|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
DWARFISM IN LOW CHILL HIGHBUSH BLUEBERRY
DAVID H. BAQUERIZO
A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
MASTER OF SCIENCE
U NIVE RS ITY OF FLORI DA
David H. Baquerizo
This thesis is dedicated to Karen, David Manuel, Gabriel Roberto, Ana Faustina
and Maria Teresa.
I thank Dr. Paul Lyrene, my chairman professor, for his guidance during this
project, and for the opportunity of learning about breeding blueberries and
I thank my advisors, Dr. Wayne Sherman, Dr. Mark Bassett and Dr. Ramon
Littell, for all their contributions to this study.
I also thank my family for their support and all the nice people at the
University of Florida. Because of people like them, life is beautiful and full of joy.
TABLE OF CONTENTS
ACKNOWLEDGMENTS .........._._ ...... .___ ...............iv....
LIST OF TABLES .........._._ ...... .___ ...............vii...
LIST OF FIGURES .........._._ ...... .___ ...............viii...
ABSTRACT .........._._ ...... .___ ...............ix....
1 INTRODUCTION................ ............. 1
2 L ITE RATU RE REVI EW. ........._._._.. ...... ............... 4..
Morphology of Dwarf Plants ........._.._... .......... ....____ ...........4
Genetics and Physiology of Plant Dwarfism ........._._............ ............. 5
Highbush Blueberry Domestication ...._._._._ ........____ ........___........ 7
3 MATERIALS AND METHODS ........._._.. ......___ ....._.. ............1
Morphological Studies ...._ ......_____ ........___ .. ...... .....1
Multiple Comparison Analysis................ ............... 11
Internode length ............. ...... ._ ....._ _. ............1
Leaf area........ ... ... .........__ ............_ .. .... ....._ .......... 1
Height of the plant or length of the longest shoot .......................... 14
Logistic Regression Analysis ......._._.............._ ........._ ........ 14
Inheritance Studies ..............__......._ ....._._ .... ........1
Field Observations................ ............. 15
Controlled Crosses ..........._.._ ....._ ............... 16....
4 RESULTS AND DISCUSSION............... ............... 18
Morphological Studies ...._ ......_____ ........___ .. ...... .....1
Multiple Comparison Studies ...._ ......_____ ...... ... ..__......... 18
Logistic Regression Analysis ...._ ......_____ ...... ... ..__......... 23
Inheritance Studies ............. .....__ ............... 30...
5 CONCLUSIONS................. ............. 38
APPENDIX CATEGORICAL DATA ANALYSIS................. ............... 39
LITERATURE CITED ............. ...... ._ ............... 44....
BIOGRAPHICAL SKETCH ............. ...... .__ ....___ .............4
LIST OF TABLES
1 Length of three-internodes (mean+~2 SE) of dwarf and normal plants
(average of 10 measurements) ................. ................. 20......... ...
2 Results of the t-test for the length of three internodes of dwarf plants vs.
normal plants ................. ................. 20..............
3 Individual leaf area (mean +2 SE)................ .................. 21
4 Results of the t-test for the leaf area of dwarf plants vs. normal plants...... 21
5 Frequency of plants in different categories of length and branch number
for each plant type. ........... ....._ ............... 25.
6 Model 1 and Model 2 equations with logit, odds of dwarf and probability
of dwarf, for Model 2 only. ...._ ......_____ ......._ __ ...........2
7 Model 4 and Model 5 equations with logit, odds of dwarf and probability
of dwarf, for Model 5 only. ...._ ......_____ ......._ __ ...........2
8 Results of fitting five logistic regression models to the dwarf data ............29
9 Segregation ratios and chi-square tests for normal to dwarf ratios (3:1
and 11:1) for dwarf-segregating normal stature phenotype crosses from
the 2002, 2003 and 2004 high density plots. ........... __... ......__........ 31
10 Tetrasomic inheritance frequencies following chromosomal segregation .. 33
11 Segregation ratios and chi-square test for normal to dwarf ratios in dwarf
x dwarf controlled crosses................ ................ 34
12 Segregation ratios and chi-square test for normal to dwarf ratios in dwarf
x normal controlled crosses. ................. ................. 35.............
13 Segregation ratios and chi-square test for normal to dwarf ratios in
normal x normal controlled crosses. ................. ................ ......... 35
14 Cross table of all the controlled crosses per genotype evaluated in this
study. ................. ................. 36..............
LIST OF FIGURES
1 The author with dwarf and normal highbush blueberry, May 2005....._...... 18
2 Dotplot of length of three internodes (mm) by clone ........._...... .............. 19
3 Dotplot of leaf area (mm2) by clone................ ................. 22
4 Scatter plot of three year old dwarf and normal plant heights (cm) from
the 2003 Stage 2 nursery................ ................ 23
5 Scatter plot of height versus number of branches for dwarf and normal
plants 10 month old. ........._._ ....__._ .. ............... 24
Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science
DWARFISM IN LOW CHILL HIGHBUSH BLUEBERRY
David H. Baquerizo
Chair: Paul M. Lyrene
Major Department: Horticultural Science
Plant dwarfism was studied in the highbush blueberry (Vaccinium
corymbosum complex hybrid) breeding program at the University of Florida
between the fall of 2002 and spring of 2005. Morphological studies included
comparisons among dwarf and normal populations for plant height, internode
length, leaf area and number of sprouts or branching. Inheritance studies were
conducted by crossing dwarf x normal, dwarf x dwarf and normal x normal plants,
and by making field observations on more than 10,000 seedlings from normal x
normal crosses grown in high-density seedling nurseries in the breeding
All of the studied morphological traits were significantly different between
normal and dwarf populations. Plant height, internode length and leaf area of
normal plants were 2.5, 1.7 and 2.7 times that of dwarf plants respectively.
Dwarf plants were also characterized by a high number of branches when
compared to normal. Based on a logistic regression analysis, plants that had
more than 5 branches and were less than 16 cm tall at 10 months had a 99.4 %
probability of being dwarf. Conversely, the probability of a dwarf plant was 0.9%
for plants with fewer than three branches and a height exceeding 16 cm.
The fertility of the studied dwarf plants was normal. The genotype of dwarf
plants appears to be simplex (Aaaa), with the nulliplex form being lethal. Most
crosses between normal plants that segregated dwarfs segregated in an 11:1
and 27:8 normal to dwarf ratio, supporting the hypothesis that triplex and duplex
plants have normal growth habit. Triple to duplex cross produces dwarf to
normal ratios of 11:1. Duplex to duplex cross segregates in a 27:8 normal to
dwarf ratio. Segregation ratios in a few dwarf to dwarf and dwarf to normal
crosses did not fit the proposed model.
Mature plant height is influenced by both environment and genotype.
Environmental influences can significantly reduce plant height and produce
dwarfed plants, as in the case of the bonsai, which, due to constant pruning and
small containers, yields dwarfed plants (Garvey, 1985). Environmentally induced
dwarfism is not passed to its progeny, and it can be tedious work to keep the
plant dwarf. A dwarf genotype achieves its dwarf phenotype without unusual
environmental influence and may pass this trait to its progeny.
Dwarf genotypes are of interest to most breeding programs, because the
shorter stature plants can have advantages in common horticultural practices.
Dwarf genotypes can reduce the cost of controlling tree size (a significant
operational cost for normal height plants), and can maximize the use of available
space. Dwarf plants can also be of horticultural and ornamental use when height
is a limitation in protected agriculture or in indoor landscapes.
Dwarf plants have been important in agriculture. The spectacular increase
in grain yield during the "Green Revolution," particularly in wheat and rice, can be
attributed largely to the dwarf traits introduced into the cultivars (Hedden, 2003).
Horticultural crops that benefit from short sized cultivars or rootstocks include
apples (Atkinson and Else, 2001, Johnson et al., 2001), cherries (Edin et al.,
1996, Franken-Bembenek, 1996), bananas and raspberries (Keep, 1969).
Advantages offered by dwarf plants include easier harvest, less need for
trimming and pruning to control height, and suitability for high-density planting
Dwarf trees may also be useful for studying the genetic control of
physiological factors affecting shoot growth (Sommer et al., 1999). To date,
there has been much published research that used dwarf mutations to study
plant regulators. Curtis et al. (2000) discovered a feedback control of GA 20-
oxidase gene expression in Solanum dulcamara by over-expressing the pumpkin
gene CmGA20ox1 that induced semi-dwarf plants in S. dulcamara. Molecular
and genetic studies of dwarf mutants of arabidopsis, tomato and pea have
helped explain the role of brassinosteroid biosynthesis and regulation (Li and
Dwarf growth habit has been previously observed and studied in highbush
blueberry (Vaccinium corymbosum complex hybrids). In 1984, Draper et al.
studied dwarf selections and concluded that their data appeared not to fit
tetraploid genetic ratios for a single locus; furthermore plant height in crosses
involving dwarfs appeared to follow a continuous distribution. In the University of
Florida blueberry breeding program, dwarf plants have been observed
segregating from southern highbush crosses in which both parents had normal
stature. These dwarf seedlings are characterized by reduced height and by
compact and multiple-sprouted canopy, similar to the dwarf Vaccinium ashei (FL
78-66) described by Garvey and Lyrene (1987).
The objective of this research was to study the inheritance and morphology
of different southern highbush dwarf phenotypes observed in the University of
Florida blueberry breeding program.
L ITE RATU RE REVI EW
Morphology of Dwarf Plants
Dwarf plants are characterized by reduced height when compared to
normal plants. Dwarfness results from either smaller andlor fewer cells (Gale
and Youssefian, 1985) and thus, shorter andlor fewer internodes. Bindloss
(1942) reported fewer cell divisions in the stem of a dwarf Lycopersicon
esculentum L. compared to a normal plant. Pelton (1964), as cited by Garvey
(1985), reported precocious secondary cell wall thickening in the dwarf
columbine (Aguilegea vulgares L. cultivar 'Compacta'). This was believed to
result in smaller cells and dwarf plants.
Other morphological traits that have been described in dwarf plants include
smaller canopy (Fideghelli et al., 2003); higher number of sprouts (Wareing and
Phillis, 1978; Draper et al., 1984; Garvey and Lyrene, 1987), smaller leaves
(Draper et al., 1984), smaller reproductive organs, smaller fruits, and smaller root
weight and depth (Gale and Youssefian, 1985).
In blueberries, dwarf plants have been observed and studied. Draper and
colleagues in 1984 reported dwarf selections of V. corymbosum with shorter
internodes and smaller leaves. They also mentioned a bushy appearance to
each shoot caused by high sprouting. Garvey and Lyrene (1987) observed and
studied dwarf selections of V. ashei. The dwarf plants were described as short-
saturated and compact-growing.
Genetics and Physiology of Plant Dwarfism
Reports on the inheritance of dwarfing genes are numerous and diverse.
There are good examples of simple inheritance like the recessive dw gene in
peach (Hansche et al., 1986), the monogenic recessive dwarfing genes in
raspberry: fr, n and dw (Knight and Scott, 1964; Jennings, 1967) and the single
dominant gene for compact habit in 'Wijcik' apple (Lapins and Watkins, 1973).
Keep (1969) mentioned the digenic 'sturdy dwarf' and 'crumpled dwarf' in
There are also complex inheritances that can not be explained by a known
inheritance ratio. Examples include the Vaccinium dwarfs studied by Draper and
colleagues (1984) and by Garvey and Lyrene (1987). These dwarf phenotypes
might be the product of several interacting genes.
Aneuploidy can also affect the stature of plants and cause dwarfs. With
aneuploids, fertility is significantly reduced because the abnormal chromosome
number affects meiosis, producing some non-viable gametes.
Three major groups of hormones are most often reported in association
with dwarf phenotypes. The most commonly mentioned group is gibberellins
(GA), followed by auxins and brassinosteroids. A fourth group, cytokinins has
been implicated in some dwarfs.
Gibberellins are usually associated with shoot and cell elongation, internode
length and other plant developmental processes like fruit enlargement (Berhow,
2000). Gibberellin was named after the fungus Gibberella fujikuroi, which causes
elongation in infected rice seedlings. Foliar application of GA (either GA3 Or GA4)
in apple (Mlalus domestica Borkh) decreased shoot number but increased total
shoot length and total bud number (Kurshid et al., 1997).
Mutations affecting GA synthesis, deactivation and reception are usually
identified via a shoot elongation screen (Ross et al. 1997). Short plant (dwarf)
mutants can be categorized as responsive or non-responsive depending on their
response to exogenous gibberellins.
Ladizinsky (1997) described a dwarf phenotype in Lens characterized by
short internodeh, short leaf axis and smaller convex leaflets. This dwarf, when
sprayed with GA, responds positively by elongation of the internode and leaf-
axis. Dwarfs that respond in this way to GA applications are referred to as GA
responsive or GA sensitive dwarfs. Goldman and Watson (1997) described a
monogenic dwarf mutant in red beet (Beta vulgaris L. subsp. Vulgaris) that is also
sensitive to GA.
Mackenzie-House et al. (1998) gives a good example of a non responsive
dwarf mutant. Application of GA to Pisum sativum L. plants that carry the Irs
mutation reduces internode length, GA synthesis and cell elongation.
Borner et al. (1999) studied two dwarfing genes in barley (Hordeum
vulgare), the recessive gai and gal dwarfing genes. Both were on chromosome
2H and both reduced plant height, but the gal phenotype was sensitive to
exogenous GA, whereas the gai phenotype was insensitive.
Auxins are also directly involved in regulation of stem elongation (Little et
al., 2003), and they might be present at lower levels in dwarf plants than in
normal tall plants (Yang et al., 1993). An abrupt growth response was observed
by Yang et al. (1996) in dwarf mutants (GAl-deficient le) of light-grown pea
(Pisum sativum L.) after applying IAA. The lag time was only 20 minutes, and
the plants reached a growth rate up to ten times higher than the control. The
elongation was in the older elongating internodes. Gibberellins also caused an
elongation response (mainly in less than 25% of expanded internodes). The
authors concluded that auxins and gibberellins control separate processes that
together contribute to stem elongation. A deficiency in either leads to a dwarf
Brassinosteroids have also been observed to cause dwarf plants, mainly
due to their role throughout plant growth and development. Yin et al. (2002) and
Schaller (2003) reported that plants with defective brassinosteroid biosynthesis
and perception have cell elongation defects and severe dwarfism.
Since prolific sprouting is characteristic of many dwarf types and is the
result of the growth of many axillary buds (Wareing and Phillis, 1978), and since
the interaction among auxins and cytokinins has been reported to control apical
dominance (Sachs and Thimann, 1967), cytokinins should also be considered
when investigating plant dwarfism mechanisms, because cytokinins can disrupt
apical dominance and cause the multiple sprouting or bushy branches reported
with the dwarf phenotype by Draper et al. (1984).
Highbush Blueberry Domestication
The domestication of highbush blueberries (complex hybrids of Vaccinium
corymbosum L.) started early in the 20th century with the work of Frederick
Coville, a botanist who made the first selections for breeding purposes in the first
US blueberry breeding program (Coville, 1937).
From the beginning, highbush blueberries were hybridized with other
species in vaccinium section Cyanococcus (Moore, 1966). Some of the other
species used to improve the highbush blueberry included lowbush blueberry (V.
angustifolium Ait.) and the myrtle blueberry of Florida (V. myrsinites Lam.). Early
in the breeding of blueberries it was noticed that some species did not hybridize,
and it was determined by the cytological work of Longley in 1927 (cited by
Coville, 1937) that the primary cause was the difference in ploidy diploids when
crossed with tetraploid species did not hybridize, or if they did, they produced a
few low vigor plants (Coville, 1937).
Draper and Hancock (2003) mentioned the work of Darrow and Sharpe,
who selected a V. darrowi (Camp) plant that they found near Tampa, Florida.
They named it Florida 4B, and this plant has been used extensively in the
southern highbush blueberry breeding programs to reduce the high chill
requirement of the northern highbush (Sharpe, 1954; Sharpe and Darrow, 1959;
Sharpe and Sherman, 1971; Lyrene and Sherman, 1984). It is important to
mention that this diploid plant hybridizes with the tetraploid V. corymbosum
because it produces unreduced gametes (Lyrene, Vorsa and Ballington, 2003).
Other interspecific crosses involving tetraploid cultivated blueberries with
non tetraploid species included hexaploid V. ashei (Darrow, 1949; Lyrene and
Sherman, 1984) and diploid V. elliottii (Lyrene and Sherman, 1983; Lyrene and
The rabbiteve blueberry (V. ashei) was grown commercially in Florida
starting around 1893, and by the late 1920's approximately two- to three-
thousand acres were in cultivation (James, 1924; Clayton, 1925; Mowry and
Camp, 1928; Lyrene and Sherman, 1979). By 1930, the industry had declined
rapidly, mainly because the quality of the Florida blueberries was low compared
to New Jersey and Michigan blueberries. The northern blueberries were
produced on clonally-propagated cultivars based on V. corymbosum developed
by the U.S. Department of Agriculture (USDA) (Lyrene and Sherman, 1977). To
support the blueberry industry in the southeast, brdeding efforts with V. ashei
were started in 1940 in Tifton, Georgia (Brightwell, 1971) and with highbush
blueberries in Florida in 1948 (Sharpe and Sherman, 1971).
The Florida blueberry breeding program has focused mainly on the
improvement of the tetraploid V. corymbosum, rather than on the hexaploid V.
ashei. Lyrene and Sherman (1977) mentioned various reasons for this, including
the fact that Georgia already had an active rabbiteve breeding program, and
none of the Vaccinium species in Florida can be easily crossed with V. ashei
without producing pentaploids. Further, early ripening (the most significant
advantage for Florida) and low chilling were not readily available in V. ashei
Low-chill highbush blueberries from Florida, based on V. corymbosum and
V. darrowi hybrids, has resulted in an early-season blueberry industry in Florida
and Georgia. To date, the annual shipment of fresh-market blueberries from
Florida is about 4 million pounds and is increasing yearly. The estimated
wholesale value (farm gate value) is about $20 million per year. Almost all early-
season varieties grown in Florida came from the University of Florida blueberry
breeding program (Dr. Paul Lyrene, personal communication).
MATERIALS AND METHODS
It was noticed in the Florida tetraploid highbush breeding program that
some crosses between two plants of normal stature segregated dwarfs. Two of
these dwarf plants, 00-266 and 00-08, were selected by Dr. Paul Lyrene before
the study reported here was carried out, because of their desirable traits for
breeding (early leafing and high fruit quality). Other dwarf selections were made
during the study period from the high density seedling nurseries of 2002 and
2003. The high density nursery (Stage One) consisted of about 120 different
crosses of normal stature plants, about 90 seedlings per cross, producing a
highly diverse blueberry population.
Two analyses were carried out for morphological traits with the objective of
contrasting normal and dwarf types. The first was a multiple comparison analysis
for particular traits (e.g. internode length). The second was a logistic regression
analysis that modeled the response (the probability of being a dwarf) for given
parameters (i.e. height and sprouting).
Multiple Comparison Analysis
To classify and distinguish dwarfs from normal plants, internode length, leaf
area, and plant height were measured from different clones representing the
dwarf and normal types.
The dwarf plants studied for internode length and leaf area measurements
were selected from the high density nursery (Stage One) of 2003, except for 00-
266 and 00-08 which had been selected and propagated previously by Dr. Paul
Lyrene. Dwarf plants were selected subjectively by visual inspection, but the
dwarf plants appeared to be qualitatively different from their normal full-siblings,
and there were few or no plants whose classification was not obvious. The
normal plants selected for contrast with the dwarfs also had a diverse
background and were used previously in the breeding program (some of them
are known cultivars, e.g. 'Emerald' and 'Jewel')
All plants used in this study were at least one year old. They were grown in
black 3-liter pots filled with peat, outside in full sun, watered as needed, and
fertilized with Tracite 20-20-20 with minor elements (Helena Corp.) about once a
month during the growing season. The measurements were taken at the end of
The plants for the height study were selected from the high density seedling
nursery of 2002. From these seedlings, approximately 36 plants were selected
as dwarfs in the spring of 2003 and were transplanted to fallow ground at one
end of the nursery to keep them from being shaded by taller plants.
The height of these 36 dwarf plants as well as of 36 randomly selected tall
plants from the Stage Two nursery of 2003 was determined. The Stage Two
nursery of 2003 was the group of selected plants (based on their desired
breeding attributes) from the high density nursery of 2002 after the unselected
plants were removed.
A one side t-test for two populations (dwarf vs. normal) was conducted to
determine if there were statistical differences between the two populations for the
traits studied. Analyses of variance and multiple comparisons were also
conducted. Tukey's Wprocedure with alpha=0.05 was performed to determine if
there were statistical differences between the means of each clone for the
The average lengths of three-internode stem segments were determined for
12 clones: six dwarf (00-08, 00-266, 03-105, 03-112, 03-115 and 03-118) and six
normal ('Emerald', 'Jewel', 00-204, 00-206, 00-59, 98-325). The three internodes
measured started at the fourth node counting from the tip of a randomly selected
stem and ending at the seventh node from the tip. The fourth internode was
selected as the starting point to avoid measuring internodes that were still
elongating. Internode length was measured for ten stem segments of each
The leaf area (LA) of dwarf and normal types was estimated by measuring
leaf length (L) and leaf width (W), then calculating the area using the formula for
a rhombus (LA= W L / 2). Five leaves were measured for each of 18 clones,
nine dwarf (00-08, 00-266, 03-105, 03-112, 03-114, 03-115, 03-116, 03-117, 03-
118) and nine normal type ('Emerald', 'Jewel', 95-174, 97-118, 98-325, 00-59,
00-116, 00-206, 00-204). The measured leaves were mature and picked at
Height of the plant or length of the longest shoot
The height of the plants was measured from the soil level to the tip of the
Logistic Regression Analysis
Plants used in this study originated from seed sown in pots of peat in June
2003. In August, the seeds were germinated in a controlled temperature
chamber at about 10oC with continuous illumination. After germination was at
about 50%, the seedlings were moved to a greenhouse. They were transplanted
at 2cm by 2cm spacing to trays of peat in September 2003. They were grown in
a greenhouse until May 2004, being watered daily by hand and fertilized every 3
weeks with Tracite 20-20-20. At the end of May, when the plants were about 10
months old, they were visually classified into dwarf and normal phenotypes.
Measurements were taken for each class as described below.
Categorical data analysis was conducted using SAS (the SAS System V.9),
to model the log of the probability that the plant was dwarf given the predictor
parameters: length of the longest shoot (measured as described previously) and
branching (the number of sprouts or branches on the longest shoot, divided by
the length of that shoot measured in cm), by a multiple logistic regression
analysis. Both predictors were treated as categorical variables. Scores were
assigned to each predictor category, and backward elimination of predictors was
conducted to select the most appropriate model as described by Agresti (1996).
(For detailed information on logistic regressions see Appendix).
The benefit of this analysis is that it allows the study of various parameters
(i.e. length of the longest shoot and branching) as well as their interaction, as
predictors for a categorical response (i.e. dwarf vs. normal blueberries). It was
noticed that dwarf blueberries had high sprouting and low stature. Both
parameters were predictors of dwarf blueberry plants. These parameters were
also preferred over others, because they were the easiest to measure.
Field observations from crosses of the University of Florida blueberry
breeding program were made in the fall of 2002, 2003 and 2004 to study the
inheritance of dwarfness. Controlled crosses between selected dwarf plants and
normal plants were carried out in a greenhouse in the fall of 2002 and were
evaluated during the summer and fall of 2004. The evaluation consisted of
classifying the progeny plants as either normal or dwarf by their physical
appearance (the normal being taller and with normal branching, while the dwarf
smaller and with high branching) to obtain inheritance ratios.
Normal to dwarf ratios from field observations of the breeding program and
from the controlled crosses were analyzed statistically by a chi-square test to see
how well they fit various hypothesized ratios.
The field observations were made in Stage One high density nurseries
planted in 2002, 2003 and 2004. All of the clones used for the crosses were of
normal stature phenotype. Stage One is the first field stage in the blueberry
breeding program before any selection is done. For each nursery, the seeds
were planted in December. The seedlings were transplanted to trays of peat and
grown in a greenhouse until May. Then, they were transplanted to a fumigated
field nursery (Stage One) at a spacing of 15 cm between plants and 45 cm
between rows (about 15 plants per m2). The evaluations were carried out in
October 2002, November 2003 and November 2004, after the plants had been
growing in the field nursery for 10 to 11 months.
Dwarf plants were selected from the 2002 Stage One high density nursery
based on their vigor and attractive architecture, except for 00-266 and 00-08,
which had been selected previously. The plants were prepared for winter
crossing by keeping them in the greenhouse and not allowing them to enter
dormancy. Controlled pollination as described by Galletta (1975) was started in
January and continued until the end of March. Various cross combinations were
tried: dwarf to dwarf (00-266 x 00-08, 00-266 x 03-105, 00-08 x 03-105 and 03-
112 x 00-08), dwarf to normal (00-266 x 01-21, 00-266 x 'Emerald', 00-08 x
'Emerald') normal to dwarf (01-21 x 03-112, 'Jewel' x 00-266), normal to normal
(03-120 x 'Southern Belle', 03-54 x 'Santa Fe', 03-73 x 'Jewel', 'Emerald' x
'Sapphire') and self (00-266 and 00-08).
The mature fruits were harvested, and the seeds were extracted following
the method used in the blueberry breeding program. The berries were processed
in a food blender with water for a few seconds, after which most of the seeds
were obtained by washing away the flesh and skin of the berries. The seeds
were then dried at room temperature and stored in a refrigerator at 7oC until they
were sown in pots with peat moss and germinated in a chamber with a
temperature of about 10oC in the summer of 2003. The seedlings started to
germinate in early August, and were then transplanted to trays, 48 seedlings per
tray, in September. A total of 96 seedlings, two trays filled to capacity, were
grown for each controlled cross in the greenhouse. Each cross was evaluated
for dwarf to normal ratios in February, when the plants were about 6 months old
and still growing in greenhouse trays.
Only the dwarf plants were transplanted to the 2004 Stage One high density
nursery. A follow up evaluation was performed in October 2004 to check for
possible short normal plants that could have been erroneously classified.
RESULTS AND DISCUSSION
Multiple Comparison Studies
The studied dwarf plants were short, with a compact look similar to the
descriptions of dwarf blueberries given by Draper et al. (1983), and Garvey and
Lyrene (1987) (Figure 1). Among the dwarf plants observed in the field, leaf size,
plant height and branching were variable, just as these characteristics are
variable among seedling blueberries that are not dwarfs.
Figure 1. The author with dwarf and normal highbush blueberry, May 2005.
The internodes of the dwarfs were significantly shorter than those of normal
plants (P = 0.0001, see Table 1 and 2). The average dwarf internode was
somewhat over half the normal length. Nevertheless, dwarfs 00-08 and 00-266
were not significantly different from normal 00-206, and only dwarf 00-266 was
not significantly different from normal cultivar 'Jewel' (Figure 2).
80 -- Dwarf clones a Normal clones
S 70 -- h
S 601 o 8 9
efg ef a
cd cde Q
30 ac be 8
Figure 2. Dotplot of length of three internodes (mm) by clone. The group means
are indicated by horizontal red lines and each circle represents an
observation. The first 6 plants were considered dwarf and the last six
normal. The letters on top of each clone represent the statistical
grouping according to Tukey's W procedure.
Leaf area was also smaller in the dwarfs compared to normal plants (Table
3 and 4). Nevertheless, leaf areas of dwarf genotypes: 03-114, 03-105, 00-08
and 00-266 were not statistically different from the normal genotypes 98-325, 97-
118, 95-174, 00-59, and 00-204 (Figure 3). Also dwarfs 03-105, 00-08 and 00-
266 were not statistically different from the normal cultivar 'Jewel'.
Table 1. Length of three-internodes (mean i2 SE) of dwarf and normal plants
(average of 10 measurements).
Pedigree Typey Three-internode length (mm)
(Mean i2 SEz)
00-08 D 29.76 2.32
00-266 D 32.24 & 2.18
03-105 D 27.52 & 2.32
03-1 12 D 25. 12 13.01
03-1 15 D 20.45 12.93
03-1 18 D 17.47 *1 .90
Emerald N 53.41 17.80
Jewel N 41 .30 & 2.68
98-325 N 41.51 & 3.45
00-59 N 46. 18 & 4.65
00-206 N 36.99 & 3.82
00-204 N 42.98 & 4.91
' D= Dwarf, N= Normal
z SE= SDI(n1/2)
Table 2. Results of the t-test for the length of three internodes of dwarf plants vs.
Typez N Mean Std.Dev. SE Mean
D 6 25.43 5.62 2.3
N 6 43.73 5.60 2.3
z D= Dwarf, N= Normal
95% Cl for p dwarf type p normal type: (-25.6, -11.0)
Ho: p dwarf type = p normal type; Ha: p dwarf type < p normal
t = -5.65 P-value = 0.0002 DF = 9
Table 3. Individual leaf area (mean 1 2 SE). Average of 5 measurements of
dwarf and normal plants.
Clone TypeY Leaf area (mm )
(Mean i2 SEz)
00-08 D 507.1 1150.2
00-266 D 565.0 &114.0
03-1 05 D 443.91 161.4
03-1 12 D 21 1.9 154.6
03-1 14 D 382.2 195.8
03-1 15 D 21 1.6 162.6
03-1 16 D 199.8 172.6
03-1 17 D 145.8 122.8
03-1 18 D 136.8 143.2
Emerald N 1242.2 &149. 0
Jewel N 780.31170.0
95-1 74 N 672.31133.0
97-1 18 N 620.0 1202.0
98-325 N 683.8 193.0
00-59 N 761.2 &172.4
00-1 16 N 1079.0 & 226.0
00-206 N 953.0 & 330.0
00-204 N 668.6 & 46.0
yD= Dwarf, N= Normal
z SE= SDI(n1/2)
Table 4. Results of the t-test for the leaf area of dwarf plants vs. normal plants
Type NY Meanz Std.Dev. SE Mean
Dwarf 9 312 164 55
Normal 9 829 215 72
Y'N number of observations
z Square mm
95% Cl for p dwarf type p normal type: (-711,-324)
Ho: p dwarf type = p normal type; Ha: p dwarf type < p normal
T = -5.74 P-value = 0.0000 DF = 14
Dwarf clones Normal clones
1500 -( g O
B def T 8 o
bcd 8 cde 0 0 0
abcd a o a d
abcd o e
-g. abc a
O500 o a o
b o cde
a~~ o a ab_ b
Clone I l l ll llll
OOO OOO OO 00
Figure 3. Dotplot of leaf area (mm2) by clone. Group means are indicated by
horizontal red lines and each circle represents an observation. The
letters on top of each clone represent the statistical grouping according
to Tukey's W procedure.
Plant height was significantly different among dwarf and normal plants.
Three year old normal plants were 2.5 times taller than three year old dwarf
plants, both from the 2003 Stage 2 nursery (Figure 4). The analysis of variance
reported an F value of 309.1 with one degree of freedom for plant type (P value=
140.0 -P O Op
O O 0 0O 95% Cl
S120.0 pg OO O O O 124.2 & 6.9
,a 80.0 4O **
S60.0 *. of + + s 95% Cl
S40.0 -~ 4 4 ~* 4 4* 49.8 & 4.8
0.0 r, ar
0 10 20 30 40 oNra
Figure 4. Scatter plot of three year old dwarf and normal plant heights (cm) from
the 2003 Stage 2 nursery. Plant numbers are the sequence in which
the plants were measured.
Logistic Regression Analysis
The total number of plants observed (373) had a large range for both
variables (44.0 cm for shoot length and 18.9 cm for number of branches), which
was due mainly to the differences between dwarf and normal types. Dwarf plants
were shorter, with a mean of 11.8 & 0.5 cm and a 95% confidence interval (CI) of
4.9 to 18.7 cm, whereas the mean length of normal plants was 22.1 & 0.7 cm with
a 95% Cl of 11.1 to 33.0 cm. Dwarf plants also had more sprouts, with a median
of six branches compared to a median of three branches for normal plants.
In order to determine the number of categories to use for each predictor
and to assign their ranges, a scatter plot of shoot length versus branch number
was made (Figure 5). Two categories were assigned for shoot length plants
with shoots less than or equal to sixteen centimeters and plants with shoots
longer than sixteen centimeters, and three categories for branch number plants
that had three branches or fewer, those that had between four and five branches
and those that had more than five branches. The categories were chosen in
such a way that each category would include both dwarf and normal plants so
that the chi square approximation would be valid (Table 5).
0 2 6 8 10 12 14 16 18 20
Figure 5. Scatter plot of height versus number of branches for dwarf and normal
plants 10 month old.
Using these categories, a logistic regression analysis was carried out to
model the odds of a plant being a dwarf (probability of dwarf I probability of
normal plants) and to test the main effects of the two predictors, shoot length and
branch number, as well as to test for interaction effects. For information on
logistic regression analysis see Appendix.
on effects. For information on
logistic regression analysis see Appendix.
Table 5. Frequency of plants in different categories of length and branch number
for each plant type.
Length of tallest shoot (cm)
(total count of shoots) Dwarf plants Normal plants
I 16 >16 116 >16
<4 27 2 22 106
4-5 45 6 6 51
>5 88 10 1 9
First, using the univariate procedure in SAS, the categories were given
scores. For branches, the scores were: 3, 4, and 8, which were the medians for
each group. The scores for the length groups were the means for each group,
11.6 and 22.9. Multiple logistic regression analysis was performed using all
variables, as well as all interactions. A second model was tested using only the
length measurements and number of branches as predictors, ignoring potential
interactions. The models are given below (see Table 6 and 8). The first model
had a good fit with G2=199.89 and df=367. The second model also had a good
fit with G2=199.93 and df-369. The likelihood ratio statistics for the interacting
terms showed there were no significant interactions between length and
branches. The differences in deviances between the two models were small,
0.04 with two degrees of freedom (P>X22 = 0.980). Thus, these differences were
not significant and it was concluded that dropping the interaction terms would
have no effect on the ability of the model to predict the log of the odds of a plant
being a dwarf.
Both predictors, height and branch number, were significant. The likelihood-
ratio test gave probability values <0.0001 for both predictors. Thus Model 2 was
the better model for indicating whether a plant was dwarf or normal.
Table 6. Model 1 and Model 2 equations with logit, odds of dwarf and probability
of dwarf, for Model 2 only.
MODEL1: Logit P(Y=dwarf) = 0.1054 4.0757 [b43] 2.2454 [41b15] + 4.3720  0.1969 [b<=3]*
MODEL 2: Logit P(Y=dwarf) = 0.137 4.127 [b13] 2.299 [41b15] + 4.198 
Logit (Y=dwarf) Branches (b) 53 41b15 Branches (b) > 5
Odds of dwarf
Probabiliy of dwarf
Length (1) 5 16 0.208 2.036 4.335
1.23 7.66 76.32
0.552 0.885 0.987
Length (1) > 16 -3.990 -2.162 0.137
0.02 0.103 1.15
0.018 0.115 0.534
Using the simpler model (Model 2), it was found that the odds of getting a
dwarf plant, if the number of branches was three or fewer, was 1.6% of the odds
of getting a dwarf plant when the number of branches was greater than five, with
a 95% confidence of 0.5% to 4.9%. Furthermore, the odds of a dwarf plant with
four or five branches was 10% the odds of a dwarf plant with more than five
branches (95% Cl 3.5% to 28.5%). The odds of getting a dwarf with three or
fewer branches was 16% the odds of having a dwarf with four or five branches.
Finally, a plant was 66.5 times (95% Cl 28.2 to 157.1i) more likely to be a dwarf if
its height was less than 16 cm than if its height was more than 16 cm.
Table 6 shows that the probability that a plant is dwarf increased as branch
count increased for a given length category, whereas the estimated probability of
dwarf types decreased when plant height was taller than sixteen centimeters for
a given branch category.
During the analysis of this model it was realized that the odds of a dwarf
with fewer than three branches was 1.6% the odds of a dwarf with more than five
branches. This number seemed rather small. The odds of getting a dwarf with
four and five branches was only 10% the odds of getting dwarfs with more than
five branches. It was hypothesized that a simpler model would fit, using only two
categories for branches those with more than five branches and those with five
or fewer branches. This model was labeled Model 3. However, Model 3 fit the
data poorly. The deviance of this model was 20. 13 with only three degrees of
freedom and various large residuals, which illustrates the poor fit of this model to
Going back to the second model, the residuals were studied. There were
three observations with large residuals, which may have had an effect on the
original model. As seen in Table 7 and Table 8, after removing these outliers,
the model was again fitted one with interaction terms (Model 4) and one without
interaction terms (Model 5). The observations with the large residuals were two
relatively tall plants that had very few branches, which had been classified as
dwarfs. From the previous analysis of these data, observations like that seemed
very unlikely. Thus, these strong outliers may have shifted the model. The third
outlier was an observation of a plant that had classified as normal but was short
and had numerous branches. After fitting the data to the remaining observations
it was again found through the likelihood-ratio test that the interacting terms were
insignificant and could be dropped from the model (G2 Of model without
interacting terms = 173.30; G2 Of model with interacting terms =169.01, thus the
difference = 4.29 with 2 degrees of freedom and a P>X22= 0.117). The deviance
of Model 5, the model that deleted the three outliers and did not include of
interactions, was equal to 173.30 with 366 degrees of freedom, and the model
was declared to be a good fit for the data.
Using Model 5, the odds of getting a dwarf with three or fewer branches
was 0.74% the odds of getting a dwarf with more than five branches (95% Cl:
0.2%, 2.8%). The probability of getting a dwarf with four or five branches was
6.8% the odds of getting a dwarf with more than five branches (95% Cl: 6.0%,
7.6%). The odds of getting a dwarf with three or fewer branches were 10.9% the
odds of getting a dwarf with four or five branches, similar to the results from
Table 7. Model 4 and Model 5 equations with logit, odds of dwarf and probability
of dwarf, for Model 5 only.
MODEL 4: Logit P(Y=dwarf) = 0.1054 28.4709 [b13] 2.2454 [41b15] + 28.2597
[I516]+0.3106[b13]* [1i16] 24.1 048[411b55]*[I116]
MODEL 5: Logit P (Y=dwarf) = 0.266 4.911 [b13] 2.69 [41b15] + 4. 81 
Logit (Y=dwarf) Branches (b) 53 41b15 Branches (b) > 5
Odds of dwarf
Probability of dwarf
Length (1) 5 16 0.125 2.346 5.036
1.133 10.444 153.853
0.531 0.913 0.994
Length (1) > 16 -4.685 -2.464 0.226
0.009 0.085 1.254
0.009 0.078 0.556
The odds of getting a dwarf with shoot length equal to or less than sixteen
centimeters was 122.73 times the odds of getting a dwarf with shoot length
greater than sixteen centimeters (95% Cl: 42.68, 353.2). For both predictors,
removing the outliers gave stronger evidence that the higher the number of
branches and the shorter the plants, the higher the probability of a dwarf
phenotype, supporting the field observation that dwarf plants are short and with a
higher than normal number of sprouts.
This model has an outcome similar to that of Model 2, but Model 5 showed
slightly stronger evidence for what has been observed in the field. Both models
had high deviances, showing both are good fits for the data. However, Model 5
has a slightly stronger deviance, as such; the p-value for the intercept of the
model is smaller than in Model 2. Thus, Model 5 is a slightly better fit for this
data. More importantly, there are no significant residuals when Model 5 is
applied to the data, omitting the three outliers. Therefore the best fit for the data
is Model 5.
Table 8. Results of fitting five logistic regression models to the dwarf data
Model Deviance DF P> X2 Models Difference P>X2
(G2) Fit of com pared
1 199.89 367 1
2 199.93 369 1 (2) -(1) 0.04 (df-2) 0.980
3 20.13 3 0.0002
4 169.01 364 1
5 173.30 366 1 (5) -(4) 4.29 (df-2) 0.117
Not a good fit for the model
It has been shown through categorical analysis that length of the longest
shoot and number of branches, when used together, are good predictors of
dwarfs in blueberry plants. More specifically, analysis of the data has shown that
a plant with more than five branches whose longest shoot is less than sixteen
centimeters long has a 99.4 % probability of being dwarf. Conversely, the
probability of having a dwarf plant with fewer than three branches and measuring
more than sixteen centimeters is 0.9%.
These results are supported by previous observations suggesting that low
or lack of apical dominance, which causes high branching, is associated with
dwarfness in Rubus sturdy dwarf (Keep, 1969) and in highbush blueberry,
Vaccinium corymbosum (Draper et al., 1984).
In the fall of 2002, 2003 and 2004, the high density plots of the blueberry
breeding program at the University of Florida's Plant Science Research and
Education Unit, in Citra, Florida, were studied for dwarf plants.
Each of these three plots consisted of seedlings from about 150 crosses,
with 90 seedlings per cross. The parents for each plant consisted of about 200
different southern highbush cultivars and advanced selections and the parents
differed for each plot. The parents were all highly heterozygous, and the
seedling populations were segregating for many characteristics.
Twenty-five crosses that were segregating dwarfs were identified, and
segregation ratios were determined for each cross. Each progeny population
was examined for fit to a 3:1 and 11:1 normal:dwarf segregation ratio. The
rationale for testing these particular ratios is given below. The dwarf-segregating
populations studied could be classified into two groups. The first seven crosses
in Table 9 fit an inheritance ratio of 3:1 fairly well. The last 18 fit a ratio of 11:1.
Table 9. Segregation ratios and chi-square tests for normal to dwarf ratios (3: 1
and 11:1) for dwarf-segregating normal stature phenotype crosses
from the 2002, 2003
high density plots.
X21 P value
0. 11 0. 44
0. 14 0.712
2. 18 0. 140
2. 18 0. 140
01-21 x 96-32
01-20 x Nui
00-43 x S.Belle
Windsor x 98-336
02-38 x 98-325
98-405 x 95-115
NC 2925 x 03-124
97-130 x 93-204
97-61 x 99-220
98-18 x 97-390
01-64 x 97-142
02-20 x 00-61
01-129 x 97-41
02-69 x 00-14
03-47 x S.Belle
03-61 x 90-4
03-01 x S.Belle
03-12 x Emerald
98-406 x Jewel
03-50 x 03-126
02-86 x Sapphire
Sapphire x 95-209-B
Jewel x 02-22
S.Belle x 00-206
03-103 x Santa Fe
Y AAaa x AAaa and the reciprocal cross
z AAaa x AAAa and the reciprocal cross
The dwarf phenotype seems to be a recessive trait with monogenic
inheritance. Since crosses between two normal plants segregated dwarf plants,
the dwarf genotype cannot include the triplex form (AAAa) because this would
imply that one of the parents was a dwarf which was not the case. This limits the
possibilities of dwarfs segregating from normal plants to the duplex (AAaa) and
simplex (Aaaa) forms, because the nulliplex (aaaa) form cannot occur in the
progeny of a triplex parent that shows chromosome segregation preferentially
over random chromatid and maximum equational segregations.
If the duplex genotype is dwarf, then the only combination of normal-
statured parents that would segregate dwarfs would be that of crossing two
triplex (AAAa). This would produce the duplex genotype of a dwarf with a 0.25
frequency, giving a 3: 1 normal to dwarf ratio (Table 10). The fact that 18 crosses
did not follow a 3: 1 frequency implies that the duplex genotype has a normal
instead of dwarf phenotype.
Since duplex plants have normal stature, the nulliplex form is possible for a
dwarf segregating from a cross between two normal plants. The genotypic
combinations of normal phenotypes that could produce dwarfs are as follows:
AAAa x AAaa, AAaa x AAaa and the reciprocal crosses.
Dwarfs produced by crossing a triplex with a duplex will be simplex and
would be expected in a ratio of 11:1 normal to dwarf phenotype. A duplex times
another duplex can produce dwarfs in a 3: 1 ratio, the possible genotypes for
dwarf being the simplex and nulliplex at frequencies of 0.22 (8/36) and 0.03
(1/36) respectively (Table 10).
To further test this hypothesis, two dwarf plants (00-266 and 00-08) were
self-pollinated, intercrossed with two other dwarf selections (03-105 and 03-112)
and backcrossed to normal types 'Emerald', 'Jewel' and 01-21 (Table 11 and
Table 12). See Table 14 for a cross table of all the controlled crosses evaluated
in this study.
The two dwarfs that were self-pollinated (00-266 and 00-08) segregated
some normal type plants, indicating that they are not nulliplex and are probably
simplex (Aaaa) for the normal allele. The 1:3 normal to dwarf segregation did not
fit very well when the two clones were self pollinated, so the possibility that the
nulliplex is lethal was tested. For this case a 1:2 segregation ratio was expected,
and both 00-266 and 00-08 had low chi-square values for this ratio, 0.09 and
0.01 respectively, with very high probabilities (Table 11).
Table 10. Tetrasomic inheritance frequencies following chromosomal
segregation. In italics, normal to dwarf segregation ratios when dwarf
is either a simplex or nulliplex. In parenthesis, normal to dwarf
segregation ratios when the nulliplex is lethal.
AAAA x aaan
AAAA x Aaan
AAAA x AAna
AAAA x AAAa
AAAa x aaan
AAAa x Aaan
AAAa x AAna
AAAa x AAAa
AAna x aaan
AAna x Aaan
AAna x AAna
Aaan x aaan
Aaan x Aaan
The other two dwarfs studied (03-105 and 03-112) also appeared to be
simplex because they segregated normal types when crossed with the putative
simplex 00-266 and 00-08. Nevertheless, neither of these dwarf to dwarf crosses
followed the expected ratios 1:3 or 1:2 (Table 11). 03-105, when crossed with
00-266 and 00-08, produced 29% and 40% of the expected number of dwarf
plants expected for a 1:3 segregation ratio. 03-112 also segregated more normal
plants than expected. For each of these three cases, the chi-square statistics
clearly indicate that they fit neither a 1:3 ratio nor a 1:2.
Table 11. Segregation ratios and chi-square test for normal to dwarf ratios in
dwarf x dwarf controlled crosses.
Pedigree Normal Dwarf N/D 1:3y 1:2z
(N) (D) X2 P. value X2 P. value
00-266 00-266 30 56 0.54 4.48 0.034 0.09 0. 760
00-08 00-08 32 63 0.51 3.82 0.051 0.01 0. 942
00-08 03-1 05 43 51 0.84 73.8 0.000 38.00 0. 000
00-266 03-1 05 50 43 1.16 38.0 0.000 15.70 0. 000
03-112 00-08 42 48 0.88 22.5 0.000 7.20 0. 007
00-266 00-08 40 54 0.74 18.0 0.000 4.69 0. 030
SAaaa x Aaaa
z Aaaa x Aaaa when the nulliplex is lethal
The cross between putative simplex plants 00-266 and 00-08 also
segregated more than the expected number of normal plants for a 1:3 ratio (three
times the expected count for normal phenotypes). The ratio of normal to dwarf
did not fit a 1:2 ratio, with X2= 4.69 and a probability of 0.030, so the genotypes
of these selections is undetermined.
The crosses between dwarf and normal plants gave ratios that could be
more easily explained than the crosses described above. As seen in Table 12,
for the crosses of dwarf x normal, all but one cross, 01-21 x 03-112, gave
seedlings that fit a 3:1 ratio, which was the expected for a simplex times a triplex
(01-21 was a putative duplex based on the high-density plot ratios, see Table 9).
01-21 x 03-112, was expected to have more dwarfs (i.e. 50% of the total
population) than the 39% observed. The chi-square test indicated that the
segregation ratio in this cross was a poor fit to a 1:1 segregation (P= 0.025).
Possibly, as was speculated before, the nulliplex form is lethal. In this case the
expected segregation would be 6:5. The chi-square test for a 6:5 segregation
supports this speculation ( y=, 1.85) with P=0.174. The contradiction is that 00-
266 x 01-21 (a putative simplex times a putative duplex) fit a 3:1 ratio and not the
expected 6:5 if indeed the nulliplex is lethal or a 1:1 otherwise.
Table 12. Segregation ratios and chi-square test for normal to dwarf ratios in
dwarf x normal controlled crosses.
Pedigree Normal Dwarf N/D 3:1x 1:1'
(N) (D) X2 P. value X2 P. value
00-266 (D) 01-21 (N) 63 24 2.63 0.31 0.577 17.50 0.000
01-21 (N) 03-112 (D) 59 37 1.59 9.39 0.002 5.04 0.025
Jewel (N) 00-266 (D) 75 19 3.95 1.15 0.284 33.40 0.000
00-266 (D) Emerald (N) 72 19 3.79 0.82 0.364 30.90 0.000
00-08 (D) Emerald (N) 78 18 4.33 2.00 0. 157 37.50 0.000
x Aaaa x AAAa
Y Aaaa x AAaa
Table 13. Segregation ratios and chi-square test for normal to dwarf ratios in
normal x normal controlled crosses.
Pedigree Normal Dwarf N/D 3:1x 11:1'
(N) (D) X21 P. value X21 P. value
03-1 20 S. Belle 90 6 15.00 18.00 0.000 0.55 0.460
03-54 Santa Fe 90 5 18.00 19.70 0.000 1.17 0.279
03-73 Jewel 86 10 8.60 10.90 0.001 0.55 0.460
Emerald Sapphire 83 12 6.92 7.75 0.005 2.30 0. 130
x AAaa x AAaa
Y AAAa x AAaa
In the normal x normal crosses (see Table 13), some dwarfs were
observed. Since the genotype of 'Jewel' and 'Emerald' is triplex, the genotype of
the plants crossed with them can be determined. In the case of 03-73 crossed
with the triplex 'Jewel', the progeny fit an 11:1 ratio, indicating that 03-73 is
duplex. 'Sapphire' is also duplex because it also fits an 11:1 ratio when crossed
with Emerald. The cross 03-120 x 'Southern Belle' also followed an 11:1 ratio,
'Southern Belle' being the duplex and 03-120 the triplex.
Table 14. Cross table of all the controlled crosses per genotype evaluated in this
Aaaa AAaa AAAa AA-
Aaaa 00-266 x 00-266 00-266 x 01-21 00-266 x Emerald
00-08 x 00-08 00-08 x Emerald
00-08 x 03-105
00-266 x 03-105
03-112 x 00-08
00-266 x 00-08
AAaa 01-21 x 03-112 03-73 x Jewel
AAAa Jewel x 00-266 Emerald x Sapphire
03-120 x S.Belle
AA 03-54 x Santa Fe
The results found have a few contradictions, in that some of the crosses
among the putative simplex dwarfs did not follow the expected ratio (Table 11),
and the dwarf x normal cross 00-266 x 01-21 did not follow the expected 6:5 ratio
(Table 12). For the cross 01-21 with the dwarf 03-112, the expected 6:5 ratio
It appears that the nulliplex is indeed lethal, but this could not be definitively
proved because of the aforementioned contradictions. It also appears likely that
the duplex form has a normal phenotype; the dwarf phenotypes observed are
Draper et al. (1984) and Garvey and Lyrene (1987) mentioned in their work
that the inheritance of dwarfism was complex, and they attributed its complexity
to multiple genes. The analysis of the data presented in this research suggests a
simpler inheritance. For most of the cases it supports monogenic inheritance
with tetrasomic segregation. The few contradictions are probably due to
heterozygosity at other loci, and to the fact that the blueberry germplasm studied
was the product of several interspecific crosses that involved lowbush
blueberries, V. darrowi (the most likely source for the dwarf genes). The
possibility of aneuploidy as a cause for dwarfness has been mentioned in the
literature, but in the case of the controlled crosses involving the dwarfs: 00-08,
00-266, 03-112 and 03-105, the fertility was normal, suggesting euploidy.
Internode length, leaf area and plant height were significantly different
between dwarf and normal plants. The average internode length of normal
plants was 1.7 times longer than that of dwarf plants. Leaf area was smaller for
dwarfs when compared to normal (37.6% the area of a normal plant). When
seedlings were three years old, the height of normal plants was 2.5 times taller
than the height of dwarf plants.
Dwarf plants were characterized by their high number of branches and low
stature. The probability of a dwarf plant from a six month old population is 99.4%
for plants with more than 5 branches and height less than 16 cm. Conversely,
the probability of a dwarf plant is 0.9% for plants with fewer than three branches
and height exceeding 16 cm.
The fertility of the dwarf plants studied in controlled crosses (clones 00-266,
00-08, 03-112 and 03-105) was normal, indicating that aneuploidy was not the
reason for the dwarf phenotype.
The genotype of the dwarf plants appears to be simplex (Aaaa), with the
nulliplex genotype being lethal. Normal plants that segregated dwarfs when
crossed are either triplex (AAAa) or duplex (AAaa). Triple crossed to duplex
and the reciprocal give 11:1 normal to dwarf ratio. Duplex crossed to duplex
gives a 27:8 normal to dwarf ratio.
CATEGORICAL DATA ANALYSIS
Categorical data is a type of data that is measured on a scale that consists
of a set of categories (i.e. small, medium or large size of clothing; dwarf or
normal height) where only one category applies to each subject.
There could be nominal and ordinal categorical variables. Nominal
variables refer to those variables that have unordered scales like race (black,
white, hispanic, other) and party affiliation (republican, democrat, independent,
other), where the order of listing the categories is irrelevant. Ordinal variables
refer to those variables that have order like size of clothing (small, medium,
large) and height (short, intermediate and tall), and the statistical analysis should
depend on that order. Further, statistical methods designed for ordinal variables
cannot be used for nominal variables, whereas statistical methods for nominal
variables can be used for ordinal variables, but the information about the order is
not used, resulting in the loss of power of the test.
In any breeding program, especially after crosses have been made, the
understanding of the genetic inheritance of particular characters is important and
in some cases necessary to the success of the program.
Inheritance ratios as between dwarf and normal plants can be tested for an
inheritance hypothesis by chi-square (X2) Statistics as proposed by Karl Pearson
in 1900. The object of this test is to see if the observed ratios correspond to the
expected or hypothesized ones (Watts, 1980).
In a multinomial experiment in which each trial can result in one of k
outcomes, the expected number of outcomes of type i in n trials is nf7; where [7; is
the probability that a single trial results in outcome i (Ott and Longnecker, 2001).
As proposed by Karl Pearson in 1900, the following test statistic can be
used to test the specified probabilities: X2= E [(n; E;)21 E; ], where n; represents
the number of trials resulting in outcome i and E; represents the number of trials
expected to result in outcome i when the hypothesized probabilities represent the
actual probabilities assigned to each outcome (Ott and Longnecker, 2001).
If the hypothesized probabilities are correct, the observed cell counts n;
should not deviate greatly from the expected cell count E;, and the computed
value of X2 should be small. Conversely, when one or more of the hypothesized
cell probabilities are incorrect, X2should be large.
The distribution of the X2 value can be approximated by a chi-square
distribution provided that the expected cell counts E; are fairly large. Cochran
(1954) indicates that the approximation should be adequate if no E; is less than
1, and at least 80% of all the Eis are greater than five. Ideally all Eis should be
greater than 5.
The chi-square goodness-of-fit test based on k specified cell probabilities
will have k-1 degrees of freedom (df). For inheritance studies lexica, df equals
the number of observed categories minus one, and the formula for calculating the
chi-square value is: =2 t [(observed ratio expected)2/ 9Xpected]
Logistic Regression Analysis a summary of Agresti's book on
categorical data analysis (1996).
Logistic regression models are a type of General Linear Model (GLM) used
to analyze categorical data when the response variable has only two categories
(binary response). Thus, its distribution is specified by probabilities of success
P(Y=1) and of failure P(Y=0) with binomial distribution.
Because the relation between the probability of success and the
observations are usually nonlinear, logistic models are better than linear models.
For instance, a fixed change in independent variable X may have less impact
when the probability of success is near 0 or 1 than when it is near the middle; a
good example would be the probability of doubling the yield when adding x
amount of fertilizer in different fertility-type soils (rich, medium, poor). The rich
soils will have a very low probability, close to zero, because the nutrients
available in the soil might be already maximizing the yield potential, thus the
fertilizer is not necessary. The poor soils will have high probabilities, close to
one, because the fertilizer amendment will significantly improve the nutrients
available for sustaining double the yield. The steeper change will occur in soils
of medium fertility (the transition between non-likely to respond and always
responding) because their response is more variable, less uniform with some
observations that will double the yield and some that will not. Overall, if the data
are plotted with the fertility-type soil categories (rich, medium, poor) on the x-axis,
and the probability of success on the y-axis, the curve will have an "S" shape.
The most important function having this S-shaped curve has the model
form: log [P(Y=1)/P(Y=0)] = ac + (Sx, where [P(Y=1)/P(Y=0)] is the odds of the
response. In this study, the odds of a plant being dwarf.
This function is called the logistic regression function. The random
component for the determination of success or failure is binomial. The link
function is the logit transformation log [P(Y=1)/P(Y=0)], symbolized by logit
P(Y=1). The logit is the natural parameter of the binomial distribution, and
because of this it is a canonical link.
To determine if a model is fitting the data set well, goodness-of-fit statistics
and residual analysis are useful. When the fitted values are relatively large
(exceeding 5) and the number of settings (categories) is fixed, Pearson (X2) and
likelihood-ratio (G2) gOodness-of-fit statistics have approximate chi-squared
distributions, and the df equals the number of response counts minus the number
of model parameters.
The deviance is the likelihood-ratio statistic for comparing model M to the
saturated model; it is the statistic for testing the hypothesis that all parameters
that are in the saturated model but not in model M equal zero. For the logit
transformation it has the same form as the G2 likelihood-ratio goodness-of fit
statistic for model M.
For two models, where M~o is a special case of M?, given that the more
complex model holds, the likelihood-ratio statistic for testing that the simpler
model holds is: Devianceo Deviance?. For larger samples, this is approximately
a chi-squared statistic, with df equal to the difference between the residual df
values for the separate models. This test works well for comparing two models,
even when the overall goodness-of-fit test is poor for each model.
In multiple logistic regressions, the significance of a predictor can be tested
by the difference in deviance of the model with the predictor and a simpler model
without the predictor, this way the effect of predictors and their interactions can
be tested to determine if they are significant or if they shouldn't be in the model.
A backward elimination consists in testing different models by the previous
method, starting from the most complex model with all the interactions and
moving backward to the simplest model possible.
Once the simplest model has been determined, various interpretations can
be obtained from it. The odds and probabilities are very useful to study and
analyze the data; also confidence intervals (CI) for the odds, and comparisons
among the odds of the two categories, i.e. the odds of a dwarf plant vs. the odds
of a normal plant, can show trends and enhance the study of the data. For
detailed information on how to interpret logistic regressions the book on
categorical data from Agresti (1996) is recommended.
Agresti, A. 1996. Categorical data analysis. Wiley. New York, USA.
Atkinson, C., Else, M. 2001. Understanding how rootstocks dwarf fruit trees. The
compact fruit tree 34 #2: 46-49.
Berhow, M.A. 2000. Effects of early plant growth regulator treatments on
flavonoid levels in grapefruit. Plant Growth Regulation 30: 225-232.
Bindloss, E. 1942. A developmental analysis of cell length as related to stem
length. Am. J. Bot. 29: 179-188.
Borner, A., Korzun, V., Malyshev, S., Ivandic, V., Graner, A. 1999. Molecular
mapping of two dwarfing genes differing in their GA response on
chromosome 2H of barley. Theor. Appl. Genet. 99: 670-675.
Brightwell, W.T. 1971. Rabbiteve blueberries. Ga. Agric. Exp. Sta. Bul. 100.
Clayton, H.G. 1925. Florida making a good beginning with blueberries. Proc. Fla.
State Hort. Soc. 38: 200-201.
Cochran, W. 1954. Some methods for strengthening the common X2 test.
Biometrics 10: 417-451.
Coville, F.V. 1937. Improving the wild blueberry. USDA Yearbook of Agriculture
Curtis, I.S., Ward, D.A., Thomas, S.G., Phillips, A.L., Davey, M.R., Power, J.B.,
Lowe, K.C., Croker, S.J., Lewis, M.J., Magness, S.L. 2000. Induction of
dwarfism in transgenic Solanum dulcamara by over expression of a
gibberellin 20 oxidase cDNA from pumpkin. The Plant Journal: For Cell and
Molecular Biology 23: 329-338.
Darrow, G.M., 1949. Polyploidy in fruit improvement. Proc. Am. Soc. Hort. Sci.
Draper, A.D., Chandler, C.K., Galletta, G.J. 1984. Dwarfed plants in some
blueberry (Vaccinium) seedling populations. Acta Hort. 146: 63-68.
Draper, A.D., Hancock, J. 2003. Florida 4B: Native blueberry with exceptional
breeding value. J. Am. Pom. Soc. 57(4): 138-141.
Edin, M., Garcin, A., Lichou, J., Jourdain, J.M. 1996. Influence of dwarfing cherry
rootstocks on fruit production. Acta Hort. 410: 239-246.
Fideghelli, C., Sartori, A., Grassi, F. 2003. Fruit tree size architecture. Acta Hort.
Franken-Bembenek, S. 1996. Cherry hybrid rootstocks developed at giessen:
new experiences and evaluation. Acta Hort. (ISHS) 410: 377-384.
Gale, M.D., Youssefian, S. 1985. Dwarfing genes in wheat. In: Progress in Plant
Breeding, Ed. by Russell G.E., Butterworths, London, p: 1-35.
Galletta, G.J. 1975. Blueberries and Cranberries. In: J. Janick and J.N. Moore
(eds.). Advances in Fruit Breeding. Purdue Univ. Press, West Lafayette,
Indiana. Pp 154-196.
Garvey, E.J. 1985. Dwarfism, self-incompatibil ity, and female sterility in
Vaccinium ashei (Reade). PhD Diss. University of Florida.
Garvey, E.J., Lyrene, P.M. 1987. Inheritance of compact growth habit in
rabbiteve blueberry. J. Am. Soc. Hort. Sci. 112: 1004-1008.
Goldman, I.L., Watson, J.F. 1997. Inheritance and phenocopy of a red beet
gibberellic acid deficient dwarf mutant. J. Am. Soc. Hort. Sci. 122: 315-318.
Hansche, P.E., Beres, W., Darnell, R. 1986. Yield potential of peach trees
dwarfed by the dw gene. Hortscience 21:1452-1453.
Hedden, P. 2003. The genes of the Green Revolution. Trends in Genetics 19: 5-
James, C.B. 1924. Blueberries in Northwest Florida. Proc. Fla. State Hort. Soc.
Jennings, D.L. 1967. Balanced lethals and polymorphism in Rubus idaeus.
Heredity 2: 465-479.
Johnson, W.C., Aldwinckle, H.S., Cummins, J.N., Forsline, P.L., Holleran, H.T.,
Norelli, J.L., Robinson, T.L. 2001. The new usda-ars/cornell university apple
rootstock breeding and evaluation program. Acta Hort. (ISHS) 557: 35-40.
Keep, E. 1969. Dwarfing in the raspberry Rubus idaeus L. Euphytica 18:256-276.
Khurshid, T., McNeil, D.L., Trought, M.C.T. 1997. Effect of foliar-applied
gibberellins and soil-applied paclobutrazol on reproductive and vegetative
growth of 'Braeburn' apple trees growing under a high-density planting
system. N. Z. J. Crop. Hort. Sci. 25: 49-58.
Knight, R.J., Scott D.H. 1964. Effects of temperatures on self- and cross-
pollination and fruiting of four highbush blueberry varieties. Proc. Amer.
Soc. Hort. Sci. 85:302-306.
Ladizinsky, G. 1997. Dwarfing genes in the genus Lens Mill. Theor. Appl. Genet.
Lapins, K.O., Watkins, R. 1973. Genetics of compact growth habit. Rep. E.
Malling Res. Stn. for 1972: 136.
Li, J., Chory, J. 1999. Brassinosteroid actions in plants. J. Exp. Bot. 50: 275-282.
Little, C.H.A., MacDonald, J.E. 2003. Effects of exogenous gibberellin and auxin
on shoot elongation and vegetative bud development in seedlings of Pinus
sylvestris and Picea glauca. Tree Physiol. 23: 73-83.
Lyrene, P.M., Sherman, W.B. 1977. Breeding blueberries for Florida:
accomplishments and goals. Proc. Fla. State Hort. Soc. 90: 215-217.
Lyrene, P.M., Sherman, W.B. 1979. The rabbiteve blueberry industry in Florida -
1887 to 1930 with notes on the current status of abandoned plantations.
Econ. Bot. 33: 237-243.
Lyrene, P.M., Sherman, W.B. 1983. Mitotic instability and 2n gamete production
in Vaccinium corymbosum x V elliottii hybrids. J. Am. Soc. Hort. Sci. 108:
Lyrene, P.M., Sherman, W.B. 1984. Breeding early-ripening blueberries for
Florida. Proc. Fla. State Hort. Soc. 97: 322-325.
Lyrene, P.M., Sherman, W.B. 1985. Breeding blueberry cultivars for the central
Florida Peninsula. Proc. Fla. State Hort. Soc. 98: 158-162.
Lyrene, P.M., Vorsa, N., Ballington, J.R. 2003. Polyploidy and sexual
polyploidization in the genus Vaccinium. Euphytica 133: 27-36.
Mackenzie-Hose, A.K., Sherriff, L.J., Ross, J.J., Reid, J.B. 1998. Internode length
in Pisum. The Irs mutation reduces gibberellin response and levels. Physiol.
Plantarum 103: 485-490.
Moore, J. N. 1966. Breeding. In: P. Eck and N. Childers (eds.). Blueberry culture.
Rutgers University, Piscataway, New Jersey, USA.
Mowry, H., Camp, A. F. 1928. Blueberry culture in Florida. Fla. Agr. Exp. Sta.
Ott, R.L., Longnecker, M. 2001. An introduction to statistical methods and data
analysis. Duxbury, Thomson learning. Pacific Grove, California, USA.
Ross, J.J., Murfet, I.C., Reid, J.B. 1997. Gibberellin mutants. Physiol. Plantarum
Sachs, T., Thimann, K.V. 1967. The role of auxins and cytokinins in the release
of buds from dominance. Am. J. Bot. 54: 136-144.
Schaller, H. 2003. The role of sterols in plant growth and development. Prog.
Lipid Res. 42: 163-175.
Sharpe, R.H. 1954. Horticultural development of Florida blueberries. Proc. Fla.
State Hort. Soc. 66: 188-190.
Sharpe, R.H., Darrow, G.M. 1959. Breeding blueberries for the Florida climate.
Proc. Fla. State Hort. Soc. 72:308-311.
Sharpe, R.H., Sherman, W.B. 1971. Breeding blueberries for low chilling
requirement. Hortscience 6: 145-147.
Sommer, H.E., Wetzstein, H.Y., Brown, C.L. 1999. Shoot growth and histological
response of dwarf sweetgum to gibberellin. Journal of Horticultural Science
& Biotechnology 74: 618-621.
Wareing, P.F., Phillips, I.D. 1978. The control of growth and differentiation in
plants. 2nd ed. Pergamon Press, New York.
Watts, L. 1980. Flower & vegetable plant breeding. Grower Books, London,
Yang, T., Davies, P.J., Reid, J.B. 1996. Genetic dissection of the relative roles of
auxin and gibberellin in the regulation of stem elongation in intact light-
grown peas. Plant Physiol. 110: 1029-1034.
Yang, T., Law, D.M., Davies, P.J. 1993. Magnitude and kinetics of stem
elongation induced by exogenous indole-3-acetic acid in intact light-grown
pea seedlings. Plant Physiol. 102: 717-724.
Yin, Y., Cheong, H., Friedrichsen, D., Zhao, Y., Hu, J., Mora-Garcia, S., Chory, J.
2002. A crucial role for the putative Arabidopsis topoisomerase VI in plant
growth and development. Proc. Natl. Acad. Sci. USA 99: 10191-10196.
David Humberto Baquerizo, born in Guayaquil, Ecuador, in 1978, is the
oldest son of Tito and Meche Baquerizo the parents of six children. From a
young age he was delighted by nature. His grandparents "Maruja" and Angel
Zambrano, a humble couple from the small town of Manta with a love for
agriculture and farm-life, taught him to appreciate simple rural living.
David learned to see nature as a sign of God's love towards men as he was
taught by the Salesians and Jesuits at the schools he attended. This inspired
David to venture into studying horticulture.
In 1996 he finished high school at Colegio Javier and moved to Costa Rica
to study agriculture at Escuela de Agricultura de la Regio~n Tropical Humeda
(EARTH College) located in the heart of the humid tropics, in Limon province.
He graduated with honors in December 1999 after four years of living among
howler monkeys and eating "gallo pinto." He returned to Ecuador with his wife
Karen and newborn son David Manuel.
In August 2000, the Baquerizo family emigrated to the US in search of
economic stability, because Ecuador was in the middle of a depression. David
worked in South Florida in a horticultural related business until August 2002
when he started graduate studies at the University of Florida under the guidance
of Dr. Paul Lyrene. Now with four children, the Baquerizo family has become
Gator fans and enjoyed living in the beautiful city of Gainesville.