104.

Maheshwaran, G., and E.G. Williams. 1984. Direct somatic embryoid formation on immature embryos of Trifolium repens, Trifolium pratense, and Medicago sativa and rapid clonal propagation of Trifolium repens. Ann. Bot. 54:201-211.

Malmberg, R.L. 1979. Regeneration of whole plants from callus cultures of diverse genetic lines of Pisum sativum L. Planta 146:243-244.

Mather, K., and J.L. Jinks. 1971. Biometrical Genetics. Cornell University Press, Ithaca, New York.

McKently, A.H., G.A. Moore, and F.P. Gardner. 1990. In vitro plant regeneration of peanut from seed explants. Crop Sci. 30:192-196.

McKently, A.H., G.A. Moore, and F.P. Gardner. 1992. Regeneration of peanut and perennial peanut from cultured leaf tissue. Crop Sci. 31:833-837.

Mehta, U., and H.Y. Mohan Ram. 1980. Regeneration of plantlets from the cotyledons of Cajanus cajan. Indian J. Exp. Bot. 18:800-802.

Meijer, E.G.M. 1982. Shoot formation in tissue cultures of three cultivars of the tropical forage legume Stylosanthes guyanensis. Z. Pflanzenzuchtg. 89:169-172.

Murashige, T., and F. Skoog. 1962. A revised medium for rapid growth and bioassays with tobacco tissue cultures. Physiol. Plant. 15:473497.

Newman, M.C., and P.M. Dixon. 1989. Estimating mean and variance for environmental samples with below detection limit observations. Water Res. Bull. 25:905-916.

Oelck, M.M., and 0. Scheider. 1983. Genotypic differences in some legume species affecting the redifferentiation ability from callus to plants. Z. Pflanzenzuchtg. 91:312-321.






88

Ohashi, H. 1973. The Asiatic species of desmodium and its allied genera (Leguminosae). Ginkgoana. Contributions of the flora of Asia and the Pacific region. Academia Sci. Book, Inc., Tokyo.

Phillips, G.C., and G.B. Collins. 1979. In vitro tissue culture of selected legumes and plant regeneration from callus cultures of red clover. Crop Sci. 19:59-64.

Phillips, G.C., and G.B. Collins. 1980. Somatic embryogenesis from cell suspension cultures of red clover. Crop Sci. 20:323-326.

Quesenberry, K.H, M.R. McKellar, and D.E. Moon. 1989. Evaluation and hybridization of germplasm in the Desmodium heterocarpon Desmodium ovalifolium species complex. In Proc. XVI Intl. Grassland Congress. Nice, France. p. 251-252.

Quesenberry, K.H., D.S. Wofford, P.A. Krottje, and R.L. Smith. 1992. Production of transgenic red clover plants using Agrobacteriummediated DNA transfer. Proceedings of the Twelfth Trifolium Conference. University of Florida, Gainesville, Florida. p. 5-10.

Quimio, C.A., and F.J. Zapata. 1990. Diallel analysis of callus induction and green-plant regeneration in rice anther culture. Crop Sci. 30:188-192.

Reisch, B., and E.T. Bingham. 1980. The genetic control of bud formation from callus cultures of diploid alfalfa. Plant Sci. Lett. 20:71-77.

Rotar, P.P. and U. Urata. 1967. Cytological studies in the genus Desmodium species: some chromosome counts. Amer. J. Bot. 54:1-4.

Rowe, K.E., and W.L. Alexander. 1980. Computations for estimating the genetic parameters in joint-scaling tests. Crop Sci. 20:109-110.

Santos, A.V.P. dos, E.G. Cutter, and M.R. Davey. 1983. Origin and development of somatic embryos in Medicago sativa L. (alfalfa). Protoplasma 117:107-115.

SAS Institute. 1994. JMP Statistics and Graphics Guide. SAS Institute Inc., Cary, North Carolina.






89

Sarria, R., A.Calderon, A.M. Thro, E. Torres, J.Mayer, and W.M. Rocca. 1994. Agrobacterium mediated transformation of Stylosanthes guianensis and production of transgenic plants. Plant Sci. 96:119127.

Saunders, J.W., G.L. Hosfield, and A. Levi. 1987. Morphogenetic effects of 2,4-dichlorophenoxyacetic acid on pinto beans (Phaseolus vulgaris L.) leaf explants in vitro. Plant Cell Rep. 6:4649.

Schenk, R.U., and A.C. Hildebrandt. 1972. Medium and techniques for induction and growth of monocotyledononous and dicotyledonous plant cell cultures. Can. J. Bot. 50:199-204.

Schubert, B.G. 1963. Desmodium: Preliminary studies IV. Journal of the Arnold Arboretum 44:284-297.

Schultze-Kraft, R., and G. Benavides. 1988. Germplasm collection and preliminary evaluation of Desmodium ovalifolium Wall. Gen. Resources Comm. 12:1-20.

Schultze-Kraft, R., and S. Pattanavibul. 1985. Collecting native forage legumes in eastern Thailand. IBPGR Newsletter. 9:4-5.

Scowcroft, W.R., and J.A. Adamson. 1976. Organogenesis from callus cultures of the legume, Stylosanthes hamata. Plant Sci. Lett. 7:3942.

Sellars, R.M., G.M. Southward, and G.C. Phillips. 1990. Adventitious somatic embryogenesis from cultured immature zygotic embryos of peanut and soybean. Crop Sci. 30:408-414.

Simmonds, N.W. 1979. Principles of crop improvement. Longman Inc., New York.

Swanson, E.B., and D.T. Tomes. 1980. Plant regeneration from cell culture of Lotus corniculatus and the selection and characterization of 2,4-D tolerant cell lines. Can. J. Bot. 58:1205-1209.

Takahasi, M. 1952. Tropical forage legume and browse plants research in Hawaii. Proc. 6th Int. Grass. Cong. 1411-1416.





90

Tisserat, B., E.B. Esan, and T. Murashige. 1979. Somatic embryogenesis in angiosperms. Hortic. Rev. 1:1-78.

Vasil, I.K. 1987. Developing cell and tissue culture systems for the improvement of cereal and grass crops. J. Plant Physiol. 128:193218.

Viera, M.L., B. Jones, E.C. Cocking, and M.R. Davey. 1990. Plant regeneration from protoplasts isolated from seedling cotyledons of Stylosanthes guianensis, S. macrocephala and S. scabra. Plant Cell Rep. 9:289-292.

Wan, Y., E.L. Sorensen, and G.H. Liang. 1988. Genetic control of in vitro regeneration in alfalfa. Euphytica 39:3-9.

Walker, K.A., M.L. Wendeln, and E.G. Jaworski. 1979. Organogenesis in callus tissue of Medicago sativa. The temporal separation of induction processes from differentiation processes. Plant Sci. Lett. 16:23-30.

Webb, K.J., M.F. Fay, and P.J. Dale. 1987. An investigation of morphogenesis within the genus Trifolium. Plant Cell, Tissue, Organ Culture 11:37-46.

White, D.W.R., and D. Greenwood. 1987. Transformation of the forage legume Trifolium repens L. using binary Agrobacterium vectors. Plant Mol. Bio. 8:461-469.

Wofford, D.S., K.H. Quesenberry, and D.D. Baltensperger. 1992a. In vitro culture responses of alyceclover genotypes on four media systems. Crop Sci. 32:261-265.

Wofford, D.S., K.H. Quesenberry, and D.D. Baltensperger. 1992b. Tissue culture regeneration of desmodium. Crop Sci. 32:266-268.

Younge, O.R., D.L. Plucknett, and P.P. Rotar. 1964. Culture and yield performance of Desmodium intortum and D. canum in Hawaii. Tech. Bull. No. 59. Hawaii Agric. Exp. Stat., Hawaii.

Zar, J.H. 1984. Biostatistical analysis. 2nd Ed. Prentice Hall, Englewood Cliffs, New Jersey.











APPENDIX A
ANALYSIS OF VARIANCE FOR BETWEEN- AND WITHIN-FAMILY
VARIANCES FOR CALLUS GROWTH

Between- and within-family variances were estimated by analysis of variance with F3 family as the class variable. All ANOVA analyses yielded highly significant results. Expected mean square for families equals (oa2 + nob2), where ao2 is within-family variance (estimated by MSE), n is harmonic mean of number of individuals per family, and ab2 is between-family variance. This equation is solved for Ob2 = (MSmodel MSE)/n. The following table presents MSmodel, MSE, ob2, Fw2, and n for each cross and each scale.




Cross Scale MSmodel MSE(= (w2) ab2 n

501 Linear 0.9901 0.0623 0.0669 15.21 501 Logarithmic 1.4962 0.0859 0.9271 501 Square Root 583.95 32.856 36.462 "

507 Linear 0.3166 0.0361 0.0220 12.73 507 Logarithmic 0.9910 0.1132 0.0689 507 Square Root 273.78 29.392 19.199











91











APPENDIX B
DERIVATION OF VARIANCE PARTITIONING MATRICES FOR CALLUS GROWTH


Coefficients for variance partitioning of the callus growth were based on those given by Mather and Jinks (1971, 1977). The following general coefficients were given for the populations treated in the present study:




Observational Sampling Component D H EW EB Variance


V1F2 1/2 1/4 1 0 0 V1F3 1/2 1/16 0 1 1/n V2F3 V2F3 1/4 1/8 1 0 0 W1F23 1/2 1/8 0 0 0 E1 0 0 1 0 0



Since there is only one entry for sampling variance and the sampling variance for V1 F3 is a function of V2F3, this column can be combined into the other columns. In the case of cross 501, n (the harmonic mean of the number of individuals per F3 family) is equal to 15.21. Therefore, after combining the sampling variance column, the coefficient for D for V1F3 becomes 1/2 + ((1/15.21) x 1/4), or 0.516435. The other entries in the V1F3 row are similarly altered. Since EB is a function of Ew, this column can also be 92




93


eliminated. The coefficient for Ew for V1 F3 becomes the sum of the contribution to this column from the V2F3 row (1/15.21 x 1) plus the contribution from the Eb column in the V1F3 row (1/15.21), or 0.1314. The resulting table is shown below:



D H EW


V1 F2 1/2 1/4 1 V1 F3 0.516435 0.0707175 0.1314803 V2F3 1/4 1/8 1 W1 F23 1/2 1/8 0 El 0 0 1




Derivation of coefficients for cross 507 are similar, but n for this cross is 12.73.




Full Text
40
To examine the effect of the three scales on normality, the scales were
applied to populations of each of the three parental lines and to F2 and F3
populations. The parental populations are theoretically normally distributed,
while the F2 and F3 populations are expected to be normally distributed if little
or no unidirectional dominant or epistatic gene action is present. The later
generation populations are included here because, as a result of a temporarily
limited seed supply, the size of the tested parental populations was quite small
(20 < n < 36). Univariate statistics for the various populations are presented in
Table 3-2. The linear scale tends to produce rightward (positive) skewing, while
the log transformed populations are moderately left-skewed (negative), and the
square root transformed populations are slightly left-skewed. The Shapiro-Wilk
W statistic is used to test for normality in populations as small as ten
observations; a low probability associated with a calculated value of W
indicates a significant deviation from normality (SAS Institute, 1989; Gilbert,
1987). None of the W values for the parental populations indicated deviation
from normality.
Results for F2 and F3 populations are similar to those for the parental
populations, but are somewhat clearer. The linear scaling results in gross
leftward skewing and yields highly significant W statistics. This may be
attributable to genetic effects if there is a strong preponderance of negatively
acting dominant genes in these populations. Strong, unidirectional epistatic
effects, which could result in a relatively small proportion of the populations
possessing the necessary combination of alleles for rapid callus growth, is
another possible genetic cause for the observed skewing. However, the fact
that the leftward skewing occurs in all populations can be interpreted to indicate
that the linear scale is inadequate and that with proper scaling the F2 and F3


49
should be possible in selecting for callus growth among segregating
generations of this material.
The nature of the regeneration variable. The callus growth trait was
amenable to quantitative analysis due to the fact that this trait could be
expressed in terms of a continuous, metric variable for each individual plant.
This was not the case for the regeneration trait. In an ideal study of the genetics
of regeneration, each plant can be evaluated in terms of a metric variable such
as percent responding explants or number of shoots per explant. Because
hypocotyl explants had to be used in the present study, only one explant existed
per plant, so the data could not be expressed as percent responding explants.
Expression of regeneration as number of shoots per explant was impossible
because the majority of explants produced no buds or shoots. Instead, the trait
was expressed on a visual rating scale of one through seven. A rating of one
represented no evidence of regeneration. Two represented a slight indication
of regeneration in the form of localized deep green coloration. Three
represented formation of a single, well-defined but nonelongated bud. Four
indicated multiple bud formation, usually with some elongation. A rating of five
indicated one or two well elongated shoots. Six indicated between three and
five elongated shoots, and a rating of seven indicated more than five elongated
shoots.
A significant number of individuals in all populations showed no sign of
regeneration and received scores of one. All individuals in the nonregenerating
parent populations (genotypes 144 and 13083) received scores of one. This
type of data structure can be described as "threshold" or "censored" data.
Falconer (1981) discussed threshold traits at length. He assumed that there
exists an underlying, continuous scale of proclivity (or as Falconer describes it,
"liability") towards a certain conditionin the case of the present work, tendency


85
Foroughi-Wehr, B., W. Friedt, and G. Wenzel. 1982. On the genetic
improvement of androgenetic haploid formation in Hordeum vulgare
L. Theor. Appl. Genet. 62:233-239.
Fraley, R.T., S.G. Rogers, and R.B. Horsch. 1986. Genetic
transformation in higher plants. CRC Critical Reviews in Plant
Sciences 4:1-46.
Frankenberger, E.A., P.M. Hasegawa, and E.C. Tigchelaar. 1981. Diallel
analysis of shoot-forming capacity among selected tomato
genotypes. Z. Pflanzenphysiol. 102:233-242.
Gamborg, O.L., R.A. Miller, and K. Ojima. 1968. Nutrient requirements
of suspension cultures of soybean root cells. Exp. Cell Res. 50:151-
158.
Ghazi, T.D., H.V. Cheema, and M.W. Nabors. 1986. Somatic
embryogenesis and plant regeneration from embryogenic callus of
soybean, Glycine max. L. Plant Cell Rep. 5:452-456.
Gilbert, R.O. 1987. Statistical methods for environmental pollution
monitoring. Van Nostrand Reinhold Company, Inc., New York.
Haccius, B. 1978. Question of unicellular origin of nonzygotic
embryos in callus cultures. Phytomorphology 28:74-81.
Hazra, S., S.S. Sathaye, and A.F. Mascarenhas. 1989. Direct somatic
embryogenesis in peanut (Arachis hypogea). Biotechnology 7:949-
951.
Hills, F.J., and T.M Little. 1978. Agricultural experimentation design
and analysis. John Wiley and Sons, New York.
Horvath, K.B., and R.R. Smith. 1979. Plant regeneration from callus
culture of red and crimson clover. Plant Sci. Lett. 16:231-237.
Imrie, B.C., R.M. Jones, and P.C. Kerridge. 1983. Desmodium. In R.L.
Burt, P.P. Rotar, J.L. Walker, and M.W. Silvey (eds.) The role of
Centrosema, Desmodium, and Stylosanthes in improving tropical
pastures. Westview Tropical Agriculture Series No. 6. Westview
Press, Boulder, Colorado.


102
Family
F3
Individual
Regeneration Score
Progeny Parent
Callus Weight
Progeny Pare
507-16
14
1
1
90
126
507-16
15
2
1
120
126
507-16
16
1
1
150
126
507-16
17
2
1
90
126
507-16
18
3
1
180
126
507-17
2
3
2
380
344
507-17
3
1
2
370
344
507-17
4
3
2
180
344
507-17
5
5
2
340
344
507-17
7
1
2
150
344
507-17
8
1
2
90
344
507-17
10
1
2
40
344
507-17
11
4
2
130
344
507-17
12
1
2
100
344
507-17
13
2
2
200
344
507-17
14
2
2
430
344
507-17
15
2
2
40
344
507-17
17
1
2
240
344
507-17
19
2
2
120
344
507-17
20
1
2
120
344
507-17
21
1
2
70
344
507-17
22
1
2
250
344
507-17
23
1
2
230
344
507-17
24
1
2
70
344
507-18
1
2
1
300
358
507-18
2
1
1
60
358
507-18
3
2
1
340
358
507-18
4
2
1
130
358
507-18
5
1
1
150
358
507-18
6
1
1
380
358
507-18
7
1
1
90
358
507-18
8
2
1
190
358
507-18
9
1
1
320
358
507-18
10
4
1
790
358
507-18
11
1
1
340
358
507-18
12
2
1
190
358
507-18
13
1
1
200
358
501-1
1
1
3
100
160
501-1
2
2
3
120
160
501-1
3
3
3
150
160
501-1
4
1
3
150
160
501-1
5
1
3
120
160
501-1
7
1
3
200
160
501-1
8
3
3
210
160
501-1
9
3
3
280
160
501-1
10
4
3
170
160
501-1
11
1
3
70
160


69
Table 3-8. Continued.
Population Raw Data Uncensored Data
507 F3 Families:
Mean
Variance
Mean
Variance
N
N>1
507-1
1.571
0.8791
0.9004
2.974
14
6
507-2
1.769
2.190


13
3
507-3
1.500
0.3330


4
2
507-4
3.850
2.976
3.700
2.036
20
20
507-5
3.706
2.220
3.667
1.672
17
16
507-6
1.667
1.466


6
2
507-7
1.545
0.6406
1.635
0.7582
22
8
507-8
2.000
1.999
1.319
5.197
13
6
507-9
1.500
0.5770
1.438
0.953
14
5
507-10
1.273
0.4182


11
2
507-11
2.733
2.065
2.641
2.577
15
12
507-12
1.650
0.5553
1.753
0.5090
20
10
507-13
2.625
0.8383
2.625
0.6933
8
7
507-14
1.714
1.214
1.220
2.965
21
8
507-15
2.000
0.9091
2.356
1.188
12
7
507-16
1.611
0.6050
1.680
0.668
18
8
507-17
1.789
1.397
1.056
4.177
19
8
507-18
1.615
0.7564
1.386
1.365
13
6


74
Regeneration Score
Figure 3-5. Regeneration score histogram for genotype IRFL 6123.


CHAPTER 1
INTRODUCTION
The last two decades have seen a surge of interest in biodiversity within
the agricultural research community. In order to exploit diverse agricultural
environments in an efficient, environmentally benign manner, the full range of
available genetic resources must be examined and utilized. Extensive research
effort is aimed at developing new crops and identifying new regions of
application for existing crops. Desmodium (Desmodium spp.) is one of several
exotic forage legumes that show potential for use in the pastures of Florida and
the adjacent southeastern United States.
Desmodium is a large genus consisting primarily of perennial herbs and
shrubs. The genus is distributed in tropical and subtropical regions worldwide,
with a probable center of origin in Southeast Asia and a secondary center of
diversity in Mexico (Ohashi, 1973; Schubert, 1963). Classification within this
genus is difficult due to continuity of taxonomic characters among species, and
estimates of the number of species range from 200 (Takahashi, 1952) to 500
(Younge et al., 1964). Most of the species studied have chromosome numbers
of 2n=2x=22, and are predominantly self-pollinated (Chow and Crowder, 1972,
1973; Rotar and Uruta, 1967).
The agronomic role of the genus Desmodium has been reviewed by Imrie
et al. (1983). Several desmodium species are grown in significant acreages in
tropical or subtropical pastures. D. intortum (Mill.) Urb. (greenleaf desmodium)
and D. uncinatum (Jacq.) DC. (silverleaf desmodium) are widely used in humid
subtropical regions of Australia, Africa, and South America.
1


68
Table 3-8. Univariate statistics for regeneration score utilizing raw and
uncensored data.
Population Raw Data Uncensored Data
Mean
Variance
Mean
Variance
N
N>1
6123
4.053
1.240
4.075
1.080
38
37
501 F2 (all)
1.868
1.076
1.679
1.774
57
29
507 F2 (all)
1.994
1.952
1.278
4.351
80
43
501 F2 (parents only)
1.636
0.8552
2.078
0.4886
11
4
507 F2 (parents only)
2.267
2.924
1.553
6.529
17
8
501 F3
1.594
0.8158
1.185
1.882
165
67
507 F3
2.058
1.807
1.554
3.733
260
136
501 F3 families:
501-1
2.000
1.166
2.116
1.143
13
7
501-2
1.462
0.4359
1.520
0.533
13
6
501-3
1.333
0.4334
1.391
0.7310
21
5
501-4
2.083
1.356
1.982
1.948
12
7
501-5
2.417
2.629
2.045
3.069
12
8
501-6
1.440
0.5067
1.391
0.832
25
8
501-7
1.600
0.7115
1.612
0.986
10
5
501-8
1.214
0.1813


14
3
501-9
1.923
0.4102
2.032
0.3152
13
11
501-10
1.2143
0.1813


14
3
501-11
1.389
0.7222
-0.0665
4.920
18
4


ACKNOWLEDGEMENTS
I gratefully acknowledge Dr. D.S. Wofford for his sustained support,
guidance, and encouragement throughout my course of study. Thanks are also
extended to Dr. K.H. Quesenberry, who has generously provided insight and
assistance in conducting this project. I would like to thank the other members of
my committee, Dr. R.L. Smith, Dr. P.M. Lyrene, and Dr. G.A. Moore for their
interest in my program and their readiness to assist and to share their expertise.


71
Cross 501 Linear Cross 507 Linear
3.00-
Cross 501 Logarithmic
LL?
1.75-
Cross 507 Logarithmic

2.75-,
/
y'
/
/
2.50-
/
y*
a? 2.25-
/
A

^
y = 0.653x + 0.784
2.00-
\X
/ y = 0.769x + 0.387
* r2 = 0.400
4 r2 = 0.602
till
1.75 -
ni*r1 1
2.00 2.25 2.50 2.75 3.00 3.25
1.75 2.00 2.25 2.50 2.75 3.00
Cross 501 Square Root Cross 507 Square Root
Figure 3-2. Regression of F3 family mean callus weight on F2 parental callus
weight under linear, logarithmic, and square root scales.


I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in
scope and quality, as a dissertation for the degree of Doctor of Philosophy.
UUVIVJ viiwi u,
Associate Professor of Agronomy
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in
scope and quality, as a dissertation for the degree of Doctor of Philosophy.
Professor of Agronomy
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in
scope and quality, as a dissertation for the degree of Doctor of Philosophy.
Rex L. Smith
Professor of Agronomy
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in
scope and quality, as a dissertation for the degree of Doctor of Philosophy.
Paul M. Lyrene fJ
Professor of Horticultural Science
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in
scope and quality, as a dissertation for the degree of Doctor of Philosophy.
CL
Gloria A. Moore
Professor of Horticultural Science


54
containing no genetic variance (since genetic factors can also cause deviations
from normality) and having a mean significantly higher than the censoring
threshold. The regenerating parental population (genotype 6123) is such a
population, and a histogram of this population is presented in Figure 3-5. The
distribution is not seriously skewed, nor is it bimodal, suggesting that there is
little scale-induced non-normality.
Genetic analysis of the regeneration trait. Due to the difficulties presented
above, the types of analysis that can be performed on the regeneration trait are
severely limited. The joint scaling test cannot be used because this test
requires an estimate of the mean for each parent. Neither of the
nonregenerating parent populations (genotypes 144 and 13083) gave any
evidence of regeneration, so meaningful estimates of mean regeneration score
could not be obtained for these parents. This reiterates the case for analyzing
the uncensored rather than raw regeneration data. All calli for each
nonregenerating parental genotype received regeneration scores of one, so
based on the raw data the two genotypes have equal mean regeneration
scores. There is no reason to believe that both of the nonregenerating parents
have the same genetic proclivity for regeneration. The scores merely indicate
that neither genotype has enough proclivity to produce any visible evidence of
regeneration under the culture protocol used.
The iterative variance partitioning method used for the callus growth trait
could not be applied to the regeneration trait. This method requires good
estimates of the uncertainty associated with variance estimates and it was felt
that these were lacking for the regeneration trait. Mather and Jinks (1971)
described a simpler, nonweighted, noniterative variance partitioning method.
Preliminary efforts were made to apply this method to the regeneration trait.
Both raw and uncensored data were examined in this way, despite the


APPENDIX A
ANALYSIS OF VARIANCE FOR BETWEEN- AND WITHIN-FAMILY
VARIANCES FOR CALLUS GROWTH
Between- and within-family variances were estimated by analysis
of variance with F3 family as the class variable. All ANOVA
analyses yielded highly significant results. Expected mean square
for families equals (aw2 + nab2), where ow2 is within-family
variance (estimated by MSE), n is harmonic mean of number of
individuals per family, and ab2 is between-family variance. This
equation is solved for ob2 = (MSmode| MSE)/n. The following table
presents MSmode|, MSE, ab2, aw2, and n for each cross and each scale.
Cross
Scale
M^model
MSE(= aw2)
ab2
n
501
Linear
0.9901
0.0623
0.0669
15.21
501
Logarithmic
1.4962
0.0859
0.9271
11
501
Square Root
583.95
32.856
36.462
11
507
Linear
0.3166
0.0361
0.0220
12.73
507
Logarithmic
0.9910
0.1132
0.0689
ll
507
Square Root
273.78
29.392
19.199
ll
91


APPENDIX C
RESULTS OF SECOND AND THIRD ITERATIONS OF VARIANCE PARTITIONING
FOR CALLUS GROWTH


88
Ohashi, H. 1973. The Asiatic species of desmodium and its allied
genera (Leguminosae). Ginkgoana. Contributions of the flora of Asia
and the Pacific region. Academia Sci. Book, Inc., Tokyo.
Phillips, G.C., and G.B. Collins. 1979. In vitro tissue culture of
selected legumes and plant regeneration from callus cultures of red
clover. Crop Sci. 19:59-64.
Phillips, G.C., and G.B. Collins. 1980. Somatic embryogenesis from
cell suspension cultures of red clover. Crop Sci. 20:323-326.
Quesenberry, K.H, M.R. McKellar, and D.E. Moon. 1989. Evaluation and
hybridization of germplasm in the Desmodium heterocarpon -
Desmodium ovalifolium species complex. In Proc. XVI Inti. Grassland
Congress. Nice, France, p. 251-252.
Quesenberry, K.H., D.S. Wofford, P.A. Krottje, and R.L. Smith. 1992.
Production of transgenic red clover plants using Agrobacterium-
mediated DNA transfer. Proceedings of the Twelfth Trifolium
Conference. University of Florida, Gainesville, Florida, p. 5-10.
Quimio, C.A., and F.J. Zapata. 1990. Diallel analysis of callus
induction and green-plant regeneration in rice anther culture. Crop
Sci. 30:188-192.
Reisch, B., and E.T. Bingham. 1980. The genetic control of bud
formation from callus cultures of diploid alfalfa. Plant Sci. Lett.
20:71-77.
Rotar, P.P. and U. Urata. 1967. Cytological studies in the genus
Desmodium species: some chromosome counts. Amer. J. Bot. 54:1-4.
Rowe, K.E., and W.L. Alexander. 1980. Computations for estimating
the genetic parameters in joint-scaling tests. Crop Sci. 20:109-110.
Santos, A.V.P. dos, E.G. Cutter, and M.R. Davey. 1983. Origin and
development of somatic embryos in Medicago sativa L. (alfalfa).
Protoplasma 117:107-115.
SAS Institute. 1994. JMP Statistics and Graphics Guide. SAS
Institute Inc., Cary, North Carolina.


28
Figure 2-3. Multifoliate shoots produced by genotype IRFL 6123.


55
truncation-related biases in the raw data. Preliminary variance partitioning
yielded no significant (or even nearly significant) estimates for any parameters.
In the case of cross 501, this appeared to be due to an excessively high within-
family variance in the Fg and extremely low among-family Fg variance and Fg*
Fg covariance. For cross 507, the primary confounding factor appeared to be
an overly high Fg variance. In both cases these factors made it impossible to
obtain solutions to the simultaneous linear equations established in the
variance partitioning matrix.
The remaining option for genetic analysis of the regeneration trait is
parent-offspring regression. Regressions of Fg family mean on Fg parent for the
two crosses are presented in Figure 3-6. Cross 507 shows a strong parent
offspring relationship, with a moderately regression coefficient, or heritability.
The coefficient of determination (r2) is higher for the raw data than for the
uncensored data, but the slope is greater for the uncensored data. In the case
of cross 501, the regression is not significant. One reason for the poor
regression observed for cross 501 is the very small sample size, especially the
low number of parents with regeneration scores higher than one. Another
reason is the broad range of mean offspring regeneration scores for parents
with regeneration scores of one. It is likely that the "true" regeneration scores
for some of the parents (particularly the parent at the far lower left of the plot) lie
somewhat below the censoring threshold. If it were possible to measure
parental regeneration scores of less than one, the observations at the lower left
corner of the plot might be distributed farther to the left, resulting in a stronger
regression.
The previous discussion of both the callus growth and regeneration traits
contained the implicit assumption that the traits were quantitatively inherited.
For any genetic system there exists a continuum of possibilities, ranging from


33
have treated both callus growth and regeneration frequency as quantitatively
inherited traits. Diallel analysis of androgenesis in both wheat (Lazar et al.,
1984) and triticale (Charmet and Bernard, 1984) revealed highly significant
GCA, SCA, and reciprocal effects. In both cases, the GCA effect was
predominant, and the wheat study reported a narrow sense heritability of 0.6 for
plantlet formation. Earlier work involving a different set of wheat lines (Bullock
et al., 1982) found no significant reciprocal effects, underscoring the fact that the
results of any genetic study are highly specific to the germplasm under
consideration. Two studies conducted on the inheritance of androgenic ability
in barley yielded similarly disparate results. A study involving inbred lines and
reciprocal F-| hybrids (Foroughti-Wehr et al., 1982) found significant differences
among reciprocal crosses, while a similar study (Dunwell et al., 1987) involving
some of the same germplasm found no reciprocal differences. The latter work
included F2 and backcross generations and concluded that plantlet production
from anther culture was highly complex genetically, with large epistatic effects.
The work of Dunwell et al. (1987) also points out the importance of clearly
defining the response variable. When response was defined as number of
green plants produced per anther (the most common practice), the dominance
effect was large and positive, with no significant additive effect detected. When
response was expressed as percent responding anthers, however, the
dominance effect was negative and a large, positive additive effect was seen.
This study also reported that variation among different spikes on the same plant
exceeded variation due to genetic factors. This suggests that physiological
status of the anther can be a major determinant of capacity for androgenesis. A
diallel analysis in rice (Quimio and Zapata, 1990) found that callus production
and number of regenerated plants per anther were both influenced primarily by
additive genetic effects, with no significant reciprocal effects. Variance (Vr) and


27
Figure 2-2. Unifoliate shoots produced by genotype IRFL 6123.


18
column span a wide range, no significant differences were found. Failure to
detect statistically significant differences may be due to the low number of
observations (10 per treatment) and the large number of nonresponding
explants.
Direct regeneration. Direct regeneration from embryonic and seedling
explants without an intervening callus phase has been demonstrated in several
legume species (Ammirato., 1983). In an effort to achieve this in desmodium,
the induction treatments applied to callus in the previous experiment were
applied to newly excised seedling hypocotyls of three genotypes. Results were
disappointing. Genotype UF 20 produced small quantities of callus on all
treatments, but showed no sign of regeneration. No growth whatsoever was
seen in UF 144. At 2,4-D concentrations of 0.1 and 0.3 mg L"1, IRFL 6123
hypocotyls showed no growth. At the two higher 2,4-D concentrations IRFL
6123 produced roots from the radicle end of nearly every explant. When these
were transferred to shoot elongation medium, tiny clusters of budlike structures
less than 1 mm long formed at the end of the hypocotyl opposite the roots, but
failed to develop further.
Effect of duration of shoot bud induction treatment. Since IRFL 6123 was
the single responding genotype in the experiments described above, this
experiment and the remainder of the chapter focus on optimizing protocols for
this genotype. This experiment was designed not to enhance regeneration
response, but rather to determine to what extent shoot bud induction period can
be reduced without a corresponding reduction in regeneration.
Results presented in Table 2-2 show an increase in the mean number of
shoot buds per explant through 14 days of induction and a leveling off
thereafter. Despite the higher number of observations per treatment (21) than in
the previous experiments, no significant differences were seen among the 3, 7,


31
seedling hypocotyl-derived callus was subject to primarily additive genetic
effects. Narrow sense heritabilities ranged from 0.25 to 0.54, depending on
culture medium. Analysis of a complete diallel in tomato (Frankenberger et al.,
1981) revealed that number of shoots produced per leaf disc explant was
controlled predominantly by additive genetic effects, and the narrow sense
heritability for this trait was estimated to be 0.98. A similar study in pigeonpea
(Kumar et al., 1985) showed that callus growth was controlled largely by
additive gene effects, while the number of shoots regenerated per cotyledon
explant was subject to primarily nonadditive effects. Komatsuda et al. (1990)
performed a detailed diallel analysis of callus proliferation and shoot
regeneration from immature embryo explants in barley. Narrow sense
heritabilities were approximately 0.7 for both traits. Significant dominant effects
were detected for both traits, and significant epistatic effects were observed for
shoot regeneration.
Evidence for qualitative inheritance of regeneration capacity has been
reported in tomato, alfalfa and petunia. Koornneef et al. (1987) examined
regeneration from leaf disc-derived callus in the progeny of a cross between
tomato (Lycopersicon esculentum) and L. peruvianum. Segregation ratios in
F2, Fg, and backcross generations indicated that the trait was controlled by two
dominant, complementary loci contributed by L. peruvianum.
Bingham et al. (1975) found that the frequency of regeneration from
hypocotyl-derived callus in alfalfa could be increased from 12% to 67% in two
cycles of recurrent phenotypic selection. Reisch and Bingham (1980) examined
several F-j, F2, and backcross populations produced from crosses between
regenerating and nonregenerating clones of diploid alfalfa. Segregation ratios
were consistent with a two gene model in which both genes were dominant and
the presence of the dominant allele at either locus resulted in low frequency


23
Table 2-2. Response of genotype IRFL 6123 to duration of shoot bud induction
treatment.
Length of induction
period, days
Mean number
of buds per
explant t
Mean percent
responding
explants, % t
0
0.43 c
28 b
3
1.04 b
67 a
7
1.14 b
38 ab
14
2.52 a
57 a
28
2.50 a
61 a
t Mean separation by Tukey's HSD (a=0.05).


104
Family F3 Regeneration Score Callus Weight
Individual Progeny Parent Progeny Parent
501-4
10
1
1
1740
1352
501-4
11
3
1
1380
1352
501-4
12
4
1
230
1352
501-5
3
1
3
80
382
501-5
7
6
3
230
382
501-5
8
3
3
170
382
501-5
9
2
3
130
382
501-5
10
1
3
110
382
501-5
11
2
3
150
382
501-5
12
2
3
230
382
501-5
13
1
3
50
382
501-5
17
5
3
120
382
501-5
18
2
3
90
382
501-5
20
3
3
130
382
501-5
21
1
3
100
382
501-6
1
1
1
360
284
501-6
2
1
1
200
284
501-6
3
1
1
130
284
501-6
4
2
1
120
284
501-6
5
1
1
110
284
501-6
6
2
1
220
284
501-6
7
1
1
380
284
501-6
8
1
1
610
284
501-6
9
1
1
320
284
501-6
10
3
1
520
284
501-6
11
1
1
240
284
501-6
12
1
1
530
284
501-6
13
1
1
360
284
501-6
14
1
1
150
284
501-6
15
2
1
990
284
501-6
16
1
1
290
284
501-6
18
2
1
1030
284
501-6
19
1
1
760
284
501-6
20
1
1
830
284
501-6
22
1
1
480
284
501-6
23
1
1
130
284
501-6
24
2
1
500
284
501-6
25
3
1
240
284
501-6
27
3
1
390
284
501-6
28
1
1
300
284
501-7
2
3
2
180
313
501-7
3
1
2
180
313
501-7
4
1
2
80
313
501-7
5
1
2
50
313
501-7
6
3
2
330
313
501-7
7
1
2
40
313


56
absolute control by a single gene, through control by one or a few major genes
in conjunction with a number of minor, or "modifier" genes, to classical
quantitative inheritance involving control by many genes with approximately
equal effects. The large error variances and non-ideal data structure of this
study make it impossible to formally distinguish among these possibilities.
However, several general observations and considerations combine to suggest
that both of the traits examined are controlled at least to some degree by
multiple genes.
First, both traits exhibited transgressive segregation. Callus weights of
many F3 individuals in both crosses greatly exceeded that of any parental
individual, and four F3 individuals from cross 507 received regeneration scores
of sevena score never achieved by the regenerating parent (Figure 3-7;
Appendix D). Transgressive segregation in crosses between inbred lines can
be attributable to either overdominance or the presence of positive-acting
alleles at different loci in each parent. True overdominance is so rare that many
authorities doubt its existence (Simmonds, 1979). If single-gene
overdominance existed for the regeneration trait, approximately one quarter of
F2 individuals would carry the overdominant genotype, and some proportion of
these would likely exhibit the overdominant (transgressive) phenotype. The
transgressive phenotype was not observed in the F2. Thus, there appear to be
at least two loci controlling the regeneration trait. In view of the moderately
large size of the F2 populations (57 for cross 501 and 80 for cross 507), the
absence of the transgressive phenotype in this generation suggests that more
than two loci may be involved.
A second indication that at least two loci control the regeneration trait is the
large difference in regeneration scores between Crosses 501 and 507, despite


93
eliminated. The coefficient for Ew for V-| P3 becomes the sum of the contribution
to this column from the V2f3 row (1/15.21 x 1) plus the contribution from the Eb
column in the Vip3 row (1/15.21), or 0.1314. The resulting table is shown
below:
D
H
£
LU
V1F2
1/2
1/4
1
V1F3
0.516435
0.0707175
0.1314803
V2F3
1/4
1/8
1
W1F23
1/2
1/8
0
E1
0
0
1
Derivation of coefficients for cross 507 are similar, but n for this cross is 12.73.


25
Table 2-4. Response of genotype IRFL 6123 to picloram and benzyladenine
levels in the shoot elongation medium.
Picloram BA Mean frequency responding explants t
cone. cone. Unifoliate Multifoliate
mg L"1 %
0.012
0.2
25
a
38
a
0.012
0.6
38
a
8
b
0.04
0.2
33
a
4
b
0.04
0.6
29
a
4
b
0.002
2.0
17
a
17
b
t Mean separation by Tukey's HSD (a=0.05).


12
cotyledon node tissue. Cotyledon explants consisted of the distal two-thirds of
the cotyledon, placed abaxial surface upward.
All culture protocols used L2 basal medium (Phillips and Collins, 1980)
solidified with 0.8 % (w:v) Phytagar (Gibco, Inc.; Island City, N.Y.). Growth
regulators were coautoclaved with the basal medium for 20 minutes at 121 C.
Cultures were maintained at 27 C with a 16/8 h light/dark cycle at an
illumination of approximately 100 pE m"2 sec'1. Callus induction and shoot
bud induction was carried out in 60 mm disposable petri dishes with two or
three explants per dish. For shoot elongation experiments, six calli were placed
in each 100 mm petri dish.
All experiments utilized completely randomized designs. For number of
shoots per explant, each explant comprised an observation. For percentage
responding explants, each petri dish comprised one observation. Prior to
statistical analysis, data on number of buds per explant were transformed to the
square root of the quantity, observed value plus one half, in order to render
variance independent of mean. Percentage data were transformed to arcsine of
the square root of the observed percentage.
Evaluation of explant sources. Twenty-one leaf discs, petioles, hypocotyls,
and cotyledons from each of genotypes UF 20, UF 144, IRFL 6123, and Cl AT
13083 were cultured in a three-step protocol based on that described by
Phillips and Collins (1980). Explants were placed on L2-based callus induction
medium containing 0.06 mg L"1 picloram (4-amino-3,5,6-trichloropicolinic acid)
and 0.1 mg L'1 BA (6-benzylaminopurine) for 28 days. Calli were then weighed
and transferred to shoot induction medium containing 1.0 mg L'1 2,4-D and
2.0 mg L'1 adenine, again for 28 days. Following the shoot induction treatment,
calli with or without visible shoot buds were transferred to shoot elongation


62
Table 3-3. Model coefficients used to construct joint scaling tests.
Population
Coefficient
m [d] [h]
Parent 1
1 -1 0
Parent 2
1 1 0
f2
1 0 0.5
f3
1 0 0.25
1
0
0.25


70
Figure 3-1. Relationship of Fg family means and variances for callus growth
under (a) linear, (b) logarithmic, and (c) square root scales.


42
Because only three parental populations were examined, meaningful
regression of mean versus variance for parental populations is not possible.
However, simple examination of parental means and variances (Table 3-2) can
shed further light on scale effects. If variance is independent of mean, the
variances of the three parental populations should be similar in spite of the
differences in means. The effects are similar to those seen in the F3
populations, but the square root transformation stands out in that it yields a
smaller relative range in variances (maximum variance 48% greater than
minimum) than either the linear (177%) or logarithmic (78%) scales.
In summary, a general statistical examination of the callus growth trait
reveals the clear inadequacies of linear scaling. Logarithmic and square root
scales are similar in their conformity to the statistical requirements of normality
and independent variance.
Joint scaling test for callus growth. A more genetically oriented approach
to scaling was developed by Cavalli (1952) and elaborated upon by Mather and
Jinks (1971). Known as the joint scaling test, in its simplest form this technique
involves estimating the effects of additive and dominant gene action on
generation means. This is done by establishing a series of simultaneous
equations, one for each generation for which data are available. The left side of
each equation is the observed generation mean. The right side consists of
three terms, each of which is composed of an unknown and an associated
coefficient. The unknowns are model parameters: the grand mean (essentially
the genetic-neutral component of the various generation means, designated as
m), additive genetic effect (designated [d]), and dominant genetic effect
(designated [h]). Coefficients correspond to the theoretical causal contributions
made by these parameters to each generation mean. Since there are three
unknowns, data from at least four generations (each parent represents one


7
failure of the radicle to produce a root (Tisserat et al., 1978; Sellars et al., 1990).
The situation can be further complicated by complete failure of the somatic
embryo to germinate, sometimes followed by adventitious bud production from
embryo tissue (Saunders et al., 1987; Vasil, 1987). Due to the frequent
occurrence of these and other developmental abnormalities during somatic
embryogenesis, histological examination is necessary to conclusively
distinguish between embryogenesis and organogenesis. Somatic embryos
are characterized by two critical features: a bipolar morphology with discrete
coleoptile and coleorhiza, and a lack of vascular connections to the surrounding
tissue (Haccuis, 1978).
A successful regeneration protocol depends on choosing the appropriate
type of tissue to initiate the in vitro culture. In easily cultured nonlegumes such
as carrot, almost any part of the plant taken at any developmental stage can
serve as an explant source (Ammirato, 1983). In most legumes, however, the
choice is quite limited. As a general rule, regenerable cultures are most easily
obtained from immature tissue. Somatic embryogenesis from immature
embryos or portions of embryos has been reported for white clover, peanut, and
soybean (Maheshwaran and Williams, 1984; Sellars et al., 1990). Mature
embryos have been used as explant sources for mung bean, pigeon pea,
peanut, and soybean (Mathews and Rao, 1984; Mehta and Mohan Ram, 1980;
McKently et al., 1990).
Various types of seedling tissue have been used as explant sources for a
wide range of legume species. Seedling hypocotyls are perhaps the most
widely utilized explant source for legumes, having been used for alfalfa, red
clover, pea, pigeon pea, and several other species (Bingham et al., 1975;
Phillips and Collins, 1979; Kumar et al., 1984; Oelck and Schneider, 1983).
Epicotyls or cotyledons have been used for pea, red clover, alfalfa, and pigeon


Table 3-4. Parameter estimates and associated probabilities obtained from joint scaling tests for callus growth.
Cross
Scale
Parameter
m
[d]
[h]
Whole Model
estimate
prob>F
estimate
prob>F
estimate
prob>F
X2
prob>x2
501
Linear
0.497
0.023
0.210
0.057
-0.417
0.091
0.481
0.45
507
Linear
0.401
0.247
0.149
0.567
-0.364
0.637
53.97
<0.001
501
Logarithmic
2.611
0.009
0.210
0.112
-0.672
0.139
2.086
0.24
507
Logarithmic
2.542
0.053
0.228
0.504
-0.582
0.601
89.17
<0.001
501
Square root
21.26
0.020
4.849
0.091
-11.73
0.128
1.360
0.36
507
Square root
19.20
0.148
4.313
0.555
-9.679
0.655
81.66
<0.001
CT>
CO


73
(a)
Normal Quantile
(b)
Normal Quantile
Figure 3-4. Normal quantile plots of a hypothetical (a) intact and (b) censored
population.


101
Family
f3
Individual
Regeneration Score
Progeny Parent
Callus Weight
Progeny Pare
507-14
1
3
2
640
220
507-14
2
1
2
160
220
507-14
3
1
2
630
220
507-14
4
2
2
120
220
507-14
5
1
2
540
220
507-14
6
1
2
730
220
507-14
7
1
2
800
220
507-14
8
1
2
720
220
507-14
9
1
2
580
220
507-14
10
4
2
270
220
507-14
11
1
2
420
220
507-14
12
1
2
380
220
507-14
13
1
2
250
220
507-14
14
4
2
650
220
507-14
15
1
2
170
220
507-14
16
2
2
580
220
507-14
17
4
2
540
220
507-14
18
1
2
360
220
507-14
19
2
2
330
220
507-14
20
1
2
920
220
507-14
21
2
2
420
220
507-15
1
3
1
240
141
507-15
3
2
1
240
141
507-15
5
2
1
120
141
507-15
6
1
1
70
141
507-15
7
1
1
200
141
507-15
8
1
1
200
141
507-15
9
3
1
130
141
507-15
10
1
1
150
141
507-15
11
3
1
130
141
507-15
12
3
1
140
141
507-15
14
1
1
70
141
507-15
15
3
1
100
141
507-16
1
2
1
160
126
507-16
2
1
1
60
126
507-16
3
1
1
130
126
507-16
4
3
1
160
126
507-16
5
2
1
180
126
507-16
6
3
1
210
126
507-16
7
1
1
190
126
507-16
8
1
1
120
126
507-16
9
2
1
110
126
507-16
10
1
1
210
126
507-16
11
1
1
160
126
507-16
12
1
1
70
126
507-16
13
1
1
90
126


3
distinguished by a leaf length/width ratio of 3.0 or greater, as opposed to 1.0 to
2.5 in ssp. heterocarpon and D. ovalifolium (Quesenberry et al., 1989), and
by a more erect growth habit.
Recent advances in biotechnology have extended the potential for genetic
improvement of both new and established crop species. Plant breeding has
traditionally relied upon genetic variation existing within the species of interest
or in closely related, sexually compatible species. Current genetic
transformation methods allow desirable genes to be moved between distantly
related or nonrelated species, thereby greatly expanding the potential gene
pool for any specific crop (Fraley et al., 1986). Among forage legumes,
Agrobacterium-med\a\ed genetic transformation has been used to insert
reporter genes into alfalfa (Deak et al., 1986), white clover (White and
Greenwood, 1987), and red clover (Quesenberry et al., 1992), and
Stylosanthes (Sarria et al. 1994). Future application of transformation
techniques to forage legumes may involve introduction of specific genes for
pathogen resistance, herbicide resistance, or forage quality.
Because genetic transformation occurs on a single cell basis, obtaining
transformed plants requires that whole plants be regenerated from single cells,
or from small cell aggregates. A protocol for efficient in vitro plantlet
regeneration is therefore highly desirable for transformation work. Besides
serving as a tool for genetic transformation, cell and tissue culture
methodologies can be useful for in vitro screening of germplasm for resistance
to pathogens, herbicides, and environmental stresses (Hughes, 1983).
Published work on the in vitro culture of desmodium is sparse. Angeloni et
al. (1988) regenerated D. affine and D. incanum from shoot tip cultures using
a modified MS medium (Murashige and Skoog, 1962), but failed to obtain
regeneration from leaf or anther explants. Using seedling hypocotyl explants


28-day callus induction step with 0.06 mg L1 picloram (4-amino-3,5,6-
trichloropicolinic acid) and 0.1 mg L1 BA (6-benzylaminopurine), a 14-day
shoot bud induction step with 1.0 mg L"1 2,4-D and 2.0 mg L'1 adenine, and a
28-day shoot elongation step with 0.012 mg L1 picloram and 0.2 mg L'1 BA.
Under this protocol, shoot production occurred in 63% of explants.
Extensive efforts at hybridizing regenerating and nonregenerating
genotypes yielded two crosses. Callus growth and regeneration capacity were
evaluated in the parental lines, the F2 and F3 generations. Based on joint
scaling tests and variance partitioning, neither trait was found to fit a simple
additive-dominant genetic model. Both traits were moderately to highly
heritable, as determined by parent-offspring regression. Heritability of callus
growth ranged from 0.52 to 0.77, depending on the cross and on the numerical
scale employed. Heritability of the regeneration trait ranged from 0.14 to 0.46.
Members of two F3 families exhibited much more vigorous and prolific
regeneration than the regenerating parental genotype.
VII


Results of variance partitioning third iteration:
Component
Cross
Scale
D
H
Ew
Whole Model
Estimate
Prob>F
Estimate
Prob>F
Estimate
Prob>F
F
Prob>F
501
Linear
0.1372
0.4024
-0.0692
0.8090
0.0323
0.1528
28.57
0.1365
507
Square Root
23.90
0.1999
46.37
0.2830
18.21
0.0500
336.9
0.0400


4
from six diverse desmodium genotypes, Wofford et al. (1992b) evaluated two
tissue culture protocols that had been previously demonstrated effective for
legumes. Regeneration was only observed in a single genotype of D.
heterocarpon ssp. angustifolium, using an L2 based protocol originally
developed for use with Trifolium (Phillips and Collins, 1980).
If in vitro techniques are to be most effectively employed for the genetic
improvement of desmodium, it will first be necessary to broaden the range of
regenerating genotypes. This may be accomplished either by modification of
the culture protocol or by breeding for regeneration ability. Nonregenerating
genotypes can often be induced to regenerate by manipulating growth regulator
levels or any of several other culture parameters. In many species, however,
the majority of genotypes have resisted exhaustive efforts to induce
regeneration (Flick et al., 1983; Ammirato, 1983). Breeding for regeneration
ability has been effective for other forage crop species, including alfalfa (Reisch
and Bingham, 1980) and red clover (Quesenberry et al. 1992). The efficiency of
such a breeding effort is dependent on the genetic basis of the regeneration
trait. Investigations of the inheritance of this trait in several higher plant species
has shown that the trait may be either qualitatively or quantitatively controlled,
and can be subject to cytoplasmic effects (Keyes et al., 1980; Kumar et al.,
1985; Wan et al., 1988).
The remainder of this dissertation addresses both of these approaches to
broadening the range of regenerable desmodium lines. Chapter 2 examines a
variety of culture protocols with the objectives of inducing regeneration in a
wider range of genotypes and optimizing the response of established
regenerator genotypes. In Chapter 3 the genetic basis of callus growth and
regeneration is examined through the production of hybrids between
regenerating and nonregenerating lines and evaluation of the resulting F2 and


LIST OF FIGURES
Figure page
2-1 Response in genotype IRFL 6123 to 2,4-D concentration in shoot
induction medium 26
2-2 Unifoliate shoots produced by genotype IRFL 6123 27
2-3 Multifoliate shoots produced by genotype IRFL 6123 28
2-4 Spatulate shoots produced by genotype IRFL 6123 in the presence
of gibberellic acid (GA3) 29
3-1 Relationship of F3 family means and variances for callus growth
under (a) linear, (b) logarithmic, and (c) square root scales 70
3-2 Regression of F3 family mean callus weight on F2 parental callus
weight under linear, logarithmic, and square root scales. ... 71
3-3 Two hypothetical distributions with three response classes and two
thresholds 72
3-4 Normal quantile plots of a hypothetical (a) intact and (b) censored
population 73
3-5 Regeneration score histogram for genotype IRFL 6123 .... 74
3-6 Regression of F3 family mean regeneration score on F2 parent
regeneration score 75
3-7 Profuse regeneration exhibited by members of two F3 families
derived from cross 507 76
3-8 Regression of callus weight on regeneration score for combined F3
populations 77
v


47
The method of Mather and Jinks employs an iterative approach for
determining model weights. Weights for each of the matrix rows are initially
established as the reciprocals of the theoretical variances of each observed
variance or covariance. The variance of a variance is equal to 2(V2)/df, where
V is the observed variance and df is the number of degrees of freedom on which
the observed variance is based. The variance of a covariance is equal to
(W2 + V-|V2)/df, where W is the observed covariance, V-| and V2 are the
variances of the populations from which the covariance was obtained, and df is
the degrees of freedom upon which the covariance is based. The model is then
solved, and predicted (Mather and Jinks use the term expected) observational
variances are obtained. These are used to derive new weigh estimates, and
the model is solved again. This is continued until no improvement in model fit is
obtained.
Coefficients used in variance partitioning for the callus growth trait are
presented in Table 3-5. Observational variance components and initial weights
are presented in Table 3-6. Results of the first iterations are given in Table 3-7.
In all cases, second and third iterations failed to improve model fits. (Results of
later iterations are presented in Appendix C.) The linear and logarithmic scales
failed to produce significant whole-model F tests for either cross. The square
root scale produced significant whole-model tests for both crosses. For cross
501, estimated additive genetic variance was large, and significantly different
from zero (assuming a rather generous a of 0.1), while dominant variance was
smaller and not significantly different from zero. Cross 507 yielded no
significant genetic variance estimates.
The analysis presented thus far gives a rather unclear and somewhat
contradictory picture of the genetic mechanism controlling the callus growth trait
in the two crosses. Both the joint scaling and variance partitioning approaches


21
performed to examine the effect of increasing picloram concentration and
manipulating picloram/BA ratios. The most effective treatment from the previous
experiment was included as a check. Results of this experiment are presented
in Table 2-4. Just as in the previous experiment, unifoliate shoots substantially
outnumbered the more vigorous multifoliate shoots. In contrast to the previous
experiment, all of the leaves produced were relatively normal in morphology.
No significant differences among treatments were observed for unifoliate shoot
production, but the 0.012 mg L"1 picloram, 0.2 mg L'1 BA treatment yielded
significantly more multifoliate shoots (38%) than the other treatments, including
the best treatment from the previous experiment. It appears from these two
experiments that when auxin level is insufficient, high cytokinin levels can serve
a partial compensatory function in stimulating elongation. Superior elongation
response can be obtained with a somewhat higher auxin level than that used in
the original experiment in combination with fairly low level of cytokinin.
Histological examination of bud formation. As noted above, the method
employed to induce regeneration in this study is consistent with an
embryogenic pathway. The regeneration response appeared on a gross
morphologic level to be organogenic. Histological examination revealed a
mixture of these two regeneration mechanisms. At one and two weeks after
transfer to induction medium, a low frequency of somatic embryos with clearly
bipolar morphology and no vascular connections to the surrounding callus
could be seen. These were mixed with a much higher frequency of clearly
organogenic structures possessing a budlike morphology and definite vascular
connections to surrounding callus. At three weeks and later stages embryos
could no longer be found, while shoot buds at various stages of development
were easily observed. Thus, it appears that regeneration in this system is
predominantly via organogenesis.


Frequency Frequency
72
(a)
Figure 3-3. Two hypothetical distributions with three response classes and two
thresholds.


LIST OF TABLES
Table page
2-1 Response of genotype IRFL 6123 to 2,4-D and kinetin levels in the
shoot bud induction medium 22
2-2 Response of genotype IRFL 6123 to duration of shoot bud induction
treatment 23
2-3 Response of genotype IRFL 6123 to benzyladenine and gibberellic
acid levels in the shoot elongation medium 24
2-4 Response of genotype IRFL 6123 to picloram and benzyladenine
levels in the shoot elongation medium 25
3-1 Number of F2 plants, F3 plants, and F3 families from which in vitro
data were collected 59
3-2 Univariate statistics for parental, F2, and F3 populations in linear,
logarithmic, and square root scales 60
3-3 Model coefficients used to construct joint scaling tests 62
3-4 Estimates of m, [d], and [h] obtained from joint scaling tests for callus
growth 63
3-5 Coefficients used for variance partitioning for callus growth. 64
3-6 Observational variance components and initial weights used for
variance partitioning for callus growth 65
3-7 Estimates of causal variance components and associated
probabilities obtained from variance partitioning for callus growth. 67
3-8 Regeneration score means and variances derived from raw and
uncensored data sets 68
IV


45
positive effect for cross 507. The discrepancy between the dominance effect
predicted by the relationship of the progeny means to the mid-parent mean and
that predicted by the relationship between and F3 means accounts for the
generally poor joint scaling test results. These observations suggest that some
factor not considered in the model may be acting to depress both and F3
means relative to parental means. This factor could be a maternal or
cytoplasmic effectnuclear genes from parents 144 and 13083 may interact in
a negative manner with cytoplasmic genes from 6123 (the female parent in both
crosses). The depressed callus growth observed in cross 510 is consistent with
this explanation. An alternative cause would be negative epistatic interactions
between nuclear genes originating from the different parents. If either
explanation is correct, then the results of the additive-dominant joint scaling
tests are invalid. Unfortunately, there are insufficient generations available to
allow construction of an additive-dominant-maternal or additive-dominant-
epistatic joint scaling test.
Variance partitioning for the callus growth trait. Variance partitioning
methods are not subject to all of the limitations described for the joint scaling
test. Specifically, variance partitioning is not affected by maternal factors that
affect means, by distribution of positive-acting alleles between parents, nor by
dominance acting in different directions at different loci. A sophisticated method
of deriving genetic variance parameters from observed variances and
covariances in segregating generations derived from crosses of inbred lines is
presented by Mather and Jinks (1971). The technique is similar in many ways
to the joint scaling test, but observed variances and theoretical causal variance
components are utilized to construct the model. As with the joint scaling test,
fairly complex genetic models are possible, but, due to data limitations, the
present discussion will be limited to a simple additive-dominant model. In the


98
Family
f3
Individual
Regeneration Score
Progeny Parent
Callus Weight
Progeny Pare
507-4
10
5
6
100
180
507-4
11
3
6
60
180
507-4
13
2
6
470
180
507-4
14
7
6
200
180
507-4
15
5
6
730
180
507-4
16
2
6
140
180
507-4
19
5
6
430
180
507-4
20
4
6
110
180
507-4
21
3
6
150
180
507-4
22
7
6
320
180
507-4
23
3
6
210
180
507-4
24
5
6
270
180
507-5
1
1
5
590
975
507-5
2
6
5
890
975
507-5
3
3
5
510
975
507-5
4
3
5
430
975
507-5
5
4
5
450
975
507-5
6
4
5
650
975
507-5
7
3
5
480
975
507-5
8
4
5
1120
975
507-5
11
4
5
290
975
507-5
14
3
5
430
975
507-5
15
6
5
700
975
507-5
16
2
5
140
975
507-5
17
3
5
990
975
507-5
18
7
5
400
975
507-5
19
3
5
230
975
507-5
20
4
5
780
975
507-5
21
3
5
680
975
507-6
1
1
2
20
184
507-6
2
1
2
20
184
507-6
3
2
2
100
184
507-6
4
1
2
20
184
507-6
5
1
2
150
184
507-6
6
4
2
590
184
507-7
1
3
1
310
285
507-7
2
1
1
170
285
507-7
3
3
1
470
285
507-7
4
1
1
150
285
507-7
8
3
1
380
285
507-7
9
2
1
150
285
507-7
10
1
1
580
285
507-7
11
1
1
640
285
507-7
12
1
1
130
285
507-7
14
2
1
660
285
507-7
15
1
1
50
285


90
Tisserat, B., E.B. Esan, and T. Murashige. 1979. Somatic
embryogenesis in angiosperms. Hortic. Rev. 1:1-78.
Vasil, I.K. 1987. Developing cell and tissue culture systems for the
improvement of cereal and grass crops. J. Plant Physiol. 128:193-
218.
Viera, M.L., B. Jones, E.C. Cocking, and M.R. Davey. 1990. Plant
regeneration from protoplasts isolated from seedling cotyledons of
Stylosanthes guianensis, S. macrocephala and S. scabra. Plant Cell
Rep. 9:289-292.
Wan, Y., E.L. Sorensen, and G.H. Liang. 1988. Genetic control of in
vitro regeneration in alfalfa. Euphytica 39:3-9.
Walker, K.A., M.L. Wendeln, and E.G. Jaworski. 1979. Organogenesis in
callus tissue of Medicago sativa. The temporal separation of
induction processes from differentiation processes. Plant Sci. Lett.
16:23-30.
Webb, K.J., M.F. Fay, and P.J. Dale. 1987. An investigation of
morphogenesis within the genus Trifolium. Plant Cell, Tissue, Organ
Culture 11:37-46.
White, D.W.R., and D. Greenwood. 1987. Transformation of the forage
legume Trifolium repens L. using binary Agrobacterium vectors.
Plant Mol. Bio. 8:461-469.
Wofford, D.S., K.H. Quesenberry, and D.D. Baltensperger. 1992a. In
vitro culture responses of alyceclover genotypes on four media
systems. Crop Sci. 32:261-265.
Wofford, D.S., K.H. Quesenberry, and D.D. Baltensperger. 1992b. Tissue
culture regeneration of desmodium. Crop Sci. 32:266-268.
Younge, O.R., D.L. Plucknett, and P.P. Rotar. 1964. Culture and yield
performance of Desmodium intortum and D. canum in Hawaii. Tech.
Bull. No. 59. Hawaii Agrie. Exp. Stat., Hawaii.
Zar, J.H. 1984. Biostatistical analysis. 2nd Ed. Prentice Hall,
Englewood Cliffs, New Jersey.


IN VITRO MORPHOGENESIS AND INHERITANCE OF IN VITRO
TRAITS IN DESMODIUM
By
PETER A. KROTTJE
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1995

ACKNOWLEDGEMENTS
I gratefully acknowledge Dr. D.S. Wofford for his sustained support,
guidance, and encouragement throughout my course of study. Thanks are also
extended to Dr. K.H. Quesenberry, who has generously provided insight and
assistance in conducting this project. I would like to thank the other members of
my committee, Dr. R.L. Smith, Dr. P.M. Lyrene, and Dr. G.A. Moore for their
interest in my program and their readiness to assist and to share their expertise.

TABLE OF CONTENTS
page
ACKNOWLEDGMENTS ii
LIST OF TABLES iv
LIST OF FIGURES v
ABSTRACT vi
CHAPTERS
1 INTRODUCTION 1
2 OPTIMIZATION OF IN VITRO REGENERATION PROTOCOL 6
Materials and Methods 11
Results and Discussion 15
3 INHERITANCE OF IN VITRO REGENERATION AND
ASSOCIATED CHARACTERS 30
Materials and Methods 35
Results and Discussion 37
4 SUMMARY AND CONCLUSIONS 78
REFERENCES 83
APPENDICES
A DERIVATION OF BETWEEN- AND WITHIN-FAMILY
VARIANCES FOR CALLUS GROWTH 91
B DERIVATION OF VARIANCE PARTITIONING MATRICES
FOR CALLUS GROWTH 92
C RESULTS OF LATER ITERATIONS OF CALLUS GROWTH
VARIANCE PARTITIONING 94
D RAW CALLUS GROWTH AND REGENERATION DATA 97
BIOGRAPHICAL SKETCH 107

LIST OF TABLES
Table page
2-1 Response of genotype IRFL 6123 to 2,4-D and kinetin levels in the
shoot bud induction medium 22
2-2 Response of genotype IRFL 6123 to duration of shoot bud induction
treatment 23
2-3 Response of genotype IRFL 6123 to benzyladenine and gibberellic
acid levels in the shoot elongation medium 24
2-4 Response of genotype IRFL 6123 to picloram and benzyladenine
levels in the shoot elongation medium 25
3-1 Number of F2 plants, F3 plants, and F3 families from which in vitro
data were collected 59
3-2 Univariate statistics for parental, F2, and F3 populations in linear,
logarithmic, and square root scales 60
3-3 Model coefficients used to construct joint scaling tests 62
3-4 Estimates of m, [d], and [h] obtained from joint scaling tests for callus
growth 63
3-5 Coefficients used for variance partitioning for callus growth. 64
3-6 Observational variance components and initial weights used for
variance partitioning for callus growth 65
3-7 Estimates of causal variance components and associated
probabilities obtained from variance partitioning for callus growth. 67
3-8 Regeneration score means and variances derived from raw and
uncensored data sets 68
IV

LIST OF FIGURES
Figure page
2-1 Response in genotype IRFL 6123 to 2,4-D concentration in shoot
induction medium 26
2-2 Unifoliate shoots produced by genotype IRFL 6123 27
2-3 Multifoliate shoots produced by genotype IRFL 6123 28
2-4 Spatulate shoots produced by genotype IRFL 6123 in the presence
of gibberellic acid (GA3) 29
3-1 Relationship of F3 family means and variances for callus growth
under (a) linear, (b) logarithmic, and (c) square root scales 70
3-2 Regression of F3 family mean callus weight on F2 parental callus
weight under linear, logarithmic, and square root scales. ... 71
3-3 Two hypothetical distributions with three response classes and two
thresholds 72
3-4 Normal quantile plots of a hypothetical (a) intact and (b) censored
population 73
3-5 Regeneration score histogram for genotype IRFL 6123 .... 74
3-6 Regression of F3 family mean regeneration score on F2 parent
regeneration score 75
3-7 Profuse regeneration exhibited by members of two F3 families
derived from cross 507 76
3-8 Regression of callus weight on regeneration score for combined F3
populations 77
v

Abstract of Dissertation Presented to the Graduate School of the
University of Florida in Partial Fulfillment of the Requirements for
the Degree of Doctor of Philosophy
IN VITRO MORPHOGENESIS AND INHERITANCE OF IN VITRO TRAITS IN
DESMODIUM
By
Peter A. Krottje
August, 1995
Chairman: David S. Wofford
Major Department: Agronomy
Desmodium, Desmodium sp., is a forage legume widely cultivated in the
tropics and of growing importance in the southern United States. Previous work
aimed at application of biotechnological methods to this crop had obtained
limited in vitro regeneration from seedling hypocotyl explants in a single
genotype. The objective of the present work was to examine the potential for
improving regeneration response in desmodium through optimization of culture
protocols and through genetic improvement.
Efforts at identifying alternative explant sources focused on seedling
cotyledons, mature leaf disks, and mature petioles. Regeneration could not be
induced from any of these explant types. Several combinations of 2,4-D and
kinetin concentrations in the shoot bud induction medium were examined in an
unsuccessful effort to obtain regeneration from hypocotyl explants of recalcitrant
genotypes. Direct regeneration from hypocotyl explants without an intervening
callus growth step was also unsuccessful. Several treatments were
investigated with the aim of enhancing regeneration response in the previously
identified regenerating genotype. The best protocol identified consisted of a
VI

28-day callus induction step with 0.06 mg L1 picloram (4-amino-3,5,6-
trichloropicolinic acid) and 0.1 mg L1 BA (6-benzylaminopurine), a 14-day
shoot bud induction step with 1.0 mg L"1 2,4-D and 2.0 mg L'1 adenine, and a
28-day shoot elongation step with 0.012 mg L1 picloram and 0.2 mg L'1 BA.
Under this protocol, shoot production occurred in 63% of explants.
Extensive efforts at hybridizing regenerating and nonregenerating
genotypes yielded two crosses. Callus growth and regeneration capacity were
evaluated in the parental lines, the F2 and F3 generations. Based on joint
scaling tests and variance partitioning, neither trait was found to fit a simple
additive-dominant genetic model. Both traits were moderately to highly
heritable, as determined by parent-offspring regression. Heritability of callus
growth ranged from 0.52 to 0.77, depending on the cross and on the numerical
scale employed. Heritability of the regeneration trait ranged from 0.14 to 0.46.
Members of two F3 families exhibited much more vigorous and prolific
regeneration than the regenerating parental genotype.
VII

CHAPTER 1
INTRODUCTION
The last two decades have seen a surge of interest in biodiversity within
the agricultural research community. In order to exploit diverse agricultural
environments in an efficient, environmentally benign manner, the full range of
available genetic resources must be examined and utilized. Extensive research
effort is aimed at developing new crops and identifying new regions of
application for existing crops. Desmodium (Desmodium spp.) is one of several
exotic forage legumes that show potential for use in the pastures of Florida and
the adjacent southeastern United States.
Desmodium is a large genus consisting primarily of perennial herbs and
shrubs. The genus is distributed in tropical and subtropical regions worldwide,
with a probable center of origin in Southeast Asia and a secondary center of
diversity in Mexico (Ohashi, 1973; Schubert, 1963). Classification within this
genus is difficult due to continuity of taxonomic characters among species, and
estimates of the number of species range from 200 (Takahashi, 1952) to 500
(Younge et al., 1964). Most of the species studied have chromosome numbers
of 2n=2x=22, and are predominantly self-pollinated (Chow and Crowder, 1972,
1973; Rotar and Uruta, 1967).
The agronomic role of the genus Desmodium has been reviewed by Imrie
et al. (1983). Several desmodium species are grown in significant acreages in
tropical or subtropical pastures. D. intortum (Mill.) Urb. (greenleaf desmodium)
and D. uncinatum (Jacq.) DC. (silverleaf desmodium) are widely used in humid
subtropical regions of Australia, Africa, and South America.
1

2
D. heterocarpon (L.) DC. (carpon desmodium) has been released as the
cultivar 'Florida' for pastures in the southeastern United States (Kretchmer et al.,
1976). This species flowers earlier in the fall than either greenleaf or silverleaf,
and is thus adapted to areas where occasional early frosts threaten seed
production. D. ovalifolium Guill and Perr. is grown in Southeast Asia and has
attracted attention elsewhere for its tolerance to acid, low fertility soils and
shaded conditions (Schultz-Kraft and Pattanavibul, 1985). Other desmodium
species with agronomic potential include D. heterophyllum (Willd.) DC., D.
sandwicense E. May., and D. barbatum Benth.
Most desmodium cultivars are simply selections from germplasm
collections, and genetic improvement of this crop has received minimal
attention until the last decade. A desmodium breeding program was initiated at
the University of Florida in the late 1980's, with the primary breeding objectives
of improved forage quality, rapid establishment, increased seed production, and
resistance to root knot nematodes. The University of Florida program utilizes
germplasm from Desmodium heterocarpon ssp. heterocarpon, D. hetero
carpon ssp. angustifolium (Craig) Ohashi, and D. ovalifolium. The species
heterocarpon and ovalifolium are closely related and yield fertile progeny
upon intercrossing (Quesenberry et al., 1989). Some authors include the
species ovalifolium within species heterocarpon (Ohashi, 1983), but the two
are morphologically distinct and are generally regarded as separate entities by
agronomic researchers (Imrie et al., 1983; Schultze-Kraft and Benavides, 1988).
D. heterocarpon produces elongated inflorescences and glabrous to slightly
pubescent seed pods, while D. ovalifolium has compact inflorescences and
bears heavily pubescent pods. Leaves are opaque in ovalifolium, glabrous to
slightly pubescent in heterocarpon ssp. heterocarpon, and coriacious in
heterocarpon ssp. angustifolium. Subspecies angustifolium is further

3
distinguished by a leaf length/width ratio of 3.0 or greater, as opposed to 1.0 to
2.5 in ssp. heterocarpon and D. ovalifolium (Quesenberry et al., 1989), and
by a more erect growth habit.
Recent advances in biotechnology have extended the potential for genetic
improvement of both new and established crop species. Plant breeding has
traditionally relied upon genetic variation existing within the species of interest
or in closely related, sexually compatible species. Current genetic
transformation methods allow desirable genes to be moved between distantly
related or nonrelated species, thereby greatly expanding the potential gene
pool for any specific crop (Fraley et al., 1986). Among forage legumes,
Agrobacterium-med\a\ed genetic transformation has been used to insert
reporter genes into alfalfa (Deak et al., 1986), white clover (White and
Greenwood, 1987), and red clover (Quesenberry et al., 1992), and
Stylosanthes (Sarria et al. 1994). Future application of transformation
techniques to forage legumes may involve introduction of specific genes for
pathogen resistance, herbicide resistance, or forage quality.
Because genetic transformation occurs on a single cell basis, obtaining
transformed plants requires that whole plants be regenerated from single cells,
or from small cell aggregates. A protocol for efficient in vitro plantlet
regeneration is therefore highly desirable for transformation work. Besides
serving as a tool for genetic transformation, cell and tissue culture
methodologies can be useful for in vitro screening of germplasm for resistance
to pathogens, herbicides, and environmental stresses (Hughes, 1983).
Published work on the in vitro culture of desmodium is sparse. Angeloni et
al. (1988) regenerated D. affine and D. incanum from shoot tip cultures using
a modified MS medium (Murashige and Skoog, 1962), but failed to obtain
regeneration from leaf or anther explants. Using seedling hypocotyl explants

4
from six diverse desmodium genotypes, Wofford et al. (1992b) evaluated two
tissue culture protocols that had been previously demonstrated effective for
legumes. Regeneration was only observed in a single genotype of D.
heterocarpon ssp. angustifolium, using an L2 based protocol originally
developed for use with Trifolium (Phillips and Collins, 1980).
If in vitro techniques are to be most effectively employed for the genetic
improvement of desmodium, it will first be necessary to broaden the range of
regenerating genotypes. This may be accomplished either by modification of
the culture protocol or by breeding for regeneration ability. Nonregenerating
genotypes can often be induced to regenerate by manipulating growth regulator
levels or any of several other culture parameters. In many species, however,
the majority of genotypes have resisted exhaustive efforts to induce
regeneration (Flick et al., 1983; Ammirato, 1983). Breeding for regeneration
ability has been effective for other forage crop species, including alfalfa (Reisch
and Bingham, 1980) and red clover (Quesenberry et al. 1992). The efficiency of
such a breeding effort is dependent on the genetic basis of the regeneration
trait. Investigations of the inheritance of this trait in several higher plant species
has shown that the trait may be either qualitatively or quantitatively controlled,
and can be subject to cytoplasmic effects (Keyes et al., 1980; Kumar et al.,
1985; Wan et al., 1988).
The remainder of this dissertation addresses both of these approaches to
broadening the range of regenerable desmodium lines. Chapter 2 examines a
variety of culture protocols with the objectives of inducing regeneration in a
wider range of genotypes and optimizing the response of established
regenerator genotypes. In Chapter 3 the genetic basis of callus growth and
regeneration is examined through the production of hybrids between
regenerating and nonregenerating lines and evaluation of the resulting F2 and

5
F3 generations. Chapter 4 presents a summary and conclusions, and
discusses the implications of this work within the broad context of genotype by
environment interaction under in vitro conditions.

CHAPTER 2
OPTIMIZATION OF IN VITRO REGENERATION PROTOCOL
Although numerous members of the family Leguminoseae have been
regenerated in vitro, the family is regarded to be difficult with respect to in vitro
regeneration (Flick et al., 1983). Regeneration frequency is low for many
species, and specific culture requirements can vary widely both among and
within species (Phillips and Collins, 1983).
Regeneration in angiosperms can be accomplished via either of two
conceptually distinct pathwaysorganogenesis or somatic embryogenesis.
Both types of regeneration can be induced either directly from the initial explant
source, or from callus or suspension cells generated in culture. Organogenesis
usually involves production of a shoot meristem followed by shoot elongation
and rooting. Shoot and root regeneration are discrete processes occurring in
response to specific culture conditions, particularly the types and concentrations
of plant growth regulators in the medium. In somatic embryogenesis, shoot and
root meristems are produced simultaneously in a process similar to zygotic
embryo development. Maturation and germination of somatic embryos can be
induced by manipulating plant growth regulator levels, or may occur
spontaneously under constant culture conditions (Hazra et al., 1989).
In practice, the distinction between organogenesis and embryogenesis is
not always so clear (Ammirato, 1983). Ideally, somatic embryos should closely
resemble sexual embryos and possess clearly identifiable shoot, root, and
cotyledonary primordia that subsequently develop into their respective organs.
Deviations from this ideal situation include abnormal or absent cotyledons and
6

7
failure of the radicle to produce a root (Tisserat et al., 1978; Sellars et al., 1990).
The situation can be further complicated by complete failure of the somatic
embryo to germinate, sometimes followed by adventitious bud production from
embryo tissue (Saunders et al., 1987; Vasil, 1987). Due to the frequent
occurrence of these and other developmental abnormalities during somatic
embryogenesis, histological examination is necessary to conclusively
distinguish between embryogenesis and organogenesis. Somatic embryos
are characterized by two critical features: a bipolar morphology with discrete
coleoptile and coleorhiza, and a lack of vascular connections to the surrounding
tissue (Haccuis, 1978).
A successful regeneration protocol depends on choosing the appropriate
type of tissue to initiate the in vitro culture. In easily cultured nonlegumes such
as carrot, almost any part of the plant taken at any developmental stage can
serve as an explant source (Ammirato, 1983). In most legumes, however, the
choice is quite limited. As a general rule, regenerable cultures are most easily
obtained from immature tissue. Somatic embryogenesis from immature
embryos or portions of embryos has been reported for white clover, peanut, and
soybean (Maheshwaran and Williams, 1984; Sellars et al., 1990). Mature
embryos have been used as explant sources for mung bean, pigeon pea,
peanut, and soybean (Mathews and Rao, 1984; Mehta and Mohan Ram, 1980;
McKently et al., 1990).
Various types of seedling tissue have been used as explant sources for a
wide range of legume species. Seedling hypocotyls are perhaps the most
widely utilized explant source for legumes, having been used for alfalfa, red
clover, pea, pigeon pea, and several other species (Bingham et al., 1975;
Phillips and Collins, 1979; Kumar et al., 1984; Oelck and Schneider, 1983).
Epicotyls or cotyledons have been used for pea, red clover, alfalfa, and pigeon

8
pea (Malmberg, 1979; Phillips and Collins, 1980; Lupotto, 1983; Kumar et al.,
1984).
Among mature tissue types, regeneration has been obtained from leaf
explants in alfalfa and peanut (Oelck and Scheider, 1983; McKently et al.,
1992). Mature petioles have served as explant sources in various clover
species (Choo, 1988), and mature stems in birdsfoot trefoil and Stylosanthes
(Swanson and Tomes, 1980; Meijer, 1981).
Several different basal medium formulations have been used for tissue
culture of legume species. The majority of work has utilized Murashige and
Skoog's (1962) MS medium (Malmberg, 1979; Kao and Michayluk, 1981; Hazra
et al., 1989; McKently et al., 1990). Gamborg et al. (1968) developed the B5
medium for tissue culture of soybean, and this medium has proven useful for
several other legume species, including red clover (Quesenberry et al., 1992)
and birdsfoot trefoil (Swanson and Tomes, 1980). The B5 medium differs from
MS primarily by its greatly reduced ammonium nitrogen level and a lower
calcium concentration. Schenk and Hildebrandt's (1972) SH medium is similar
to B5, but contains much higher levels of inositol. This medium has been used
for Stylosanthes (Scowcroft and Adamson, 1976) and crimson clover (Horvath
et al., 1979). Phillips and Collins' (1979) L2 medium was developed for red
clover culture and has subsequently found application in the culture of peanut
and soybean (Sellars et al., 1990), and alyceclover (Wofford et al., 1992a). This
medium is somewhat higher in calcium than MS, B5, or SH, and also differs
from these media in that it lacks nicotinic acid. Blaydes' (1966) medium
contains less nitrate than the above media, and has been employed for
soybean, alfalfa (Bingham et al., 1975), and pigeonpea (Kumar et al., 1984).
It is generally acknowledged that plant growth regulator levels are of
critical importance in plant tissue culture. For angiosperms in general,

9
organogenesis can often be induced by "reversal transfer" of callus from a high
auxin, low cytokinin to a low auxin, high cytokinin medium (Dodds and Roberts,
1985). Somatic embryos can often be induced on media containing high levels
of auxinparticularly 2,4-Dsometimes in combination with low levels of
cytokinin (Ammirato, 1983). Reviews of legume work (Allavena, 1983;
Ammirato, 1983; Flick et al., 1983; Phillips and Collins, 1983) indicate that the
above generalities are applicable. Earlier work favored use of kinetin to induce
organogenesis, but more recently, BA (6-benzylaminopurine) generally in the
concentration range of 0.5 to 5.0 mg L"2, has been widely used (Webb et al.,
1987; Vieira et al., 1990; McKently et al., 1990). Somatic embryogenesis in
legumes has been induced with 2,4-D concentrations ranging from
0.001 mg L1 (Phillips and Collins, 1980) to 80 mg L'1 (Saunders et al., 1987).
In callus based regeneration systems, there is strong evidence that culture
conditions for the initial establishment of callus can strongly influence the
potential for subsequent organogenesis or embryogenesis (Saunders et al.,
1985; Tisserat et al., 1978). This issue has received relatively little attention in
legumes, and is generally approached empirically. Callus induction and
regeneration can sometimes occur on the same medium (Santos et al., 1983;
Bovo et al., 1986), but most work has employed a separate callus induction step
(Phillips and Collins, 1979; Walker et al., 1979; Flick et al., 1983).
Little has been published specifically on tissue culture of desmodium.
Angeloni et al (1988) produced multiple shoots from from shoot tip cultures of
D. incanum on MS medium supplemented with 1.0 mg L'1 NAA
(naphthalenacetic acid), 0.1 mg L'1 BA, and 1.0 mg L'1 GA (gibberellic acid).
Shoot tip culture is regarded as meristem cloning rather than true regeneration,
however. Wofford et al. (1992b) evaluated regeneration from seedling
hypocotyl explants of six genotypes of D. heterocarpon and D. ovalifolium

10
under two protocolsan MS-based procedure intended to induce
organogenesis and an L2-based procedure originally used by Collins and
Phillips (1982) to induce embryogenesis in red clover. The MS procedure
consisted of a callus induction step with 2.0 mg L'^ IAA (indole-3-acetic acid)
and 1.0 mg L'1 kinetin, shoot induction with 0.1 mg L"1 IAA and 4.0 mg L'1 BA,
and rooting on medium lacking plant growth regulators. The L2 protocol
consisted of callus induction with 0.06 mg L'1 picloram (4-amino-3,5,6-
trichloropicolinic acid) and 0.1 mg L'^ BA, embryo induction with 0.01 mg
2,4-D and 2.0 mg L"^ adenine, and embryo germination with 0.002 mg L'1
picloram and 0.2 mg L1 BA. The MS protocol resulted in production of shoot
meristems in five of six genotypes, but in all cases shoots failed to elongate.
The L2 protocol yielded shoot meristems in two genotypes and whole plants
were obtained in one genotype of D. heterocarpon ssp. angustifolium. The
response on L2 superficially resembled organogenesis, but histological
examination was not performed.
This chapter will evaluate a number of variations on the L2-based
procedure utilized by Wofford et al. (1992b). Primary objectives were to
determine if regeneration could be obtained in a wider range of desmodium
genotypes and to optimize the response of previously identified regenerating
genotypes. A bewildering range of culture variables can potentially influence
regeneration response. The work presented here attempts to examine some of
these variables in an orderly, stepwise manner in which the treatments in each
experiment are based on the results of the previous experiment. The work
begins with an examination of explant sources, proceeds to an investigation of
various shoot bud induction treatments, and then examines several shoot
elongation treatments. The chapter concludes with histological data presented
with the objective of clarifying the mode of regeneration in this system.

11
Materials and Methods
Germplasm. Four genotypes were selected to represent a wide range of in
vitro responses. In previous work (Wofford et al., 1992b) IRFL 6123 had been
identified as a strong regenerator, CIAT 13083 as a possible regenerator, and
UF 20 and UF 144 as nonregenerators. IRFL 6123 is classified as D. hetero-
carpon ssp. angustifolium, UF 20 represents D. heterocarpon ssp.
heterocarpon, UF 144 represents D. ovalifolium, and CIAT 13083 possesses
characteristics intermediate between D. heterocarpon ssp. heterocarpon and
D. ovalifolium.
General protocol. Young, fully expanded leaves from greenhouse-grown
plants were used to obtain leaf disc and petiole explants. Leaves and petioles
were sterilized by immersion in 70% ethanol for 30 seconds followed by
immersion in a 15% (v:v) Chlorox solution for one minute and several rinses
with sterile deionized water. A stainless steel cork borer was used to cut 5 mm
leaf discs, each containing a portion of midvein. Leaf discs were cultured with
abaxial surface upward. Petioles were cut into sections approximately 10 mm
in length.
For hypocotyl and cotyledon explants, seeds were scarified and sterilized
by a 12-minute immersion in concentrated sulfuric acid followed by several
rinses in sterile deionized water. Seeds were then placed in petri dishes
containing SGL medium (Collins and Phillips, 1982) and incubated at 27 C.
Hypocotyls and cotyledons were excised and placed onto callus induction
medium after the hypocotyls reached a length of 7 to 10 mm. This occurred
between 7 and 10 days after scarification. Hypocotyl explants were excised to a
length of approximately 5 mm, with care taken to avoid radicle tissue and

12
cotyledon node tissue. Cotyledon explants consisted of the distal two-thirds of
the cotyledon, placed abaxial surface upward.
All culture protocols used L2 basal medium (Phillips and Collins, 1980)
solidified with 0.8 % (w:v) Phytagar (Gibco, Inc.; Island City, N.Y.). Growth
regulators were coautoclaved with the basal medium for 20 minutes at 121 C.
Cultures were maintained at 27 C with a 16/8 h light/dark cycle at an
illumination of approximately 100 pE m"2 sec'1. Callus induction and shoot
bud induction was carried out in 60 mm disposable petri dishes with two or
three explants per dish. For shoot elongation experiments, six calli were placed
in each 100 mm petri dish.
All experiments utilized completely randomized designs. For number of
shoots per explant, each explant comprised an observation. For percentage
responding explants, each petri dish comprised one observation. Prior to
statistical analysis, data on number of buds per explant were transformed to the
square root of the quantity, observed value plus one half, in order to render
variance independent of mean. Percentage data were transformed to arcsine of
the square root of the observed percentage.
Evaluation of explant sources. Twenty-one leaf discs, petioles, hypocotyls,
and cotyledons from each of genotypes UF 20, UF 144, IRFL 6123, and Cl AT
13083 were cultured in a three-step protocol based on that described by
Phillips and Collins (1980). Explants were placed on L2-based callus induction
medium containing 0.06 mg L"1 picloram (4-amino-3,5,6-trichloropicolinic acid)
and 0.1 mg L'1 BA (6-benzylaminopurine) for 28 days. Calli were then weighed
and transferred to shoot induction medium containing 1.0 mg L'1 2,4-D and
2.0 mg L'1 adenine, again for 28 days. Following the shoot induction treatment,
calli with or without visible shoot buds were transferred to shoot elongation

13
medium with 0.002 mg L1 picloram and 0.2 mg L1 BA for another 28 days.
Shoot bud production and callus appearance were evaluated regularly
throughout the culture period.
Effect of 2.4-D level and kinetin level on shoot bud induction. Seedling
hypocotyls from genotypes UF 20, UF 144, and IRFL 6123 were cultured for
twenty eight days on callus induction medium as described above. Calli were
then transferred to the following L2-based shoot bud induction media:
(1) 0.1 mg L1 2,4-D, 2.0 mg L'1 adenine
(2) 0.3 mg L"1 2,4-D, 2.0 mg L"1 adenine
(3) 1.0 mg L'1 2,4-D, 2.0 mg L"1 adenine
(4) 3.0 mg L"1 2,4-D, 2.0 mg L"1 adenine
(5) 0.1 mg L'1 2,4-D, 0.15 mg L'1 kinetin, 2.0 mg L'1 adenine
(6) 1.0 mg L'1 2,4-D, 0.15 mg L'1 kinetin, 2.0 mg L"1 adenine
The 28-day shoot induction treatments were followed by a 28-day shoot
elongation treatment as described for the previous experiment. Ten explants
(two per dish) were used per genotype-treatment combination.
Induction of direct regeneration from hypocotyl explants without an
intervening callus induction step. This experiment was carried out identically to
the previous experiment with the exception that explants were placed directly
onto the six shoot induction treatments rather than onto callus induction
medium. Explants were rated after 28 days of shoot bud induction treatments
and after 28 days on elongation medium.
Effect of duration of shoot bud induction treatment. Twenty-eight-day
hypocotyl-derived calli from IRFL 6123 were transferred to shoot bud induction
medium containing 2.0 mg L'1 adenine and either 1.0 or 0.1 mg L*1 2,4-D for 0,

14
3, 7, 14, or 28 days. Following the induction treatment, calli were transferred to
the shoot elongation medium described above. The 28-day induction treatment
was followed by 28 days on elongation medium and the 14-, 7-, 3-, and 0-day
induction treatments were followed by 42, 49, 53, or 56 days on elongation
medium, respectively, to yield a total of 84 days in culture for each treatment.
Calli were evaluated for bud formation at frequent intervals throughout the
study. Twenty-one explants (three per dish) were used for each of the nine
treatments.
Effect of BA and GA3 (aibberellic acidt levels on shoot elongation.
Hypocotyls from genotype IRFL 6123 were cultured on the callus growth
medium described above for 28 days followed by 28 days on bud induction
medium containing 2.0 mg L'1 adenine and 1.0 mg L'1 2,4-D. Twenty-four calli,
each containing at least one well-formed shoot bud or small bud cluster, were
transferred to each of the following elongation media:
(1) 0.002 mg L1 picloram, 0.2 mg L'1 BA, 0 mg L1 GA3
(2) 0.002 mg L1 picloram, 0.2 mg L'1 BA, 0.2 mg L'1 GA3
(3) 0.002 mg L'1 picloram, 0.6 mg L'1 BA, 0 mg L'1 GA3
(4) 0.002 mg L1 picloram, 0.6 mg L'^ BA, 0.2 mg L'1 GA3
(5) 0.002 mg L"1 picloram, 2.0 mg L1 BA, 0 mg L'1 GA3
(6) 0.002 mg L'1 picloram, 2.0 mg L'1 BA, 0.2 mg L'^ GA3
Shoot elongation was visually evaluated after 28 days of elongation treatment.

15
Effect of picloram/BA ratio on shoot bud elongation. This experiment was
conducted in the same manner as the previous experiment, but with the
following elongation treatments:
(1) 0.012 mg L"1 picloram, 0.2 mg L"1 BA
(2) 0.012 mg L'1 picloram, 0.6 mg L'1 BA
(3) 0.04 mg L*1 picloram, 0.2 mg L'1 BA
(4) 0.04 mg L1 picloram, 0.6 mg L'1 BA
An additional elongation treatment of 0.002 mg L'1 picloram and
2.0 mg L'1 BA was included as a check.
Histological examination. After twenty-eight days of callus growth medium,
calli of IRFL 6123 were transferred to shoot induction medium containing
1.0 mg L"1 2,4-D and 2.0 mg L_1 adenine. Calli were removed for sectioning
after 1, 2, and 3 weeks on induction medium. Specimens were embedded in
Tissue-Tek O.T.C. Compound (Miles, Inc., Elkhart, IN) and sectioned on a CTF
Microtome-Cryostat (International Equipment Co., Needham, MA). Sections
eight microns in thickness were observed under the light microscope without
staining.
Results and Discussion
Evaluation of explant sources. Wofford et al. (1992b) showed that seedling
hypocotyls are satisfactory explant sources for in vitro regeneration in
desmodium. The use of hypocotyls presents certain practical problems,
however. Since the seedling must be dissected in order to place the hypocotyl

16
into culture, it is difficult or impractical to obtain both in vitro and in planta data
from a single individual. Because an individual seedling possesses only a
single hypocotyl, it is impossible to obtain replicated data on in vitro response
unless quantities of genetically uniform seed are available. In the hope of
eliminating these difficulties, an experiment was conducted to determine if
regenerable callus could be produced using cotyledon, leaf disc, or petiole
explants from genotypes IRFL 6123, CIAT 13083, or UF 20, with hypocotyl
explants included as a check.
Results were discouraging. Cotyledons from IRFL 6123 and CIAT 13083
produced small amounts of necrotic callus while UF 20 cotyledons produced
fairly abundant, deep green callus that showed no signs of regeneration. Leaf
disc explants followed a similar pattern, with IRFL 6123 and CIAT 13083
producing meager quantities of nonregenerating callus primarily from major
veins, and UF 20 producing larger quantities of nonregenerating callus.
Petioles yielded somewhat greater callus mass than either cotyledons or leaf
discs, but still failed to regenerate.
Hypocotyls yielded the greatest mass of callus from all genotypes. Ten of
21 hypocotyl-derived calli from IRFL 6123 produced shoot buds and elongated
shoots were obtained from three of these. Shoot regeneration was observed
from one UF 144 hypocotyl and one CIAT 13083 hypocotyl. In both cases, a
single shoot bud appeared to form directly from explant tissue. This response
could not be repeated in later experiments. It is possible that in both cases
regeneration was the result of a small portion of meristematic tissue from the
cotyledonary node being inadvertently included in the excised hypocotyl
explants. Thus, shoots may have resulted from meristem cloning rather than
true regeneration.

17
This work was rather limited in that the various explants were only tested
under a single culture protocol. While it is possible that a different protocol
might be more effective in inducing regenerable callus from leaf, petiole, or
cotyledon explants, it is also possible that these explants would prove
unresponsive to a wide range of protocols. In view of this uncertainty, further
investigation of alternative explant sources was abandoned, and the remainder
of this dissertation deals only with hypocotyl explants.
Effect of 2.4-D level and kinetin on shoot induction. The selection of
treatments for this experiment were based on reports that regeneration in other
legume species can be sensitive to 2,4-D concentration (Phillips and Collins,
1980; Saunders et al., 1987) and that low concentrations of kinetin in
combination with 2,4-D can stimulate regeneration (Sellars et al., 1990).
None of the 2,4-D treatments, either with or without kinetin, resulted in
shoot production in genotypes UF 20 or UF 144. Response of IRFL 6123 to
shoot induction treatments is presented in Table 2-1. The presence of kinetin in
the induction medium resulted in a slight browning of the callus and complete
inhibition of shoot formation. In the treatments lacking kinetin, a clear response
to 2,4-D concentration was observed. Analysis of variance (ANOVA) indicates
that with respect to mean number of shoot buds per explant, 1.0 mg L'1 2,4-D
was superior to 0.1 mg L'1, but not different from either 0.3 or 3.0 mg L~1.
Linear regression analysis of shoot buds per explant on 2,4-D level yields a
significant quadratic relationship with a maximum between 1.0 and 3.0 mg L'1
(Figure 2-1). Thus, although ANOVA fails to indicate clear superiority of the
1.0 mg L~1 treatment, regression suggests that this treatment is close to the
optimum 2,4-D level, and this is the concentration adopted for later work.
Consistency of regeneration response, as indicated by percent responding
explants, is presented in the last column of Table 2-1. Although values in this

18
column span a wide range, no significant differences were found. Failure to
detect statistically significant differences may be due to the low number of
observations (10 per treatment) and the large number of nonresponding
explants.
Direct regeneration. Direct regeneration from embryonic and seedling
explants without an intervening callus phase has been demonstrated in several
legume species (Ammirato., 1983). In an effort to achieve this in desmodium,
the induction treatments applied to callus in the previous experiment were
applied to newly excised seedling hypocotyls of three genotypes. Results were
disappointing. Genotype UF 20 produced small quantities of callus on all
treatments, but showed no sign of regeneration. No growth whatsoever was
seen in UF 144. At 2,4-D concentrations of 0.1 and 0.3 mg L"1, IRFL 6123
hypocotyls showed no growth. At the two higher 2,4-D concentrations IRFL
6123 produced roots from the radicle end of nearly every explant. When these
were transferred to shoot elongation medium, tiny clusters of budlike structures
less than 1 mm long formed at the end of the hypocotyl opposite the roots, but
failed to develop further.
Effect of duration of shoot bud induction treatment. Since IRFL 6123 was
the single responding genotype in the experiments described above, this
experiment and the remainder of the chapter focus on optimizing protocols for
this genotype. This experiment was designed not to enhance regeneration
response, but rather to determine to what extent shoot bud induction period can
be reduced without a corresponding reduction in regeneration.
Results presented in Table 2-2 show an increase in the mean number of
shoot buds per explant through 14 days of induction and a leveling off
thereafter. Despite the higher number of observations per treatment (21) than in
the previous experiments, no significant differences were seen among the 3, 7,

19
14, and 28 day treatments with respect to percent responding explants. The
zero day treatment yielded significantly fewer responding explants than all
except the 7 day treatment. The zero day treatment is essentially a reversal
transfer from high auxin, low cytokinin to low auxin, high cytokinin medium.
Reversal transfer is a widely utilized method for inducing organogenesis. The
rather poor response to this treatment relative to the 2,4-D treatments suggests
that regeneration may be occurring via somatic embryogenesis rather than
organogenesis. Histological evidence on this question will be presented later
in this chapter.
The fact that a percentage of IRFL 6123 calli produce shoot buds under all
treatments is evidence of this genotype's strong predisposition to regenerate.
The failure of many of the explants to produce shoot buds under any of the
treatments and the lack of significant differences among many of the treatments
underscores the high level of variability often encountered in tissue culture
work, and may also be an indication that we have not yet foundand may
never findthe optimal regeneration protocol for this recalcitrant species.
Effect of BA and GAg levels on shoot elongation. Both shoot meristem
formation and shoot elongation are difficult to induce in desmodium. In
previous work (Wofford et al., 1992b) as well as in the experiments described
above, a large percentage of shoot buds failed to develop into plants, either due
to poor elongation or failure to root. Three levels of BA were evaluated in the
first of two experiments intended to improve elongation response. Gibberellic
acid has been reported to stimulate development and elongation of shoot
meristems in cultured soybean and pigeonpea (Ghazi et al., 1986; Kumar et al.,
1983). To determine if this compound might be useful for desmodium, each BA
level was tested both with and without the addition of 0.2 mg L'1 GAg.

20
Results are presented in Table 2-3. The two right-hand columns of this
table represent two distinct types of elongation response. In the majority of
cases shoot development ceased at the single leaf stage and at a length of one
cm or less (Figure 2-2). These unifoliate shoots exhibited low vigor and could
be rooted only with difficulty by transfer to basal medium supplemented with
1.0 mg L'1 indoleacetic acid (IAA). A smaller proportion of buds developed into
multifoliate shoots (Figure 2-3). The more vigorous multifoliate shoots often
spontaneously formed roots after two to three weeks in elongation medium, or
could easily be induced to root upon transfer to basal medium lacking plant
growth regulators.
Percentage of buds showing the unifoliate elongation response decreased
with increasing BA level. In addition, a qualitative difference could be seen
between the treatments with GA3 and without. In the presence of GA3,
unifoliate shoots tended to bear abnormal, spatulate leaves (Figure 2-4).
Because the degree of abnormality was variable and graded continuously into
normal leaf development, abnormal and normal unifoliate elongation are not
differentiated in Table 2-3.
Significantly more multifoliate shoots were produced on the medium
containing 2.0 mg L'1 BA and lacking GA3 than on the other treatments. The
21% multifoliate shoot production obtained with this treatment was nonetheless
rather disappointing, and the following experiment was designed in an effort to
improve on this figure.
Effect of picloram/BA ratio on shoot elongation. All of the treatments in the
above experiment followed the precedent of Phillips and Collins (1980) and
Wofford et al. (1992b) in utilizing the relatively low picloram concentration of
0.002 mg L1. Since neither increasing BA level nor addition of GA3 produced
dramatic improvement of elongation response, a second experiment was

21
performed to examine the effect of increasing picloram concentration and
manipulating picloram/BA ratios. The most effective treatment from the previous
experiment was included as a check. Results of this experiment are presented
in Table 2-4. Just as in the previous experiment, unifoliate shoots substantially
outnumbered the more vigorous multifoliate shoots. In contrast to the previous
experiment, all of the leaves produced were relatively normal in morphology.
No significant differences among treatments were observed for unifoliate shoot
production, but the 0.012 mg L"1 picloram, 0.2 mg L'1 BA treatment yielded
significantly more multifoliate shoots (38%) than the other treatments, including
the best treatment from the previous experiment. It appears from these two
experiments that when auxin level is insufficient, high cytokinin levels can serve
a partial compensatory function in stimulating elongation. Superior elongation
response can be obtained with a somewhat higher auxin level than that used in
the original experiment in combination with fairly low level of cytokinin.
Histological examination of bud formation. As noted above, the method
employed to induce regeneration in this study is consistent with an
embryogenic pathway. The regeneration response appeared on a gross
morphologic level to be organogenic. Histological examination revealed a
mixture of these two regeneration mechanisms. At one and two weeks after
transfer to induction medium, a low frequency of somatic embryos with clearly
bipolar morphology and no vascular connections to the surrounding callus
could be seen. These were mixed with a much higher frequency of clearly
organogenic structures possessing a budlike morphology and definite vascular
connections to surrounding callus. At three weeks and later stages embryos
could no longer be found, while shoot buds at various stages of development
were easily observed. Thus, it appears that regeneration in this system is
predominantly via organogenesis.

22
Table 2-1. Response of genotype IRFL 6123 to 2,4-D and kinetin levels in the
shoot bud Induction medium.
2,4-D Kinetin
cone. cone.
mg L1
Mean number
of buds per
explant "I"
Mean percent
responding
explants, %t
0.1
0
0.5 b
30
a
0.3
0
2.2 ab
70
a
1.0
0
3.3 a
80
a
3.0
0
1.6 ab
50
a
0.1
0.15
0 c
0
b
1.0
0.15
0 c
0
b
t Mean separation by Tukey's honestly significant difference (HSD a=0.05).

23
Table 2-2. Response of genotype IRFL 6123 to duration of shoot bud induction
treatment.
Length of induction
period, days
Mean number
of buds per
explant t
Mean percent
responding
explants, % t
0
0.43 c
28 b
3
1.04 b
67 a
7
1.14 b
38 ab
14
2.52 a
57 a
28
2.50 a
61 a
t Mean separation by Tukey's HSD (a=0.05).

Table 2-3. Response of genotype IRFL 6123 to benzyladenine and
gibberellic acid levels in the shoot elongation medium.
24
BA GA3 Mean frequency responding explants t
cone. cone. Unifoliate Multifoliate
mg L'1 %
0.2
0
50
a
0
a
0.2
0.2
54
a
0
a
0.6
0
33
a
4
a
0.6
0.2
38
a
0
a
2.0
0
13
b
21
b
2.0
0.2
13
b
0
a
t Mean separation by Tukey's HSD (a=0.05).

25
Table 2-4. Response of genotype IRFL 6123 to picloram and benzyladenine
levels in the shoot elongation medium.
Picloram BA Mean frequency responding explants t
cone. cone. Unifoliate Multifoliate
mg L"1 %
0.012
0.2
25
a
38
a
0.012
0.6
38
a
8
b
0.04
0.2
33
a
4
b
0.04
0.6
29
a
4
b
0.002
2.0
17
a
17
b
t Mean separation by Tukey's HSD (a=0.05).

26
10-
8-
co
o 6 -\
o
-C
CO
*4
o
k
0
E
3
y = -1,697x 2 + 5.405x + 0.333
r2 = 0.302
X
<
2-
V
/
/
/
'
2,4-D Concentration, mg/L
Figure 2-1. Response of genotype IRFL 6123 to 2,4-D concentration in shoot
induction medium.

27
Figure 2-2. Unifoliate shoots produced by genotype IRFL 6123.

28
Figure 2-3. Multifoliate shoots produced by genotype IRFL 6123.

29
Figure 2-4. Spatulate shoots produced by genotype IRFL 6123 in the presence
of gibberellic acid (GA3).

CHAPTER 3
INHERITANCE OF IN VITRO REGENERATION AND ASSOCIATED
CHARACTERS
The high genotype-specificity of in vitro regeneration in desmodium has
been clearly demonstrated in the previous chapter. An understanding of the
genetic control of regeneration is desirable for the efficient application of
biotechnological methods to this crop. The mode of inheritance of regeneration
determines whether it is feasible to transfer the trait to agronomically desirable
nonregenerating lines and, if so, the appropriate method of accomplishing this.
The genetic basis of regeneration and associated in vitro traits has been
studied in few higher plant species. In most studies regeneration was
accomplished through organogenesis, and organogenesis is assumed in the
current discussion unless otherwise stated. Regardless of whether
organogenic or embryogenic response was examined, regeneration in most
species was treated as a quantitatively inherited trait, although evidence
supporting this assumption was often not presented. Several studies have
been directed at partitioning genetic variance components and determining
heritabilities for the regeneration trait in different crops. Buiatti et al. (1974)
performed a diallel analysis, without reciprocals, of callus growth and shoot
regeneration from flower petal explants in cauliflower. Additive gene effects
were high for both traits, and narrow sense heritability estimates were 0.81 for
callus growth and 0.09 for percent explants forming shoots. The low heritability
for the latter trait resulted from extremely high epistatic and error effects. Based
on North Carolina Design II matings among twenty four red clover genotypes,
Keyes et al. (1980) determined that embryogenic regeneration frequency from
30

31
seedling hypocotyl-derived callus was subject to primarily additive genetic
effects. Narrow sense heritabilities ranged from 0.25 to 0.54, depending on
culture medium. Analysis of a complete diallel in tomato (Frankenberger et al.,
1981) revealed that number of shoots produced per leaf disc explant was
controlled predominantly by additive genetic effects, and the narrow sense
heritability for this trait was estimated to be 0.98. A similar study in pigeonpea
(Kumar et al., 1985) showed that callus growth was controlled largely by
additive gene effects, while the number of shoots regenerated per cotyledon
explant was subject to primarily nonadditive effects. Komatsuda et al. (1990)
performed a detailed diallel analysis of callus proliferation and shoot
regeneration from immature embryo explants in barley. Narrow sense
heritabilities were approximately 0.7 for both traits. Significant dominant effects
were detected for both traits, and significant epistatic effects were observed for
shoot regeneration.
Evidence for qualitative inheritance of regeneration capacity has been
reported in tomato, alfalfa and petunia. Koornneef et al. (1987) examined
regeneration from leaf disc-derived callus in the progeny of a cross between
tomato (Lycopersicon esculentum) and L. peruvianum. Segregation ratios in
F2, Fg, and backcross generations indicated that the trait was controlled by two
dominant, complementary loci contributed by L. peruvianum.
Bingham et al. (1975) found that the frequency of regeneration from
hypocotyl-derived callus in alfalfa could be increased from 12% to 67% in two
cycles of recurrent phenotypic selection. Reisch and Bingham (1980) examined
several F-j, F2, and backcross populations produced from crosses between
regenerating and nonregenerating clones of diploid alfalfa. Segregation ratios
were consistent with a two gene model in which both genes were dominant and
the presence of the dominant allele at either locus resulted in low frequency

32
regeneration, while the presence of both dominant alleles yielded a high
regeneration frequency. Crosses between two different sets of nonregenerating
clones yielded no regenerating progeny, suggesting that inheritance was not
quantitative.
Wan et al. (1988) examined the genetics of regeneration from petiole
explants in tetraploid alfalfa, employing a different culture protocol from that of
Reisch and Bingham. Based on data from several F-| and S-| populations, it
was concluded that regeneration was controlled by a two gene system distinct
from that described by Reisch and Bingham. Duplicate recessive epistasis was
postulated; that is, the presence of dominant alleles at both loci was necessary
for any regeneration to occur. Variation in the frequency of regeneration among
regenerating clones was attributed to either dosage effects or the presence of
modifier genes.
Dulieu (1991) obtained rather tentative evidence of single gene control of
regeneration from hypocotyl explants in petunia. Two backcross populations
yielded approximately 1:3 ratios of "high regenerating" to combined "low
regenerating" and "nonregenerating" plants, but the distinction between high
and low level regeneration was somewhat arbitrary, and quantitative
inheritance could not be ruled out.
All of the studies discussed above have dealt with regeneration from
sporophytic (2n) tissue. The process of androgenesis, or embryogenesis from
tissue derived from gametophytic (n) pollen mother cells, is markedly distinct
from regeneration from sporophytic tissue. However, the two processes are
sufficiently similar that work conducted on the genetics of androgenesis may
provide insight into the genetics of regeneration from 2n tissue. Several papers
have been published on the genetic control of androgenesis in cereals, and all

33
have treated both callus growth and regeneration frequency as quantitatively
inherited traits. Diallel analysis of androgenesis in both wheat (Lazar et al.,
1984) and triticale (Charmet and Bernard, 1984) revealed highly significant
GCA, SCA, and reciprocal effects. In both cases, the GCA effect was
predominant, and the wheat study reported a narrow sense heritability of 0.6 for
plantlet formation. Earlier work involving a different set of wheat lines (Bullock
et al., 1982) found no significant reciprocal effects, underscoring the fact that the
results of any genetic study are highly specific to the germplasm under
consideration. Two studies conducted on the inheritance of androgenic ability
in barley yielded similarly disparate results. A study involving inbred lines and
reciprocal F-| hybrids (Foroughti-Wehr et al., 1982) found significant differences
among reciprocal crosses, while a similar study (Dunwell et al., 1987) involving
some of the same germplasm found no reciprocal differences. The latter work
included F2 and backcross generations and concluded that plantlet production
from anther culture was highly complex genetically, with large epistatic effects.
The work of Dunwell et al. (1987) also points out the importance of clearly
defining the response variable. When response was defined as number of
green plants produced per anther (the most common practice), the dominance
effect was large and positive, with no significant additive effect detected. When
response was expressed as percent responding anthers, however, the
dominance effect was negative and a large, positive additive effect was seen.
This study also reported that variation among different spikes on the same plant
exceeded variation due to genetic factors. This suggests that physiological
status of the anther can be a major determinant of capacity for androgenesis. A
diallel analysis in rice (Quimio and Zapata, 1990) found that callus production
and number of regenerated plants per anther were both influenced primarily by
additive genetic effects, with no significant reciprocal effects. Variance (Vr) and

34
covariance (Wr) analysis indicated that regeneration was largely controlled by
partially recessive genes.
The limited number of studies that have addressed phenotypic or genetic
correlations between in vitro regeneration and various in vitro and in planta
characters have yielded both expected and unexpected results. Oelck and
Scheider (1983) found that in Melilotus officinalis, Trifolium pratense, and T.
resupinatum, the ability to form shoots from callus was correlated with the
production of side shoots from in vitro shoot tip cultures. Kern et al. (1986)
observed a similar correlation between capacity for somatic embryogenesis and
in vitro axillary shoot development in soybean. These correlations may reflect a
generalized tendency among responding genotypes to form shoots, or may
simply indicate a tolerance to the specific culture conditions employed in the
study. It has been observed in alfalfa (Bingham et al., 1975; Brown and
Atanassov, 1985) and alyceclover (Wofford et al., 1992a) that regenerating
genotypes often have a creeping growth habit and readily produce adventitious
shoots. Thus it appears that in at least some cases the ability to regenerate in
vitro reflects a general proclivity for shoot production both in vitro and in planta.
Oelck and Scheider (1983) suggested that a general tendency to produce
adventitious shoots may be a useful indicator in preliminary screenings of
germplasm for regenerating lines.
Perhaps surprisingly, there appears to be little correlation between callus
growth rate and regeneration frequency ( Baroncelli et al., 1974; Kumar et al.,
1985; Lazar et al., 1984). However, shoot regeneration from callus can be
restricted in genotypes with very poor callus growth (Bingham et al., 1975). It is
well established that callus appearance frequently bears a strong relationship
to regeneration potential (Bingham et al., 1975; Ketchum et al., 1987; Delieu

35
1991), but due to the difficulty of quantifying callus appearance the relationship
has not been subjected to statistical analysis.
The primary objective of the work described in this chapter was to
determine the mode of genetic control of in vitro regeneration in desmodium. A
secondary objective was investigation of the mode of inheritance of callus
growth and examination of the relationship between callus growth and
regeneration.
Materials and Methods
Production of hybrids. Two regenerating and three nonregenerating lines
were selected for crossing. The regenerators, IRFL 6123 and IRFL 6128 are
classified as D. heterocarpon ssp. angustifolium, and are morphologically
distinguishable from the nonregenerating genotypes by their lanceolate,
coriacious leaves, upright growth habit, elongated racemes, and glabrous seed
pods. Lines 6123 and 6128 are very similar except for a distinct leaf mark on
6123 that is absent on 6128. All of the selected nonregeneratorsUF 20,
UF 144, and Cl AT 13083have ovate or obovate, noncoriacious leaves,
spreading growth habits, compact racemes, and pubescent pods. Genotype
UF 20 ('Florida' carpon) represents D. heterocarpon ssp. heterocarpon and is
distinguished by its thin, slightly pubescent leaves and prominent leaf marks.
Genotype UF 144 is classified as D. ovalifolium, and possesses thicker,
glabrous leaves. Genotype CIAT 13083 is intermediate between D.
heterocarpon and D. ovalifolium.
Parent plants were moved from the field to the greenhouse in April, 1989.
To induce flowering, a ten hour day length was simulated by covering the plants
with a tarpaulin from approximately 7:00 P.M. until 9:00 A.M. Pollinations were

36
done in June and July. Donor pollen was obtained by tripping flowers using a
toothpick with a small piece of fine sandpaper glued to one end in such a way
that the anthers and stigma would strike the sandpaper and pollen would
adhere. Pollen was then transferred to recipient flowers by tripping in a similar
manner. Recipient flowers were not emasculated because the flowers were
extremely sensitive to handling, and emasculation tended to result in flower
abscission. Pollinations were carried out at a time of day when the flowers had
fully opened, but self-tripping had not yet occurred. The specific time varied
according to night temperature and daytime cloud cover, but was generally
between 9:00 A.M. and noon. At least one hundred flowers for each possible
combination of regenerator and nonregenerator parent, including reciprocals,
were pollinated in this manner.
Seeds were harvested in August, 1989 and germinated and planted in the
greenhouse the following winter. Hybrids were clearly identifiable based on the
morphological characteristics described above. Flowering of hybrids was
induced in the summer of 1990. Plants were allowed to self-pollinate, but
flowers were hand tripped to increase seed set.
Evaluation of callus growth and regeneration in the parental. Fo and Fg
generations. The F2 seed harvested from the original hybrids was scarified and
germinated on SGL medium (Collins and Phillips, 1982) as described in
Chapter 2. Hypocotyls were excised and placed onto callus induction medium
and epicotyls were returned to SGL medium for rooting. Rooted epicotyls were
transferred to potting soil after ten days and grown to maturity in the greenhouse
where they were allowed to self-pollinate to produce the Fg generation.
Hypocotyls from both F2 and Fg populations were cultured according to
the optimal protocol established in Chapter 2. Initial callus production was on
L2 medium supplemented with 0.06 mg L"1 picloram and 0.2 mg L'1 BA for

37
28 days. Shoot buds were induced on L2 with 1.0 mg L"1 2,4-D and 2.0 mg L'1
adenine for 14 days, and shoot elongation was induced on L2 with
0.012 mg L"^ picloram and 0.2 mg L^ BA for 28 days. Shoots were rooted on
L2 lacking plant growth regulators. Hypocotyls from parental lines were
cultured simultaneously with the F2 and F3 populations to serve as checks.
Callus fresh weights were measured at the time of transfer from callus
medium to bud induction medium. Each callus was visually scored for shoot
bud formation and elongation at the end of both the bud induction and bud
elongation steps. The scoring system and statistical transformations used will
be discussed in the results.
Statistical analysis. General statistical analyses were conducted using
JMP Version 3.0, the SAS Institute general statistics product for Macintosh
computers (SAS Institute, 1994). Means and variances of truncated data sets
were estimated using UNCENSOR Version 3.0 (Newman and Dixon, 1989).
Results and Discussion
Results of crossing efforts were disappointing. The approximately 2200
flowers that were cross-pollinated produced only 590 seeds. Poor seedset was
likely due in part to inadequate greenhouse ventilation, which resulted in
occasional daytime temperatures sometimes exceeding 38 C. Hybrid yield
was further reduced by the low ratio of hybrids to seifs in the seeds that were
produced. Approximately 550 of the 590 seeds germinated, but only eight of
the resulting plants were identified as hybrids. Hybrids could be readily
identified by leaf shape and growth habit, both of which were intermediate
between the parental types. Two hybrid plants were progeny of the cross IRFL
6123 x CIAT 13083 (hereafter referred to as cross 501), three were from cross

38
IRFL 6123 x UF 144 (cross 507), and three were from CIAT 13083 x IRFL 6128
(cross 510). Since genotypes 6123 and 6128 are nearly identical, both in gross
morphology and in vitro performance, crosses 501 and 510 are essentially
reciprocals.
Hybrids were selfed through the F3 generation. Table 3-1 summarizes
numbers of individuals and number of families from which callus growth and
regeneration data were collected. In the case of cross 510, data collection was
terminated at the F2 generation. Severe morphological abnormalities
appeared in the F-j of this cross, and persisted through F3. The three F-j
seedlings appeared normal through approximately 6 weeks of age, after which
new growth showed severely stunted leaves and internodes. Stunting was
evident from the early seedling stage in many F2 and F3 individuals, and
appeared at late seedling stages in all individuals. Stunting was accompanied
by extremely poor flower production and seedset. The F2 of cross 510 also
exhibited a depressed callus growth rate relative to the two other crosses and to
the parental lines with a high incidence of callus necrosis and death. Since
cross 510 is essentially the reciprocal of cross 501, the abnormalities may be
the result of interactions between the cytoplasm of CIAT 13083 and nuclear
genes of IRFL 6127. (Evidence of maternal effects in the other crosses will be
discussed later in this chapter.) Whatever the cause, the observed
abnormalities have the potential effect of masking expression of the in vitro
characters of interest, as well as exerting confounding selective pressures in the
F2 and F3 generations. Therefore, genetic analysis was not attempted on cross
510, and the remainder of this chapter deals exclusively with crosses 501 and
507.
Scaling of the callus growth trait. The logical first step in genetic analysis
of any trait is to identify an appropriate numerical scale; that is, to determine if

39
mathematical scale transformation is necessary. This task can be approached
in at least two general ways. A purely statistical approach, described in any
number of statistics texts (Little and Hills, 1978; Zar, 1984), seeks a scale that
yields a data structure satisfying the general assumptions of parametric
statisticsin this case, normally distributed error terms with error variances
independent of means. A more genetically oriented (and somewhat more
complex) approach is not directly concerned with statistical assumptions, but
stresses instead whether a scale will facilitate partitioning of genetic variation
into underlying genetic causes, or components; e.g., additive, dominant, and
epistatic gene action (Mather and Jinks, 1971). Implicit in this second
approach is the statistical assumption that factor levels act in an additive
fashion. Both approaches will be presented below.
Three scales were examined in the present study. The simplest is the
untransformed linear scale, which in this case is callus weight in grams. This
scale possesses the virtue of allowing various statistics and parameters to be
reported in grams, permitting easy interpretation. A logarithmic scale was also
investigated. While this scale results in somewhat non-intuitive statistical
outputs, the scale is theoretically more applicable to analysis of growth
variables than is a linear scale. The logarithmic scale asks the question, how
many times has the cell mass doubled, while the linear scale asks how many
mass units have been added to the starting mass. The logarithmic
transformation is frequently used to stabilize variances in cases where standard
deviation is proportional to mean in the untransformed scale. The third scale
tested was the square root transformation. This scale is intermediate between
the linear and logarithmic scales in terms of skewing of distributions, and is
useful when variance is proportional to mean.

40
To examine the effect of the three scales on normality, the scales were
applied to populations of each of the three parental lines and to F2 and F3
populations. The parental populations are theoretically normally distributed,
while the F2 and F3 populations are expected to be normally distributed if little
or no unidirectional dominant or epistatic gene action is present. The later
generation populations are included here because, as a result of a temporarily
limited seed supply, the size of the tested parental populations was quite small
(20 < n < 36). Univariate statistics for the various populations are presented in
Table 3-2. The linear scale tends to produce rightward (positive) skewing, while
the log transformed populations are moderately left-skewed (negative), and the
square root transformed populations are slightly left-skewed. The Shapiro-Wilk
W statistic is used to test for normality in populations as small as ten
observations; a low probability associated with a calculated value of W
indicates a significant deviation from normality (SAS Institute, 1989; Gilbert,
1987). None of the W values for the parental populations indicated deviation
from normality.
Results for F2 and F3 populations are similar to those for the parental
populations, but are somewhat clearer. The linear scaling results in gross
leftward skewing and yields highly significant W statistics. This may be
attributable to genetic effects if there is a strong preponderance of negatively
acting dominant genes in these populations. Strong, unidirectional epistatic
effects, which could result in a relatively small proportion of the populations
possessing the necessary combination of alleles for rapid callus growth, is
another possible genetic cause for the observed skewing. However, the fact
that the leftward skewing occurs in all populations can be interpreted to indicate
that the linear scale is inadequate and that with proper scaling the F2 and F3

41
will in fact be approximately normally distributed. Further evidence regarding
this hypothesis will be presented later in this chapter.
Skewness and W statistics for the F2 and F3 populations fail to clearly
indicate whether the logarithmic or the square root transformation yields a more
normal distribution. As in the case of the parental populations, the log
transformed populations are skewed to the right, while the square root
transformed populations are skewed to the left. In the case of F3 populations,
the magnitude of skewing is substantially less for the logarithmic than for the
square root scales. The only non-significant W statistic was obtained from the
square root transformed cross 507 F2 population.
If a variable is properly scaled, populations with high means should, in
general, have variances no greater than those of populations with lower means.
This can be tested by regressing F3 family means versus variances, as
presented in Figure 3-1. The linear scale exhibits a strong, positive relationship
between mean and variance. Logarithmic transformation results in a weak,
negative relationship, and the square root transformation produces a weak,
positive relationship. It is feasible that the expectation of independence of
means and variances can be confounded by genetic effects, since different F3
families may possess differing amounts of genetic variance. For example, in the
not unlikely situation that a large, positive dominance effect exists, families with
high means will be those whose F2 parents had high levels of heterozygosity
for genes controlling callus growth. These families will also exhibit elevated
variances due to the presence of segregating genes. This genetic effect can
only be large, however, if total genetic variance is much greater than
environmental variance. It will be shown later in this chapter that this is not the
case. Therefore, the strong relationship between means and variances for F3
families provides further evidence of the inadequacy of the linear scale.

42
Because only three parental populations were examined, meaningful
regression of mean versus variance for parental populations is not possible.
However, simple examination of parental means and variances (Table 3-2) can
shed further light on scale effects. If variance is independent of mean, the
variances of the three parental populations should be similar in spite of the
differences in means. The effects are similar to those seen in the F3
populations, but the square root transformation stands out in that it yields a
smaller relative range in variances (maximum variance 48% greater than
minimum) than either the linear (177%) or logarithmic (78%) scales.
In summary, a general statistical examination of the callus growth trait
reveals the clear inadequacies of linear scaling. Logarithmic and square root
scales are similar in their conformity to the statistical requirements of normality
and independent variance.
Joint scaling test for callus growth. A more genetically oriented approach
to scaling was developed by Cavalli (1952) and elaborated upon by Mather and
Jinks (1971). Known as the joint scaling test, in its simplest form this technique
involves estimating the effects of additive and dominant gene action on
generation means. This is done by establishing a series of simultaneous
equations, one for each generation for which data are available. The left side of
each equation is the observed generation mean. The right side consists of
three terms, each of which is composed of an unknown and an associated
coefficient. The unknowns are model parameters: the grand mean (essentially
the genetic-neutral component of the various generation means, designated as
m), additive genetic effect (designated [d]), and dominant genetic effect
(designated [h]). Coefficients correspond to the theoretical causal contributions
made by these parameters to each generation mean. Since there are three
unknowns, data from at least four generations (each parent represents one

43
generation) are needed to obtain a non-singular solution and a measurement of
statistical significance of the solution. Each equation is weighted in proportion
to the level of certainty associated with each generation mean, which is
equivalent to the reciprocal of the variance of the mean. As originally
described, a solution is obtained through cumbersome matrix manipulations.
This is computationally identical to performing weighted least-squares
regression using the observed means as the dependent variable and the model
coefficients as independent variables (Rowe and Alexander, 1980). The partial
regression coefficients obtained in this way are equivalent to the unknowns in
the above-described equations. The adequacy of solutions obtained by this
method is assessed in two ways. First, each predicted model parameter has an
associated F test for the hypothesis that the parameter differs from zero.
Secondly, the whole model can be tested by examining the error sum of
squares (SSE). In this particular form of analysis, SSE exhibits a x2
distribution, and the reported SSE can be compared to tabular x2 values to
determine goodness of fit (Rowe and Alexander, 1980).
Six joint scaling tests were conducted in order to test each of the three
scales using data from each cross. Model coefficients used to construct the
tests are given in Table 3-3. Observed means and weights used in the tests are
are based on the means and variances presented in Table 3-2. Predicted
values of m, [d], and [h], and the results of goodness of fit testing are given in
Table 3-4. Adequate goodness of fit is obtained only for the linear and square
root scales and only for cross 501. In both cases, a large, negative dominance
effect is indicated.
The joint scaling test suffers from limitations that necessitate cautious
interpretation of results. One serious limitation is that the test assumes that one
parent possesses most of the positively acting alleles and the other possesses

44
the majority of negatively acting alleles, and that dominance acts primarily in a
single direction. Judging from the large difference in callus growth rates among
the parents in the current work, the first assumption is probably at least partially
valid. The validity of the second assumption is difficult to assess. A second
problem with the joint scaling test is that two separate hypotheses are
simultaneously tested: that the scale is adequate, and that the genetics of the
trait in question can be adequately described by an additive-dominant genetic
model. When an adequate goodness of fit is not obtained, the joint scaling test
cannot by itself distinguish which of these hypotheses is incorrect. It is possible
to partially solve this problem by adapting the test to more complex genetic
models involving epistatic and perhaps other types of genetic effects.
Unfortunately, more complex models require data from more generations and
are dependant on an increasing number of rather dubious assumptions about
gene interactions (Mather and Jinks, 1971). As a result of these limitations,
spurious positive results are quite possible, particularly when the test involves
only a single degree of freedom. Considering the statistical evidence for the
inadequacy of the linear scale, it is likely that the good fit obtained when this
scale was applied to cross 501 is in fact spurious.
Examination of the data presented in Table 3-2 suggests a factor that may
confound the joint scaling test in the present case. In both crosses, and for all
scales, the F2 and F3 means are significantly lower than the mid-parent mean,
and are fairly close to one another. In the case of cross 501, the F3 mean is
higher than the F2 mean, while in cross 507 the F3 mean is lower. If a simple
additive-dominant model is assumed, the deviation of the F2 and F3 from the
mid-parent mean suggests a large, negative dominance effect, just as indicated
by the joint scaling test. However, the relationship between F2 and F3 means
suggests a small dominance effecta negative effect for cross 501, and a

45
positive effect for cross 507. The discrepancy between the dominance effect
predicted by the relationship of the progeny means to the mid-parent mean and
that predicted by the relationship between and F3 means accounts for the
generally poor joint scaling test results. These observations suggest that some
factor not considered in the model may be acting to depress both and F3
means relative to parental means. This factor could be a maternal or
cytoplasmic effectnuclear genes from parents 144 and 13083 may interact in
a negative manner with cytoplasmic genes from 6123 (the female parent in both
crosses). The depressed callus growth observed in cross 510 is consistent with
this explanation. An alternative cause would be negative epistatic interactions
between nuclear genes originating from the different parents. If either
explanation is correct, then the results of the additive-dominant joint scaling
tests are invalid. Unfortunately, there are insufficient generations available to
allow construction of an additive-dominant-maternal or additive-dominant-
epistatic joint scaling test.
Variance partitioning for the callus growth trait. Variance partitioning
methods are not subject to all of the limitations described for the joint scaling
test. Specifically, variance partitioning is not affected by maternal factors that
affect means, by distribution of positive-acting alleles between parents, nor by
dominance acting in different directions at different loci. A sophisticated method
of deriving genetic variance parameters from observed variances and
covariances in segregating generations derived from crosses of inbred lines is
presented by Mather and Jinks (1971). The technique is similar in many ways
to the joint scaling test, but observed variances and theoretical causal variance
components are utilized to construct the model. As with the joint scaling test,
fairly complex genetic models are possible, but, due to data limitations, the
present discussion will be limited to a simple additive-dominant model. In the

46
present case, five observational variances or covariances are available: pooled
parental variance (E-j), F2 variance (V-||r2), among-family F3 variance (V1F3),
within-family F3 variance (V2p3), and ^2*^3 covariance (Wip23)- Among- and
within-family F3 variances are obtained by analysis of variance using F2 parent
as the grouping variable, followed by partitioning of
variances based on expected mean squares (Appendix A). The F2-F3
covariance is the covariance between F2 parents and F3 family means.
The causal variance components described by Mather and Jinks are
somewhat complex. In addition to additive and dominant genetic components
(designated D and H, respectively), two types of environmental variance and a
so-called sampling variance are described. The first environmental variance
(Ew) is simply the error variance among individuals. The second (Eg) is the
error among plots or rearing environments and is applicable only to V-jp3 in the
present study. Since explants from the different F3 families were randomly
distributed among petri dishes, which were randomly distributed within a single
incubator, there is no intrinsic plot error. Instead, Eg is equivalent to Eyy
divided by the harmonic mean of the number of individuals per F3 family
(Mather and Jinks; 1971). Since Eg is a function of Eyy, it can be combined into
Eyy, thereby eliminating a column from the model and gaining a degree of
freedom. The sampling variance applies only to among-group variances. It is
similar in concept to Eg, but is a function of within-group genetic variances
rather than error variances. Therefore, in the case of V-| p3 the sampling
variance is equivalent to V2p3 divided by the harmonic mean of the number of
individuals per F3 family. Sampling variance can be algebraically included into
the model row for V1F3 and does not have an associated model column.
Derivation of model matrices is described in detail in Appendix B.

47
The method of Mather and Jinks employs an iterative approach for
determining model weights. Weights for each of the matrix rows are initially
established as the reciprocals of the theoretical variances of each observed
variance or covariance. The variance of a variance is equal to 2(V2)/df, where
V is the observed variance and df is the number of degrees of freedom on which
the observed variance is based. The variance of a covariance is equal to
(W2 + V-|V2)/df, where W is the observed covariance, V-| and V2 are the
variances of the populations from which the covariance was obtained, and df is
the degrees of freedom upon which the covariance is based. The model is then
solved, and predicted (Mather and Jinks use the term expected) observational
variances are obtained. These are used to derive new weigh estimates, and
the model is solved again. This is continued until no improvement in model fit is
obtained.
Coefficients used in variance partitioning for the callus growth trait are
presented in Table 3-5. Observational variance components and initial weights
are presented in Table 3-6. Results of the first iterations are given in Table 3-7.
In all cases, second and third iterations failed to improve model fits. (Results of
later iterations are presented in Appendix C.) The linear and logarithmic scales
failed to produce significant whole-model F tests for either cross. The square
root scale produced significant whole-model tests for both crosses. For cross
501, estimated additive genetic variance was large, and significantly different
from zero (assuming a rather generous a of 0.1), while dominant variance was
smaller and not significantly different from zero. Cross 507 yielded no
significant genetic variance estimates.
The analysis presented thus far gives a rather unclear and somewhat
contradictory picture of the genetic mechanism controlling the callus growth trait
in the two crosses. Both the joint scaling and variance partitioning approaches

48
indicate that cross 507 cannot be adequately described by an additive-
dominant genetic model. Based on the joint scaling test, the genetics of cross
501 may be characterized by a large, negative dominance effect and a
moderate additive effect. An alternate interpretation is that large, negative
maternal or epistatic effects exist in both crosses. In contrast, variance
partitioning indicates a large additive genetic variance and a much smaller
dominant variance in cross 501.
A general shortcoming of the types of analysis presented to this point is the
large number of simplifying genetic assumptions required. The results derived
from variance partitioning may be more reliable than those of the joint scaling
test, due to the fewer genetic assumptions required by this method. Many
assumptions, such as absence of linkage and equal magnitude of effect from
each contributing locus, are common to both techniques. The generally poor
model fits obtained by both methods may be due to failure of the experimental
material to conform to these assumptions, or to the presence of significant
nonadditive, nondomininant genetic effects. It is likely that experimental error
also contributes to poor model fit.
Parent-offspring regression is a simpler, more empirical method of
analysis, relatively free of genetic assumptions. Regressions of F3 family
means on F2 parents are presented in Figure 3-2. The slopes (regresin
coefficients) represent, by definition, heritabilities. It should be noted, however,
that in the case of selfing, parent-offspring regression is in part a function of
dominant and epistatic genetic effects, but these effects play a much smaller
role in the regression than does additive genetic variance. Regression
coefficients are remarkably high, regardless of cross or scale, ranging from
0.524 to 0.769. The practical implication of this is that substantial progress

49
should be possible in selecting for callus growth among segregating
generations of this material.
The nature of the regeneration variable. The callus growth trait was
amenable to quantitative analysis due to the fact that this trait could be
expressed in terms of a continuous, metric variable for each individual plant.
This was not the case for the regeneration trait. In an ideal study of the genetics
of regeneration, each plant can be evaluated in terms of a metric variable such
as percent responding explants or number of shoots per explant. Because
hypocotyl explants had to be used in the present study, only one explant existed
per plant, so the data could not be expressed as percent responding explants.
Expression of regeneration as number of shoots per explant was impossible
because the majority of explants produced no buds or shoots. Instead, the trait
was expressed on a visual rating scale of one through seven. A rating of one
represented no evidence of regeneration. Two represented a slight indication
of regeneration in the form of localized deep green coloration. Three
represented formation of a single, well-defined but nonelongated bud. Four
indicated multiple bud formation, usually with some elongation. A rating of five
indicated one or two well elongated shoots. Six indicated between three and
five elongated shoots, and a rating of seven indicated more than five elongated
shoots.
A significant number of individuals in all populations showed no sign of
regeneration and received scores of one. All individuals in the nonregenerating
parent populations (genotypes 144 and 13083) received scores of one. This
type of data structure can be described as "threshold" or "censored" data.
Falconer (1981) discussed threshold traits at length. He assumed that there
exists an underlying, continuous scale of proclivity (or as Falconer describes it,
"liability") towards a certain conditionin the case of the present work, tendency

50
to regeneratethat can only be detected when the level of proclivity rises
above a given threshold.
Distributions of this type cannot be rendered normal by any mathematical
transformation, since we lack any information about differences among those
individuals that fall below the observable threshold, and no transformation can
restore this missing information. For example, no matter what transformation is
used, the two nonregenerating parental populations will always have equal
means, and variances of zero. Fortunately, useful analysis can still be
performed on this type of data, provided it can be assumed that the populations
under consideration are normally distributed.
In Falconer's treatment, threshold traits can be manifested at either two or
three discrete levels, or classes. This can be done with the present data by
collapsing the rating scale into two (nonregeneration versus regeneration), or
three (nonregeneration versus partial regeneration versus full regeneration)
classes. While little useful analysis can be done when only two classes are
present, meaningful analysis is possible in situations with three classes,
contingent on the assumption that the underlying variable is normally
distributed. For any population, if the percentage of individuals falling into each
of three threshold classes is known, then the mean and variance of the
population, expressed in threshold units, can be deduced. The principle is
illustrated in Figure 3-3, adapted from Falconer (1981), but based on actual F3
data. The classes are separated by the two vertical lines on each graph,
nonregenerators lying to the left of both lines, partial regenerators lying
between the lines, and full regenerators lying to the right. While both
populations contain 68% full regenerators, population (a) contains more partial
regenerators and fewer nonregenerators than population (b). As a result,

51
population (a) has a lower standard deviation and lower mean than
population (b).
An attractive advantage of analyzing data based on three ordinal classes
with two thresholds is that scaling problems are avoided. This is because only
one real unit exists on the x axisthe unit separating the two thresholds.
Values either to the left or to the right of both thresholds can never be
experimentally measured, and are simply extrapolations of the central threshold
unit. A serious disadvantage of the three class approach is that populations that
do not contain at least one observation in each of the three classes cannot be
analyzed. Many of the F3 families in the present study contain no fully
regenerating individuals, and therefore cannot be analyzed using this
approach.
While the methods described by Falconer for analysis of threshold traits
cannot be directly applied to data with more than three classes of observations,
it is possible to analyze data with an unlimited number of classes by applying
techniques developed for censored data sets. Censored data are similar to
threshold data in that both data types are subject to a threshold below which no
information is known (Schneider, 1986). The difference is that in censored
data there is only one threshold, and data above this threshold are of a
continuous, or at least semicontinuous nature. In both cases, data lying above
the detection threshold are used to make inferences about data below the
threshold; therefore, assumptions of normality are necessary in both cases. By
treating the seven-category regeneration rating scale as a quasi-continuous
variable, the data collected in the present study are amenable to analysis as
censored data.
Several statistical techniques have been developed for "uncensoring"
censored data sets, and, as in many fields of statistics, the relative strengths and

52
weaknesses of each are subject to considerable debate among statisticians.
The most intuitively appealing method is termed regression order statistics.
This method is graphically illustrated in Figure 3-4. Figure 3-4 (a) is a normal
probability plot of a hypothetical, approximately normally distributed population.
The Y-axis is observed value of the variable of interest, and the X-axis is normal
quantile score (a transformed Z-score scale). The points represent the
individual observations, and the line represents the simple linear regression
line obtained from the individual data points. The slope and intercept of this line
can be used to approximate the variance and mean of the population. Figure
3-4 (b) is a similar plot of the same population with a censoring threshold
applied such that approximately 30% of the observations yield no observable
response. These observations, of course, all have the same values for both
axes, and therefore all lie atop one another at the lowest left data point on the
graph. The result is a truncated plot. The regression line is derived only from
those observations lying above the censoring threshold, but is similar to the line
obtained from the uncensored population, and would yield variance and mean
estimates close to the actual parameters of the original population. The
regression order statistics method is a numeric analog of Figure 3-4 (b). That is,
it performs a normal quantile regression on a truncated population in order to
approximate the mean and variance of the intact population.
The accuracy of population parameters estimated by uncensoring through
regression order analysis is limited by the number of observations lying above
the censoring threshold. However, reasonably accurate results can be
obtained even when the majority of observations are below the threshold, as
long as several observations lie above the threshold. Regression order
analysis was used to estimate means and variances of all populations in the
current study that had at least four individuals with regeneration ratings higher

53
than one. Although this is a less than ideal approach, there appears to be no
acceptable alternative. The option of simply analyzing the raw regeneration
scoreseither transformed or otherwiseis undesirable because this entails
accepting the clearly incorrect assumption that all individuals receiving
regeneration scores of one have the same proclivity for regeneration. The
contrasting options, then, are to ignore the left-hand tails of many of the
population distributions, or to use an acknowledgedly problematic method of
reconstructing these tails. Means and variances determined by uncensoring, as
well as means and variances of the raw data are presented in Table 3-8. In
general, means derived from regression order analysis are lower than those
obtained from the raw data, and variances are higher. This is not unexpected,
since truncation of the low end of a data setas occurs in the raw datawill
artificially raise the mean and lower the variance. For the raw data, a significant
(a=0.05), positive relationship exists between F3 family variance and family
mean, due at least in part to the fact that the lower the mean, the more severely
truncated the distribution. This relationship is absent from the uncensored data
because the truncated lower ends of the distributions have been reconstructed.
The unusually high variances observed for some F3 families may be due to
segregation of major genes within these families, resulting in a somewhat
bimodal distribution and exaggerated variance. This effect is most pronounced
in the uncensored data. Unfortunately, population sizes for the F3 families are
too small to clearly distinguish between spurious bimodality and bimodality
resulting from genetic causes. Bimodality was not observed in the F2
populations, nor in the complete F3 populations.
The accuracy of the variance estimates presented in Table 3-8 can be
compromised by non-normality resulting from inadequate scaling. Major
scaling problems can be detected by examining the distribution of a population

54
containing no genetic variance (since genetic factors can also cause deviations
from normality) and having a mean significantly higher than the censoring
threshold. The regenerating parental population (genotype 6123) is such a
population, and a histogram of this population is presented in Figure 3-5. The
distribution is not seriously skewed, nor is it bimodal, suggesting that there is
little scale-induced non-normality.
Genetic analysis of the regeneration trait. Due to the difficulties presented
above, the types of analysis that can be performed on the regeneration trait are
severely limited. The joint scaling test cannot be used because this test
requires an estimate of the mean for each parent. Neither of the
nonregenerating parent populations (genotypes 144 and 13083) gave any
evidence of regeneration, so meaningful estimates of mean regeneration score
could not be obtained for these parents. This reiterates the case for analyzing
the uncensored rather than raw regeneration data. All calli for each
nonregenerating parental genotype received regeneration scores of one, so
based on the raw data the two genotypes have equal mean regeneration
scores. There is no reason to believe that both of the nonregenerating parents
have the same genetic proclivity for regeneration. The scores merely indicate
that neither genotype has enough proclivity to produce any visible evidence of
regeneration under the culture protocol used.
The iterative variance partitioning method used for the callus growth trait
could not be applied to the regeneration trait. This method requires good
estimates of the uncertainty associated with variance estimates and it was felt
that these were lacking for the regeneration trait. Mather and Jinks (1971)
described a simpler, nonweighted, noniterative variance partitioning method.
Preliminary efforts were made to apply this method to the regeneration trait.
Both raw and uncensored data were examined in this way, despite the

55
truncation-related biases in the raw data. Preliminary variance partitioning
yielded no significant (or even nearly significant) estimates for any parameters.
In the case of cross 501, this appeared to be due to an excessively high within-
family variance in the Fg and extremely low among-family Fg variance and Fg*
Fg covariance. For cross 507, the primary confounding factor appeared to be
an overly high Fg variance. In both cases these factors made it impossible to
obtain solutions to the simultaneous linear equations established in the
variance partitioning matrix.
The remaining option for genetic analysis of the regeneration trait is
parent-offspring regression. Regressions of Fg family mean on Fg parent for the
two crosses are presented in Figure 3-6. Cross 507 shows a strong parent
offspring relationship, with a moderately regression coefficient, or heritability.
The coefficient of determination (r2) is higher for the raw data than for the
uncensored data, but the slope is greater for the uncensored data. In the case
of cross 501, the regression is not significant. One reason for the poor
regression observed for cross 501 is the very small sample size, especially the
low number of parents with regeneration scores higher than one. Another
reason is the broad range of mean offspring regeneration scores for parents
with regeneration scores of one. It is likely that the "true" regeneration scores
for some of the parents (particularly the parent at the far lower left of the plot) lie
somewhat below the censoring threshold. If it were possible to measure
parental regeneration scores of less than one, the observations at the lower left
corner of the plot might be distributed farther to the left, resulting in a stronger
regression.
The previous discussion of both the callus growth and regeneration traits
contained the implicit assumption that the traits were quantitatively inherited.
For any genetic system there exists a continuum of possibilities, ranging from

56
absolute control by a single gene, through control by one or a few major genes
in conjunction with a number of minor, or "modifier" genes, to classical
quantitative inheritance involving control by many genes with approximately
equal effects. The large error variances and non-ideal data structure of this
study make it impossible to formally distinguish among these possibilities.
However, several general observations and considerations combine to suggest
that both of the traits examined are controlled at least to some degree by
multiple genes.
First, both traits exhibited transgressive segregation. Callus weights of
many F3 individuals in both crosses greatly exceeded that of any parental
individual, and four F3 individuals from cross 507 received regeneration scores
of sevena score never achieved by the regenerating parent (Figure 3-7;
Appendix D). Transgressive segregation in crosses between inbred lines can
be attributable to either overdominance or the presence of positive-acting
alleles at different loci in each parent. True overdominance is so rare that many
authorities doubt its existence (Simmonds, 1979). If single-gene
overdominance existed for the regeneration trait, approximately one quarter of
F2 individuals would carry the overdominant genotype, and some proportion of
these would likely exhibit the overdominant (transgressive) phenotype. The
transgressive phenotype was not observed in the F2. Thus, there appear to be
at least two loci controlling the regeneration trait. In view of the moderately
large size of the F2 populations (57 for cross 501 and 80 for cross 507), the
absence of the transgressive phenotype in this generation suggests that more
than two loci may be involved.
A second indication that at least two loci control the regeneration trait is the
large difference in regeneration scores between Crosses 501 and 507, despite

57
the fact that the crosses share a regenerating female parent. If ability to
regenerate were conferred entirely by alleles originating from one or two loci in
the regenerating parent, it would be expected that the best regenerators among
the F2 and F3 would be those bearing the fixed genotype of the regenerating
parent. The best regenerators from each cross would then be expected to
perform approximately equally, regardless of the identity of the nonregenerating
parent. This was not observed. Instead, the best regenerators from cross 507
consistently outperformed those of cross 501 in both the F2 and F3. The
transgressive phenotype occurred only in cross 507, suggesting that the
nonregenerating parent in this cross contributes at least one positive-acting
allele. The issue of quantitative versus qualitative inheritance could be clarified
by culturing of a large number of progeny from several high-regenerating F3
plants.
Some final insights can be obtained by examining the relationship
between regeneration and callus growth. Figure 3-8 presents a scattergram of
regeneration score versus logarithm of callus weight for the combined F3
population. A very weak, but significant correlation exists, and perhaps as
meaningful is the observation that the high regeneration scores are generally
associated with moderate callus growth. This apparent clustering is in part an
illusion stemming from the smaller population sizes for the higher regeneration
scores. However, there is a significant (a=0.05) decrease in population
variances for the callus growth trait as regeneration score increases, indicating
that the clustering is real. The association can be either causal, accidental, or
both. In both crosses, callus growth was greater in the nonregenerating parent.
The absence of regeneration in the high growth calli may therefore be the result
of genetic linkage between the two traitsan accident of the distribution of
genetic material among the parents. The lack of regeneration among low

58
growth calli cannot be attributed to linkage. A likely explanation is that a certain
degree of in vitro vigor, as manifested by at least a moderate callus growth rate,
promotes regeneration. This would constitute pleiotropy in a broad sense. An
alternative possibility is that there exist genes that act in a physiologically
pleiotropic manner; that is, there may be genes that either promote or inhibit
specific metabolic pathways that are necessary for both callus growth and
regeneration. To experimentally distinguish between these subtly different
hypotheses is far beyond the scope of the present study, or of any study
undertaken to date on the genetics of in vitro traits.

59
Table 3-1.
Number of F2 plants, F3
data were collected.
plants, and F3
families from which in vitro
Cross
F2 plants
F3 plants
F3 families
501
65
181
11
507
91
291
18
510
52
t
t
t Cross 510 was discontinued after F2.

Table 3-2. Univariate statistics for callus weight for parental, F2, and F3 populations in linear, logarithmic and
square root scales.
Population
Scalet
Mean
Variance
Skewness
W
Prob n
144 (Parental)
li
0.7442
0.0648
0.247
0.953
0.288
26
lo
2.844
0.0268
-0.702
0.934
0.102
26
sq
26.87
22.89
-0.196
0.985
0.372
26
6123 (Parental)
li
0.2897
0.0172
0.38
0.947
0.119
35
lo
2.413
0.0477
-0.528
0.95
0.15
35
sq
16.57
15.44
-0.021
0.964
0.369
35
13083 (Parental)
li
0.7215
0.0541
-0.114
0.958
0.514
20
lo
2.833
0.0261
-0.93
0.906
0.055
20
sq
26.49
20.74
-0.503
0.943
0.293
20
501 (F2)
li
0.301
0.071
2.111
0.794
<0.001
65
lo
2.319
0.162
-0.432
0.954
0.038
65
sq
15.93
47.88
0.766
0.942
0.008
65
507 (F2)
li
0.366
0.069
1.506
0.877
<0.001
92
lo
2.445
0.123
-0.65
0.9953
0.009
92
sq
17.97
43.58
0.431
0.969
0.15
92
t li = linear (grams); lo = log-j o(nnilligrams); sq = square root (milligrams)
05
o

Table 3-2 continued.
Population
Scalet
Mean
Variance
Skewness
W
Prob n
501 (Fg)
li
0.380
0.1138
1.729
0.830
<0.001
181
lo
2.412
0.1643
-0.242
0.9612
0.001
181
sq
17.81
63.47
0.691
0.940
<0.001
181
507 (Fg)
li
0.254
0.0526
1.509
0.817
<0.001
291
lo
2.231
0.1647
-0.161
0.963
<0.001
291
sq
14.49
43.72
0.737
0.924
<0.001
291
t li = linear (grams); lo = logi0(milligrams); sq = square root (milligrams)

62
Table 3-3. Model coefficients used to construct joint scaling tests.
Population
Coefficient
m [d] [h]
Parent 1
1 -1 0
Parent 2
1 1 0
f2
1 0 0.5
f3
1 0 0.25
1
0
0.25

Table 3-4. Parameter estimates and associated probabilities obtained from joint scaling tests for callus growth.
Cross
Scale
Parameter
m
[d]
[h]
Whole Model
estimate
prob>F
estimate
prob>F
estimate
prob>F
X2
prob>x2
501
Linear
0.497
0.023
0.210
0.057
-0.417
0.091
0.481
0.45
507
Linear
0.401
0.247
0.149
0.567
-0.364
0.637
53.97
<0.001
501
Logarithmic
2.611
0.009
0.210
0.112
-0.672
0.139
2.086
0.24
507
Logarithmic
2.542
0.053
0.228
0.504
-0.582
0.601
89.17
<0.001
501
Square root
21.26
0.020
4.849
0.091
-11.73
0.128
1.360
0.36
507
Square root
19.20
0.148
4.313
0.555
-9.679
0.655
81.66
<0.001
CT>
CO

64
Table 3-5. Coefficients used for variance partitioning for callus growth.
Cross
Observational
Component
Causal Component
df
D
H
EW
501
ViF2
0.5
0.25
1
64
V1F3
0.5164
0.07076
0.1314
10
v2f3
0.25
0.125
1
170
WiF23
0.5
0.125
0
10
El
0
0
1
53
507
ViF2
0.5
0.25
1
91
V1F3
0.5196
0.07233
0.1571
17
v2f3
0.25
0.125
1
273
W1F23
0.5
0.125
0
17
El
0
0
1
59

65
Table 3-6. Observational variance components and initial weighs
used for variance partitioning for callus growth.
Cross
Scale
Component
Variance
Weight
df
501
Linear
ViF2
0.07076
12780
64
V1F3
0.06100
2688
1 0
v2f3
0.06232
43780
170
W2f23
0.08426
582.6
1 0
El
0.03040
57370
53
501
Logarithmic
ViF2
0.1618
2445
64
VlF3
0.09271
1163
1 0
v2f3
0.08591
23030
170
W2F23
0.05925
813.8
1 0
El
0.03994
33210
53
501
Square Root
ViF2
4788
0.02791
64
ViF3
36.46
0.007521
1 0
v2f3
32.86
0.1575
170
w2f23
36.93
0.002687
1 0
El
17.37
0.1757
53

66
Table 3-6. Continued.
Cross
Scale
Component
Variance
Weight
df
507
Linear
ViF2
0.06941
19100
91
v-|F3
0.02204
3500
1 7
v2f3
0.03618
208600
273
W2f23
0.02148
13890
1 7
El
0.03746
42040
59
507
Logarithmic
ViF2
0.1227
6044
91
V1F3
0.06896
3574
1 7
v2f3
0.1132
21300
273
W2F23
0.04864
2612
1 7
El
0.03879
39210
59
507
Square Root
ViF2
43.72
0.04760
91
v-|F3
19.19
0.04642
1 7
v2f3
29.39
0.3160
273
w2f23
15.64
0.02542
1 7
El
18.23
0.1701
59

Table 3-7. Estimates of causal variance components and associated probabilities obtained from variance partitioning
for callus growth.
Component
Cross
Scale
D
H
Ew
Whole Model
Estimate
LL
A
S
CL
Estimate
Prob>F
Estimate
Prob>F
F
Prob>F
501
Linear
0.1258
0.4033
-0.0578
0.8239
0.0324
0.1428
28.57
0.1365
507
Linear
0.0313
0.6534
0.0271
0.9055
0.0284
0.2982
11.50
0.2127
501
Logarithmic
0.1273
0.4244
0.1535
0.6266
0.3881
0.1241
48.08
0.1055
507
Logarithmic
0.0594
0.7487
0.3036
0.5601
0.0438
0.2364
11.64
0.2114
501
Square Root
70.39
0.0592
-17.50
0.4434
17.38
0.0189
1944
0.0189
507
Square Root
23.61
0.2057
46.27
0.2916
18.24
0.0512
332.2
0.0403
O
-vi

68
Table 3-8. Univariate statistics for regeneration score utilizing raw and
uncensored data.
Population Raw Data Uncensored Data
Mean
Variance
Mean
Variance
N
N>1
6123
4.053
1.240
4.075
1.080
38
37
501 F2 (all)
1.868
1.076
1.679
1.774
57
29
507 F2 (all)
1.994
1.952
1.278
4.351
80
43
501 F2 (parents only)
1.636
0.8552
2.078
0.4886
11
4
507 F2 (parents only)
2.267
2.924
1.553
6.529
17
8
501 F3
1.594
0.8158
1.185
1.882
165
67
507 F3
2.058
1.807
1.554
3.733
260
136
501 F3 families:
501-1
2.000
1.166
2.116
1.143
13
7
501-2
1.462
0.4359
1.520
0.533
13
6
501-3
1.333
0.4334
1.391
0.7310
21
5
501-4
2.083
1.356
1.982
1.948
12
7
501-5
2.417
2.629
2.045
3.069
12
8
501-6
1.440
0.5067
1.391
0.832
25
8
501-7
1.600
0.7115
1.612
0.986
10
5
501-8
1.214
0.1813


14
3
501-9
1.923
0.4102
2.032
0.3152
13
11
501-10
1.2143
0.1813


14
3
501-11
1.389
0.7222
-0.0665
4.920
18
4

69
Table 3-8. Continued.
Population Raw Data Uncensored Data
507 F3 Families:
Mean
Variance
Mean
Variance
N
N>1
507-1
1.571
0.8791
0.9004
2.974
14
6
507-2
1.769
2.190


13
3
507-3
1.500
0.3330


4
2
507-4
3.850
2.976
3.700
2.036
20
20
507-5
3.706
2.220
3.667
1.672
17
16
507-6
1.667
1.466


6
2
507-7
1.545
0.6406
1.635
0.7582
22
8
507-8
2.000
1.999
1.319
5.197
13
6
507-9
1.500
0.5770
1.438
0.953
14
5
507-10
1.273
0.4182


11
2
507-11
2.733
2.065
2.641
2.577
15
12
507-12
1.650
0.5553
1.753
0.5090
20
10
507-13
2.625
0.8383
2.625
0.6933
8
7
507-14
1.714
1.214
1.220
2.965
21
8
507-15
2.000
0.9091
2.356
1.188
12
7
507-16
1.611
0.6050
1.680
0.668
18
8
507-17
1.789
1.397
1.056
4.177
19
8
507-18
1.615
0.7564
1.386
1.365
13
6

70
Figure 3-1. Relationship of Fg family means and variances for callus growth
under (a) linear, (b) logarithmic, and (c) square root scales.

71
Cross 501 Linear Cross 507 Linear
3.00-
Cross 501 Logarithmic
LL?
1.75-
Cross 507 Logarithmic

2.75-,
/
y'
/
/
2.50-
/
y*
a? 2.25-
/
A

^
y = 0.653x + 0.784
2.00-
\X
/ y = 0.769x + 0.387
* r2 = 0.400
4 r2 = 0.602
till
1.75 -
ni*r1 1
2.00 2.25 2.50 2.75 3.00 3.25
1.75 2.00 2.25 2.50 2.75 3.00
Cross 501 Square Root Cross 507 Square Root
Figure 3-2. Regression of F3 family mean callus weight on F2 parental callus
weight under linear, logarithmic, and square root scales.

Frequency Frequency
72
(a)
Figure 3-3. Two hypothetical distributions with three response classes and two
thresholds.

73
(a)
Normal Quantile
(b)
Normal Quantile
Figure 3-4. Normal quantile plots of a hypothetical (a) intact and (b) censored
population.

74
Regeneration Score
Figure 3-5. Regeneration score histogram for genotype IRFL 6123.

75
Cross 501 Raw Data
Cross 501
Uncensored
Cross 507 Raw Data
Cross 507
Uncensored
Figure 3-6. Regression of F3 family mean regeneration score on F2 parent
regeneration score.

76
Figure 3-7. Profuse regeneration exhibited by members of two F3
families derived from cross 507

Log Callus Weight
77
Regeneration Score
Figure 3-8. Regression of callus weight on regeneration score for combined F3
populations.

78
CHAPTER 4
SUMMARY AND CONCLUSIONS
The point of departure for this dissertation was a study conducted by
Wofford et al. (1992b). Wofford's study identified a three stage (callus induction,
shoot induction, and shoot elongation) protocol utilizing the L2-based media of
Phillips and Collins (1979) as suitable for tissue culture of desmodium
hypocotyls, and obtained limited regeneration from a single genotypeIRFL
6123. The objective of the current work has been to examine the potential for
enhancing in vitro response in desmodium through improvements in culture
protocol and through breeding.
Chapter 2 focused on optimization of culture protocols. The work was not
intended to be an exhaustive investigation of potential methods for enhancing
in vitro response in desmodium. Given the wide range of variables that can
determine the effectiveness of tissue culture protocols and the limited prior work
with this crop, such an investigation would be beyond the scope of any single
study. Instead, the intent was to cast a broad netto determine if the poor,
highly genotype-specific regeneration observed in desmodium could be
significantly improved by applying a wide range of techniques similar to those
that had proven successful in other legumes. Accordingly, within each general
culture strategy (e.g., alternative explant sources, manipulation of auxin-
cytokinin ratio, substitution of different auxin sources) only a small number of
specific culture protocols were examined. For the same reason, number of
replications was often fairly low. Thus, failure, for example, to obtain
regeneration from leaf disks or petioles does not indicate that regeneration from
78

79
these explant types can not be achievedonly that none of the methods tested
yielded strongly encouraging results.
The first general culture strategy investigated was the use of alternative
explant sources. In addition to the previously proven hypocotyl explants, leaf
disks, petioles, and seedling cotyledons were tested. One previously
established regenerating genotype (IRFL 6123) and two nonregenerating
genotypes ( CIAT 13083 and UF 20) were tested. Regeneration was not
obtained from any of the alternative explant sources under the single culture
protocol tested. The only repeatable regeneration response was from
hypocotyls of IRFL 6123, as had been previously reported (Wofford et al.,
1992b).
Two more strategies for broadening the range of regenerating genotypes
were examined. The first was manipulation of 2,4-D concentration, with and
without added kinetin, in the shoot induction medium. This experiment failed to
induce regeneration in recalcitrant genotypes, but succeeded in identifying an
optimal 2,4-D concentration for regeneration in IRFL 6123. In addition, kinetin
was found to inhibit regeneration in this genotype. The last attempt at inducing
regeneration in recalcitrant genotypes involved direct regeneration from
hypocotyl explants. This approach failed to yield regeneration even from
IRFL 6123.
The remainder of Chapter 2 dealt with enhancing the regeneration
response in IRFL 6123. The first experiment demonstrated that reducing the
shoot induction period from the original 28 days to 14 days had no effect on
regeneration. Two additional experiments investigated the effects of growth
regulators in the shoot elongation step. An optimum ratio of 0.012 mg L'"*
picloram to 0.2 mg L BA was identified.

80
Histological examination of regenerating cultures revealed vascular
connections between shoots and the surrounding callus tissue, indicating that
regeneration was of an organogenic nature. However, early in the shoot
induction period, small structures resembling somatic embryos could
occasionally be observed. The combination of growth regulators used in the
shoot induction step is similar to that used for somatic embryogenesis in other
species, so it is possible that the regeneration observed in this study was some
aberrant form of somatic embryogenesis.
It was clearly established in Chapter 2 that regeneration in desmodium is
highly dependent on genotype and that broadening the range of regenerating
genotypes through modification of culture protocols is difficult or impossible.
Chapter 3 focused on the genetics of callus growth and regeneration. An
extensive effort at crossing regenerating with nonregenerating genotypes
yielded only three crosses501, 507, and 510. Cross 510 proved useless for
the analysis at hand due to severe internodal stunting and other morphological
abnormalities in the F-| and F2 generations. This may have been the result of
cytoplasmic effects, since cross 510 was essentially the reciprocal of cross 507,
which showed no abnormalities. Examination of the genetics and physiology of
this phenomenon might produce very interesting results.
The remaining two crosses were selfed through the F3 generation.
Analysis of callus growth and and regeneration in these crosses produced
many ambiguous results, but also yielded useful insights. A large amount of
effort was devoted to scaling of the callus growth trait. General statistical
considerations indicated that a simple linear scale was inadequate for
describing this trait. Logarithmic and square root transformation were both
more satisfactory than the linear scale, but it was not clear which of these
transformations was superior.

81
Formal genetic analysis of callus growth was approached via the joint
scaling test and variance partitioning. Neither method yielded conclusive
results. This may have been due to violation of the genetic assumptions
inherent in both methods, or may have been because both methods
inappropriately attempted to describe the trait with a simple additive-dominant
genetic model. Parent offspring regression yielded heritabilities ranging from
0.52 to 0.77, depending on cross and scale. It appears, then, that callus growth
in this germplasm is controlled to a significant extent by additive genetic effects,
but that there may in addition be nonadditive, nondominant genetic effects that
act to confound the joint scaling test and variance partitioning.
The regeneration trait presented unusual analytical difficulties. Many calli
showed no evidence of regeneration. As a result the majority of population
distributions were severely truncated. Truncated (or "censored") distributions
cannot be rendered normal by any mathematical transformation. A procedure
was described by which truncated population distributions can be
reconstructed, or "uncensored." While this procedure has apparently not been
previously applied to genetic analysis, it is conceptually similar to the threshold
trait approach described by Falconer (1981). The method is less than ideal, but
is preferable to attempting to analyze raw, severely censored data. Constraints
imposed by the data structure and by the uncensoring technique made it
impossible to conduct meaningful joint scaling tests or variance partitioning
analyses. Parent-offspring regression yielded relatively high heritability
estimates for cross 5070.416 for raw data; 0.460 for uncensored data. No
significant regression was obtained for cross 501. It appears likely that this
failure is due to shortcomings of the data structure, and that regeneration is in
fact weakly to moderately heritable in this cross.

82
Vasil (1987) has suggested that the genetics of regeneration is irrelevant,
and that with sufficient insight into the in vitro physiology of a species
regeneration can be induced in even the most difficult genotype. While this may
be true in theory, this work suggests that such an assumption can be impractical
or even counterproductive. Approximately equal effort was directed at culture
protocol and genetic approaches and the latter was found to be by far the more
productive path. It is noteworthy that the original regenerating parent
regenerated under a variety of culture conditions, suggesting a general genetic
proclivity to regenerate. A more experienced investigator with greater
resources may or may not have obtained better results from culture protocol
optimization, but success may well have been very costly in terms of time, effort,
and financial resources. The genetic approach has been relatively
straightforward and strikingly successful. The magnitude of this success was
particularly apparent from visual examination of regenerating calli. Several F3
calli, including that shown in Figure 3-7, grossly outperformed the parental
genotype, continuing to produce vigorous shoots through repeated subcultures,
and ultimately yielding dozens of shoots.
This study has demonstrated that regeneration in desmodium can be
greatly improved by conventional crossing techniques. Limited light was shed
on the genetic basis of regeneration in this crop, and many questions remain
unanswered. It is hoped that the F3 material produced in this study may prove
useful to any investigator who might desire to perform regeneration-dependent
work with this crop.

REFERENCES
Ammirato, P.V. 1983. Basic techniques of plant cell culture:
Embryogenesis. In D.A. Evans, W.R. Sharp, P.V. Ammirato, and Y.
Yamada (eds.) Handbook of plant cell culture. Vol. 1. Techniques for
propagation and breeding. Macmillan Publishing Co., New York.
Angeloni, P.N., H.Y. Rey, and L.A. Mroginski. 1988. Cultivo in vitro de
tejidos de Desmodium incanum y D. affine (Leguminosae).
Phyton:48:71 -76.
Bingham, E.T., L.V. Hurley, D.M. Kaatz, and J.W. Saunders. 1975.
Breeding alfalfa which regenerates from callus tissue in culture.
Crop Sci. 15:719-721.
Blaydes, D.F. 1966. Interaction of kinetin and various inhibitors in
the growth of soybean tissue. Physiol. Plant. 19:748-753.
Bovo, O.A., L.A. Mroginski, and H.Y. Rey. 1986. Regeneration of plants
from callus tissue of the pasture legume Lotononis bainessi. Plant
Cell Rep. 5:295-297.
Brown, D.C.W., and A. Atanassov. 1985. Role of genetic background in
somatic embryogenesis in Medicago. Plant Cell, Tissue, Organ
Culture 4:111-122.
Buiatti, M., S. Baroncelli, A. Bennici, M. Pagliai, and R. Tesi. 1974.
Genetics of growth and differentiation in vitro of Brassica olercea
var. botrytis. Z. Pflanzenzuchtg. 72:269-274.
Bullock, W.P., and P.S. Baenziger. 1982. Anther culture of wheat
(Triticum aestivum L.) FTs and their reciprocal crosses. Theor.
Appl. Genet. 62:155-159.
Cavalli, L. 1952. An analysis of linkage in quantitative inheritance.
In E.C.R. Reeve and C.H. Waddington (eds.) Quantitative inheritance.
HMSO, London.
83

Charmet, G., and S. Bernard. 1984. Diallel analysis of androgenetic
plant production in hexaploid Triticale X. triticosecale, Wittmack.
Theor. Appl. Genet. 69:55-61.
84
Choo, T.M. 1988. Plant regeneration in zigzag clover (Trifolium
medium L). Plant Cell Rep. 7:246-248.
Chow, K.H., and L.V. Crowder. 1972. Hybridization of Desmodium
canum (Gmel.) Schin. and Thell. and D. uncinatum (Jacq.) DC. Crop
Sci. 12:784-785.
Chow, K.H., and L.V. Crowder. 1973. Hybridization of Desmodium
species. Euphytica 22:339-404.
Collins, G.B., and G.C. Phillips. 1982. In vitro tissue culture and plant
regeneration in Trifolium pratense L. In E.D. Earle and Y Demardy
(eds.) Variation in plants regenerated from cells and tissue culture.
Praeger Sci. Pub. New York.
Deak, M., G.G. Kiss, C. Koncz, and D. Dudits. 1986. Transformation of
Medicago by Agrobacterium mediated gene transfer. Plant Cell Rep.
5:97-100.
Dodds, J.H. 1995. Experiments in plant tissue culture. 3rd Ed.
Cambridge University Press, Cambridge.
Dulieu, H. 1991. Inheritance of regeneration capacity in the genus
Petunia. Euphytica 53:173-181.
Dunwell, J.M., R.J. Francis, and W. Powell. 1987. Anther culture of
Hordeum vulgare L.: a genetic study of microspore callus production
and differentiation. Theor. Appl. Genet. 74:60-64.
Falconer, D.S. 1981. Introduction to quantitative genetics, 2nd Ed.
Longman Inc., New York.
Flick, C.E., D.A. Evans, and W.R. Sharp. 1983. Basic techniques of plant
cell culture: organogenesis. In D.A. Evans, W.R. Sharp, P.V. Ammirato,
and Y. Yamada (eds.) Handbook of plant cell culture. Vol. 1.
Techniques for propagation and breeding. Macmillan Publishing Co.,
New York. p. 13-81.

85
Foroughi-Wehr, B., W. Friedt, and G. Wenzel. 1982. On the genetic
improvement of androgenetic haploid formation in Hordeum vulgare
L. Theor. Appl. Genet. 62:233-239.
Fraley, R.T., S.G. Rogers, and R.B. Horsch. 1986. Genetic
transformation in higher plants. CRC Critical Reviews in Plant
Sciences 4:1-46.
Frankenberger, E.A., P.M. Hasegawa, and E.C. Tigchelaar. 1981. Diallel
analysis of shoot-forming capacity among selected tomato
genotypes. Z. Pflanzenphysiol. 102:233-242.
Gamborg, O.L., R.A. Miller, and K. Ojima. 1968. Nutrient requirements
of suspension cultures of soybean root cells. Exp. Cell Res. 50:151-
158.
Ghazi, T.D., H.V. Cheema, and M.W. Nabors. 1986. Somatic
embryogenesis and plant regeneration from embryogenic callus of
soybean, Glycine max. L. Plant Cell Rep. 5:452-456.
Gilbert, R.O. 1987. Statistical methods for environmental pollution
monitoring. Van Nostrand Reinhold Company, Inc., New York.
Haccius, B. 1978. Question of unicellular origin of nonzygotic
embryos in callus cultures. Phytomorphology 28:74-81.
Hazra, S., S.S. Sathaye, and A.F. Mascarenhas. 1989. Direct somatic
embryogenesis in peanut (Arachis hypogea). Biotechnology 7:949-
951.
Hills, F.J., and T.M Little. 1978. Agricultural experimentation design
and analysis. John Wiley and Sons, New York.
Horvath, K.B., and R.R. Smith. 1979. Plant regeneration from callus
culture of red and crimson clover. Plant Sci. Lett. 16:231-237.
Imrie, B.C., R.M. Jones, and P.C. Kerridge. 1983. Desmodium. In R.L.
Burt, P.P. Rotar, J.L. Walker, and M.W. Silvey (eds.) The role of
Centrosema, Desmodium, and Stylosanthes in improving tropical
pastures. Westview Tropical Agriculture Series No. 6. Westview
Press, Boulder, Colorado.

Kao, K.N., and M.R. Michayluk. 1981. Embryoid formation in alfalfa
cell suspensions from differentiated plants. In Vitro 17:645-648.
86
Kerns, H.R., U.B. Barwale, M.M. Meyer, Jr., and J.M. Widholm. 1986.
Correlation of cotyledonary node shoot proliferation and somatic
embryoid development in suspension cultures of soybean (Glycine
max L. Merr.). Plant Cell Rep. 5:140-143.
Ketchum, J.L.F., O.L. Gamborg, G.E. Hamming, and M.W. Nabors. 1987.
Progress report, tissue culture for crops project. Colorado State
University Press., Ft. Collins, Colorado.
Keyes, G.J., G.B. Collins, and N.L. Taylor. 1980. Genetic variation in
tissue cultures of red clover. Theor. Appl. Genet. 58:265-271.
Komatsuda, T., S. Enomoto, and K. Nakajima. 1989. Genetics of callus
proliferation and shoot differentiation in barley. J. Hered. 80:345-
350.
Koornneef, M., C.J. Hanhart, and L. Martinelli. 1987. A genetic
analysis of cell culture traits in tomato. Theor. Appl. Genet. 74:633-
641.
Kretchmer, A.E., Jr., J.B. Brolmann, G.H. Snyder, and S.W. Coleman.
1976. 'Florida' Carpon desmodium, a perennial tropical legume for
use in south Florida. Soil and Crop Sci. Soc. of Fla. Proc. 35:25-30.
Kumar, A.S., T.P. Reddy, and G.M. Reddy. 1983. Plantlet regeneration
from different callus cultures of pigeonpea (Cajanus cajan L). Plant
Sci. Lett. 32:271-278.
Kumar, A.S., T.P. Reddy, and G.M. Reddy. 1984. Multiple shoots from
cultured explants of pigeonpea and atylosia species. SABRAO J.
16:101-105.
Kumar, A.S., T.P. Reddy, and G.M. Reddy. 1985. Genetic analysis of
certain in vitro and in vivo parameters in pigeonpea (Cajanus cajan
L.). Theor. Appl. Genet. 70:151-156.
Lazar, M.D., P.S. Baenziger, and G.W. Schaeffer. 1984. Combining
abilities and heritability of callus formation and plantlet
regeneration in wheat (Triticum aestivum L.) anther cultures. Theor.
Appl. Genet. 68:131-134.

87
Lupotto, E. 1983. Propagation of an embryogenic culture of Medicago
sativa L. Z. Pflanzenphysiol. 111:95-104.
Maheshwaran, G., and E.G. Williams. 1984. Direct somatic embryoid
formation on immature embryos of Trifolium repens, Trifolium
pratense, and Medicago sativa and rapid clonal propagation of
Trifolium repens. Ann. Bot. 54:201-211.
Malmberg, R.L. 1979. Regeneration of whole plants from callus
cultures of diverse genetic lines of Pisum sativum L. Planta
146:243-244.
Mather, K., and J.L. Jinks. 1971. Biometrical Genetics. Cornell
University Press, Ithaca, New York.
McKently, A.H., G.A. Moore, and F.P. Gardner. 1990. In vitro plant
regeneration of peanut from seed explants. Crop Sci. 30:192-196.
McKently, A.H., G.A. Moore, and F.P. Gardner. 1992. Regeneration of
peanut and perennial peanut from cultured leaf tissue. Crop Sci.
31:833-837.
Mehta, U., and H.Y. Mohan Ram. 1980. Regeneration of plantlets from
the cotyledons of Cajanus cajan. Indian J. Exp. Bot. 18:800-802.
Meijer, E.G.M. 1982. Shoot formation in tissue cultures of three
cultivars of the tropical forage legume Stylosanthes guyanensis. Z.
Pflanzenzuchtg. 89:169-172.
Murashige, T., and F. Skoog. 1962. A revised medium for rapid growth
and bioassays with tobacco tissue cultures. Physiol. Plant. 15:473-
497.
Newman, M.C., and P.M. Dixon. 1989. Estimating mean and variance for
environmental samples with below detection limit observations.
Water Res. Bull. 25:905-916.
Oelck, M.M., and O. Scheider. 1983. Genotypic differences in some
legume species affecting the redifferentiation ability from callus to
plants. Z. Pflanzenzuchtg. 91:312-321.

88
Ohashi, H. 1973. The Asiatic species of desmodium and its allied
genera (Leguminosae). Ginkgoana. Contributions of the flora of Asia
and the Pacific region. Academia Sci. Book, Inc., Tokyo.
Phillips, G.C., and G.B. Collins. 1979. In vitro tissue culture of
selected legumes and plant regeneration from callus cultures of red
clover. Crop Sci. 19:59-64.
Phillips, G.C., and G.B. Collins. 1980. Somatic embryogenesis from
cell suspension cultures of red clover. Crop Sci. 20:323-326.
Quesenberry, K.H, M.R. McKellar, and D.E. Moon. 1989. Evaluation and
hybridization of germplasm in the Desmodium heterocarpon -
Desmodium ovalifolium species complex. In Proc. XVI Inti. Grassland
Congress. Nice, France, p. 251-252.
Quesenberry, K.H., D.S. Wofford, P.A. Krottje, and R.L. Smith. 1992.
Production of transgenic red clover plants using Agrobacterium-
mediated DNA transfer. Proceedings of the Twelfth Trifolium
Conference. University of Florida, Gainesville, Florida, p. 5-10.
Quimio, C.A., and F.J. Zapata. 1990. Diallel analysis of callus
induction and green-plant regeneration in rice anther culture. Crop
Sci. 30:188-192.
Reisch, B., and E.T. Bingham. 1980. The genetic control of bud
formation from callus cultures of diploid alfalfa. Plant Sci. Lett.
20:71-77.
Rotar, P.P. and U. Urata. 1967. Cytological studies in the genus
Desmodium species: some chromosome counts. Amer. J. Bot. 54:1-4.
Rowe, K.E., and W.L. Alexander. 1980. Computations for estimating
the genetic parameters in joint-scaling tests. Crop Sci. 20:109-110.
Santos, A.V.P. dos, E.G. Cutter, and M.R. Davey. 1983. Origin and
development of somatic embryos in Medicago sativa L. (alfalfa).
Protoplasma 117:107-115.
SAS Institute. 1994. JMP Statistics and Graphics Guide. SAS
Institute Inc., Cary, North Carolina.

89
Sarria, R., A.Calderon, A.M. Thro, E. Torres, J.Mayer, and W.M. Rocca.
1994. Agrobacterium mediated transformation of Stylosanthes
guianensis and production of transgenic plants. Plant Sci. 96:119-
127.
Saunders, J.W., G.L. Hosfield, and A. Levi. 1987. Morphogenetic
effects of 2,4-dichlorophenoxyacetic acid on pinto beans
(Phaseolus vulgaris L.) leaf explants in vitro. Plant Cell Rep. 6:46-
49.
Schenk, R.U., and A.C. Hildebrandt. 1972. Medium and techniques for
induction and growth of monocotyledononous and dicotyledonous
plant cell cultures. Can. J. Bot. 50:199-204.
Schubert, B.G. 1963. Desmodium: Preliminary studies IV. Journal of
the Arnold Arboretum 44:284-297.
Schultze-Kraft, R., and G. Benavides. 1988. Germplasm collection and
preliminary evaluation of Desmodium ovalifolium Wall. Gen.
Resources Comm. 12:1-20.
Schultze-Kraft, R., and S. Pattanavibul. 1985. Collecting native
forage legumes in eastern Thailand. IBPGR Newsletter. 9:4-5.
Scowcroft, W.R., and J.A. Adamson. 1976. Organogenesis from callus
cultures of the legume, Stylosanthes hamata. Plant Sci. Lett. 7:39-
42.
Sellars, R.M., G.M. Southward, and G.C. Phillips. 1990. Adventitious
somatic embryogenesis from cultured immature zygotic embryos of
peanut and soybean. Crop Sci. 30:408-414.
Simmonds, N.W. 1979. Principles of crop improvement. Longman Inc.,
New York.
Swanson, E.B., and D.T. Tomes. 1980. Plant regeneration from cell
culture of Lotus corniculatus and the selection and characterization
of 2,4-D tolerant cell lines. Can. J. Bot. 58:1205-1209.
Takahasi, M. 1952. Tropical forage legume and browse plants
research in Hawaii. Proc. 6th Int. Grass. Cong. 1411-1416.

90
Tisserat, B., E.B. Esan, and T. Murashige. 1979. Somatic
embryogenesis in angiosperms. Hortic. Rev. 1:1-78.
Vasil, I.K. 1987. Developing cell and tissue culture systems for the
improvement of cereal and grass crops. J. Plant Physiol. 128:193-
218.
Viera, M.L., B. Jones, E.C. Cocking, and M.R. Davey. 1990. Plant
regeneration from protoplasts isolated from seedling cotyledons of
Stylosanthes guianensis, S. macrocephala and S. scabra. Plant Cell
Rep. 9:289-292.
Wan, Y., E.L. Sorensen, and G.H. Liang. 1988. Genetic control of in
vitro regeneration in alfalfa. Euphytica 39:3-9.
Walker, K.A., M.L. Wendeln, and E.G. Jaworski. 1979. Organogenesis in
callus tissue of Medicago sativa. The temporal separation of
induction processes from differentiation processes. Plant Sci. Lett.
16:23-30.
Webb, K.J., M.F. Fay, and P.J. Dale. 1987. An investigation of
morphogenesis within the genus Trifolium. Plant Cell, Tissue, Organ
Culture 11:37-46.
White, D.W.R., and D. Greenwood. 1987. Transformation of the forage
legume Trifolium repens L. using binary Agrobacterium vectors.
Plant Mol. Bio. 8:461-469.
Wofford, D.S., K.H. Quesenberry, and D.D. Baltensperger. 1992a. In
vitro culture responses of alyceclover genotypes on four media
systems. Crop Sci. 32:261-265.
Wofford, D.S., K.H. Quesenberry, and D.D. Baltensperger. 1992b. Tissue
culture regeneration of desmodium. Crop Sci. 32:266-268.
Younge, O.R., D.L. Plucknett, and P.P. Rotar. 1964. Culture and yield
performance of Desmodium intortum and D. canum in Hawaii. Tech.
Bull. No. 59. Hawaii Agrie. Exp. Stat., Hawaii.
Zar, J.H. 1984. Biostatistical analysis. 2nd Ed. Prentice Hall,
Englewood Cliffs, New Jersey.

APPENDIX A
ANALYSIS OF VARIANCE FOR BETWEEN- AND WITHIN-FAMILY
VARIANCES FOR CALLUS GROWTH
Between- and within-family variances were estimated by analysis
of variance with F3 family as the class variable. All ANOVA
analyses yielded highly significant results. Expected mean square
for families equals (aw2 + nab2), where ow2 is within-family
variance (estimated by MSE), n is harmonic mean of number of
individuals per family, and ab2 is between-family variance. This
equation is solved for ob2 = (MSmode| MSE)/n. The following table
presents MSmode|, MSE, ab2, aw2, and n for each cross and each scale.
Cross
Scale
M^model
MSE(= aw2)
ab2
n
501
Linear
0.9901
0.0623
0.0669
15.21
501
Logarithmic
1.4962
0.0859
0.9271
11
501
Square Root
583.95
32.856
36.462
11
507
Linear
0.3166
0.0361
0.0220
12.73
507
Logarithmic
0.9910
0.1132
0.0689
ll
507
Square Root
273.78
29.392
19.199
ll
91

APPENDIX B
DERIVATION OF VARIANCE PARTITIONING MATRICES FOR CALLUS
GROWTH
Coefficients for variance partitioning of the callus growth were based on those
given by Mather and Jinks (1971, 1977). The following general coefficients
were given for the populations treated in the present study:
Observational
Component
D
H
EW
eb
Sampling
Variance
V1F2
1/2
1/4
1
0
0
V1F3
1/2
1/16
0
1
Vn V2F3
V2F3
1/4
1/8
1
0
0
W1F23
1/2
1/8
0
0
0
Ei
0
0
1
0
0
Since there is only one entry for sampling variance and the sampling variance
for V-| p3 is a function of V2F3, this column can be combined into the other
columns. In the case of cross 501, n (the harmonic mean of the number of
individuals per F3 family) is equal to 15.21. Therefore, after combining the
sampling variance column, the coefficient for D for V-)p3 becomes
V2 + ((1/15.21) x 1/4), or 0.516435. The other entries in the V-| p3 row are
similarly altered. Since Eg is a function of Ew, this column can also be
92

93
eliminated. The coefficient for Ew for V-| P3 becomes the sum of the contribution
to this column from the V2f3 row (1/15.21 x 1) plus the contribution from the Eb
column in the Vip3 row (1/15.21), or 0.1314. The resulting table is shown
below:
D
H
£
LU
V1F2
1/2
1/4
1
V1F3
0.516435
0.0707175
0.1314803
V2F3
1/4
1/8
1
W1F23
1/2
1/8
0
E1
0
0
1
Derivation of coefficients for cross 507 are similar, but n for this cross is 12.73.

APPENDIX C
RESULTS OF SECOND AND THIRD ITERATIONS OF VARIANCE PARTITIONING
FOR CALLUS GROWTH

Results of variance partitioning second iteration:
Component
Cross
Scale
D
H
Ew
Whole Model
Estimate
LL
A
-Q
O
Q.
Estimate
Prob>F
Estimate
ProtF
F
Prob>F
501
Linear
0.1362
0.3918
-0.0672
0.8024
0.0334
0.1578
27.73
0.1385
507
Linear
0.0289
0.7176
0.0411
0.8616
0.0138
0.2744
9.16
0.2372
501
Logarithmic
0.1327
0.3790
0.1501
0.6365
0.4253
0.1218
49.48
0.1040
507
Logarithmic
0.0637
0.7024
0.3337
0.5073
0.0445
0.2465
14.89
0.1877
501
Square Root
70.02
0.0642
-18.68
0.4518
17.37
0.0206
1641
0.0181
507
Square Root
23.89
0.1992
46.39
0.2828
18.21
0.0503
335.7
0.0401
co
cn

Results of variance partitioning third iteration:
Component
Cross
Scale
D
H
Ew
Whole Model
Estimate
Prob>F
Estimate
Prob>F
Estimate
Prob>F
F
Prob>F
501
Linear
0.1372
0.4024
-0.0692
0.8090
0.0323
0.1528
28.57
0.1365
507
Square Root
23.90
0.1999
46.37
0.2830
18.21
0.0500
336.9
0.0400

APPENDIX D
RAW CALLUS GROWTH AND REGENERATION DATA
Family
f3
Individual
Regeneration Score
Progeny Parent
Callus Weight
Progeny Pare
507-1
1
1
1
80
129
507-1
2
1
1
140
129
507-1
3
1
1
220
129
507-1
4
1
1
100
129
507-1
5
1
1
40
129
507-1
6
1
1
140
129
507-1
7
1
1
40
129
507-1
8
2
1
290
129
507-1
9
2
1
90
129
507-1
10
4
1
120
129
507-1
11
1
1
130
129
507-1
12
1
1
130
129
507-1
13
2
1
180
129
507-1
14
3
1
190
129
507-2
1
1
3
130
265
507-2
2
4
3
440
265
507-2
3
1
3
370
265
507-2
4
1
3
290
265
507-2
5
4
3
350
265
507-2
6
1
3
850
265
507-2
7
1
3
710
265
507-2
8
1
3
60
265
507-2
9
1
3
110
265
507-2
10
1
3
70
265
507-2
11
1
3
40
265
507-2
12
5
3
30
265
507-2
13
1
3
90
265
507-3
1
2
2
70
191
507-3
2
2
2
340
191
507-3
3
1
2
90
191
507-3
4
1
2
100
191
507-4
1
3
6
180
180
507-4
2
3
6
40
180
507-4
4
2
6
340
180
507-4
5
2
6
90
180
507-4
6
3
6
190
180
507-4
7
2
6
760
180
507-4
8
4
6
250
180
507-4
9
7
6
170
180
97

98
Family
f3
Individual
Regeneration Score
Progeny Parent
Callus Weight
Progeny Pare
507-4
10
5
6
100
180
507-4
11
3
6
60
180
507-4
13
2
6
470
180
507-4
14
7
6
200
180
507-4
15
5
6
730
180
507-4
16
2
6
140
180
507-4
19
5
6
430
180
507-4
20
4
6
110
180
507-4
21
3
6
150
180
507-4
22
7
6
320
180
507-4
23
3
6
210
180
507-4
24
5
6
270
180
507-5
1
1
5
590
975
507-5
2
6
5
890
975
507-5
3
3
5
510
975
507-5
4
3
5
430
975
507-5
5
4
5
450
975
507-5
6
4
5
650
975
507-5
7
3
5
480
975
507-5
8
4
5
1120
975
507-5
11
4
5
290
975
507-5
14
3
5
430
975
507-5
15
6
5
700
975
507-5
16
2
5
140
975
507-5
17
3
5
990
975
507-5
18
7
5
400
975
507-5
19
3
5
230
975
507-5
20
4
5
780
975
507-5
21
3
5
680
975
507-6
1
1
2
20
184
507-6
2
1
2
20
184
507-6
3
2
2
100
184
507-6
4
1
2
20
184
507-6
5
1
2
150
184
507-6
6
4
2
590
184
507-7
1
3
1
310
285
507-7
2
1
1
170
285
507-7
3
3
1
470
285
507-7
4
1
1
150
285
507-7
8
3
1
380
285
507-7
9
2
1
150
285
507-7
10
1
1
580
285
507-7
11
1
1
640
285
507-7
12
1
1
130
285
507-7
14
2
1
660
285
507-7
15
1
1
50
285

99
Family F3
Individual
Regeneration Score
Progeny Parent
Callus Weight
Progeny Parent
507-7
16
1
507-7
17
1
507-7
18
1
507-7
19
1
507-7
20
3
507-7
21
2
507-7
22
1
507-7
23
1
507-7
24
1
507-7
25
2
507-7
26
1
507-8
1
1
507-8
2
2
507-8
3
1
507-8
4
1
507-8
6
1
507-8
7
1
507-8
8
5
507-8
9
4
507-8
10
2
507-8
11
1
507-8
12
2
507-8
13
4
507-8
14
1
507-9
1
3
507-9
2
1
507-9
3
1
507-9
4
1
507-9
5
1
507-9
6
1
507-9
8
3
507-9
9
2
507-9
10
1
507-9
13
1
507-9
14
1
507-9
15
2
507-9
16
1
507-9
19
2
507-10
2
1
507-10
3
2
507-10
4
3
507-10
5
1
507-10
8
1
507-10
9
1
507-10
10
1
1
330
285
1
230
285
1
110
285
1
240
285
1
800
285
1
110
285
1
530
285
1
280
285
1
300
285
1
150
285
1
90
285
1
960
400
1
590
400
1
130
400
1
440
400
1
440
400
1
400
400
1
400
400
1
360
400
1
200
400
1
220
400
1
490
400
1
120
400
1
470
400
1
420
380
1
140
380
1
520
380
1
390
380
1
910
380
1
850
380
1
130
380
1
900
380
1
250
380
1
1110
380
1
200
380
1
400
380
1
170
380
1
180
380
1
140
106
1
80
106
1
380
106
1
360
106
1
20
106
1
40
106
1
20
106

100
Family F3 Regeneration Score Callus Weight
Individual
Progeny
Parent
Progeny
Parent
507-10
11
1
1
20
106
507-10
12
1
1
30
106
507-10
13
1
1
70
106
507-10
14
1
1
50
106
507-11
1
2
5
160
360
507-11
2
1
5
120
360
507-11
3
1
5
110
360
507-11
4
5
5
320
360
507-11
5
3
5
260
360
507-11
6
5
5
370
360
507-11
7
2
5
200
360
507-11
8
3
5
180
360
507-11
9
2
5
80
360
507-11
10
1
5
130
360
507-11
11
5
5
210
360
507-11
12
3
5
250
360
507-11
13
2
5
380
360
507-11
14
2
5
130
360
507-11
15
4
5
290
360
507-12
1
1
1
190
92
507-12
2
3
1
260
92
507-12
3
1
1
90
92
507-12
5
1
1
80
92
507-12
6
1
1
110
92
507-12
8
1
1
340
92
507-12
10
1
1
140
92
507-12
11
1
1
100
92
507-12
13
1
1
120
92
507-12
14
2
1
120
92
507-12
16
2
1
160
92
507-12
18
1
1
100
92
507-12
20
2
1
70
92
507-12
21
3
1
170
92
507-12
22
2
1
60
92
507-12
23
2
1
80
92
507-12
24
1
1
50
92
507-12
25
3
1
100
92
507-12
26
2
1
140
92
507-13
1
4
2
480
382
507-13
2
3
2
210
382
507-13
3
2
2
270
382
507-13
4
2
2
290
382
507-13
5
2
2
220
382
507-13
6
2
2
170
382
507-13
7
4
2
900
382
507-13
8
2
2
150
382

101
Family
f3
Individual
Regeneration Score
Progeny Parent
Callus Weight
Progeny Pare
507-14
1
3
2
640
220
507-14
2
1
2
160
220
507-14
3
1
2
630
220
507-14
4
2
2
120
220
507-14
5
1
2
540
220
507-14
6
1
2
730
220
507-14
7
1
2
800
220
507-14
8
1
2
720
220
507-14
9
1
2
580
220
507-14
10
4
2
270
220
507-14
11
1
2
420
220
507-14
12
1
2
380
220
507-14
13
1
2
250
220
507-14
14
4
2
650
220
507-14
15
1
2
170
220
507-14
16
2
2
580
220
507-14
17
4
2
540
220
507-14
18
1
2
360
220
507-14
19
2
2
330
220
507-14
20
1
2
920
220
507-14
21
2
2
420
220
507-15
1
3
1
240
141
507-15
3
2
1
240
141
507-15
5
2
1
120
141
507-15
6
1
1
70
141
507-15
7
1
1
200
141
507-15
8
1
1
200
141
507-15
9
3
1
130
141
507-15
10
1
1
150
141
507-15
11
3
1
130
141
507-15
12
3
1
140
141
507-15
14
1
1
70
141
507-15
15
3
1
100
141
507-16
1
2
1
160
126
507-16
2
1
1
60
126
507-16
3
1
1
130
126
507-16
4
3
1
160
126
507-16
5
2
1
180
126
507-16
6
3
1
210
126
507-16
7
1
1
190
126
507-16
8
1
1
120
126
507-16
9
2
1
110
126
507-16
10
1
1
210
126
507-16
11
1
1
160
126
507-16
12
1
1
70
126
507-16
13
1
1
90
126

102
Family
F3
Individual
Regeneration Score
Progeny Parent
Callus Weight
Progeny Pare
507-16
14
1
1
90
126
507-16
15
2
1
120
126
507-16
16
1
1
150
126
507-16
17
2
1
90
126
507-16
18
3
1
180
126
507-17
2
3
2
380
344
507-17
3
1
2
370
344
507-17
4
3
2
180
344
507-17
5
5
2
340
344
507-17
7
1
2
150
344
507-17
8
1
2
90
344
507-17
10
1
2
40
344
507-17
11
4
2
130
344
507-17
12
1
2
100
344
507-17
13
2
2
200
344
507-17
14
2
2
430
344
507-17
15
2
2
40
344
507-17
17
1
2
240
344
507-17
19
2
2
120
344
507-17
20
1
2
120
344
507-17
21
1
2
70
344
507-17
22
1
2
250
344
507-17
23
1
2
230
344
507-17
24
1
2
70
344
507-18
1
2
1
300
358
507-18
2
1
1
60
358
507-18
3
2
1
340
358
507-18
4
2
1
130
358
507-18
5
1
1
150
358
507-18
6
1
1
380
358
507-18
7
1
1
90
358
507-18
8
2
1
190
358
507-18
9
1
1
320
358
507-18
10
4
1
790
358
507-18
11
1
1
340
358
507-18
12
2
1
190
358
507-18
13
1
1
200
358
501-1
1
1
3
100
160
501-1
2
2
3
120
160
501-1
3
3
3
150
160
501-1
4
1
3
150
160
501-1
5
1
3
120
160
501-1
7
1
3
200
160
501-1
8
3
3
210
160
501-1
9
3
3
280
160
501-1
10
4
3
170
160
501-1
11
1
3
70
160

103
Family F3 Regeneration Score Callus Weight
Individual Progeny Parent Progeny Parent
501-1
12
2
3
70
160
501-1
13
3
3
160
160
501-1
14
1
3
90
160
501-2
1
1
1
1240
984
501-2
2
1
1
1110
984
501-2
3
2
1
860
984
501-2
4
3
1
1000
984
501-2
5
1
1
890
984
501-2
6
2
1
600
984
501-2
7
1
1
1300
984
501-2
8
1
1
740
984
501-2
9
2
1
350
984
501-2
10
1
1
750
984
501-2
11
1
1
1870
984
501-2
12
1
1
550
984
501-2
13
2
1
620
984
501-3
1
1
3
210
298
501-3
2
1
3
320
298
501-3
3
1
3
420
298
501-3
4
1
3
370
298
501-3
5
2
3
440
298
501-3
6
3
3
140
298
501-3
7
2
3
90
298
501-3
8
2
3
610
298
501-3
9
1
3
190
298
501-3
10
1
3
130
298
501-3
13
1
3
310
298
501-3
14
1
3
130
298
501-3
15
1
3
560
298
501-3
16
1
3
330
298
501-3
17
1
3
510
298
501-3
18
1
3
130
298
501-3
19
1
3
830
298
501-3
20
1
3
130
298
501-3
21
3
3
160
298
501-3
22
1
3
150
298
501-3
23
1
3
530
298
501-4
1
3
1
590
1352
501-4
2
2
1
710
1352
501-4
3
1
1
190
1352
501-4
4
1
1
580
1352
501-4
5
2
1
920
1352
501-4
6
4
1
1140
1352
501-4
7
2
1
1680
1352
501-4
8
1
1
240
1352
501-4
9
1
1
540
1352

104
Family F3 Regeneration Score Callus Weight
Individual Progeny Parent Progeny Parent
501-4
10
1
1
1740
1352
501-4
11
3
1
1380
1352
501-4
12
4
1
230
1352
501-5
3
1
3
80
382
501-5
7
6
3
230
382
501-5
8
3
3
170
382
501-5
9
2
3
130
382
501-5
10
1
3
110
382
501-5
11
2
3
150
382
501-5
12
2
3
230
382
501-5
13
1
3
50
382
501-5
17
5
3
120
382
501-5
18
2
3
90
382
501-5
20
3
3
130
382
501-5
21
1
3
100
382
501-6
1
1
1
360
284
501-6
2
1
1
200
284
501-6
3
1
1
130
284
501-6
4
2
1
120
284
501-6
5
1
1
110
284
501-6
6
2
1
220
284
501-6
7
1
1
380
284
501-6
8
1
1
610
284
501-6
9
1
1
320
284
501-6
10
3
1
520
284
501-6
11
1
1
240
284
501-6
12
1
1
530
284
501-6
13
1
1
360
284
501-6
14
1
1
150
284
501-6
15
2
1
990
284
501-6
16
1
1
290
284
501-6
18
2
1
1030
284
501-6
19
1
1
760
284
501-6
20
1
1
830
284
501-6
22
1
1
480
284
501-6
23
1
1
130
284
501-6
24
2
1
500
284
501-6
25
3
1
240
284
501-6
27
3
1
390
284
501-6
28
1
1
300
284
501-7
2
3
2
180
313
501-7
3
1
2
180
313
501-7
4
1
2
80
313
501-7
5
1
2
50
313
501-7
6
3
2
330
313
501-7
7
1
2
40
313

105
Family
f3
Individual
Regeneration Score
Progeny Parent
Callus Weight
Progeny Pare
501-7
8
2
2
90
313
501-7
9
1
2
140
313
501-7
10
1
2
410
313
501-7
11
2
2
900
313
501-8
1
1
1
130
206
501-8
2
1
1
330
206
501-8
3
1
1
250
206
501-8
4
2
1
340
206
501-8
5
1
1
90
206
501-8
6
1
1
390
206
501-8
7
2
1
220
206
501-8
8
1
1
670
206
501-8
9
1
1
90
206
501-8
10
1
1
190
206
501-8
11
1
1
210
206
501-8
12
1
1
360
206
501-8
13
2
1
610
206
501-8
14
1
1
430
206
501-9
1
2
4
550
365
501-9
2
1
4
390
365
501-9
3
3
4
160
365
501-9
4
2
4
130
365
501-9
5
2
4
660
365
501-9
6
1
4
90
365
501-9
7
2
4
430
365
501-9
8
2
4
520
365
501-9
9
3
4
540
365
501-9
10
1
4
480
365
501-9
11
2
4
260
365
501-9
12
2
4
630
365
501-9
13
2
4
730
365
501-10
1
1
1
700
345
501-10
2
1
1
190
345
501-10
3
2
1
290
345
501-10
4
1
1
830
345
501-10
5
1
1
420
345
501-10
6
1
1
480
345
501-10
7
1
1
720
345
501-10
8
1
1
590
345
501-10
9
1
1
600
345
501-10
10
2
1
410
345
501-10
11
2
1
660
345
501-10
12
1
1
840
345
501-10
13
1
1
450
345
501-10
14
1
1
710
345
501-11
1
1
1
240
133

106
Family F3 Regeneration Score Callus Weight
Individual
Progeny
Parent
Progeny
Parent
501-11
2
2
1
480
133
501-11
3
4
1
260
133
501-11
4
3
1
140
133
501-11
5
1
1
80
133
501-11
7
1
1
140
133
501-11
8
1
1
440
133
501-11
9
1
1
270
133
501-11
10
1
1
200
133
501-11
11
2
1
540
133
501-11
12
1
1
420
133
501-11
13
1
1
270
133
501-11
14
1
1
330
133
501-11
15
1
1
370
133
501-11
16
1
1
180
133
501-11
17
1
1
260
133
501-11
18
1
1
120
133
501-11
19
1
1
180
133

BIOGRAPHICAL SKETCH
Peter Krottje was born in Haskell, New Jersey, in 1951. He attended
Fordham University, receiving a B.S. in biology in 1973. He received an M.S. in
soil science from the University of Florida in 1980. He has been very fortunate.
107

I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in
scope and quality, as a dissertation for the degree of Doctor of Philosophy.
UUVIVJ viiwi u,
Associate Professor of Agronomy
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in
scope and quality, as a dissertation for the degree of Doctor of Philosophy.
Professor of Agronomy
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in
scope and quality, as a dissertation for the degree of Doctor of Philosophy.
Rex L. Smith
Professor of Agronomy
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in
scope and quality, as a dissertation for the degree of Doctor of Philosophy.
Paul M. Lyrene fJ
Professor of Horticultural Science
I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in
scope and quality, as a dissertation for the degree of Doctor of Philosophy.
CL
Gloria A. Moore
Professor of Horticultural Science

This dissertation was submitted to the Graduate Faculty of the College
of Agriculture and to the Graduate School and was accepted as partial
fulfillment of the requirements for the degree of Doctor of Philosophy.
August 1995
Dean, College of Agriculture
Dean, Graduate School



11
Materials and Methods
Germplasm. Four genotypes were selected to represent a wide range of in
vitro responses. In previous work (Wofford et al., 1992b) IRFL 6123 had been
identified as a strong regenerator, CIAT 13083 as a possible regenerator, and
UF 20 and UF 144 as nonregenerators. IRFL 6123 is classified as D. hetero-
carpon ssp. angustifolium, UF 20 represents D. heterocarpon ssp.
heterocarpon, UF 144 represents D. ovalifolium, and CIAT 13083 possesses
characteristics intermediate between D. heterocarpon ssp. heterocarpon and
D. ovalifolium.
General protocol. Young, fully expanded leaves from greenhouse-grown
plants were used to obtain leaf disc and petiole explants. Leaves and petioles
were sterilized by immersion in 70% ethanol for 30 seconds followed by
immersion in a 15% (v:v) Chlorox solution for one minute and several rinses
with sterile deionized water. A stainless steel cork borer was used to cut 5 mm
leaf discs, each containing a portion of midvein. Leaf discs were cultured with
abaxial surface upward. Petioles were cut into sections approximately 10 mm
in length.
For hypocotyl and cotyledon explants, seeds were scarified and sterilized
by a 12-minute immersion in concentrated sulfuric acid followed by several
rinses in sterile deionized water. Seeds were then placed in petri dishes
containing SGL medium (Collins and Phillips, 1982) and incubated at 27 C.
Hypocotyls and cotyledons were excised and placed onto callus induction
medium after the hypocotyls reached a length of 7 to 10 mm. This occurred
between 7 and 10 days after scarification. Hypocotyl explants were excised to a
length of approximately 5 mm, with care taken to avoid radicle tissue and


Table 2-3. Response of genotype IRFL 6123 to benzyladenine and
gibberellic acid levels in the shoot elongation medium.
24
BA GA3 Mean frequency responding explants t
cone. cone. Unifoliate Multifoliate
mg L'1 %
0.2
0
50
a
0
a
0.2
0.2
54
a
0
a
0.6
0
33
a
4
a
0.6
0.2
38
a
0
a
2.0
0
13
b
21
b
2.0
0.2
13
b
0
a
t Mean separation by Tukey's HSD (a=0.05).


9
organogenesis can often be induced by "reversal transfer" of callus from a high
auxin, low cytokinin to a low auxin, high cytokinin medium (Dodds and Roberts,
1985). Somatic embryos can often be induced on media containing high levels
of auxinparticularly 2,4-Dsometimes in combination with low levels of
cytokinin (Ammirato, 1983). Reviews of legume work (Allavena, 1983;
Ammirato, 1983; Flick et al., 1983; Phillips and Collins, 1983) indicate that the
above generalities are applicable. Earlier work favored use of kinetin to induce
organogenesis, but more recently, BA (6-benzylaminopurine) generally in the
concentration range of 0.5 to 5.0 mg L"2, has been widely used (Webb et al.,
1987; Vieira et al., 1990; McKently et al., 1990). Somatic embryogenesis in
legumes has been induced with 2,4-D concentrations ranging from
0.001 mg L1 (Phillips and Collins, 1980) to 80 mg L'1 (Saunders et al., 1987).
In callus based regeneration systems, there is strong evidence that culture
conditions for the initial establishment of callus can strongly influence the
potential for subsequent organogenesis or embryogenesis (Saunders et al.,
1985; Tisserat et al., 1978). This issue has received relatively little attention in
legumes, and is generally approached empirically. Callus induction and
regeneration can sometimes occur on the same medium (Santos et al., 1983;
Bovo et al., 1986), but most work has employed a separate callus induction step
(Phillips and Collins, 1979; Walker et al., 1979; Flick et al., 1983).
Little has been published specifically on tissue culture of desmodium.
Angeloni et al (1988) produced multiple shoots from from shoot tip cultures of
D. incanum on MS medium supplemented with 1.0 mg L'1 NAA
(naphthalenacetic acid), 0.1 mg L'1 BA, and 1.0 mg L'1 GA (gibberellic acid).
Shoot tip culture is regarded as meristem cloning rather than true regeneration,
however. Wofford et al. (1992b) evaluated regeneration from seedling
hypocotyl explants of six genotypes of D. heterocarpon and D. ovalifolium


13
medium with 0.002 mg L1 picloram and 0.2 mg L1 BA for another 28 days.
Shoot bud production and callus appearance were evaluated regularly
throughout the culture period.
Effect of 2.4-D level and kinetin level on shoot bud induction. Seedling
hypocotyls from genotypes UF 20, UF 144, and IRFL 6123 were cultured for
twenty eight days on callus induction medium as described above. Calli were
then transferred to the following L2-based shoot bud induction media:
(1) 0.1 mg L1 2,4-D, 2.0 mg L'1 adenine
(2) 0.3 mg L"1 2,4-D, 2.0 mg L"1 adenine
(3) 1.0 mg L'1 2,4-D, 2.0 mg L"1 adenine
(4) 3.0 mg L"1 2,4-D, 2.0 mg L"1 adenine
(5) 0.1 mg L'1 2,4-D, 0.15 mg L'1 kinetin, 2.0 mg L'1 adenine
(6) 1.0 mg L'1 2,4-D, 0.15 mg L'1 kinetin, 2.0 mg L"1 adenine
The 28-day shoot induction treatments were followed by a 28-day shoot
elongation treatment as described for the previous experiment. Ten explants
(two per dish) were used per genotype-treatment combination.
Induction of direct regeneration from hypocotyl explants without an
intervening callus induction step. This experiment was carried out identically to
the previous experiment with the exception that explants were placed directly
onto the six shoot induction treatments rather than onto callus induction
medium. Explants were rated after 28 days of shoot bud induction treatments
and after 28 days on elongation medium.
Effect of duration of shoot bud induction treatment. Twenty-eight-day
hypocotyl-derived calli from IRFL 6123 were transferred to shoot bud induction
medium containing 2.0 mg L'1 adenine and either 1.0 or 0.1 mg L*1 2,4-D for 0,


22
Table 2-1. Response of genotype IRFL 6123 to 2,4-D and kinetin levels in the
shoot bud Induction medium.
2,4-D Kinetin
cone. cone.
mg L1
Mean number
of buds per
explant "I"
Mean percent
responding
explants, %t
0.1
0
0.5 b
30
a
0.3
0
2.2 ab
70
a
1.0
0
3.3 a
80
a
3.0
0
1.6 ab
50
a
0.1
0.15
0 c
0
b
1.0
0.15
0 c
0
b
t Mean separation by Tukey's honestly significant difference (HSD a=0.05).


50
to regeneratethat can only be detected when the level of proclivity rises
above a given threshold.
Distributions of this type cannot be rendered normal by any mathematical
transformation, since we lack any information about differences among those
individuals that fall below the observable threshold, and no transformation can
restore this missing information. For example, no matter what transformation is
used, the two nonregenerating parental populations will always have equal
means, and variances of zero. Fortunately, useful analysis can still be
performed on this type of data, provided it can be assumed that the populations
under consideration are normally distributed.
In Falconer's treatment, threshold traits can be manifested at either two or
three discrete levels, or classes. This can be done with the present data by
collapsing the rating scale into two (nonregeneration versus regeneration), or
three (nonregeneration versus partial regeneration versus full regeneration)
classes. While little useful analysis can be done when only two classes are
present, meaningful analysis is possible in situations with three classes,
contingent on the assumption that the underlying variable is normally
distributed. For any population, if the percentage of individuals falling into each
of three threshold classes is known, then the mean and variance of the
population, expressed in threshold units, can be deduced. The principle is
illustrated in Figure 3-3, adapted from Falconer (1981), but based on actual F3
data. The classes are separated by the two vertical lines on each graph,
nonregenerators lying to the left of both lines, partial regenerators lying
between the lines, and full regenerators lying to the right. While both
populations contain 68% full regenerators, population (a) contains more partial
regenerators and fewer nonregenerators than population (b). As a result,


29
Figure 2-4. Spatulate shoots produced by genotype IRFL 6123 in the presence
of gibberellic acid (GA3).


51
population (a) has a lower standard deviation and lower mean than
population (b).
An attractive advantage of analyzing data based on three ordinal classes
with two thresholds is that scaling problems are avoided. This is because only
one real unit exists on the x axisthe unit separating the two thresholds.
Values either to the left or to the right of both thresholds can never be
experimentally measured, and are simply extrapolations of the central threshold
unit. A serious disadvantage of the three class approach is that populations that
do not contain at least one observation in each of the three classes cannot be
analyzed. Many of the F3 families in the present study contain no fully
regenerating individuals, and therefore cannot be analyzed using this
approach.
While the methods described by Falconer for analysis of threshold traits
cannot be directly applied to data with more than three classes of observations,
it is possible to analyze data with an unlimited number of classes by applying
techniques developed for censored data sets. Censored data are similar to
threshold data in that both data types are subject to a threshold below which no
information is known (Schneider, 1986). The difference is that in censored
data there is only one threshold, and data above this threshold are of a
continuous, or at least semicontinuous nature. In both cases, data lying above
the detection threshold are used to make inferences about data below the
threshold; therefore, assumptions of normality are necessary in both cases. By
treating the seven-category regeneration rating scale as a quasi-continuous
variable, the data collected in the present study are amenable to analysis as
censored data.
Several statistical techniques have been developed for "uncensoring"
censored data sets, and, as in many fields of statistics, the relative strengths and


BIOGRAPHICAL SKETCH
Peter Krottje was born in Haskell, New Jersey, in 1951. He attended
Fordham University, receiving a B.S. in biology in 1973. He received an M.S. in
soil science from the University of Florida in 1980. He has been very fortunate.
107


Table 3-2 continued.
Population
Scalet
Mean
Variance
Skewness
W
Prob n
501 (Fg)
li
0.380
0.1138
1.729
0.830
<0.001
181
lo
2.412
0.1643
-0.242
0.9612
0.001
181
sq
17.81
63.47
0.691
0.940
<0.001
181
507 (Fg)
li
0.254
0.0526
1.509
0.817
<0.001
291
lo
2.231
0.1647
-0.161
0.963
<0.001
291
sq
14.49
43.72
0.737
0.924
<0.001
291
t li = linear (grams); lo = logi0(milligrams); sq = square root (milligrams)


34
covariance (Wr) analysis indicated that regeneration was largely controlled by
partially recessive genes.
The limited number of studies that have addressed phenotypic or genetic
correlations between in vitro regeneration and various in vitro and in planta
characters have yielded both expected and unexpected results. Oelck and
Scheider (1983) found that in Melilotus officinalis, Trifolium pratense, and T.
resupinatum, the ability to form shoots from callus was correlated with the
production of side shoots from in vitro shoot tip cultures. Kern et al. (1986)
observed a similar correlation between capacity for somatic embryogenesis and
in vitro axillary shoot development in soybean. These correlations may reflect a
generalized tendency among responding genotypes to form shoots, or may
simply indicate a tolerance to the specific culture conditions employed in the
study. It has been observed in alfalfa (Bingham et al., 1975; Brown and
Atanassov, 1985) and alyceclover (Wofford et al., 1992a) that regenerating
genotypes often have a creeping growth habit and readily produce adventitious
shoots. Thus it appears that in at least some cases the ability to regenerate in
vitro reflects a general proclivity for shoot production both in vitro and in planta.
Oelck and Scheider (1983) suggested that a general tendency to produce
adventitious shoots may be a useful indicator in preliminary screenings of
germplasm for regenerating lines.
Perhaps surprisingly, there appears to be little correlation between callus
growth rate and regeneration frequency ( Baroncelli et al., 1974; Kumar et al.,
1985; Lazar et al., 1984). However, shoot regeneration from callus can be
restricted in genotypes with very poor callus growth (Bingham et al., 1975). It is
well established that callus appearance frequently bears a strong relationship
to regeneration potential (Bingham et al., 1975; Ketchum et al., 1987; Delieu


87
Lupotto, E. 1983. Propagation of an embryogenic culture of Medicago
sativa L. Z. Pflanzenphysiol. 111:95-104.
Maheshwaran, G., and E.G. Williams. 1984. Direct somatic embryoid
formation on immature embryos of Trifolium repens, Trifolium
pratense, and Medicago sativa and rapid clonal propagation of
Trifolium repens. Ann. Bot. 54:201-211.
Malmberg, R.L. 1979. Regeneration of whole plants from callus
cultures of diverse genetic lines of Pisum sativum L. Planta
146:243-244.
Mather, K., and J.L. Jinks. 1971. Biometrical Genetics. Cornell
University Press, Ithaca, New York.
McKently, A.H., G.A. Moore, and F.P. Gardner. 1990. In vitro plant
regeneration of peanut from seed explants. Crop Sci. 30:192-196.
McKently, A.H., G.A. Moore, and F.P. Gardner. 1992. Regeneration of
peanut and perennial peanut from cultured leaf tissue. Crop Sci.
31:833-837.
Mehta, U., and H.Y. Mohan Ram. 1980. Regeneration of plantlets from
the cotyledons of Cajanus cajan. Indian J. Exp. Bot. 18:800-802.
Meijer, E.G.M. 1982. Shoot formation in tissue cultures of three
cultivars of the tropical forage legume Stylosanthes guyanensis. Z.
Pflanzenzuchtg. 89:169-172.
Murashige, T., and F. Skoog. 1962. A revised medium for rapid growth
and bioassays with tobacco tissue cultures. Physiol. Plant. 15:473-
497.
Newman, M.C., and P.M. Dixon. 1989. Estimating mean and variance for
environmental samples with below detection limit observations.
Water Res. Bull. 25:905-916.
Oelck, M.M., and O. Scheider. 1983. Genotypic differences in some
legume species affecting the redifferentiation ability from callus to
plants. Z. Pflanzenzuchtg. 91:312-321.


59
Table 3-1.
Number of F2 plants, F3
data were collected.
plants, and F3
families from which in vitro
Cross
F2 plants
F3 plants
F3 families
501
65
181
11
507
91
291
18
510
52
t
t
t Cross 510 was discontinued after F2.


26
10-
8-
co
o 6 -\
o
-C
CO
*4
o
k
0
E
3
y = -1,697x 2 + 5.405x + 0.333
r2 = 0.302
X
<
2-
V
/
/
/
'
2,4-D Concentration, mg/L
Figure 2-1. Response of genotype IRFL 6123 to 2,4-D concentration in shoot
induction medium.


8
pea (Malmberg, 1979; Phillips and Collins, 1980; Lupotto, 1983; Kumar et al.,
1984).
Among mature tissue types, regeneration has been obtained from leaf
explants in alfalfa and peanut (Oelck and Scheider, 1983; McKently et al.,
1992). Mature petioles have served as explant sources in various clover
species (Choo, 1988), and mature stems in birdsfoot trefoil and Stylosanthes
(Swanson and Tomes, 1980; Meijer, 1981).
Several different basal medium formulations have been used for tissue
culture of legume species. The majority of work has utilized Murashige and
Skoog's (1962) MS medium (Malmberg, 1979; Kao and Michayluk, 1981; Hazra
et al., 1989; McKently et al., 1990). Gamborg et al. (1968) developed the B5
medium for tissue culture of soybean, and this medium has proven useful for
several other legume species, including red clover (Quesenberry et al., 1992)
and birdsfoot trefoil (Swanson and Tomes, 1980). The B5 medium differs from
MS primarily by its greatly reduced ammonium nitrogen level and a lower
calcium concentration. Schenk and Hildebrandt's (1972) SH medium is similar
to B5, but contains much higher levels of inositol. This medium has been used
for Stylosanthes (Scowcroft and Adamson, 1976) and crimson clover (Horvath
et al., 1979). Phillips and Collins' (1979) L2 medium was developed for red
clover culture and has subsequently found application in the culture of peanut
and soybean (Sellars et al., 1990), and alyceclover (Wofford et al., 1992a). This
medium is somewhat higher in calcium than MS, B5, or SH, and also differs
from these media in that it lacks nicotinic acid. Blaydes' (1966) medium
contains less nitrate than the above media, and has been employed for
soybean, alfalfa (Bingham et al., 1975), and pigeonpea (Kumar et al., 1984).
It is generally acknowledged that plant growth regulator levels are of
critical importance in plant tissue culture. For angiosperms in general,


CHAPTER 3
INHERITANCE OF IN VITRO REGENERATION AND ASSOCIATED
CHARACTERS
The high genotype-specificity of in vitro regeneration in desmodium has
been clearly demonstrated in the previous chapter. An understanding of the
genetic control of regeneration is desirable for the efficient application of
biotechnological methods to this crop. The mode of inheritance of regeneration
determines whether it is feasible to transfer the trait to agronomically desirable
nonregenerating lines and, if so, the appropriate method of accomplishing this.
The genetic basis of regeneration and associated in vitro traits has been
studied in few higher plant species. In most studies regeneration was
accomplished through organogenesis, and organogenesis is assumed in the
current discussion unless otherwise stated. Regardless of whether
organogenic or embryogenic response was examined, regeneration in most
species was treated as a quantitatively inherited trait, although evidence
supporting this assumption was often not presented. Several studies have
been directed at partitioning genetic variance components and determining
heritabilities for the regeneration trait in different crops. Buiatti et al. (1974)
performed a diallel analysis, without reciprocals, of callus growth and shoot
regeneration from flower petal explants in cauliflower. Additive gene effects
were high for both traits, and narrow sense heritability estimates were 0.81 for
callus growth and 0.09 for percent explants forming shoots. The low heritability
for the latter trait resulted from extremely high epistatic and error effects. Based
on North Carolina Design II matings among twenty four red clover genotypes,
Keyes et al. (1980) determined that embryogenic regeneration frequency from
30


This dissertation was submitted to the Graduate Faculty of the College
of Agriculture and to the Graduate School and was accepted as partial
fulfillment of the requirements for the degree of Doctor of Philosophy.
August 1995
Dean, College of Agriculture
Dean, Graduate School


TABLE OF CONTENTS
page
ACKNOWLEDGMENTS ii
LIST OF TABLES iv
LIST OF FIGURES v
ABSTRACT vi
CHAPTERS
1 INTRODUCTION 1
2 OPTIMIZATION OF IN VITRO REGENERATION PROTOCOL 6
Materials and Methods 11
Results and Discussion 15
3 INHERITANCE OF IN VITRO REGENERATION AND
ASSOCIATED CHARACTERS 30
Materials and Methods 35
Results and Discussion 37
4 SUMMARY AND CONCLUSIONS 78
REFERENCES 83
APPENDICES
A DERIVATION OF BETWEEN- AND WITHIN-FAMILY
VARIANCES FOR CALLUS GROWTH 91
B DERIVATION OF VARIANCE PARTITIONING MATRICES
FOR CALLUS GROWTH 92
C RESULTS OF LATER ITERATIONS OF CALLUS GROWTH
VARIANCE PARTITIONING 94
D RAW CALLUS GROWTH AND REGENERATION DATA 97
BIOGRAPHICAL SKETCH 107


Log Callus Weight
77
Regeneration Score
Figure 3-8. Regression of callus weight on regeneration score for combined F3
populations.


19
14, and 28 day treatments with respect to percent responding explants. The
zero day treatment yielded significantly fewer responding explants than all
except the 7 day treatment. The zero day treatment is essentially a reversal
transfer from high auxin, low cytokinin to low auxin, high cytokinin medium.
Reversal transfer is a widely utilized method for inducing organogenesis. The
rather poor response to this treatment relative to the 2,4-D treatments suggests
that regeneration may be occurring via somatic embryogenesis rather than
organogenesis. Histological evidence on this question will be presented later
in this chapter.
The fact that a percentage of IRFL 6123 calli produce shoot buds under all
treatments is evidence of this genotype's strong predisposition to regenerate.
The failure of many of the explants to produce shoot buds under any of the
treatments and the lack of significant differences among many of the treatments
underscores the high level of variability often encountered in tissue culture
work, and may also be an indication that we have not yet foundand may
never findthe optimal regeneration protocol for this recalcitrant species.
Effect of BA and GAg levels on shoot elongation. Both shoot meristem
formation and shoot elongation are difficult to induce in desmodium. In
previous work (Wofford et al., 1992b) as well as in the experiments described
above, a large percentage of shoot buds failed to develop into plants, either due
to poor elongation or failure to root. Three levels of BA were evaluated in the
first of two experiments intended to improve elongation response. Gibberellic
acid has been reported to stimulate development and elongation of shoot
meristems in cultured soybean and pigeonpea (Ghazi et al., 1986; Kumar et al.,
1983). To determine if this compound might be useful for desmodium, each BA
level was tested both with and without the addition of 0.2 mg L'1 GAg.


Results of variance partitioning second iteration:
Component
Cross
Scale
D
H
Ew
Whole Model
Estimate
LL
A
-Q
O
Q.
Estimate
Prob>F
Estimate
ProtF
F
Prob>F
501
Linear
0.1362
0.3918
-0.0672
0.8024
0.0334
0.1578
27.73
0.1385
507
Linear
0.0289
0.7176
0.0411
0.8616
0.0138
0.2744
9.16
0.2372
501
Logarithmic
0.1327
0.3790
0.1501
0.6365
0.4253
0.1218
49.48
0.1040
507
Logarithmic
0.0637
0.7024
0.3337
0.5073
0.0445
0.2465
14.89
0.1877
501
Square Root
70.02
0.0642
-18.68
0.4518
17.37
0.0206
1641
0.0181
507
Square Root
23.89
0.1992
46.39
0.2828
18.21
0.0503
335.7
0.0401
co
cn


10
under two protocolsan MS-based procedure intended to induce
organogenesis and an L2-based procedure originally used by Collins and
Phillips (1982) to induce embryogenesis in red clover. The MS procedure
consisted of a callus induction step with 2.0 mg L'^ IAA (indole-3-acetic acid)
and 1.0 mg L'1 kinetin, shoot induction with 0.1 mg L"1 IAA and 4.0 mg L'1 BA,
and rooting on medium lacking plant growth regulators. The L2 protocol
consisted of callus induction with 0.06 mg L'1 picloram (4-amino-3,5,6-
trichloropicolinic acid) and 0.1 mg L'^ BA, embryo induction with 0.01 mg
2,4-D and 2.0 mg L"^ adenine, and embryo germination with 0.002 mg L'1
picloram and 0.2 mg L1 BA. The MS protocol resulted in production of shoot
meristems in five of six genotypes, but in all cases shoots failed to elongate.
The L2 protocol yielded shoot meristems in two genotypes and whole plants
were obtained in one genotype of D. heterocarpon ssp. angustifolium. The
response on L2 superficially resembled organogenesis, but histological
examination was not performed.
This chapter will evaluate a number of variations on the L2-based
procedure utilized by Wofford et al. (1992b). Primary objectives were to
determine if regeneration could be obtained in a wider range of desmodium
genotypes and to optimize the response of previously identified regenerating
genotypes. A bewildering range of culture variables can potentially influence
regeneration response. The work presented here attempts to examine some of
these variables in an orderly, stepwise manner in which the treatments in each
experiment are based on the results of the previous experiment. The work
begins with an examination of explant sources, proceeds to an investigation of
various shoot bud induction treatments, and then examines several shoot
elongation treatments. The chapter concludes with histological data presented
with the objective of clarifying the mode of regeneration in this system.


Table 3-2. Univariate statistics for callus weight for parental, F2, and F3 populations in linear, logarithmic and
square root scales.
Population
Scalet
Mean
Variance
Skewness
W
Prob n
144 (Parental)
li
0.7442
0.0648
0.247
0.953
0.288
26
lo
2.844
0.0268
-0.702
0.934
0.102
26
sq
26.87
22.89
-0.196
0.985
0.372
26
6123 (Parental)
li
0.2897
0.0172
0.38
0.947
0.119
35
lo
2.413
0.0477
-0.528
0.95
0.15
35
sq
16.57
15.44
-0.021
0.964
0.369
35
13083 (Parental)
li
0.7215
0.0541
-0.114
0.958
0.514
20
lo
2.833
0.0261
-0.93
0.906
0.055
20
sq
26.49
20.74
-0.503
0.943
0.293
20
501 (F2)
li
0.301
0.071
2.111
0.794
<0.001
65
lo
2.319
0.162
-0.432
0.954
0.038
65
sq
15.93
47.88
0.766
0.942
0.008
65
507 (F2)
li
0.366
0.069
1.506
0.877
<0.001
92
lo
2.445
0.123
-0.65
0.9953
0.009
92
sq
17.97
43.58
0.431
0.969
0.15
92
t li = linear (grams); lo = log-j o(nnilligrams); sq = square root (milligrams)
05
o


89
Sarria, R., A.Calderon, A.M. Thro, E. Torres, J.Mayer, and W.M. Rocca.
1994. Agrobacterium mediated transformation of Stylosanthes
guianensis and production of transgenic plants. Plant Sci. 96:119-
127.
Saunders, J.W., G.L. Hosfield, and A. Levi. 1987. Morphogenetic
effects of 2,4-dichlorophenoxyacetic acid on pinto beans
(Phaseolus vulgaris L.) leaf explants in vitro. Plant Cell Rep. 6:46-
49.
Schenk, R.U., and A.C. Hildebrandt. 1972. Medium and techniques for
induction and growth of monocotyledononous and dicotyledonous
plant cell cultures. Can. J. Bot. 50:199-204.
Schubert, B.G. 1963. Desmodium: Preliminary studies IV. Journal of
the Arnold Arboretum 44:284-297.
Schultze-Kraft, R., and G. Benavides. 1988. Germplasm collection and
preliminary evaluation of Desmodium ovalifolium Wall. Gen.
Resources Comm. 12:1-20.
Schultze-Kraft, R., and S. Pattanavibul. 1985. Collecting native
forage legumes in eastern Thailand. IBPGR Newsletter. 9:4-5.
Scowcroft, W.R., and J.A. Adamson. 1976. Organogenesis from callus
cultures of the legume, Stylosanthes hamata. Plant Sci. Lett. 7:39-
42.
Sellars, R.M., G.M. Southward, and G.C. Phillips. 1990. Adventitious
somatic embryogenesis from cultured immature zygotic embryos of
peanut and soybean. Crop Sci. 30:408-414.
Simmonds, N.W. 1979. Principles of crop improvement. Longman Inc.,
New York.
Swanson, E.B., and D.T. Tomes. 1980. Plant regeneration from cell
culture of Lotus corniculatus and the selection and characterization
of 2,4-D tolerant cell lines. Can. J. Bot. 58:1205-1209.
Takahasi, M. 1952. Tropical forage legume and browse plants
research in Hawaii. Proc. 6th Int. Grass. Cong. 1411-1416.


46
present case, five observational variances or covariances are available: pooled
parental variance (E-j), F2 variance (V-||r2), among-family F3 variance (V1F3),
within-family F3 variance (V2p3), and ^2*^3 covariance (Wip23)- Among- and
within-family F3 variances are obtained by analysis of variance using F2 parent
as the grouping variable, followed by partitioning of
variances based on expected mean squares (Appendix A). The F2-F3
covariance is the covariance between F2 parents and F3 family means.
The causal variance components described by Mather and Jinks are
somewhat complex. In addition to additive and dominant genetic components
(designated D and H, respectively), two types of environmental variance and a
so-called sampling variance are described. The first environmental variance
(Ew) is simply the error variance among individuals. The second (Eg) is the
error among plots or rearing environments and is applicable only to V-jp3 in the
present study. Since explants from the different F3 families were randomly
distributed among petri dishes, which were randomly distributed within a single
incubator, there is no intrinsic plot error. Instead, Eg is equivalent to Eyy
divided by the harmonic mean of the number of individuals per F3 family
(Mather and Jinks; 1971). Since Eg is a function of Eyy, it can be combined into
Eyy, thereby eliminating a column from the model and gaining a degree of
freedom. The sampling variance applies only to among-group variances. It is
similar in concept to Eg, but is a function of within-group genetic variances
rather than error variances. Therefore, in the case of V-| p3 the sampling
variance is equivalent to V2p3 divided by the harmonic mean of the number of
individuals per F3 family. Sampling variance can be algebraically included into
the model row for V1F3 and does not have an associated model column.
Derivation of model matrices is described in detail in Appendix B.


36
done in June and July. Donor pollen was obtained by tripping flowers using a
toothpick with a small piece of fine sandpaper glued to one end in such a way
that the anthers and stigma would strike the sandpaper and pollen would
adhere. Pollen was then transferred to recipient flowers by tripping in a similar
manner. Recipient flowers were not emasculated because the flowers were
extremely sensitive to handling, and emasculation tended to result in flower
abscission. Pollinations were carried out at a time of day when the flowers had
fully opened, but self-tripping had not yet occurred. The specific time varied
according to night temperature and daytime cloud cover, but was generally
between 9:00 A.M. and noon. At least one hundred flowers for each possible
combination of regenerator and nonregenerator parent, including reciprocals,
were pollinated in this manner.
Seeds were harvested in August, 1989 and germinated and planted in the
greenhouse the following winter. Hybrids were clearly identifiable based on the
morphological characteristics described above. Flowering of hybrids was
induced in the summer of 1990. Plants were allowed to self-pollinate, but
flowers were hand tripped to increase seed set.
Evaluation of callus growth and regeneration in the parental. Fo and Fg
generations. The F2 seed harvested from the original hybrids was scarified and
germinated on SGL medium (Collins and Phillips, 1982) as described in
Chapter 2. Hypocotyls were excised and placed onto callus induction medium
and epicotyls were returned to SGL medium for rooting. Rooted epicotyls were
transferred to potting soil after ten days and grown to maturity in the greenhouse
where they were allowed to self-pollinate to produce the Fg generation.
Hypocotyls from both F2 and Fg populations were cultured according to
the optimal protocol established in Chapter 2. Initial callus production was on
L2 medium supplemented with 0.06 mg L"1 picloram and 0.2 mg L'1 BA for


CHAPTER 2
OPTIMIZATION OF IN VITRO REGENERATION PROTOCOL
Although numerous members of the family Leguminoseae have been
regenerated in vitro, the family is regarded to be difficult with respect to in vitro
regeneration (Flick et al., 1983). Regeneration frequency is low for many
species, and specific culture requirements can vary widely both among and
within species (Phillips and Collins, 1983).
Regeneration in angiosperms can be accomplished via either of two
conceptually distinct pathwaysorganogenesis or somatic embryogenesis.
Both types of regeneration can be induced either directly from the initial explant
source, or from callus or suspension cells generated in culture. Organogenesis
usually involves production of a shoot meristem followed by shoot elongation
and rooting. Shoot and root regeneration are discrete processes occurring in
response to specific culture conditions, particularly the types and concentrations
of plant growth regulators in the medium. In somatic embryogenesis, shoot and
root meristems are produced simultaneously in a process similar to zygotic
embryo development. Maturation and germination of somatic embryos can be
induced by manipulating plant growth regulator levels, or may occur
spontaneously under constant culture conditions (Hazra et al., 1989).
In practice, the distinction between organogenesis and embryogenesis is
not always so clear (Ammirato, 1983). Ideally, somatic embryos should closely
resemble sexual embryos and possess clearly identifiable shoot, root, and
cotyledonary primordia that subsequently develop into their respective organs.
Deviations from this ideal situation include abnormal or absent cotyledons and
6


82
Vasil (1987) has suggested that the genetics of regeneration is irrelevant,
and that with sufficient insight into the in vitro physiology of a species
regeneration can be induced in even the most difficult genotype. While this may
be true in theory, this work suggests that such an assumption can be impractical
or even counterproductive. Approximately equal effort was directed at culture
protocol and genetic approaches and the latter was found to be by far the more
productive path. It is noteworthy that the original regenerating parent
regenerated under a variety of culture conditions, suggesting a general genetic
proclivity to regenerate. A more experienced investigator with greater
resources may or may not have obtained better results from culture protocol
optimization, but success may well have been very costly in terms of time, effort,
and financial resources. The genetic approach has been relatively
straightforward and strikingly successful. The magnitude of this success was
particularly apparent from visual examination of regenerating calli. Several F3
calli, including that shown in Figure 3-7, grossly outperformed the parental
genotype, continuing to produce vigorous shoots through repeated subcultures,
and ultimately yielding dozens of shoots.
This study has demonstrated that regeneration in desmodium can be
greatly improved by conventional crossing techniques. Limited light was shed
on the genetic basis of regeneration in this crop, and many questions remain
unanswered. It is hoped that the F3 material produced in this study may prove
useful to any investigator who might desire to perform regeneration-dependent
work with this crop.


99
Family F3
Individual
Regeneration Score
Progeny Parent
Callus Weight
Progeny Parent
507-7
16
1
507-7
17
1
507-7
18
1
507-7
19
1
507-7
20
3
507-7
21
2
507-7
22
1
507-7
23
1
507-7
24
1
507-7
25
2
507-7
26
1
507-8
1
1
507-8
2
2
507-8
3
1
507-8
4
1
507-8
6
1
507-8
7
1
507-8
8
5
507-8
9
4
507-8
10
2
507-8
11
1
507-8
12
2
507-8
13
4
507-8
14
1
507-9
1
3
507-9
2
1
507-9
3
1
507-9
4
1
507-9
5
1
507-9
6
1
507-9
8
3
507-9
9
2
507-9
10
1
507-9
13
1
507-9
14
1
507-9
15
2
507-9
16
1
507-9
19
2
507-10
2
1
507-10
3
2
507-10
4
3
507-10
5
1
507-10
8
1
507-10
9
1
507-10
10
1
1
330
285
1
230
285
1
110
285
1
240
285
1
800
285
1
110
285
1
530
285
1
280
285
1
300
285
1
150
285
1
90
285
1
960
400
1
590
400
1
130
400
1
440
400
1
440
400
1
400
400
1
400
400
1
360
400
1
200
400
1
220
400
1
490
400
1
120
400
1
470
400
1
420
380
1
140
380
1
520
380
1
390
380
1
910
380
1
850
380
1
130
380
1
900
380
1
250
380
1
1110
380
1
200
380
1
400
380
1
170
380
1
180
380
1
140
106
1
80
106
1
380
106
1
360
106
1
20
106
1
40
106
1
20
106


78
CHAPTER 4
SUMMARY AND CONCLUSIONS
The point of departure for this dissertation was a study conducted by
Wofford et al. (1992b). Wofford's study identified a three stage (callus induction,
shoot induction, and shoot elongation) protocol utilizing the L2-based media of
Phillips and Collins (1979) as suitable for tissue culture of desmodium
hypocotyls, and obtained limited regeneration from a single genotypeIRFL
6123. The objective of the current work has been to examine the potential for
enhancing in vitro response in desmodium through improvements in culture
protocol and through breeding.
Chapter 2 focused on optimization of culture protocols. The work was not
intended to be an exhaustive investigation of potential methods for enhancing
in vitro response in desmodium. Given the wide range of variables that can
determine the effectiveness of tissue culture protocols and the limited prior work
with this crop, such an investigation would be beyond the scope of any single
study. Instead, the intent was to cast a broad netto determine if the poor,
highly genotype-specific regeneration observed in desmodium could be
significantly improved by applying a wide range of techniques similar to those
that had proven successful in other legumes. Accordingly, within each general
culture strategy (e.g., alternative explant sources, manipulation of auxin-
cytokinin ratio, substitution of different auxin sources) only a small number of
specific culture protocols were examined. For the same reason, number of
replications was often fairly low. Thus, failure, for example, to obtain
regeneration from leaf disks or petioles does not indicate that regeneration from
78


105
Family
f3
Individual
Regeneration Score
Progeny Parent
Callus Weight
Progeny Pare
501-7
8
2
2
90
313
501-7
9
1
2
140
313
501-7
10
1
2
410
313
501-7
11
2
2
900
313
501-8
1
1
1
130
206
501-8
2
1
1
330
206
501-8
3
1
1
250
206
501-8
4
2
1
340
206
501-8
5
1
1
90
206
501-8
6
1
1
390
206
501-8
7
2
1
220
206
501-8
8
1
1
670
206
501-8
9
1
1
90
206
501-8
10
1
1
190
206
501-8
11
1
1
210
206
501-8
12
1
1
360
206
501-8
13
2
1
610
206
501-8
14
1
1
430
206
501-9
1
2
4
550
365
501-9
2
1
4
390
365
501-9
3
3
4
160
365
501-9
4
2
4
130
365
501-9
5
2
4
660
365
501-9
6
1
4
90
365
501-9
7
2
4
430
365
501-9
8
2
4
520
365
501-9
9
3
4
540
365
501-9
10
1
4
480
365
501-9
11
2
4
260
365
501-9
12
2
4
630
365
501-9
13
2
4
730
365
501-10
1
1
1
700
345
501-10
2
1
1
190
345
501-10
3
2
1
290
345
501-10
4
1
1
830
345
501-10
5
1
1
420
345
501-10
6
1
1
480
345
501-10
7
1
1
720
345
501-10
8
1
1
590
345
501-10
9
1
1
600
345
501-10
10
2
1
410
345
501-10
11
2
1
660
345
501-10
12
1
1
840
345
501-10
13
1
1
450
345
501-10
14
1
1
710
345
501-11
1
1
1
240
133


APPENDIX B
DERIVATION OF VARIANCE PARTITIONING MATRICES FOR CALLUS
GROWTH
Coefficients for variance partitioning of the callus growth were based on those
given by Mather and Jinks (1971, 1977). The following general coefficients
were given for the populations treated in the present study:
Observational
Component
D
H
EW
eb
Sampling
Variance
V1F2
1/2
1/4
1
0
0
V1F3
1/2
1/16
0
1
Vn V2F3
V2F3
1/4
1/8
1
0
0
W1F23
1/2
1/8
0
0
0
Ei
0
0
1
0
0
Since there is only one entry for sampling variance and the sampling variance
for V-| p3 is a function of V2F3, this column can be combined into the other
columns. In the case of cross 501, n (the harmonic mean of the number of
individuals per F3 family) is equal to 15.21. Therefore, after combining the
sampling variance column, the coefficient for D for V-)p3 becomes
V2 + ((1/15.21) x 1/4), or 0.516435. The other entries in the V-| p3 row are
similarly altered. Since Eg is a function of Ew, this column can also be
92


32
regeneration, while the presence of both dominant alleles yielded a high
regeneration frequency. Crosses between two different sets of nonregenerating
clones yielded no regenerating progeny, suggesting that inheritance was not
quantitative.
Wan et al. (1988) examined the genetics of regeneration from petiole
explants in tetraploid alfalfa, employing a different culture protocol from that of
Reisch and Bingham. Based on data from several F-| and S-| populations, it
was concluded that regeneration was controlled by a two gene system distinct
from that described by Reisch and Bingham. Duplicate recessive epistasis was
postulated; that is, the presence of dominant alleles at both loci was necessary
for any regeneration to occur. Variation in the frequency of regeneration among
regenerating clones was attributed to either dosage effects or the presence of
modifier genes.
Dulieu (1991) obtained rather tentative evidence of single gene control of
regeneration from hypocotyl explants in petunia. Two backcross populations
yielded approximately 1:3 ratios of "high regenerating" to combined "low
regenerating" and "nonregenerating" plants, but the distinction between high
and low level regeneration was somewhat arbitrary, and quantitative
inheritance could not be ruled out.
All of the studies discussed above have dealt with regeneration from
sporophytic (2n) tissue. The process of androgenesis, or embryogenesis from
tissue derived from gametophytic (n) pollen mother cells, is markedly distinct
from regeneration from sporophytic tissue. However, the two processes are
sufficiently similar that work conducted on the genetics of androgenesis may
provide insight into the genetics of regeneration from 2n tissue. Several papers
have been published on the genetic control of androgenesis in cereals, and all


5
F3 generations. Chapter 4 presents a summary and conclusions, and
discusses the implications of this work within the broad context of genotype by
environment interaction under in vitro conditions.


35
1991), but due to the difficulty of quantifying callus appearance the relationship
has not been subjected to statistical analysis.
The primary objective of the work described in this chapter was to
determine the mode of genetic control of in vitro regeneration in desmodium. A
secondary objective was investigation of the mode of inheritance of callus
growth and examination of the relationship between callus growth and
regeneration.
Materials and Methods
Production of hybrids. Two regenerating and three nonregenerating lines
were selected for crossing. The regenerators, IRFL 6123 and IRFL 6128 are
classified as D. heterocarpon ssp. angustifolium, and are morphologically
distinguishable from the nonregenerating genotypes by their lanceolate,
coriacious leaves, upright growth habit, elongated racemes, and glabrous seed
pods. Lines 6123 and 6128 are very similar except for a distinct leaf mark on
6123 that is absent on 6128. All of the selected nonregeneratorsUF 20,
UF 144, and Cl AT 13083have ovate or obovate, noncoriacious leaves,
spreading growth habits, compact racemes, and pubescent pods. Genotype
UF 20 ('Florida' carpon) represents D. heterocarpon ssp. heterocarpon and is
distinguished by its thin, slightly pubescent leaves and prominent leaf marks.
Genotype UF 144 is classified as D. ovalifolium, and possesses thicker,
glabrous leaves. Genotype CIAT 13083 is intermediate between D.
heterocarpon and D. ovalifolium.
Parent plants were moved from the field to the greenhouse in April, 1989.
To induce flowering, a ten hour day length was simulated by covering the plants
with a tarpaulin from approximately 7:00 P.M. until 9:00 A.M. Pollinations were


58
growth calli cannot be attributed to linkage. A likely explanation is that a certain
degree of in vitro vigor, as manifested by at least a moderate callus growth rate,
promotes regeneration. This would constitute pleiotropy in a broad sense. An
alternative possibility is that there exist genes that act in a physiologically
pleiotropic manner; that is, there may be genes that either promote or inhibit
specific metabolic pathways that are necessary for both callus growth and
regeneration. To experimentally distinguish between these subtly different
hypotheses is far beyond the scope of the present study, or of any study
undertaken to date on the genetics of in vitro traits.


Kao, K.N., and M.R. Michayluk. 1981. Embryoid formation in alfalfa
cell suspensions from differentiated plants. In Vitro 17:645-648.
86
Kerns, H.R., U.B. Barwale, M.M. Meyer, Jr., and J.M. Widholm. 1986.
Correlation of cotyledonary node shoot proliferation and somatic
embryoid development in suspension cultures of soybean (Glycine
max L. Merr.). Plant Cell Rep. 5:140-143.
Ketchum, J.L.F., O.L. Gamborg, G.E. Hamming, and M.W. Nabors. 1987.
Progress report, tissue culture for crops project. Colorado State
University Press., Ft. Collins, Colorado.
Keyes, G.J., G.B. Collins, and N.L. Taylor. 1980. Genetic variation in
tissue cultures of red clover. Theor. Appl. Genet. 58:265-271.
Komatsuda, T., S. Enomoto, and K. Nakajima. 1989. Genetics of callus
proliferation and shoot differentiation in barley. J. Hered. 80:345-
350.
Koornneef, M., C.J. Hanhart, and L. Martinelli. 1987. A genetic
analysis of cell culture traits in tomato. Theor. Appl. Genet. 74:633-
641.
Kretchmer, A.E., Jr., J.B. Brolmann, G.H. Snyder, and S.W. Coleman.
1976. 'Florida' Carpon desmodium, a perennial tropical legume for
use in south Florida. Soil and Crop Sci. Soc. of Fla. Proc. 35:25-30.
Kumar, A.S., T.P. Reddy, and G.M. Reddy. 1983. Plantlet regeneration
from different callus cultures of pigeonpea (Cajanus cajan L). Plant
Sci. Lett. 32:271-278.
Kumar, A.S., T.P. Reddy, and G.M. Reddy. 1984. Multiple shoots from
cultured explants of pigeonpea and atylosia species. SABRAO J.
16:101-105.
Kumar, A.S., T.P. Reddy, and G.M. Reddy. 1985. Genetic analysis of
certain in vitro and in vivo parameters in pigeonpea (Cajanus cajan
L.). Theor. Appl. Genet. 70:151-156.
Lazar, M.D., P.S. Baenziger, and G.W. Schaeffer. 1984. Combining
abilities and heritability of callus formation and plantlet
regeneration in wheat (Triticum aestivum L.) anther cultures. Theor.
Appl. Genet. 68:131-134.


Abstract of Dissertation Presented to the Graduate School of the
University of Florida in Partial Fulfillment of the Requirements for
the Degree of Doctor of Philosophy
IN VITRO MORPHOGENESIS AND INHERITANCE OF IN VITRO TRAITS IN
DESMODIUM
By
Peter A. Krottje
August, 1995
Chairman: David S. Wofford
Major Department: Agronomy
Desmodium, Desmodium sp., is a forage legume widely cultivated in the
tropics and of growing importance in the southern United States. Previous work
aimed at application of biotechnological methods to this crop had obtained
limited in vitro regeneration from seedling hypocotyl explants in a single
genotype. The objective of the present work was to examine the potential for
improving regeneration response in desmodium through optimization of culture
protocols and through genetic improvement.
Efforts at identifying alternative explant sources focused on seedling
cotyledons, mature leaf disks, and mature petioles. Regeneration could not be
induced from any of these explant types. Several combinations of 2,4-D and
kinetin concentrations in the shoot bud induction medium were examined in an
unsuccessful effort to obtain regeneration from hypocotyl explants of recalcitrant
genotypes. Direct regeneration from hypocotyl explants without an intervening
callus growth step was also unsuccessful. Several treatments were
investigated with the aim of enhancing regeneration response in the previously
identified regenerating genotype. The best protocol identified consisted of a
VI


76
Figure 3-7. Profuse regeneration exhibited by members of two F3
families derived from cross 507


79
these explant types can not be achievedonly that none of the methods tested
yielded strongly encouraging results.
The first general culture strategy investigated was the use of alternative
explant sources. In addition to the previously proven hypocotyl explants, leaf
disks, petioles, and seedling cotyledons were tested. One previously
established regenerating genotype (IRFL 6123) and two nonregenerating
genotypes ( CIAT 13083 and UF 20) were tested. Regeneration was not
obtained from any of the alternative explant sources under the single culture
protocol tested. The only repeatable regeneration response was from
hypocotyls of IRFL 6123, as had been previously reported (Wofford et al.,
1992b).
Two more strategies for broadening the range of regenerating genotypes
were examined. The first was manipulation of 2,4-D concentration, with and
without added kinetin, in the shoot induction medium. This experiment failed to
induce regeneration in recalcitrant genotypes, but succeeded in identifying an
optimal 2,4-D concentration for regeneration in IRFL 6123. In addition, kinetin
was found to inhibit regeneration in this genotype. The last attempt at inducing
regeneration in recalcitrant genotypes involved direct regeneration from
hypocotyl explants. This approach failed to yield regeneration even from
IRFL 6123.
The remainder of Chapter 2 dealt with enhancing the regeneration
response in IRFL 6123. The first experiment demonstrated that reducing the
shoot induction period from the original 28 days to 14 days had no effect on
regeneration. Two additional experiments investigated the effects of growth
regulators in the shoot elongation step. An optimum ratio of 0.012 mg L'"*
picloram to 0.2 mg L BA was identified.


38
IRFL 6123 x UF 144 (cross 507), and three were from CIAT 13083 x IRFL 6128
(cross 510). Since genotypes 6123 and 6128 are nearly identical, both in gross
morphology and in vitro performance, crosses 501 and 510 are essentially
reciprocals.
Hybrids were selfed through the F3 generation. Table 3-1 summarizes
numbers of individuals and number of families from which callus growth and
regeneration data were collected. In the case of cross 510, data collection was
terminated at the F2 generation. Severe morphological abnormalities
appeared in the F-j of this cross, and persisted through F3. The three F-j
seedlings appeared normal through approximately 6 weeks of age, after which
new growth showed severely stunted leaves and internodes. Stunting was
evident from the early seedling stage in many F2 and F3 individuals, and
appeared at late seedling stages in all individuals. Stunting was accompanied
by extremely poor flower production and seedset. The F2 of cross 510 also
exhibited a depressed callus growth rate relative to the two other crosses and to
the parental lines with a high incidence of callus necrosis and death. Since
cross 510 is essentially the reciprocal of cross 501, the abnormalities may be
the result of interactions between the cytoplasm of CIAT 13083 and nuclear
genes of IRFL 6127. (Evidence of maternal effects in the other crosses will be
discussed later in this chapter.) Whatever the cause, the observed
abnormalities have the potential effect of masking expression of the in vitro
characters of interest, as well as exerting confounding selective pressures in the
F2 and F3 generations. Therefore, genetic analysis was not attempted on cross
510, and the remainder of this chapter deals exclusively with crosses 501 and
507.
Scaling of the callus growth trait. The logical first step in genetic analysis
of any trait is to identify an appropriate numerical scale; that is, to determine if


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E604ZTRQX_PA27DJ INGEST_TIME 2015-03-27T19:09:12Z PACKAGE AA00029821_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


15
Effect of picloram/BA ratio on shoot bud elongation. This experiment was
conducted in the same manner as the previous experiment, but with the
following elongation treatments:
(1) 0.012 mg L"1 picloram, 0.2 mg L"1 BA
(2) 0.012 mg L'1 picloram, 0.6 mg L'1 BA
(3) 0.04 mg L*1 picloram, 0.2 mg L'1 BA
(4) 0.04 mg L1 picloram, 0.6 mg L'1 BA
An additional elongation treatment of 0.002 mg L'1 picloram and
2.0 mg L'1 BA was included as a check.
Histological examination. After twenty-eight days of callus growth medium,
calli of IRFL 6123 were transferred to shoot induction medium containing
1.0 mg L"1 2,4-D and 2.0 mg L_1 adenine. Calli were removed for sectioning
after 1, 2, and 3 weeks on induction medium. Specimens were embedded in
Tissue-Tek O.T.C. Compound (Miles, Inc., Elkhart, IN) and sectioned on a CTF
Microtome-Cryostat (International Equipment Co., Needham, MA). Sections
eight microns in thickness were observed under the light microscope without
staining.
Results and Discussion
Evaluation of explant sources. Wofford et al. (1992b) showed that seedling
hypocotyls are satisfactory explant sources for in vitro regeneration in
desmodium. The use of hypocotyls presents certain practical problems,
however. Since the seedling must be dissected in order to place the hypocotyl


44
the majority of negatively acting alleles, and that dominance acts primarily in a
single direction. Judging from the large difference in callus growth rates among
the parents in the current work, the first assumption is probably at least partially
valid. The validity of the second assumption is difficult to assess. A second
problem with the joint scaling test is that two separate hypotheses are
simultaneously tested: that the scale is adequate, and that the genetics of the
trait in question can be adequately described by an additive-dominant genetic
model. When an adequate goodness of fit is not obtained, the joint scaling test
cannot by itself distinguish which of these hypotheses is incorrect. It is possible
to partially solve this problem by adapting the test to more complex genetic
models involving epistatic and perhaps other types of genetic effects.
Unfortunately, more complex models require data from more generations and
are dependant on an increasing number of rather dubious assumptions about
gene interactions (Mather and Jinks, 1971). As a result of these limitations,
spurious positive results are quite possible, particularly when the test involves
only a single degree of freedom. Considering the statistical evidence for the
inadequacy of the linear scale, it is likely that the good fit obtained when this
scale was applied to cross 501 is in fact spurious.
Examination of the data presented in Table 3-2 suggests a factor that may
confound the joint scaling test in the present case. In both crosses, and for all
scales, the F2 and F3 means are significantly lower than the mid-parent mean,
and are fairly close to one another. In the case of cross 501, the F3 mean is
higher than the F2 mean, while in cross 507 the F3 mean is lower. If a simple
additive-dominant model is assumed, the deviation of the F2 and F3 from the
mid-parent mean suggests a large, negative dominance effect, just as indicated
by the joint scaling test. However, the relationship between F2 and F3 means
suggests a small dominance effecta negative effect for cross 501, and a


41
will in fact be approximately normally distributed. Further evidence regarding
this hypothesis will be presented later in this chapter.
Skewness and W statistics for the F2 and F3 populations fail to clearly
indicate whether the logarithmic or the square root transformation yields a more
normal distribution. As in the case of the parental populations, the log
transformed populations are skewed to the right, while the square root
transformed populations are skewed to the left. In the case of F3 populations,
the magnitude of skewing is substantially less for the logarithmic than for the
square root scales. The only non-significant W statistic was obtained from the
square root transformed cross 507 F2 population.
If a variable is properly scaled, populations with high means should, in
general, have variances no greater than those of populations with lower means.
This can be tested by regressing F3 family means versus variances, as
presented in Figure 3-1. The linear scale exhibits a strong, positive relationship
between mean and variance. Logarithmic transformation results in a weak,
negative relationship, and the square root transformation produces a weak,
positive relationship. It is feasible that the expectation of independence of
means and variances can be confounded by genetic effects, since different F3
families may possess differing amounts of genetic variance. For example, in the
not unlikely situation that a large, positive dominance effect exists, families with
high means will be those whose F2 parents had high levels of heterozygosity
for genes controlling callus growth. These families will also exhibit elevated
variances due to the presence of segregating genes. This genetic effect can
only be large, however, if total genetic variance is much greater than
environmental variance. It will be shown later in this chapter that this is not the
case. Therefore, the strong relationship between means and variances for F3
families provides further evidence of the inadequacy of the linear scale.


Table 3-7. Estimates of causal variance components and associated probabilities obtained from variance partitioning
for callus growth.
Component
Cross
Scale
D
H
Ew
Whole Model
Estimate
LL
A
S
CL
Estimate
Prob>F
Estimate
Prob>F
F
Prob>F
501
Linear
0.1258
0.4033
-0.0578
0.8239
0.0324
0.1428
28.57
0.1365
507
Linear
0.0313
0.6534
0.0271
0.9055
0.0284
0.2982
11.50
0.2127
501
Logarithmic
0.1273
0.4244
0.1535
0.6266
0.3881
0.1241
48.08
0.1055
507
Logarithmic
0.0594
0.7487
0.3036
0.5601
0.0438
0.2364
11.64
0.2114
501
Square Root
70.39
0.0592
-17.50
0.4434
17.38
0.0189
1944
0.0189
507
Square Root
23.61
0.2057
46.27
0.2916
18.24
0.0512
332.2
0.0403
O
-vi


39
mathematical scale transformation is necessary. This task can be approached
in at least two general ways. A purely statistical approach, described in any
number of statistics texts (Little and Hills, 1978; Zar, 1984), seeks a scale that
yields a data structure satisfying the general assumptions of parametric
statisticsin this case, normally distributed error terms with error variances
independent of means. A more genetically oriented (and somewhat more
complex) approach is not directly concerned with statistical assumptions, but
stresses instead whether a scale will facilitate partitioning of genetic variation
into underlying genetic causes, or components; e.g., additive, dominant, and
epistatic gene action (Mather and Jinks, 1971). Implicit in this second
approach is the statistical assumption that factor levels act in an additive
fashion. Both approaches will be presented below.
Three scales were examined in the present study. The simplest is the
untransformed linear scale, which in this case is callus weight in grams. This
scale possesses the virtue of allowing various statistics and parameters to be
reported in grams, permitting easy interpretation. A logarithmic scale was also
investigated. While this scale results in somewhat non-intuitive statistical
outputs, the scale is theoretically more applicable to analysis of growth
variables than is a linear scale. The logarithmic scale asks the question, how
many times has the cell mass doubled, while the linear scale asks how many
mass units have been added to the starting mass. The logarithmic
transformation is frequently used to stabilize variances in cases where standard
deviation is proportional to mean in the untransformed scale. The third scale
tested was the square root transformation. This scale is intermediate between
the linear and logarithmic scales in terms of skewing of distributions, and is
useful when variance is proportional to mean.


APPENDIX D
RAW CALLUS GROWTH AND REGENERATION DATA
Family
f3
Individual
Regeneration Score
Progeny Parent
Callus Weight
Progeny Pare
507-1
1
1
1
80
129
507-1
2
1
1
140
129
507-1
3
1
1
220
129
507-1
4
1
1
100
129
507-1
5
1
1
40
129
507-1
6
1
1
140
129
507-1
7
1
1
40
129
507-1
8
2
1
290
129
507-1
9
2
1
90
129
507-1
10
4
1
120
129
507-1
11
1
1
130
129
507-1
12
1
1
130
129
507-1
13
2
1
180
129
507-1
14
3
1
190
129
507-2
1
1
3
130
265
507-2
2
4
3
440
265
507-2
3
1
3
370
265
507-2
4
1
3
290
265
507-2
5
4
3
350
265
507-2
6
1
3
850
265
507-2
7
1
3
710
265
507-2
8
1
3
60
265
507-2
9
1
3
110
265
507-2
10
1
3
70
265
507-2
11
1
3
40
265
507-2
12
5
3
30
265
507-2
13
1
3
90
265
507-3
1
2
2
70
191
507-3
2
2
2
340
191
507-3
3
1
2
90
191
507-3
4
1
2
100
191
507-4
1
3
6
180
180
507-4
2
3
6
40
180
507-4
4
2
6
340
180
507-4
5
2
6
90
180
507-4
6
3
6
190
180
507-4
7
2
6
760
180
507-4
8
4
6
250
180
507-4
9
7
6
170
180
97


75
Cross 501 Raw Data
Cross 501
Uncensored
Cross 507 Raw Data
Cross 507
Uncensored
Figure 3-6. Regression of F3 family mean regeneration score on F2 parent
regeneration score.


17
This work was rather limited in that the various explants were only tested
under a single culture protocol. While it is possible that a different protocol
might be more effective in inducing regenerable callus from leaf, petiole, or
cotyledon explants, it is also possible that these explants would prove
unresponsive to a wide range of protocols. In view of this uncertainty, further
investigation of alternative explant sources was abandoned, and the remainder
of this dissertation deals only with hypocotyl explants.
Effect of 2.4-D level and kinetin on shoot induction. The selection of
treatments for this experiment were based on reports that regeneration in other
legume species can be sensitive to 2,4-D concentration (Phillips and Collins,
1980; Saunders et al., 1987) and that low concentrations of kinetin in
combination with 2,4-D can stimulate regeneration (Sellars et al., 1990).
None of the 2,4-D treatments, either with or without kinetin, resulted in
shoot production in genotypes UF 20 or UF 144. Response of IRFL 6123 to
shoot induction treatments is presented in Table 2-1. The presence of kinetin in
the induction medium resulted in a slight browning of the callus and complete
inhibition of shoot formation. In the treatments lacking kinetin, a clear response
to 2,4-D concentration was observed. Analysis of variance (ANOVA) indicates
that with respect to mean number of shoot buds per explant, 1.0 mg L'1 2,4-D
was superior to 0.1 mg L'1, but not different from either 0.3 or 3.0 mg L~1.
Linear regression analysis of shoot buds per explant on 2,4-D level yields a
significant quadratic relationship with a maximum between 1.0 and 3.0 mg L'1
(Figure 2-1). Thus, although ANOVA fails to indicate clear superiority of the
1.0 mg L~1 treatment, regression suggests that this treatment is close to the
optimum 2,4-D level, and this is the concentration adopted for later work.
Consistency of regeneration response, as indicated by percent responding
explants, is presented in the last column of Table 2-1. Although values in this


16
into culture, it is difficult or impractical to obtain both in vitro and in planta data
from a single individual. Because an individual seedling possesses only a
single hypocotyl, it is impossible to obtain replicated data on in vitro response
unless quantities of genetically uniform seed are available. In the hope of
eliminating these difficulties, an experiment was conducted to determine if
regenerable callus could be produced using cotyledon, leaf disc, or petiole
explants from genotypes IRFL 6123, CIAT 13083, or UF 20, with hypocotyl
explants included as a check.
Results were discouraging. Cotyledons from IRFL 6123 and CIAT 13083
produced small amounts of necrotic callus while UF 20 cotyledons produced
fairly abundant, deep green callus that showed no signs of regeneration. Leaf
disc explants followed a similar pattern, with IRFL 6123 and CIAT 13083
producing meager quantities of nonregenerating callus primarily from major
veins, and UF 20 producing larger quantities of nonregenerating callus.
Petioles yielded somewhat greater callus mass than either cotyledons or leaf
discs, but still failed to regenerate.
Hypocotyls yielded the greatest mass of callus from all genotypes. Ten of
21 hypocotyl-derived calli from IRFL 6123 produced shoot buds and elongated
shoots were obtained from three of these. Shoot regeneration was observed
from one UF 144 hypocotyl and one CIAT 13083 hypocotyl. In both cases, a
single shoot bud appeared to form directly from explant tissue. This response
could not be repeated in later experiments. It is possible that in both cases
regeneration was the result of a small portion of meristematic tissue from the
cotyledonary node being inadvertently included in the excised hypocotyl
explants. Thus, shoots may have resulted from meristem cloning rather than
true regeneration.


48
indicate that cross 507 cannot be adequately described by an additive-
dominant genetic model. Based on the joint scaling test, the genetics of cross
501 may be characterized by a large, negative dominance effect and a
moderate additive effect. An alternate interpretation is that large, negative
maternal or epistatic effects exist in both crosses. In contrast, variance
partitioning indicates a large additive genetic variance and a much smaller
dominant variance in cross 501.
A general shortcoming of the types of analysis presented to this point is the
large number of simplifying genetic assumptions required. The results derived
from variance partitioning may be more reliable than those of the joint scaling
test, due to the fewer genetic assumptions required by this method. Many
assumptions, such as absence of linkage and equal magnitude of effect from
each contributing locus, are common to both techniques. The generally poor
model fits obtained by both methods may be due to failure of the experimental
material to conform to these assumptions, or to the presence of significant
nonadditive, nondomininant genetic effects. It is likely that experimental error
also contributes to poor model fit.
Parent-offspring regression is a simpler, more empirical method of
analysis, relatively free of genetic assumptions. Regressions of F3 family
means on F2 parents are presented in Figure 3-2. The slopes (regresin
coefficients) represent, by definition, heritabilities. It should be noted, however,
that in the case of selfing, parent-offspring regression is in part a function of
dominant and epistatic genetic effects, but these effects play a much smaller
role in the regression than does additive genetic variance. Regression
coefficients are remarkably high, regardless of cross or scale, ranging from
0.524 to 0.769. The practical implication of this is that substantial progress


64
Table 3-5. Coefficients used for variance partitioning for callus growth.
Cross
Observational
Component
Causal Component
df
D
H
EW
501
ViF2
0.5
0.25
1
64
V1F3
0.5164
0.07076
0.1314
10
v2f3
0.25
0.125
1
170
WiF23
0.5
0.125
0
10
El
0
0
1
53
507
ViF2
0.5
0.25
1
91
V1F3
0.5196
0.07233
0.1571
17
v2f3
0.25
0.125
1
273
W1F23
0.5
0.125
0
17
El
0
0
1
59


2
D. heterocarpon (L.) DC. (carpon desmodium) has been released as the
cultivar 'Florida' for pastures in the southeastern United States (Kretchmer et al.,
1976). This species flowers earlier in the fall than either greenleaf or silverleaf,
and is thus adapted to areas where occasional early frosts threaten seed
production. D. ovalifolium Guill and Perr. is grown in Southeast Asia and has
attracted attention elsewhere for its tolerance to acid, low fertility soils and
shaded conditions (Schultz-Kraft and Pattanavibul, 1985). Other desmodium
species with agronomic potential include D. heterophyllum (Willd.) DC., D.
sandwicense E. May., and D. barbatum Benth.
Most desmodium cultivars are simply selections from germplasm
collections, and genetic improvement of this crop has received minimal
attention until the last decade. A desmodium breeding program was initiated at
the University of Florida in the late 1980's, with the primary breeding objectives
of improved forage quality, rapid establishment, increased seed production, and
resistance to root knot nematodes. The University of Florida program utilizes
germplasm from Desmodium heterocarpon ssp. heterocarpon, D. hetero
carpon ssp. angustifolium (Craig) Ohashi, and D. ovalifolium. The species
heterocarpon and ovalifolium are closely related and yield fertile progeny
upon intercrossing (Quesenberry et al., 1989). Some authors include the
species ovalifolium within species heterocarpon (Ohashi, 1983), but the two
are morphologically distinct and are generally regarded as separate entities by
agronomic researchers (Imrie et al., 1983; Schultze-Kraft and Benavides, 1988).
D. heterocarpon produces elongated inflorescences and glabrous to slightly
pubescent seed pods, while D. ovalifolium has compact inflorescences and
bears heavily pubescent pods. Leaves are opaque in ovalifolium, glabrous to
slightly pubescent in heterocarpon ssp. heterocarpon, and coriacious in
heterocarpon ssp. angustifolium. Subspecies angustifolium is further


66
Table 3-6. Continued.
Cross
Scale
Component
Variance
Weight
df
507
Linear
ViF2
0.06941
19100
91
v-|F3
0.02204
3500
1 7
v2f3
0.03618
208600
273
W2f23
0.02148
13890
1 7
El
0.03746
42040
59
507
Logarithmic
ViF2
0.1227
6044
91
V1F3
0.06896
3574
1 7
v2f3
0.1132
21300
273
W2F23
0.04864
2612
1 7
El
0.03879
39210
59
507
Square Root
ViF2
43.72
0.04760
91
v-|F3
19.19
0.04642
1 7
v2f3
29.39
0.3160
273
w2f23
15.64
0.02542
1 7
El
18.23
0.1701
59


Charmet, G., and S. Bernard. 1984. Diallel analysis of androgenetic
plant production in hexaploid Triticale X. triticosecale, Wittmack.
Theor. Appl. Genet. 69:55-61.
84
Choo, T.M. 1988. Plant regeneration in zigzag clover (Trifolium
medium L). Plant Cell Rep. 7:246-248.
Chow, K.H., and L.V. Crowder. 1972. Hybridization of Desmodium
canum (Gmel.) Schin. and Thell. and D. uncinatum (Jacq.) DC. Crop
Sci. 12:784-785.
Chow, K.H., and L.V. Crowder. 1973. Hybridization of Desmodium
species. Euphytica 22:339-404.
Collins, G.B., and G.C. Phillips. 1982. In vitro tissue culture and plant
regeneration in Trifolium pratense L. In E.D. Earle and Y Demardy
(eds.) Variation in plants regenerated from cells and tissue culture.
Praeger Sci. Pub. New York.
Deak, M., G.G. Kiss, C. Koncz, and D. Dudits. 1986. Transformation of
Medicago by Agrobacterium mediated gene transfer. Plant Cell Rep.
5:97-100.
Dodds, J.H. 1995. Experiments in plant tissue culture. 3rd Ed.
Cambridge University Press, Cambridge.
Dulieu, H. 1991. Inheritance of regeneration capacity in the genus
Petunia. Euphytica 53:173-181.
Dunwell, J.M., R.J. Francis, and W. Powell. 1987. Anther culture of
Hordeum vulgare L.: a genetic study of microspore callus production
and differentiation. Theor. Appl. Genet. 74:60-64.
Falconer, D.S. 1981. Introduction to quantitative genetics, 2nd Ed.
Longman Inc., New York.
Flick, C.E., D.A. Evans, and W.R. Sharp. 1983. Basic techniques of plant
cell culture: organogenesis. In D.A. Evans, W.R. Sharp, P.V. Ammirato,
and Y. Yamada (eds.) Handbook of plant cell culture. Vol. 1.
Techniques for propagation and breeding. Macmillan Publishing Co.,
New York. p. 13-81.


100
Family F3 Regeneration Score Callus Weight
Individual
Progeny
Parent
Progeny
Parent
507-10
11
1
1
20
106
507-10
12
1
1
30
106
507-10
13
1
1
70
106
507-10
14
1
1
50
106
507-11
1
2
5
160
360
507-11
2
1
5
120
360
507-11
3
1
5
110
360
507-11
4
5
5
320
360
507-11
5
3
5
260
360
507-11
6
5
5
370
360
507-11
7
2
5
200
360
507-11
8
3
5
180
360
507-11
9
2
5
80
360
507-11
10
1
5
130
360
507-11
11
5
5
210
360
507-11
12
3
5
250
360
507-11
13
2
5
380
360
507-11
14
2
5
130
360
507-11
15
4
5
290
360
507-12
1
1
1
190
92
507-12
2
3
1
260
92
507-12
3
1
1
90
92
507-12
5
1
1
80
92
507-12
6
1
1
110
92
507-12
8
1
1
340
92
507-12
10
1
1
140
92
507-12
11
1
1
100
92
507-12
13
1
1
120
92
507-12
14
2
1
120
92
507-12
16
2
1
160
92
507-12
18
1
1
100
92
507-12
20
2
1
70
92
507-12
21
3
1
170
92
507-12
22
2
1
60
92
507-12
23
2
1
80
92
507-12
24
1
1
50
92
507-12
25
3
1
100
92
507-12
26
2
1
140
92
507-13
1
4
2
480
382
507-13
2
3
2
210
382
507-13
3
2
2
270
382
507-13
4
2
2
290
382
507-13
5
2
2
220
382
507-13
6
2
2
170
382
507-13
7
4
2
900
382
507-13
8
2
2
150
382


20
Results are presented in Table 2-3. The two right-hand columns of this
table represent two distinct types of elongation response. In the majority of
cases shoot development ceased at the single leaf stage and at a length of one
cm or less (Figure 2-2). These unifoliate shoots exhibited low vigor and could
be rooted only with difficulty by transfer to basal medium supplemented with
1.0 mg L'1 indoleacetic acid (IAA). A smaller proportion of buds developed into
multifoliate shoots (Figure 2-3). The more vigorous multifoliate shoots often
spontaneously formed roots after two to three weeks in elongation medium, or
could easily be induced to root upon transfer to basal medium lacking plant
growth regulators.
Percentage of buds showing the unifoliate elongation response decreased
with increasing BA level. In addition, a qualitative difference could be seen
between the treatments with GA3 and without. In the presence of GA3,
unifoliate shoots tended to bear abnormal, spatulate leaves (Figure 2-4).
Because the degree of abnormality was variable and graded continuously into
normal leaf development, abnormal and normal unifoliate elongation are not
differentiated in Table 2-3.
Significantly more multifoliate shoots were produced on the medium
containing 2.0 mg L'1 BA and lacking GA3 than on the other treatments. The
21% multifoliate shoot production obtained with this treatment was nonetheless
rather disappointing, and the following experiment was designed in an effort to
improve on this figure.
Effect of picloram/BA ratio on shoot elongation. All of the treatments in the
above experiment followed the precedent of Phillips and Collins (1980) and
Wofford et al. (1992b) in utilizing the relatively low picloram concentration of
0.002 mg L1. Since neither increasing BA level nor addition of GA3 produced
dramatic improvement of elongation response, a second experiment was


53
than one. Although this is a less than ideal approach, there appears to be no
acceptable alternative. The option of simply analyzing the raw regeneration
scoreseither transformed or otherwiseis undesirable because this entails
accepting the clearly incorrect assumption that all individuals receiving
regeneration scores of one have the same proclivity for regeneration. The
contrasting options, then, are to ignore the left-hand tails of many of the
population distributions, or to use an acknowledgedly problematic method of
reconstructing these tails. Means and variances determined by uncensoring, as
well as means and variances of the raw data are presented in Table 3-8. In
general, means derived from regression order analysis are lower than those
obtained from the raw data, and variances are higher. This is not unexpected,
since truncation of the low end of a data setas occurs in the raw datawill
artificially raise the mean and lower the variance. For the raw data, a significant
(a=0.05), positive relationship exists between F3 family variance and family
mean, due at least in part to the fact that the lower the mean, the more severely
truncated the distribution. This relationship is absent from the uncensored data
because the truncated lower ends of the distributions have been reconstructed.
The unusually high variances observed for some F3 families may be due to
segregation of major genes within these families, resulting in a somewhat
bimodal distribution and exaggerated variance. This effect is most pronounced
in the uncensored data. Unfortunately, population sizes for the F3 families are
too small to clearly distinguish between spurious bimodality and bimodality
resulting from genetic causes. Bimodality was not observed in the F2
populations, nor in the complete F3 populations.
The accuracy of the variance estimates presented in Table 3-8 can be
compromised by non-normality resulting from inadequate scaling. Major
scaling problems can be detected by examining the distribution of a population


37
28 days. Shoot buds were induced on L2 with 1.0 mg L"1 2,4-D and 2.0 mg L'1
adenine for 14 days, and shoot elongation was induced on L2 with
0.012 mg L"^ picloram and 0.2 mg L^ BA for 28 days. Shoots were rooted on
L2 lacking plant growth regulators. Hypocotyls from parental lines were
cultured simultaneously with the F2 and F3 populations to serve as checks.
Callus fresh weights were measured at the time of transfer from callus
medium to bud induction medium. Each callus was visually scored for shoot
bud formation and elongation at the end of both the bud induction and bud
elongation steps. The scoring system and statistical transformations used will
be discussed in the results.
Statistical analysis. General statistical analyses were conducted using
JMP Version 3.0, the SAS Institute general statistics product for Macintosh
computers (SAS Institute, 1994). Means and variances of truncated data sets
were estimated using UNCENSOR Version 3.0 (Newman and Dixon, 1989).
Results and Discussion
Results of crossing efforts were disappointing. The approximately 2200
flowers that were cross-pollinated produced only 590 seeds. Poor seedset was
likely due in part to inadequate greenhouse ventilation, which resulted in
occasional daytime temperatures sometimes exceeding 38 C. Hybrid yield
was further reduced by the low ratio of hybrids to seifs in the seeds that were
produced. Approximately 550 of the 590 seeds germinated, but only eight of
the resulting plants were identified as hybrids. Hybrids could be readily
identified by leaf shape and growth habit, both of which were intermediate
between the parental types. Two hybrid plants were progeny of the cross IRFL
6123 x CIAT 13083 (hereafter referred to as cross 501), three were from cross


57
the fact that the crosses share a regenerating female parent. If ability to
regenerate were conferred entirely by alleles originating from one or two loci in
the regenerating parent, it would be expected that the best regenerators among
the F2 and F3 would be those bearing the fixed genotype of the regenerating
parent. The best regenerators from each cross would then be expected to
perform approximately equally, regardless of the identity of the nonregenerating
parent. This was not observed. Instead, the best regenerators from cross 507
consistently outperformed those of cross 501 in both the F2 and F3. The
transgressive phenotype occurred only in cross 507, suggesting that the
nonregenerating parent in this cross contributes at least one positive-acting
allele. The issue of quantitative versus qualitative inheritance could be clarified
by culturing of a large number of progeny from several high-regenerating F3
plants.
Some final insights can be obtained by examining the relationship
between regeneration and callus growth. Figure 3-8 presents a scattergram of
regeneration score versus logarithm of callus weight for the combined F3
population. A very weak, but significant correlation exists, and perhaps as
meaningful is the observation that the high regeneration scores are generally
associated with moderate callus growth. This apparent clustering is in part an
illusion stemming from the smaller population sizes for the higher regeneration
scores. However, there is a significant (a=0.05) decrease in population
variances for the callus growth trait as regeneration score increases, indicating
that the clustering is real. The association can be either causal, accidental, or
both. In both crosses, callus growth was greater in the nonregenerating parent.
The absence of regeneration in the high growth calli may therefore be the result
of genetic linkage between the two traitsan accident of the distribution of
genetic material among the parents. The lack of regeneration among low


REFERENCES
Ammirato, P.V. 1983. Basic techniques of plant cell culture:
Embryogenesis. In D.A. Evans, W.R. Sharp, P.V. Ammirato, and Y.
Yamada (eds.) Handbook of plant cell culture. Vol. 1. Techniques for
propagation and breeding. Macmillan Publishing Co., New York.
Angeloni, P.N., H.Y. Rey, and L.A. Mroginski. 1988. Cultivo in vitro de
tejidos de Desmodium incanum y D. affine (Leguminosae).
Phyton:48:71 -76.
Bingham, E.T., L.V. Hurley, D.M. Kaatz, and J.W. Saunders. 1975.
Breeding alfalfa which regenerates from callus tissue in culture.
Crop Sci. 15:719-721.
Blaydes, D.F. 1966. Interaction of kinetin and various inhibitors in
the growth of soybean tissue. Physiol. Plant. 19:748-753.
Bovo, O.A., L.A. Mroginski, and H.Y. Rey. 1986. Regeneration of plants
from callus tissue of the pasture legume Lotononis bainessi. Plant
Cell Rep. 5:295-297.
Brown, D.C.W., and A. Atanassov. 1985. Role of genetic background in
somatic embryogenesis in Medicago. Plant Cell, Tissue, Organ
Culture 4:111-122.
Buiatti, M., S. Baroncelli, A. Bennici, M. Pagliai, and R. Tesi. 1974.
Genetics of growth and differentiation in vitro of Brassica olercea
var. botrytis. Z. Pflanzenzuchtg. 72:269-274.
Bullock, W.P., and P.S. Baenziger. 1982. Anther culture of wheat
(Triticum aestivum L.) FTs and their reciprocal crosses. Theor.
Appl. Genet. 62:155-159.
Cavalli, L. 1952. An analysis of linkage in quantitative inheritance.
In E.C.R. Reeve and C.H. Waddington (eds.) Quantitative inheritance.
HMSO, London.
83


52
weaknesses of each are subject to considerable debate among statisticians.
The most intuitively appealing method is termed regression order statistics.
This method is graphically illustrated in Figure 3-4. Figure 3-4 (a) is a normal
probability plot of a hypothetical, approximately normally distributed population.
The Y-axis is observed value of the variable of interest, and the X-axis is normal
quantile score (a transformed Z-score scale). The points represent the
individual observations, and the line represents the simple linear regression
line obtained from the individual data points. The slope and intercept of this line
can be used to approximate the variance and mean of the population. Figure
3-4 (b) is a similar plot of the same population with a censoring threshold
applied such that approximately 30% of the observations yield no observable
response. These observations, of course, all have the same values for both
axes, and therefore all lie atop one another at the lowest left data point on the
graph. The result is a truncated plot. The regression line is derived only from
those observations lying above the censoring threshold, but is similar to the line
obtained from the uncensored population, and would yield variance and mean
estimates close to the actual parameters of the original population. The
regression order statistics method is a numeric analog of Figure 3-4 (b). That is,
it performs a normal quantile regression on a truncated population in order to
approximate the mean and variance of the intact population.
The accuracy of population parameters estimated by uncensoring through
regression order analysis is limited by the number of observations lying above
the censoring threshold. However, reasonably accurate results can be
obtained even when the majority of observations are below the threshold, as
long as several observations lie above the threshold. Regression order
analysis was used to estimate means and variances of all populations in the
current study that had at least four individuals with regeneration ratings higher


43
generation) are needed to obtain a non-singular solution and a measurement of
statistical significance of the solution. Each equation is weighted in proportion
to the level of certainty associated with each generation mean, which is
equivalent to the reciprocal of the variance of the mean. As originally
described, a solution is obtained through cumbersome matrix manipulations.
This is computationally identical to performing weighted least-squares
regression using the observed means as the dependent variable and the model
coefficients as independent variables (Rowe and Alexander, 1980). The partial
regression coefficients obtained in this way are equivalent to the unknowns in
the above-described equations. The adequacy of solutions obtained by this
method is assessed in two ways. First, each predicted model parameter has an
associated F test for the hypothesis that the parameter differs from zero.
Secondly, the whole model can be tested by examining the error sum of
squares (SSE). In this particular form of analysis, SSE exhibits a x2
distribution, and the reported SSE can be compared to tabular x2 values to
determine goodness of fit (Rowe and Alexander, 1980).
Six joint scaling tests were conducted in order to test each of the three
scales using data from each cross. Model coefficients used to construct the
tests are given in Table 3-3. Observed means and weights used in the tests are
are based on the means and variances presented in Table 3-2. Predicted
values of m, [d], and [h], and the results of goodness of fit testing are given in
Table 3-4. Adequate goodness of fit is obtained only for the linear and square
root scales and only for cross 501. In both cases, a large, negative dominance
effect is indicated.
The joint scaling test suffers from limitations that necessitate cautious
interpretation of results. One serious limitation is that the test assumes that one
parent possesses most of the positively acting alleles and the other possesses


106
Family F3 Regeneration Score Callus Weight
Individual
Progeny
Parent
Progeny
Parent
501-11
2
2
1
480
133
501-11
3
4
1
260
133
501-11
4
3
1
140
133
501-11
5
1
1
80
133
501-11
7
1
1
140
133
501-11
8
1
1
440
133
501-11
9
1
1
270
133
501-11
10
1
1
200
133
501-11
11
2
1
540
133
501-11
12
1
1
420
133
501-11
13
1
1
270
133
501-11
14
1
1
330
133
501-11
15
1
1
370
133
501-11
16
1
1
180
133
501-11
17
1
1
260
133
501-11
18
1
1
120
133
501-11
19
1
1
180
133


80
Histological examination of regenerating cultures revealed vascular
connections between shoots and the surrounding callus tissue, indicating that
regeneration was of an organogenic nature. However, early in the shoot
induction period, small structures resembling somatic embryos could
occasionally be observed. The combination of growth regulators used in the
shoot induction step is similar to that used for somatic embryogenesis in other
species, so it is possible that the regeneration observed in this study was some
aberrant form of somatic embryogenesis.
It was clearly established in Chapter 2 that regeneration in desmodium is
highly dependent on genotype and that broadening the range of regenerating
genotypes through modification of culture protocols is difficult or impossible.
Chapter 3 focused on the genetics of callus growth and regeneration. An
extensive effort at crossing regenerating with nonregenerating genotypes
yielded only three crosses501, 507, and 510. Cross 510 proved useless for
the analysis at hand due to severe internodal stunting and other morphological
abnormalities in the F-| and F2 generations. This may have been the result of
cytoplasmic effects, since cross 510 was essentially the reciprocal of cross 507,
which showed no abnormalities. Examination of the genetics and physiology of
this phenomenon might produce very interesting results.
The remaining two crosses were selfed through the F3 generation.
Analysis of callus growth and and regeneration in these crosses produced
many ambiguous results, but also yielded useful insights. A large amount of
effort was devoted to scaling of the callus growth trait. General statistical
considerations indicated that a simple linear scale was inadequate for
describing this trait. Logarithmic and square root transformation were both
more satisfactory than the linear scale, but it was not clear which of these
transformations was superior.


65
Table 3-6. Observational variance components and initial weighs
used for variance partitioning for callus growth.
Cross
Scale
Component
Variance
Weight
df
501
Linear
ViF2
0.07076
12780
64
V1F3
0.06100
2688
1 0
v2f3
0.06232
43780
170
W2f23
0.08426
582.6
1 0
El
0.03040
57370
53
501
Logarithmic
ViF2
0.1618
2445
64
VlF3
0.09271
1163
1 0
v2f3
0.08591
23030
170
W2F23
0.05925
813.8
1 0
El
0.03994
33210
53
501
Square Root
ViF2
4788
0.02791
64
ViF3
36.46
0.007521
1 0
v2f3
32.86
0.1575
170
w2f23
36.93
0.002687
1 0
El
17.37
0.1757
53


103
Family F3 Regeneration Score Callus Weight
Individual Progeny Parent Progeny Parent
501-1
12
2
3
70
160
501-1
13
3
3
160
160
501-1
14
1
3
90
160
501-2
1
1
1
1240
984
501-2
2
1
1
1110
984
501-2
3
2
1
860
984
501-2
4
3
1
1000
984
501-2
5
1
1
890
984
501-2
6
2
1
600
984
501-2
7
1
1
1300
984
501-2
8
1
1
740
984
501-2
9
2
1
350
984
501-2
10
1
1
750
984
501-2
11
1
1
1870
984
501-2
12
1
1
550
984
501-2
13
2
1
620
984
501-3
1
1
3
210
298
501-3
2
1
3
320
298
501-3
3
1
3
420
298
501-3
4
1
3
370
298
501-3
5
2
3
440
298
501-3
6
3
3
140
298
501-3
7
2
3
90
298
501-3
8
2
3
610
298
501-3
9
1
3
190
298
501-3
10
1
3
130
298
501-3
13
1
3
310
298
501-3
14
1
3
130
298
501-3
15
1
3
560
298
501-3
16
1
3
330
298
501-3
17
1
3
510
298
501-3
18
1
3
130
298
501-3
19
1
3
830
298
501-3
20
1
3
130
298
501-3
21
3
3
160
298
501-3
22
1
3
150
298
501-3
23
1
3
530
298
501-4
1
3
1
590
1352
501-4
2
2
1
710
1352
501-4
3
1
1
190
1352
501-4
4
1
1
580
1352
501-4
5
2
1
920
1352
501-4
6
4
1
1140
1352
501-4
7
2
1
1680
1352
501-4
8
1
1
240
1352
501-4
9
1
1
540
1352


81
Formal genetic analysis of callus growth was approached via the joint
scaling test and variance partitioning. Neither method yielded conclusive
results. This may have been due to violation of the genetic assumptions
inherent in both methods, or may have been because both methods
inappropriately attempted to describe the trait with a simple additive-dominant
genetic model. Parent offspring regression yielded heritabilities ranging from
0.52 to 0.77, depending on cross and scale. It appears, then, that callus growth
in this germplasm is controlled to a significant extent by additive genetic effects,
but that there may in addition be nonadditive, nondominant genetic effects that
act to confound the joint scaling test and variance partitioning.
The regeneration trait presented unusual analytical difficulties. Many calli
showed no evidence of regeneration. As a result the majority of population
distributions were severely truncated. Truncated (or "censored") distributions
cannot be rendered normal by any mathematical transformation. A procedure
was described by which truncated population distributions can be
reconstructed, or "uncensored." While this procedure has apparently not been
previously applied to genetic analysis, it is conceptually similar to the threshold
trait approach described by Falconer (1981). The method is less than ideal, but
is preferable to attempting to analyze raw, severely censored data. Constraints
imposed by the data structure and by the uncensoring technique made it
impossible to conduct meaningful joint scaling tests or variance partitioning
analyses. Parent-offspring regression yielded relatively high heritability
estimates for cross 5070.416 for raw data; 0.460 for uncensored data. No
significant regression was obtained for cross 501. It appears likely that this
failure is due to shortcomings of the data structure, and that regeneration is in
fact weakly to moderately heritable in this cross.


14
3, 7, 14, or 28 days. Following the induction treatment, calli were transferred to
the shoot elongation medium described above. The 28-day induction treatment
was followed by 28 days on elongation medium and the 14-, 7-, 3-, and 0-day
induction treatments were followed by 42, 49, 53, or 56 days on elongation
medium, respectively, to yield a total of 84 days in culture for each treatment.
Calli were evaluated for bud formation at frequent intervals throughout the
study. Twenty-one explants (three per dish) were used for each of the nine
treatments.
Effect of BA and GA3 (aibberellic acidt levels on shoot elongation.
Hypocotyls from genotype IRFL 6123 were cultured on the callus growth
medium described above for 28 days followed by 28 days on bud induction
medium containing 2.0 mg L'1 adenine and 1.0 mg L'1 2,4-D. Twenty-four calli,
each containing at least one well-formed shoot bud or small bud cluster, were
transferred to each of the following elongation media:
(1) 0.002 mg L1 picloram, 0.2 mg L'1 BA, 0 mg L1 GA3
(2) 0.002 mg L1 picloram, 0.2 mg L'1 BA, 0.2 mg L'1 GA3
(3) 0.002 mg L'1 picloram, 0.6 mg L'1 BA, 0 mg L'1 GA3
(4) 0.002 mg L1 picloram, 0.6 mg L'^ BA, 0.2 mg L'1 GA3
(5) 0.002 mg L"1 picloram, 2.0 mg L1 BA, 0 mg L'1 GA3
(6) 0.002 mg L'1 picloram, 2.0 mg L'1 BA, 0.2 mg L'^ GA3
Shoot elongation was visually evaluated after 28 days of elongation treatment.


IN VITRO MORPHOGENESIS AND INHERITANCE OF IN VITRO
TRAITS IN DESMODIUM
By
PETER A. KROTTJE
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1995