<%BANNER%>

Genetic Architecture of Fungal Disease Traits in Loblolly Pine


PAGE 1

GENETIC ARCHITECTURE OF FUNGAL DI SEASE TRAITS IN LOBLOLLY PINE By GOGCE CEREN KAYIHAN A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006

PAGE 2

Copyright 2006 by Gogce Ceren Kayihan

PAGE 3

Dedicated to my family and my cats

PAGE 4

iv ACKNOWLEDGMENTS This study could not have become a reality without the hard work of many talented and smart people who funded, designed the projec t and physically worked in the field, so I would like to take this opport unity to thank them all. I w ould not be able to do any of this work without the help from my adviso rs Dr. Timothy L. White, Dr. John M. Davis, Dr. Dudley A. Huber, Dr. Dana Nelson and Dr. Rongling Wu who were extremely patient, giving and kind. I would like to tha nk Davis lab members who worked with me day/night and always kept thei r sense of humor to get all of us through very hot days in the greenhouse. My family, the Kayihans although thousands of miles away, had never left me alone and endured all my frustration and depression with me, I cant imagine going through this without thei r support. I would also like to mention my lovely cats Kouketsu and Suki who were my only family here for the last 5 years and Shinsetsu, Cakal, late Takashi and Tsubasa who were always in my heart. My friends, who laughed with me, cooked for me, watched weird TV shows with me and made the PhD process much lighter on my shoulders. I also want to take this opportunity to thank my therapists w ho helped me break a lot of waves. Last but certainly not the least I would like to thank Stefan Crynen for all the support he has given during the last very stressful months.

PAGE 5

v TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...........................................................................................................viii LIST OF FIGURES...........................................................................................................ix ABSTRACT.......................................................................................................................xi CHAPTER 1 INTRODUCTION........................................................................................................1 2 GENETIC DISSECTION OF FUSIFO RM RUST AND PITCH CANKER DISEASE TRAITS IN LOBLOLLY PINE..................................................................6 Introduction...................................................................................................................6 Materials and Methods.................................................................................................8 Genetic Material, Plant Propagati on, and Experimental Design...........................8 Pitch Canker: Inoculations and Data Collection...................................................9 Fusiform Rust: Inoculati ons and Data Collection...............................................11 Estimation of Genetic Parameters.......................................................................12 Genetic Correlations............................................................................................14 Results........................................................................................................................ .15 Pitch Canker Disease Resistance is Heritable.....................................................15 Two Distinct Inoculation Pr ocedures Reveal Similar Heritabilities for Lesion Length..............................................................................................................17 Disease Traits Associated with Fusifo rm Rust are Independently Inherited......17 Host Genes Underlying Resistance to P itch Canker and Fusiform Rust are Independent......................................................................................................22 Efficiency of Using Multiple Ramets per Genotype...........................................22 Discussion...................................................................................................................24 Genetic variation for pitch canker resistance......................................................24 Gall score and gall length are the most heritable fusiform rust traits..................26 Resistance to pitch canker and fusiform rust are under the control of two different mechanisms.......................................................................................28 Phenotyping for disease trait dissection in l oblolly pine.....................................29

PAGE 6

vi 3 FUSIFORM RUST RESISTANCE COSEGREGATES WITH AN FR1 -LINKED MARKER AND REVEALS VARIABLE PENETRANCE OF THE DISEASE PHENOTYPE.............................................................................................................30 Introduction.................................................................................................................30 Materials and Methods...............................................................................................32 Genetic Material..................................................................................................32 Genotyping Families 0 and 1 for Fr1 ..................................................................32 Greenhouse screen...............................................................................................33 Field.....................................................................................................................35 Data Analysis.......................................................................................................35 Genetic Correlations............................................................................................38 Asymptotic Z-test................................................................................................39 Results........................................................................................................................ .40 Inheritance of Fusiform Rust Resi stance in the Greenhouse and Field...............40 Validation of Fr1 marker.....................................................................................41 The Genetic Basis for Escape Rate..................................................................44 Discussion...................................................................................................................47 Agreement Among Greenhouse and Field Screens.............................................47 Marker-trait Cosegregation for Fusiform Rust Disease Resistance....................49 Penetrance of the Fusiform Rust Disease Phenotype..........................................50 Pathogen Infection in an Ecol ogically Relevant Setting.....................................52 4 TRANSCRIPT PROFILING REVEALS POTENTIAL MECHANISMS OF FUSIFORM RUST DISEASE DEPENDENT SHIFTS IN PINE STEM DEVELOPMENT.......................................................................................................54 Introduction.................................................................................................................54 Materials and Methods...............................................................................................55 Plant Material, Genotyping and Harvesting........................................................55 Fungal Material and Inoculation.........................................................................56 Microarray...........................................................................................................56 Statistical Analysis..............................................................................................57 Results........................................................................................................................ .60 Discussion...................................................................................................................66 Transcription Profiling Reveals Differential Gene Expression...........................67 Influences of Fusiform Rust Dis ease Development on Gene Profiles.........69 5 CONCLUSION...........................................................................................................73 APPENDIX A SAS SCRIPTS FOR MICROARRAY ANALYSIS..................................................75 B ASREML SCRIPT FOR ASYMTOTIC Z-TEST......................................................78 C HEALTHY VS. DISEASED GENE LIST.................................................................79

PAGE 7

vii D GENES THAT ARE REGULATED ACROSS TIME..............................................89 E MICROARRAY PROCEDURE..............................................................................108 Indirect Incorporation of Cy Dyes............................................................................108 Hybridization............................................................................................................112 LIST OF REFERENCES.................................................................................................116

PAGE 8

viii LIST OF TABLES Table page 2-1 Summary of the four inoculation experiments reported in this study......................10 3-1 Summary of the greenhous e and field screens report ed in this study. The 63 families and most of the clones screened were the same across the ten gall, one gall and field screens................................................................................................33 3-2 Summary of score (disease incidence) and escape rate datasets along with narrow sense heritabilities (h2) and broad sense heritability (H2) for escape rate and score in ten-gall, one-gall and field fusiform rust screens.................................40 3-3 Segregation of marker J7_485A linked to Fr1 gene in families 0 and 1 across ten-gall, one-gall and field screens (658 ramets combined) with disease phenotype. Parent number 17 is heteroz ygous for pathotype-specific resistance gene Fr1 ...................................................................................................................43

PAGE 9

ix LIST OF FIGURES Figure page 2-1 A circular mating design was used to generate the plant material. Thirty-two parents were crossed following a circul ar design, and the resulting progeny was used as the material screened for this study...............................................................9 2-2 Frequency distributions and genetic correlation fo r pitch canker lesion length.......16 2-3 Heritability estimates fo r pitch canker lesion lengths..............................................18 2-4 Frequency distributions a nd genetic correlations for fusi form rust disease traits....19 2-5 Heritability estimates and family rank scatter plots for fusiform rust disease traits......................................................................................................................... .21 2-6 No genetic correlation between pitch canke r and fusiform rust resistance. Family rankrank scatter plot based on predicte d family means for pitch canker and fusiform rust, fitted with a least squares regression line..........................................23 2-7 Efficiency is inversely proportional to the number of ramets per genotype............23 3-1 The inoculum sources used in the tengall and one-gall greenhouse trials and field screens mapped in Flor ida and Georgia along with the other areas that were assessed for virulence...............................................................................................34 3-2 Diagrams illustrating genotype (clone) based phenotyping for disease resistance, susceptibility and escape rate...................................................................................36 3-3 Scatter plot of ranks based on BLUP-predi cted family genetic values for ten-gall and field were plotted against each other (a rank of is the most resistant and the most susceptible)........................................................................................41 3-4 Distribution of percentage of galled ramets by clone in the ten-gall (A), one-gall (B) and field (C) screens. There were a total of 1471 genotypes (i.e. clones) in all the experiments and each clone was repl icated 1-5 times in each experiment...45 3-5 Random distribution of fusiform rust disease resistance perf ormance of ramets from clones that had at least one dise ased ramet in Randolph, GA field trial.........46

PAGE 10

x 4-1 Flow chart illustrating the procedure to identify significant and biologically interesting gene expression profiles. ANOVA was performed for each of the 3705 genes................................................................................................................59 4-2 Analyses of mean gene expression da ta support 27 distinct profile groups, A through AA...........................................................................................................61 4-3 Genes with the same expression profil e in all treatment combinations were predominantly induced during the first time interval...............................................62 4-4 Profile groups can be categorized into biologically interesting clusters with distinct changes in ge ne expression patterns............................................................64 4-5 Clustergram of gene profile differen ces (or contrasts) between diseased and healthy treatments....................................................................................................65

PAGE 11

xi Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy GENETIC ARCHITECTURE OF FUNGAL DI SEASE TRAITS IN LOBLOLLY PINE By Gogce Ceren Kayihan August 2006 Chair: Timothy L. White Cochair: John M. Davis Major Department: Forest Resources and Conservation In the southeastern United States, loblolly pine ( Pinus taeda L) is the most common tree species covering nearly 13.4 hectares in southern United States with over 1 billion seedlings produced every year. This popular pine species bring $30 billion and 110,000 jobs to the region. However, two endemi c fungal diseases are threatening this productive view: fusiform rust incited by Cronartium quercuum Berk. Miyable ex Shirai f. sp. fusiforme and pitch canker incited by Fusarium circinatum Nirenberg et ODonnell. Loblolly pine is not totally susceptible to these diseases and it has been shown by many researchers, using natural and artificial inoc ulations, that loblolly pine families show genetic variation in resistance to both fusiform rust and pitch canker diseases. Precision was acquired by a combination of clonal propa gation, which allows repeat observations of the same genotypes and the use of a mi xed linear model (GAREML) to adjust for environmental effects. In the first part of this study, I identified traits, clones, families, and parents that guide a genetic approach to dissecting disease traits in loblolly pine. I

PAGE 12

xii verified that pitch canke r and fusiform rust traits are he ritable and identified the disease traits that are genetically distinct from one anothe r. Second, I used DNA marker information that was developed in prev ious mapping studies to distinguish host genotypes that carry/lack the path otype-specific Fr1 allele. I tested the hypothesis that the Fr1 allele is predictive of resistance in greenhouse and field experiments. Because these studies involved clonally propaga ted materials, I also quan tified the extent to which genetic and non-genetic factors influence disease expression le vels and escape rate in greenhouse and field trials. Fina lly, I used gene expression data obtained from a very complex design of microarray experiments us ing diseased and hea lthy loblolly pine clones from a family that is segregating for Fr1 to identify genes that are differentially regulated in diseased and hea lthy individuals. I contrasted ge ne expression in diseased and healthy individuals over a time frame of 4 months. Together, th ese studies revealed the genetic architecture of fusiform rust di sease resistance in scales ranging from the population level to the molecular level.

PAGE 13

1 CHAPTER 1 INTRODUCTION Forests cover one-third of the earths terrestrial surface and provide social, economical and environmental benefits (FAO, 2006). Pine is a dominant plant species in Europe, Asia and America and has been used both as source for forest products and as a model organism to study wood formation (Lev-Yadun and Sederoff, 2000). In the southeastern United Stat es, loblolly pine ( Pinus taeda L) is the most common tree species covering nearly 13.4 hectares in southern United States (Schu ltz, 1999) with over 1 billion seedlings planted every year (McK eand et al., 2003). This popular pine species brings $30 billion and 110,000 jobs to the re gion (Schultz, 1999). In addition, loblolly pine plantation and natural forests offer habi tat for many diverse sp ecies, control erosion, improve water quality, provide recreation and sustain rural communities. As management of loblolly pine planta tions intensive to maximize product yield, new problems started to emerge in producti on of healthy loblolly pine. Among these problems, two endemic fungal diseases attrac ted the attention of many researchers and breeders: fusiform rust (incited by Cronartium quercuum Berk. Miyable ex Shirai f. sp. fusiforme) (Burdsall and Snow, 1977) and pitch canker (incited by Fusarium circinatum Nirenberg et ODonnell) (Nir enberg and ODonnell, 1998). Loss of millions of dollars (Cubbage et al., 2000) pushed br eeders and researchers to inve stigate fusiform rust, one of the most economically destru ctive diseases of the southeas tern United States. It is incited by Cronartium quercuum a biotrophic pathogen that a lternates its life cycle with pine and oak as hosts.

PAGE 14

2 Fusiform shaped galls on pine hosts are the major symptom of fusiform rust. As disease progresses through the years these infections may take the form of sunken cankers. Galls on stems decrease the wood qua lity and sometimes kill the plant (Schmidt, 1998). Both specific resistance, i.e., genefor-gene interactions (Powers, 1980; Stelzer et al., 1997; Wilcox et al., 1996), and partial resi stance in the form of short galls (Schmidt et al., 2000) have been demonstrated for the C. quercuum pine pathosystem. Loblolly pine is not totally susceptible to this dis ease and it has been show n by many researchers, using natural and artificial inoculations, that loblolly pine families show genetic variation in resistance to fusiform rust (Kuhlma n and Powers, 1988; McKeand et al., 1999). The pitch canker disease is not as economi cally important as fusiform rust in southeastern United States although it can dama ge southern pine plantations sporadically in the USA and it is an important problem for Pinus radiata in California (Storer et al., 2002). In the southeastern Unite d States seedling production can be severely hampered by this disease (Dwinell et al., 1985). The pitch canker inciting agent, F. circinatum is a necrotrophic fungus that survives on dead tissues. A successful in fection results in symptoms like resinous lesions on stems and br anches that cause seedling mortality and decreased growth rates and cr own die-back of plantation tr ees (Dwinell et al. 1985). Although genetic variation among loblolly pine families (Kuhlman et al., 1982) and clones (Dwinell and Barrowsbroaddus, 1982) has been detected, the genetic architecture of resistance has not be en thoroughly investigated. Knowing that family level genetic variati on for both diseases exists in the same species, namely, loblolly pine, provided an opportunity to investigate and contrast the nature and architecture of resistance to th e diseases incited by the biotrophic and

PAGE 15

3 necrotrophic fungi. The main difference be tween the two types of fungi is that necrotrophs survive on the d ead plant cells and biotrophs feed on living plant cells (Lewis, 1973). Thus the damage they cause is significantly different and the host defense mechanisms against them may also vary. Fo r example, biotrophic fungi are typically associated with gene-for gene systems (Glazebrook, 2005; Hammond-Kosack and Jones, 1997) and necrotrophs are often linked to qua ntitative disease resistance genes (Oliver and Ipcho, 2004). Thus, I hypothesized that re sistance and responses on the pathogen and host sides would differ for the two distinct pathosystems. Complex traits such as disease resistance can be dissected by two core activities, genotyping and phenotyping. These two different sets of data can be analyzed in two ways: by linkage, which uses QTL or linkage analysis approach; or by association, which uses linkage disequilibrium to make genotype-phenotype as sociation (Jannink et al., 2001). Association genetics is gain ing favor as an approach to identify genes that underlie complex traits (Rafalski, 2002). The a ssociation approach relies on linkage disequilibrium between marker loci and targ et trait loci, and because many unrelated individuals are examined in a single associa tion experiment, it is po ssible to evaluate the effects of many alleles across a broad sample of the population (FlintGarcia et al., 2003). In contrast QTL approaches ev aluate pairs of segregating a lleles typically within single families (Jannink et al., 2001). Loblolly pine is an ideal organism fo r association genetics because loblolly pine has natural and out crossing populations dist ributed across large areas that have high gene-flow and little population substructure (Al-Rababah and Williams, 2002; Brown et al., 2004b; Schmidtling et al., 1999). Also with loblolly pine it is possible to create large experimental populations and clonally propagate them to

PAGE 16

4 detect, verify and evaluate phenotypes and ge notypes. Loblolly pine also has desirable levels of nucleotide diversity ( =0.005; Brown et al., 2004) and limited linkage disequilibrium (<2500 bp on average, Brown et al., 2004). As a first step toward dissecting complex disease traits in loblolly pine, I have undertaken this study to evaluate a variety of disease phenotypes in a population of 32 unrelated parents mated to form approximate ly 63 full-sib families that were clonally propagated to form hundreds or thousands of clones depending upon the experiment. Although an ideal association population would contain hundreds of unrelated individuals (McLeod and Long, 1999), this population is an excellent starting point to evaluate the heritabilities and genetic relati onships among the two sets of di sease traits that can then be dissected by association or QTL mapping. An important aspect of experimental mate rial is its clonal propagation to form a hierarchy of genetic relationshi ps (parent, full-sib family, and clone) that facilitate the dissection of genetic architecture of diseas es and their genetic relations. Since microenvironmental variation for a given genotype can be calculated using its vegetative propagules, I can obtain a more precise esti mate of genetic components of disease resistance. Given that I am using the same genotypes to predict breeding values for both pitch canker and fusiform rust disease resistan ce, the values can be compared to look for correlations that will be informative when I try to understand the underlying genetic architecture. The clonally propagated material of clone s screened in the greenhouse conditions was also planted in several field sites. The fi rst year fusiform rust disease incidence data from the naturally inoculated field site can be compared with the greenhouse screen to

PAGE 17

5 confirm resistant and susceptible genotype s, since the ultimate goal of resistance screening is to identify the resi stant genotypes that will be di sease free in the plantations. Genomic mapping has identif ied the region containing Fr1 (fusiform resistance-1) conferring pathotype-specific resistance to fusiform rust (Wilcox et al., 1996). RAPD marker J7_485A was linked to the Fr1 locus in progeny of a single loblolly pine parent. Thus, the progeny that have this marker we re resistant whereas the ones without the marker were susceptible to fusiform rust incited by C. quercuum with the avirulence gene (Avr1). This genetic marker was consistently pr edictive of fusiform rust resistant trees in greenhouse (Kubisiak et al., 2005; Kuhlman et al., 1997) and field screens (Wilcox et al., 1996). Two families among 63 families that were screened for fusiform rust resistance in the greenhouse and the field were genotyped for Fr1 Thus, clones belonging to the two genotyped families can be used to verify th e resistance prediction power of the genetic marker. The genotypic information on progeny of the families that are segregating for the RAPD marker J7_485A can also be useful in molecular genetics studies. Microarray technology which became available with the la st decade (ref) can be used to identify genes regulated in response to inoculation with C. quercuum With the genetic marker information genetically resistant and suscep tible individuals can be isolated to be challenged by Fr1 avirulent strains of C. quercuum The host responses, disease development and the interactions between the host and the pathogen can be revealed at the molecular genetics level.

PAGE 18

6 CHAPTER 2 GENETIC DISSECTION OF FUSIFORM RUST AND PITCH CANKER DISEASE TRAITS IN LOBLOLLY PINE Introduction Pinus species are both economically and eco logically important. Pines grown in the southeastern United States genera te nearly half of the nation s pulpwood, with an annual harvest value of approximately $19 billion (McKeever and Howard, 1996). Loblolly pine ( Pinus taeda L.) is the most widely planted Pinus species in this region, averaging 74% of the annual seedling producti on (Carey and Kelley, 1993). In addition to plantations, loblolly pine is the predominant species on 11.7 million ha of native forest (Baker and Langdon, 1990), where it impacts the welfare of nearly 400 species of vertebrates (Schultz, 1999). Loblolly pine is a host for two endemic pathogens, Cronartium quercuum Berk. Miyable ex Shirai f. sp. fusiforme (Burds all and Snow, 1977), th e inciting agent of fusiform rust disease, and Fusarium circinatum Nirenberg et O Donnell (Nirenberg and O Donnell, 1998), the inciting agent of pitch canke r disease. Fusiform rust is one of the most destructive fungal diseas es in the southeastern United States, causing damage ranging from $25$135 million per year (Cubbage et al., 2000). The major symptom of fusiform rust disease is the formation of stem galls that lead to decreases in survival, wood quality, and growth. Genetic variation in resistance at the fam ily level has been demonstrated for fusiform rust (Kuhlma n and Powers, 1988; Mc Keand et al., 1999). Based on controlled inoculation studies carried out on specific loblolly and slash pine

PAGE 19

7 Pinus elliottii Engelm. var. elliottii ) families, specific resistancei.e., gene-for-gene interactionshas evolved (Powers, 1980; St elzer et al., 1997; Wilcox et al., 1996), as well as partial resistance in the form of s hort galls (Schmidt et al., 2000), which may be genetically distinct from specific resistance. Pitch canker is also a significant, alth ough more episodic, disease problem (Dwinell et al., 1985). Symptoms of p itch canker disease include re sinous lesions on stems and branches that cause seedling mortality, d ecreased growth rates, and crown dieback (Dwinell et al., 1985). A considerable amount of genetic variation for pitch canker resistance has been detected in loblolly pi ne families (Kuhlman et al., 1982) and clones (Dwinell and Barrows-Broaddus, 1981); however, the genetic architecture of resistance is not well understood. Our goal in this work was to obtain precise estimates of pitch canker and fusiform rust disease phenotypes expressed in l oblolly pine. Precision was acquired by a combination of clonal propagation, which a llows repeat observations of the same genotypes, and is now feasible in loblolly pine (Frampton et al., 2000), testing of over one thousand pedigreed genotypes, and the us e of a mixed linear model (GAREML) to adjust for environmental effects (Huber, 1993). In this study, I identified traits, clones, families, and parents that guide a genetic approa ch to dissecting disease traits in loblolly pine. I verified that pitch canker and fusiform rust traits are heritable and identified the disease traits that are genetica lly distinct from one another. This work creates the baseline knowledge required for identifying genes that condition phenotypes of interest, either through genetic linkage analys is within defined pedigree s, or by association in populations of unrelated ge notypes (Flint-Garcia et al ., 2003; Jannink et al., 2001).

PAGE 20

8 Materials and Methods Genetic Material, Plant Propagatio n, and Experimental Design The 63 loblolly pine families screened in this study were obtai ned from a circular mating design with some off-diagonal cros sing. Members of the Cooperative Forest Genetics Research Program at the University of Florida and the North Carolina State UniversityIndustry Cooperative Tree Improve ment Program (Figure 2-1) provided the 32 parents and generated the full-sib families and clones screened in this study. Fortynine seeds from each full-sib family were germinated and grown into hedges for clonal propagation. Maintenance of hedges and propagati on of cuttings is reported in Baltunis et al. (2005). In brief, cuttings were set in Ju ly 2001, assessed for rooting after 9 weeks, and clones with the highest rooting ability select ed for this experiment. The number of clones within families and the number of ramets (i .e., rooted cuttings) for each clone was not equal, since families did not produce the same number of clones, and clones had different rates of rooting. Cuttings assigned to a gree nhouse screen were chosen at random from the ramet pool of each available clone (Table 2-1). The screens were grouped according to the disease (fusiform rust or pitch canke r). The fusiform rust screens were conducted using two types of inoculum (a one-gall mix or a ten-gall mix), whereas both pitch canker screens used a single inoculum. The experi mental design was a randomized complete block with single-tree plots arranged in an alph a lattice with an incomplete block size of 20. The clones were replicated with a maxi mum number of five ramets per experiment. Ramets were pruned twice to stimulate s ynchronous elongation of multiple succulent

PAGE 21

9 ID 1234567891011121314151617181920212223242526272829303132 116171819 2 6 181720181922 3 3423 1919 4 6937 15 5 58 19182021 6 1510 191918 7 451164 2119 8 5056 181921 9 4026 20 10 2549 21 11 3944 1821 12 28 211921 13 354119 1921 14 4665 18 15 768 2022 16 7038 2221 17 163642 3031 18 27600 2122 19 122 2215 20 3057 212021 21 1329 19 22 8 191820 23 486720 1820 24 5155 17 25 664362 2119 26 9 1819 27 54 3132 20 28 33 5221 1422 29 32 30 12 53 31 19 32 614 Figure 2-1 A circular mating de sign was used to generate th e plant material. Thirty-two parents were crossed following a circ ular design, and the resulting progeny was used as the material screened fo r this study. The numbers in the cells above the diagonal are the number of clones used from a given cross, and the numbers below the diagonal are th e family identification numbers shoots for inoculation. The initial pruni ng occurred in March 2002, 8 months after setting, by cutting back the shoots from 10 cm to 3 cm each. The second pruning occurred 6 weeks prior to inoc ulations for both pitch canker and fusiform rust screens; shoots were succulent and 5 cm in length at the time of inoculation. After pruning, all trees were fertilized weekly with Miracle-Gro 15-30-15 unt il inoculation. Pitch Canker: Inoculations and Data Collection The larger of the two pitch canker screen s was conducted at the USDA Forest Service Resistance Screening Center in Bent Creek, North Carolina, and is referred to as

PAGE 22

10 Table 2-1 Summary of the four inoculati on experiments reported in this study Testa # Families# Clones Rangeb # Ramets # Observationsc RSC pitch canker 63 1065 731 4483 7664 UF pitch canker 60 362 124 1316 3119 Ten Gall fusiform rust 63 1360 17-31 5473 11,395 One Gall fusiform rust 63 698 230 2743 5195 aRSC USDA Forest Service Resistance Screen ing Center, UF University of Florida bNumber of clones within families cNumber of observations exceeds the number of ramets because multiple shoots were assessed on a given ramet the RSC screen in this manuscript. New gr owth (5 cm) was inoculated following the standard RSC protocol (Oak et al., 1987) with F. circinatum isolate S45 (Forest Pathology laboratory collection, Un iversity of Florida) at a density of 92,500 spores/ml. In brief, prior to spray inoculation, shoot tips were excised from two shoots on each ramet. After inoculation, plants were pl aced in a high-humidity chamber for 24 h, then transferred to a greenhouse and maintained at an average temperature of 20C for 3 months. The smaller of the two pitch canker scr eens was conducted at the University of Florida and is referred to here as the UF screen. Plants were pruned 6 weeks before inoculation with the same S45 isolate. On e shoot tip per ramet was excised, and 1 l of a 500-spores/ l solution was applied to the wound w ith a micropipette. All plants were incubated under high humidity for 24 h. The te st was kept in the greenhouse for 36 days at an average temperature of 30C. Disease symptoms were measured at 90 days (RSC) and 36 days (UF). Shoot length and lesion length were measured (in millimeters) on one shoot chosen at random from each ramet at the RSC and on the single shoot inoculated per ramet at UF.

PAGE 23

11 Both the RSC and UF pitch canker raw data sets included only one lesion-length and shoot-length measurement for each rame t. Prior to analysis, the data were standardized by experiment, us ing the phenotypic standard devi ation calculated from the appropriate linear model for the screen. Fusiform Rust: Inoculations and Data Collection Plants were pruned twice before inoculation to stimulate elongation of multiple shoots per ramet. Both rust screens were inoculated at the RSC, following standard protocols (Knighten, 1988). The ten-gall te st was inoculated at a density of 52,000 spores/ml with aeciospores pooled from a ten-gall collection obtained from a 6year-old loblolly pine plantation near Lee, Florida (designated L-10-1-99, provided by Dr. Henry Amerson, NC State Univ ersity) The one-gall test was inoculated at a density of 50,700 spores/ml with aeciospores isolated from a single gall obtained from a branch of slash pine family 84-57 near Callahan, Flor ida (designated #501, provided by Dr. Robert Schmidt, University of Florida). Data were collected from both rust scr eens 6 months after inoculation. For each ramet with multiple shoots, the number of shoot s with galls and the number of galls were counted and recorded. In addition, two shoots with single galls were randomly chosen to measure stem length, gall length, and gall width (in millimeters) for each ramet. Data collected from both the ten-gall and one-gall screens were treated identically for gall measurements. Gall measurement values were averaged by ramet if there was more than one shoot with a single gall. Gall volume was calculated from gall length and gall width data, assuming a fusiform shape: 2) ( ) 4 3 ( width length Volume

PAGE 24

12 Ramets were scored as 0 (no gall) or 1 (at least one gall) for gall score. Ramets that did not form galls were not included in th e gall length, width, a nd volume data. Gall length, width, and volume data sets were stan dardized using their respective phenotypic standard deviations calculated from the linear model. Estimation of Genetic Parameters Variance components and gene tic parameters were estimated by GAREML (Huber, 1993), which employs restricted maximu m likelihood estimation (Patterson and Thompson, 1971) and best linear unbiased prediction [(BLUP) Henderson, 1973]. The same linear model was applied to the traits meas ured in all four disease screens, since the experimental designs were iden tical. The linear model was: ijklm ikl klm kl k ij i ijklme f r family c sca gca gca r t R yl ) ( ) ( where: yijklm is the mth observation of the klth cross in the jth tray of ith rep. is the population mean. Ri is the fixed resolvable replication, i=1. t(r)ij is the random variable tray incomplete block ~NID(0, 2 t), j=1. gcak is the random variable female gene ral combining abili ty (GCA) ~NID(0, 2 gca) k=1 32. gcal is the random variable male general combining ability ~NID(0, 2 gca) l=1. scakl is the random variable specific combining ability (SCA) ~NID(0, 2 sca). c(family)klm is the random variable clone within a family ~NID(0, 2 c(family)). r*fikl is the random variable replica tion by family interaction ~NID(0, 2 r*f)). eijklm is the random variable error within the experiment ~NID(0, 2 e).

PAGE 25

13 The narrow(h2) and broad-sense (H2) he ritabilities were calculated according to Falconer and Mackay (1996): ) ( )) ( ) ( ) ( ( 2 ) ( ) ( 42 2 ) ( 2 2 2 2 2 2P V I V D V A V H P V A V hP f c sca gca P gca where: P 2 is the phenotypic variance, ) ( P V is the total phenotypic variance, ) ( A V is the additive variance, ) ( D V is the dominance variance, ) ( I V is the epistasis variance. To partition the broad sense heritability I calculated the ratio of dominance variance to total phenotypic variance (2Dh ) and the ratio of epistatic variance to total phenotypic variance (2Ih), using the following formulas: ) ( ) ( 4 2 2 2P V D V hP sca D ) ( ) ( ) 75 0 50 0 ( 2 2 2 2 ) ( 2P V I V hP sca gca f c I The broad sense heritability of clonal means (HC 2) and family means (HF 2) were calculated using the formulae below: ) / ( ) / ( ) 2 ( ) 2 ( 2 2 2 ) ( 2 2 2 ) ( 2 2 2r r He f r f c sca gca f c sca gca C

PAGE 26

14 ) / ( ) / ( ) / ( ) 2 ( ) 2 ( 2 2 2 ) ( 2 2 2 2 2c r r c He f r f C sca gca sca gca F where r is the harmonic mean of ramets per clone and c is the ha rmonic mean of clones per family. Family deviations were predicted by summing the following BLUP estimates produced by GAREML: Family deviation = Predicted female value (gcak) + Predicted male value (gcal) + Predicted specific combining ability (scakl) Genetic Correlations The genetic correlation between gall score and gall length at the parental, family, and clonal levels, and the co rrelation among screens within and across diseases were calculated on combined data se ts by adding experiment GCA (2 ge), experiment by family (2 se) and experiment by clone(family) (2 ) ( e f c) interaction factors to the linear model and using the Type B genetic correlation formula (rB; Yamada, 1962): ) ( ) (r2 2 2 PARENTAL B ge gca gca 2 2 2 2 2 2 FAMILY B 2 2 2 ) (rse ge sca gca sca gca 2 ) ( 2 2 2 ) ( 2 2 2 ) ( 2 2 C(F) B 2 2 2 ) (re f c se ge f c sca gca F C sca gca Efficiency of using multiple ramets per genot ype was calculated according to Huber et al. (1992): r H Efficiency / ) 1 (2 where r is the number of ramets per clone.

PAGE 27

15 Results The mating design shown in Figure 1, c oupled with clonal propagation, allowed predictions of clonal, family, and parental genotypic values as we ll as population-wise estimates of heritabilities and genetic correlations of disease traits for both pathosystems. A total of 27,373 phenotypic data points were collected for lesion length (pitch canker), gall score, gall length, and gall width (fusiform rust). I first present data on pitch canker phenotypes, followed by fusiform rust, and finally a comparison of pitch canker and fusiform rust resistance. Pitch Canker Disease Re sistance is Heritable The pitch canker disease screens performed at UF and RSC resulted in 89% of the ramets (i.e., rooted cuttings) showing meas urable disease symptoms in each screen. BLUP clonal values were predicted for each screen, and the resulting distributions are shown in Figure 2a. The consistency of th e disease rates and the shapes of the distributions (i.e., skewed to the right) suggest that statis tical comparisons between the RSC and UF screens are appr opriate. The genetic correlati on between the RSC and UF screens was 0.88 at the parental level, 0.76 at the family level, and 0.69 at the clonal level. A scatter plot based on family ranks is presented in Figure 2b and reflects the positive correlation between the two screens. Therefore, I conclude that parents, families, and clones performed consisten tly across screens. After comb ining the data from the two screens, the five most resistant and the fi ve most susceptible full-sib families were identified based on predicted family values and standard deviations for lesion length; these are indicated in Figure 22a by their ID number from Figure 2-1. The resistant tail contains families 50 and 48, which have parent 8 in common. The resistant tail also contains half-sib families 61 and 4, which share parent 32. Resistant family 44 is not

PAGE 28

16 related to any of the other resistant families in the tail (Figure 2-1). The susceptible tail is composed of three half-sib families. Suscep tible family 12 has parent 30 in common with family 53. Family 12 also shares parent 2 with susceptible family 69. Susceptible families 0 20 40 60 0204060 RSC Family ranks for lesion lengthUF Family ranks for lesion length 0 50 100 150 200 250 300 350 271217222732374247 Lesion length (mm)Number of clones 1 69 53 36 48 61 50 44 4 1 2A B 0 20 40 60 0204060 RSC Family ranks for lesion lengthUF Family ranks for lesion length 0 50 100 150 200 250 300 350 271217222732374247 Lesion length (mm)Number of clones 1 69 53 36 48 61 50 44 4 1 2A B Figure 2-2 Frequency distributi ons and genetic correlation fo r pitch canker lesion length. (A) Distribution of best linear unbiased prediction (BLUP)-predicted clonal values for the USDA Forest Service Resistance Screening Center [(RSC) black] and University of Florida [(UF) white] pitch canker screens. Above the distribution are the predicted means and st andard deviations of the five most susceptible and resistant families identified by their family ID number. (B) Ranks based on BLUP-predicted family values for RSC and UF were plotted against each other (a rank of 1 is the most resistant and 63 the most susceptible). A least squares regressi on line is shown af ter being forced through the origin due to a non-signif icant intercept.

PAGE 29

17 1 and 36 have parent 17 in common. Families in the resistant tail and families in the susceptible tail did not have any parents in common, indicating no genetic relationships across the classes. Two Distinct Inoculation Procedures Reve al Similar Heritabilities for Lesion Length Using the RSC, UF, and pooled data, the heritabilities based on individual tree, family, and clonal means were calculated to determine how much of the variation in lesion length could be attributed to genetic variation and to determine the precision of the predicted clonal and family means. The br oad-sense heritabilities for the clonal (HC 2) and family (HF 2) means were determined for both th e individual and pooled pitch canker screens to evaluate the precision of the cl onal and family means predicted above. HC2 and HF 2 were greater for RSC (0.75) than for UF (0.61; Figure 2-3), because the number of ramets per clone and the number of clones per family were approximately three times greater for the RSC screen compared to the UF screen (Table 1). Narrow-sense heritabilities (h2) for both the RSC and UF datasets we re 0.27. Broad-sense heritabilities (H2) were similar for both the RSC (0.43) and the UF screens (0.37) (Figure 2-3). When the RSC and UF data sets were pooled, heri tabilities were not different from that calculated for each screen individually (Figure 23). This is another indicator that results from the two screens were comparable. Disease Traits Associated with Fusifo rm Rust are Independently Inherited The two fusiform rust screens are characterized by the type of inocula used, either ten-gall or one-gall. There were 36% and 31% galled ramets for the ten-gall and one-gall screens, respectively. A diseas e incidence (referred to as score ) dataset was generated

PAGE 30

18 Figure 2-3 Heritability estimates for pitch can ker lesion lengths. The bar graph shows the heritabilities for individual ramets (R=H2) and the broad-sens e heritabilities for clonal (C=HC 2) and family (F=HF 2) means for the RSC, UF, and pooled data. Narrow-sense heritability [(h2) solid black], epistatic heritability [(hI 2) solid gray], and dominance heritability [(hD 2) white] are stacked so that the yaxis corresponding to the top of the bar is the broad-sense heritability. by designating disease-free ramets as 0 and galled ramets as 1. Fusiform rust screens for score are shown in Figure 2-4a. The distri butions for score in both screens follow a similar pattern, that is, there is a minor p eak at a mean ~0.1, and the distribution is skewed to the right. In addition to disease incidence, gall length and gall width were measured for ramets with galls. In contrast to score, the predicted clonal means for gall length revealed a normal dist ribution for both fusiform ru st screens (Figure 2-4b). Because the distributions and overall disease in cidences were similar, scaling prior to comparing the data from the two screens was not necessary fo r either trait. Genetic correlations between the two screens were calculated for sc ore and gall length in order to determine if inoculum type might im pact trait expression. The genetic correlation for score was 0.80 at the parental level, 0.83 at the family level, and 0.86 at the clonal 0.0 0.2 0.4 0.6 0.8 1.0HeritabilityR C F R C F R C F RSC UF Pooled

PAGE 31

19 0 20 40 60 0204060 Ten-gall Family ranks for gall lengthOne-gall Family ranks for gall length 0 20 40 60 0204060 Ten-gall Family ranks for gall scoreOne-gall Family ranks for gall score 0 40 80 120 160 200 0.00.10.20.30.40.50.60.70.81.0 Score (decimal)Number of clones 54 53 52 49 40 26 51 20 69 3A 0 20 40 60 80 100 1368111315182023 Gall length (mm)Number of clones 29 1 3 1 5 30 25 31 34 36 57 62B CD 0 20 40 60 0204060 Ten-gall Family ranks for gall lengthOne-gall Family ranks for gall length 0 20 40 60 0204060 Ten-gall Family ranks for gall scoreOne-gall Family ranks for gall score 0 40 80 120 160 200 0.00.10.20.30.40.50.60.70.81.0 Score (decimal)Number of clones 54 53 52 49 40 26 51 20 69 3A 0 20 40 60 80 100 1368111315182023 Gall length (mm)Number of clones 29 1 3 1 5 30 25 31 34 36 57 62B CD 0 20 40 60 0204060 Ten-gall Family ranks for gall scoreOne-gall Family ranks for gall score 0 40 80 120 160 200 0.00.10.20.30.40.50.60.70.81.0 Score (decimal)Number of clones 54 53 52 49 40 26 51 20 69 3A 0 20 40 60 80 100 1368111315182023 Gall length (mm)Number of clones 29 1 3 1 5 30 25 31 34 36 57 62B CD 0 40 80 120 160 200 0.00.10.20.30.40.50.60.70.81.0 Score (decimal)Number of clones 54 53 52 49 40 26 51 20 69 3A 0 20 40 60 80 100 1368111315182023 Gall length (mm)Number of clones 29 1 3 1 5 30 25 31 34 36 57 62B CD Figure 2-4 Frequency distributi ons and genetic correlations for fusiform rust disease traits. Distribution of BLUP-predicted clonal values for ga ll score (A) and gall length (B) in the ten-gall inoculum (b lack) and one-gall inoculum (white) screens are shown. Above the distri bution are the predicted values and standard deviations of the five most susceptible and resistant families identified by their family ID number. Ranks based on predicted family values for gall score (C) and gall length (D) (1 = resistant, 63 = susceptible) are plotted against each other. A least square s regression line is shown after being forced through the origin due to a non-significant intercept level, suggesting a general consistency in pe rformance between the ten-gall and one-gall mixes. For gall length, the genetic correla tion between the two screens was 1.00 at the parental level, 1.00 at the family level, and 0.76 at the clonal le vel, again indicating general consistency in perfor mance between the two fusiform rust screens. Despite the high genetic correlations, I did observe outlier families that performed differently in

PAGE 32

20 the two screens, suggesting some potentia lly significant genotype by inoculum interactions (Figure 2-4c, d). Relationships among families with extr eme phenotypes can reveal information regarding inheritance. For scor e, the predicted family values for the five most resistant and five most susceptible families are plotte d on the graph in Figure 4a, along with their within-family standard deviations and family ID numbers. The resist ant tail contains two half-sib family groups, that is, families 26, 40, and 49 that have parent 9 in common, and families 52 and 53 that have parent 28 in co mmon. The susceptible tail is composed of two families that are half-sibs, that is, fam ilies 3, 20, and 51 have parent 22 in common, and families 54 and 69 have parent 2 in common. Similarly for gall length, the five families with the shortest galls and the five families with the longest galls are shown above the distribution in Figure 4b. The shor t gall-forming tail includes families 31 and 62 that have parent 25 in common. The remaini ng three families in this tail are unrelated. The five families with the longest galls comp rise three half-sib families (13 and 29; 29 and 30; and 57, 13, and 57) that are related to one another through parents 21, 20, and 19, respectively. Family 15 is unrelated to the others. For both score and gall length, familial relationships within a given tail were co mmon, whereas no such genetic relationships among families in opposing tails were observed. This is consistent with both score and gall length being heritable traits. To evaluate how much of the trait variati on associated with fusiform rust can be attributed to genetic effects, heritabilities were calculated. Since the genetic correlations for score and gall length were high across inoc ula (Figure 2-4c, d), data were pooled and used for heritability calculations. Gall score was consistently more heritable than gall

PAGE 33

21 0.0 0.2 0.4 0.6 0.8 1.0Heritability-Gall score R C F R C F R C F Ten-gall One-gall Pooled A 0.0 0.2 0.4 0.6 0.8 1.0Heritability-Gall lengthR C F R C F R C F Ten-gall One-gall Pooled 0 20 40 60 0204060 Family ranks for gall lengthFamily ranks for gall scoreB C 0.0 0.2 0.4 0.6 0.8 1.0Heritability-Gall score R C F R C F R C F Ten-gall One-gall Pooled A 0.0 0.2 0.4 0.6 0.8 1.0Heritability-Gall lengthR C F R C F R C F Ten-gall One-gall Pooled 0 20 40 60 0204060 Family ranks for gall lengthFamily ranks for gall scoreB C Figure 2-5 Heritability estimates and family ra nk scatter plots for fusiform rust disease traits. Gall score (A) and gall length (B) heritabilities for R, and C and F means are shown. Heritability estimates for the ten-gall and one-gall pooled screens are given for both traits. h2 (solid black), hI 2 (solid gray), and hD 2 are stacked such that the y-axis correspo nding to the top of the bar is the H2. (C) Scatter plot of family ranks illustrates a lack of correlation between gall score and gall length traits (1 = resistant, 63 = susceptible)

PAGE 34

22 used for heritability calculations. Gall score was consistently more heritable than gall length for the one-gall, ten-gall, and pooled datasets (Figure 2-5a, b). Host Genes Underlying Resistance to Pitch Canker and Fusiform Rust are Independent Necrotrophic (i.e., F. circinatum ) and biotrophic (i.e., C. quercuum ) pathogens have distinct life history properties. This implies that host genes underlying resistance may be different for diseases incited by n ecrotrophic and biotrophic pathogens. To determine whether host responses to F. circinatum and C. quercuum are independent, I computed the genetic correlations between lesion length (pitch canker) and the various gall characteristics (fusiform rust). There were no significant correlations between lesion length and gall length (Figure 2-6), or between lesion leng th and gall score (data not shown). The estimated genetic correlation be tween lesion length and gall length were 0.00 at the parental level, 0.00 at the family le vel, and 0.02 at the clon al level. No genetic correlations were found between lesion length and gall volume or gall width (0.00 for all, data not shown). Together, these results im ply that resistance to pitch canker and resistance to fusiform rust are controlled by different host genes. Efficiency of Using Multiple Ramets per Genotype Theoretically, if the number of ramets per genotype is high enough, heritability estimates based on clonal means will be 1. To describe the relationship between the number of ramets and HC 2 for disease traits investigated in this study, the efficiencies (Huber et al., 1992) for increasing number of ramets per genotype were plotted against the number of ramets (Figure 2-7), where effici ency is calculated as the average reduction

PAGE 35

23 0 20 40 60 0204060 Fusiform rust Family ranks for gall lengthPitch canker Family ranks for lesion length Figure 2-6 No genetic correlation between pi tch canker and fusiform rust resistance. Family rankrank scatter plot based on predicted family means for pitch canker (lesion length) and fusiform rust (gall length), fitted with a least squares regression line (1= re sistant, 63= susceptible) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 12345678910 Number of rametsEfficiency (1-H2)/(r) Figure 2-7 Efficiency is inversely proportio nal to the number of ramets per genotype. Efficiency of using multiple ramets in the estimation of HC2 plotted against number of ramets for RSC-lesion lengt h (filled circles), UF-lesion length (open circles), ten-gall-gall score (fill ed squares), one-gall-gall score (open squares), ten-gall-gall length (filled tr iangles) and one-gal l-gall length (open triangles). Efficiency was calculated as (1H 2)/(number of ramets) in error per ramet. For the different disease tr aits, the error associated with a clonal means decreases at different rates, depending on the number of ramets used to represent genotypes and H2. An increase in the experime nt size above ca. five ramets per clone

PAGE 36

24 does not appreciably increase the precision of heritability estima tes, suggesting that future experiments of this type should be repl icated to approximately the same extent as this study. Discussion Loblolly pine exhibits considerable variation in resistance to both fusiform rust (Kuhlman and Powers, 1988; Powers and Z obel, 1978) and pitch canker diseases (Dwinell and Barrows-Broaddus, 1981; Kuhlma n and Cade, 1985). The pathogens that incite these diseases, the biotrophic fungus C. quercuum and the necrotrophic fungus F. circinatum have distinct life histor y strategies, reflected in the contrasting disease symptoms visible on susceptible hosts. This st udy allowed a direct comparison of the host resistance mechanisms to these distinct pathogens in a common set of host genotypes. Consequently, it was possible to compare and contrast the genetic architecture of host responses to both pathogens. Complex trait analysis requires a reliab le estimation of phenotypic values for subsequent correlations with genotype. As a first step toward dissecting complex disease traits in loblolly pine, I u ndertook this study to evaluate a variety of disease phenotypes in a clonally propagated population generated via a circular mating design. Complex pedigree structures such as these can be useful for mapping QTL (Jannink et al., 2001). Genetic variation for pitch canker resistance Pitch canker resistance was continuously distributed across clones, suggesting that resistance may behave as a complex trait. Resistance to fungal necrotrophs is often inherited as a complex trait in crop species including maize (Bubeck et al., 1993) and rice (Wang et al., 1994). Another explanation for this continuous distribution is Mendelian inheritance of resistance within families th at appears continuous when examined across

PAGE 37

25 families. If resistance were monogenic, some families would be expected to show a bimodal distribution for lesion length. To a ssess this possibility, I tested individual families for bimodal distributions of resistan ce. None of the within-family distributions was bimodal; all showed continuous distri butions. Since lesion length showed a continuous distribution within families across th e entire study, I infer that pitch canker resistance is appropriate to analyze as a complex trait. The repeatability of the pitch canker resi stance screens was high, indicated by the high genetic correlation between the two sc reens, one of which was based on handinoculation in a warm environment (UF scr een) and the other using established spray inoculation methods in a cooler environm ent (RSC screen). The stability of H2 in the pooled dataset relative to the individual sc reens also supports this conclusion. I do not expect pathogenic variation to significantly change the resistance rankings of these genotypes, even though these experiments were performed by inocul ating hosts with a single clonal isolate of F. circinatum This is because there is little evidence for specific resistance in this pathosystem; families rank consistently when challenged with different fungal isolates (G. Blakeslee, personal comm unication). The facultative nature of this pathogen presumably creates little selection pr essure for the evolution of gene-for-gene specificity in this pathosystem. Consequen tly, these clonal rankings may be robust across a broad range of pathogen isolates and predic tive of rankings expected in the clonal field trials established with these genotypes. While narrow-sense heritability is an important metric for breeding applications, our use of clonally replicated material al lowed additional heritability calculations, H2, HC 2, and HF 2 values, which take advantage of the mating and propagation designs used in

PAGE 38

26 this study. HC 2 is an appropriate metric for association and quantitative trait loci studies, because genotyping and phenotyping are both done at the clonal level. Accordingly, in the RSC screen (which involved the most genot ypes of the two pitch canker screens) ca. 75% of the variation in lesion length at the clonal mean level was due to genetic variation. Therefore, I expect lesion length to be an appropria te phenotypic trait for future QTL identification. Gall score and gall length are the most heritable fusiform rust traits Our analysis of gall score (i.e., disease incidence) revealed a non-normal, rightskewed distribution with one major peak a nd several minor peaks. The major peak of apparently resistant genotypes may reflect an over estimation of host resistance because of the use of rooted cuttings. Studi es comparing the responses of seedlings to rooted cuttings have revealed that these two types of plant material behave differently in response to pathogen challenge, with rooted cuttings showing enhan ced resistance (Foster and Anderson, 1989; Frampton et al., 2000). This enhanced resistance phenomenon has been observed in other species and is often referred to as age-dependent resistance because the developmental stage of the infected organ is the key driver of resistance, over and above the action of specific resistance genes (Kus et al., 2002). As clonal host materials become more widely used in research and plantation forestry, our understanding of this phenomenon should improve. Evidence for specific resistance in the loblolly pine C. quercuum pathosystem has been obtained using genomic mapping (Wilc ox et al., 1996) and by inference based on family rank changes in response to genetical ly distinct pathogen cultures (Kuhlman, 1992; Powers, 1980; Stelzer et al., 1997). Alt hough the overall consistency among clonal performances in our two screens was high, I observed a few family and clonal rank

PAGE 39

27 changes for particular families and genotypes between the ten-gall and one-gall inoculations (see outliers in Figure 2-4c), suggesting resistance genes in the host population interacted with specific pathotype s in the inocula. The families showing rank changes between the two inocula may pr ovide a good starting point for identifying additional resistance gene s in loblolly pine. Gall length was normally distributed and was heritable, although to a lesser extent than gall score. Gall length could only be measured on a subset of the population (i.e., on galled ramets), and this sampling effect may account in part for the reduced heritability estimates. Our rationale for measuring gall size characteristics was based on work in slash pine (Schmidt et al., 2000) suggesting that families exhibiting small (short) gall phenotypes were expressing partia l resistance to fusiform ru st, based on their lack of subsequent sporulation. Partia l resistance may be a more durable form of resistance given that it is often race nonspecific (Schmidt et al., 2000 and references therein). I observed continuous variation in gall length in loblolly pine and found no changes in the relative rankings of genotypes that formed galls in both screens as indicated by high genetic correlations (Figure 2-4d). Thus inoculum type did not appear to exert a major effect on gall length. Studies involving a number of defined pathogen cu ltures will be required to resolve the question of whether gall length is conditioned by (relatively late-acting) specific resistance factors, or if gall lengt h is a complex trait, potentially involving multiple genes with small effects. The relationship between gall score and gall length was of interest, because these are distinct phenotypes whose ge netic relationship is not we ll understood. The lack of genetic correlation between gall score and ga ll length, and the l ack of relatedness among

PAGE 40

28 families in the tail distributions for gall sc ore and gall length both suggest that distinct gene systems condition these two traits. Previo us studies have revealed that mean gall length varies substantially in loblolly pine families phenotyped as resistant based on score (Kuhlman, 1992), providing further suppor t for the conclusion that gall score and gall length are conditioned by distinct genetic mechanisms. Future identification of QTL underlying gall length should help distinguish th ese loci from resistance genes known to be associated with gall score in loblolly pine (Wilcox et al., 1996). Resistance to pitch canker and fusiform ru st are under the control of two different mechanisms The lack of genetic correlation between pi tch canker resistance and fusiform rust resistance (as measured by gall score, or gall length) is consistent with distinct genetic architectures underlying host resistance to these two diseases. Biotrophic pathogens suppress host defenses because they require living host cells for survival and nutrient uptake. Hosts resistant to bi otrophic pathogens often act ivate a localized cell death response to prevent spread of the pathogen (Thomma et al., 2001). In contrast, necrotrophic pathogens actively destroy host ce lls and utilize the re leased nutrients for survival. Therefore, a host-c ell death-response e ffective against biotrophic pathogens is postulated to benefit necrotrophic pathogens by increasing nutrient availability through accelerated host tissue destructi on. I propose that resistance to the necrotrophic pathogen F. circinatum is mechanistically distinct from resistance to the biotrophic pathogen C. quercuum due to the differing strategies employed by the two pathogens to incite disease in the host. This is supported by gene-expre ssion array data, which revealed a lack of regulation of rust-associated genes after challenge by Fu sarium (Morse et al., 2004). Although I identified families with excellent resistance to both diseases (Figure 2-6),

PAGE 41

29 disease resistance to the two pathogens shoul d be regarded as independent traits by breeders. Phenotyping for disease trait dissection in loblolly pine The work described in this manuscript ha s assigned specific phenotypic values to more than 1,000 loblolly pine genotypes, enab ling the identification of genes and alleles that condition resistance th rough association studies. Genot yping and association studies are currently underway (ADEPT project Web site, Allele Discovery of Economicallyimportant Pine Traits, http://dendrome.ucdavis.edu/ADEPT/) for candidate loci (Morse et al., 2004) thought to be invol ved in disease resistance in loblolly pine. In this study, I increased the precisio n of phenotyping by using clonally propagated genotypes and mixed linear modeling to adjust for environmental effects. Increasing the number of ramets for a given clone will incr ease the clonal mean based heritability for use in linkage or association studies. However, there is a point of diminishing returns beyond which adding more ramets does not increase precision of phenotyping. This population was an excellent star ting point to evaluate the he ritabilities and relationships among disease traits. Furthermore, it should afford an opportunity to identify QTL by linkage and linkage disequilibrium (i.e., asso ciation) mapping, which has been proposed (Wu et al., 2002) and applied with success (Far nir et al., 2002; Meuwissen et al., 2002).

PAGE 42

30 CHAPTER 3 FUSIFORM RUST RESISTANCE COSEGREGATES WITH AN FR1 -LINKED MARKER AND REVEALS VARIABLE PENETRANCE OF THE DISEASE PHENOTYPE Introduction The economic value of pine in the south eastern United States exceeds $19 billion annually with this region supplying more th an half of the nations pulpwood (McKeever and Howard, 1996). Loblolly pine ( Pinus taeda L.) is the primary pine species in the region, covering 45% of the commercial forest land (Schultz, 1999) with annual production of over 1 billion seedlings for plan ted in reforestation programs (McKeand et al., 2003). Successful plantation establishment in the southeastern United States is highly dependent on the resistance of the planting stock to fusiform rust disease, which is incited by the endemic pathogen, Cronartium quercuum Berk. Miyable ex Shirai f. sp. fusiforme (Burdsall and Snow, 1977). The major symptom of fusiform rust disease is the formation of stem galls which decrease survival, w ood quality, and growth, causing an annual loss ranging from $25$135 million (Cubbage et al., 2000). Loblolly pine families exhibit substantial genetic variation in resistance to fusiform rust disease (Kuhlman and Powers, 1988; McKeand et al., 1999) both in greenhouse and the field. Genomic mapping has identif ied the region containing Fr1 (fusiform resistance-1) conferring pathotype-specific resistance to fusiform rust (Wilcox et al., 1996). RAPD marker J7_485A was linked to the Fr1 locus in progeny of a singl e loblolly pine parent. Thus, the progeny that have this marker we re resistant whereas the ones without the

PAGE 43

31 marker were susceptible to fusiform rust incited by C. quercuum with the corresponding avirulence allele (Avr1). This genetic marker was consistently predictive of fusiform rust resistant trees in greenhouse (Kubisiak et al., 2005; Kuhl man et al., 1997), and field screens (Wilcox et al., 1996). Screening of fusiform rust disease on clonally propagated loblolly pine has revealed the existence of escapes; ramets that are genetically susceptible yet do not show any disease symptoms (Foster and Anderson, 1989; Frampton et al., 2000). The basis for an escape can be a passive fo rm of resistance; a random phenomenon where some cuttings harden off faster than others because of local e nvironment within an experimental block. Alternatively, there ma y be a genetic basis for disease resistance, which may occur, for example if certain genot ypes develop succulent shoots in response to hedging and fertilization at different rates or to different extent s than other genotypes. A genetic analysis can answer this question. Biologically a clone is suscepti ble if it has at least one dis eased ramet. In this study I used the same approach which led us to use a genotype based analysis rather than a ramet based. I used some of the DNA marker s developed in previous mapping studies to distinguish host genotypes that carry/lack the pathotype-specific Fr1 allele. I tested the hypothesis that the Fr1 allele is predictive of re sistance in greenhouse and field experiments. Because these studies involve d clonally propagated materials, I also quantified the extent to which genetic and non-genetic factors influence disease expression levels and escape rate in greenhouse and field trials.

PAGE 44

32 Materials and Methods Genetic Material All the clones that were screened in th e greenhouse and the field came from 63 fullsib loblolly pine families obtained from a circular mating design among 32 unrelated parents with some off-diagonal crossing. Th e genetic material, the propagation methods, the inoculations and the data collection were described in (Kayihan et al., 2005). The parents were from the Atlantic Coastal Plai n and Florida provenances of loblolly pine. Briefly, there were 7-21 clones per fullsib family depending on the family and experiment and approximately 4 ramets per clone. Genotyping Families 0 and 1 for Fr1 Among 32 parents used in this study, parent number 17 was recognized as heterozygous for pathotypespecific resistance gene Fr1 ( Fr1 / Fr1 ; (Wilcox et al., 1996)). Full-sib families 0 and 1 were generated by crossing parent number 17 with parents 18 and 19 (Kayihan et al., 2005) which were known to be Fr1 / Fr1 (unpublished data) and a total of 61 clones from these families were genotyped using the protocols described in Wilcox et al. (1996). The J7_470 RAPD marker is linked to the Fr1 locus and therefore could be used to predict seedling genotypes ( Fr1 / Fr1 or Fr1 / Fr1 ). The mating design (Kayihan et al., 2005) coupled with clonal propagation allowed direct assessment of marker-trait co-segregation. Because pa rents 18 and 19 are homozygous for the Fr1 allele (recessive), families 0 and 1 are test-cross progeny and segregate 1:1 for resistance to Fr1 avirulent (AvrFr1 ) inoculum. Since the matern al parent is heterozygous, megagametophyte samples were genotyped at th e onset of the study to predict seedling genotypes. At the conclusion of the greenhous e screen, foliage samples were collected

PAGE 45

33 from galled ramets that had been initially genotyped as Fr1 / Fr1 and the genotyping reactions were repeated on the foliar DNA. Greenhouse screen The experimental design was a randomized complete block with single-tree plots arranged in an alpha lattice with an incomp lete block size of twenty. Propagation of cuttings was described in Baltunis et al.( 2005). A total of 63 families were used to generate 1360 clones for the te n-gall screen and 699 clones fo r the one-gall screen (Table 3-1). The clones were replicated with a maxi mum number of five ramets per experiment (Kayihan et al., 2005). Table 3-1 Summary of the greenhouse and fiel d screens reported in this study. The 63 families and most of the clones screened were the same across the ten gall, one gall and field screens. Percentage of diseased ramets and clones are reported as a measure of infection rate. Screen# of families# of clones# of ramets% of clones galled% of ramets galled Ten gall63136054736236 One Gall6369827434931 Field6086833625126 The ten-gall test was inocul ated with aeciospores pooled from a ten-gall collection from Madison, FL (designated L-10-1-99) (F igure 3-1). The ten-gall inoculum was tested for virulence against Fr1 ; I inoculated 100 open-pollinated seedlings derived from parent 17, using 50,000 basidiospores/ml and RSC standard methods (Knighten, 1988). Ninetyfour out of ninety-six seedli ngs that were chosen for DNA an alysis were scorable for the RAPD marker J7_470 and the marker data obt ained from the megagametophytes of these seedlings were used to detect virulence against Fr1 In order to choose an inoculum with the least amount of genetic diversity for th e one-gall screen, aeciospores collected from single galls on slash pines in a fi eld site (a generous gift fr om Dr. Robert Schmidt) were assessed by Simple Sequence Repeat (SSR) markers using the methods described by

PAGE 46

34 (Kubisiak et al., 2004). The anal ysis showed that all four of the single gall samples contained at least four or more SSR haplotypes, indicati ng a minimum of four fungal pathotypes in each gall (data not shown). Th e single gall spore collection (designated #501) from Nassau, FL (Figure 3-1) was chosen from for in oculation because of its low genetic diversity, however its virulence against Fr1 was not known prior to this study. The artificial inoculation procedures are desc ribed in Kayihan et al. (2005) and ramets were assessed for the presence (1) and ab sence (0) of a gall 6 months after the inoculation. R S A M B N R S A M B N R S A M B N Figure 3-1 The inoculum sources used in th e ten-gall (Madison County, FL) and one-gall (Nassau County, FL) greenhouse trials and field screens (Randolph County, GA) mapped in Florida and Georgia al ong with the other areas that were assessed for virulence. Virulence against Fr1 was not detected in the inoculum obtained from these counties (S=Santa Rosa, A=Alachua, B=Bradford, M=Madison, R=Randolph, N=Nassau; personal communication Dr. Henry Amerson).

PAGE 47

35 Field The experimental design was a randomized complete block, w ith single-tree plots arranged in an alpha lattice with row/column family restrictions and an incomplete block size of five. A total of 868 clones from 60 families shared with the greenhouse screens were evaluated for fusiform rust disease resi stance in the field (Table 3-1). The field location was Randolph, GA at latitude 31.78N longitude 84.8W (Figure 3-1). Trees were planted in 4 replica tions each with 40 columns and 110 rows, inoculation was allowed to occur naturally and the cuttings were phenotyped for pres ence (1) or absence (0) of fusiform rust galls during the second grow ing season. This is an area classified as high hazard for fusiform rust disease based on south wide classifi cation of >30% of stems in 5 to 15 years old stands that are like ly to have galls on main stems or on live limbs that are fairly close to th e main stem (Anderson et al., 1988). Data Analysis Conceptually, all the ramets from a resi stant clone should be disease free and all ramets from a susceptible clone should be dis eased. In this study, a clone was labeled as resistant when all of the ramets from that particular clone were disease free. However, there were a considerable number of cases where only some of the ramets from a susceptible clone were galled. For the purposes of this study, a single ramet bearing one or more galls was sufficient to identify a susceptible clone. I converted ramet based score data to a clone based dataset by cla ssifying all the genotypes (i.e. clones) with one or more galls as susceptib le (1) and the ones with no galls at all resistant (0). I used families 0 and 1 as a measure of resistance, since both of these families included clones that were Fr1 / Fr1 or Fr1 / Fr1 A few cases of no marker-trait cosegregation could be explained by genetic recombination between the marker and Fr1

PAGE 48

36 To evaluate the probability of a crossove r between the molecular marker J7_485A and the Fr1 locus in a given family I used the formula; x n xp p x n x f ) 1 ( ) ( where n is the number of clones in the given family, x is the number of clones that are putati ve recombinants in the given family, p is the recombination fraction. A B C A B C Figure 3-2 Diagrams illustrating genotype (clone) based phenotyping for disease resistance, susceptibility and escape ra te. (A)A clone was declared resistant with five ramets were disease free. Re sistant genotypes were not included in calculations of escape rate. (B) An i llustration of a susceptible clone with five galled ramets (i.e., an escape rate of 0%). (C)An illustration of a susceptible clone with three galled a nd two disease free ramets (i.e., escape rate of 40%).

PAGE 49

37 I defined escape rate as the ratio of ramets that did not exhibit a gall, to the total number of ramets, given a cl one that had at least one ga lled ramet (Figure 3-2). t g t rr r r E where, Er is escape rate rt is total number of rame ts for a given clone, and rg is the number of ramets with galls for a given clone When all the ramets from a clone are disease free, without genetic marker information one can not distinguish a true ge notype-level escape fr om a clone harboring additional resistance dete rminants. Therefore, genotypes that lacked diseased ramet were deleted from this dataset. Since the escape rate dataset was formed using the percentage of disease-free ramets in otherwise suscepti ble clones, this dataset was clone based. The clone based datasets from all 63 families from the two greenhouse screens (ten-gall and one-gall) and field screen were analyzed to understa nd genetic control of both score (susceptible or re sistant) and escape rate. Va riance components and genetic parameters were estimated by GAREML (H uber, 1993) which employs restricted maximum likelihood estimation (REML) (Patte rson and Thompson, 1971) and best linear unbiased prediction (BLUP) (Henderson, 1973). This approach also aided a more valid comparison of the score and the escape rate datasets. The linear model used to analyze the escape rate and th e score dataset was: klm kl l k klme sca gca gca y where,

PAGE 50

38 klmy is the mth clone of the klth full-sib family, is the population mean, gcak is the random variable female gene ral combining ability (GCA) ~NID(0, 2 gca) k=1 to 32, gcal is the random variable male general combining ability ~NID(0, 2 gca) l=1 to 32,scakl is the random variable specific combining ability (SCA) ~NID(0, 2 sca), eklm is the random variable error within the experiment ~NID(0, 2 e). The narrow (h2) and broad (H2) sense heritabilities were calculated according to (Falconer and Mackay, 1996) based on 0, 1 da ta for disease incidence (gall score) and decimal equivalent for escape: ) ( )) ( ) ( ( 4 4 ) ( ) ( 42 2 2 2 2 2 2P V D V A V H P V A V hP sca gca P gca where: P2 is the phenotypic variance, ) ( P V is the total phenotypic variance, ) ( A V is the additive variance, ) ( D V is the dominance variance. Genetic Correlations The genetic correlation at the family level between the ten-gall and the field screens was calculated on combined data sets by adding Experiment by GCA (2ge) and

PAGE 51

39 Experiment by Family (2 se)interaction factors to the linea r model and using the Type B genetic correlation formula (rB; Yamada 1962): 2 2 2 2 2 2 FAMILY B 2 2 2 ) (rse ge sca gca sca gca Asymptotic Z-test In the field test, there was a possibility of uneven inoculation a result of natural dispersion of inocula. With the purpose of investigating this possibility I run the following model in ASREML(Gilmour et al., 2004); ijklmn ilm lmn lm m l ik ij k j i klme sca r family clone sca gca gca w r c r w c r y ) ( * is the population mean, ri is the random replication ~NID(0, 2 r), i=1 to 4, cj is the random variable colu mn incomplete block ~NID(0, 2 c), j=1 to 40, wk is the random variable row incomplete block ~NID(0, 2 w), j=1 to 110, gcal is the random variable female ge neral combining abili ty (GCA) ~NID(0, 2 gca) k=1 to 32, gcam is the random variable male general combining ability ~NID(0, 2 gca) l=1 to 32, scalm is the random variable specific combining ability (SCA) ~NID(0, 2 sca), clone(family)lmn is the random variable clone within a family ~NID(0, 2 c(family)), (r*sca) ilm is the random variable replicat ion by family interaction ~NID(0, 2 r*sca)), eijklmn is the random variable error within the experiment ~NID(0, 2 e). Asymptotic Z-test for ri, cj, (r*c)ij, (r*w)ik and (r*sca)ilm were calculated by dividing the variance of these components by co rresponding standard deviation.

PAGE 52

40 Results Inheritance of Fusiform Rust Resi stance in the Greenhouse and Field In a previous report (Kayihan et al., 2005) I calculated fusiform rust disease resistance for genotypes and disease inciden ce (score) based on the proportion of ramets that were diseased for each clone. In this study a clone with at least one diseased ramet was classified as susceptible (1) and a clone with no diseased ramet was classified as resistant (0). Clone based resistance to fusiform rust disease (score) was an equally heritable trait both in the greenhouse and the field screens. Narrow sense heritabilities (h2) for gall score were the highest in one-gall screen and the lowest in the field screen (Table 3-2). The broad sense heritabilities (H2) for both the greenhouse and the field screen were high for this trait (Table 3-2). H2 was highest in one-ga ll screen which was followed by ten-gall screen. The field screen yielded the lowest H2 among these screens. Comparison of h2 with H2 within each screen showed that general combining ability (GCA) was higher then specific combin ing ability (SCA) in all trials. Table 3-2 Summary of score (disease incide nce) and escape rate datasets along with narrow sense heritabilities (h2) and br oad sense heritability (H2) for escape rate and score in ten-gall, one-gall and field fusiform rust screens. All the families analyzed in one-gall and field were a subset of the families screened in ten-gall. 2 2 2 2# of families# of cloneshH# of families# of cloneshH Ten gall6313600.390.46624430.230.29 One gall636990.430.52593370.290.30 Field608680.310.366143900 Escape dataset Score dataset 2 2 2 2# of families# of cloneshH# of families# of cloneshH Ten gall6313600.390.46624430.230.29 One gall636990.430.52593370.290.30 Field608680.310.366143900 Escape dataset Score dataset The ten-gall and one-gall greenhouse screens we re highly correlated (Kayihan et al, 2005). According to the genetic correlations I calculated between the ten-gall and the

PAGE 53

41 field screen for score (disease incidence), the families performed consistently, yielding high genetic correlation (rf = 0.83) (Figure 3-3). 6 7 8 9 10 11 12 13 16 19 20 21 22 23 25 26 27 28 29 30 31 32 34 35 36 37 38 39 41 42 43 44 46 48 50 51 52 55 56 57 58 60 61 62 64 66 67 68 69 54 0 70 15 33 40 49 530 10 20 30 40 50 60 0102030405060 Figure 3-3 Scatter plot of ra nks based on BLUP-predicted family genetic values for tengall and field were plotted against each other (a rank of is the most resistant and the most susceptible). A least squares regression line is shown after being forced th rough the origin due to a not-significant intercept. The numbers shown are the family identification codes for the full-sib families. Validation of Fr1 marker Prior to the greenhouse screen using the ro oted cuttings, the ten-gall inoculum was tested for virulence against Fr1 with use of Fr1 / Fr1 parent 17 seedling progeny in North Carolina State University (personal communi cation Dr Henry Amerson). From the 94 seedlings that were inoculated with ten-gall inoculum, 47 were genotyped Fr1 (resistant)

PAGE 54

42 and 47 were Fr1 (susceptible). Of the 47 resistant seedlings, none were galled 9 months after inoculation and in the susceptible group 34 of 47 trees were ga lled. Hence infection in the Fr1 group was 72%, while infection in the Fr1 group was 0%. There was no evidence of virulence against Fr1 Also infection in Resistance Screening Center (RSC), Asheville, NC standard susceptible check fo r loblolly pine (10-8-3) was 76%, so the susceptible check lot and the Fr1 group had the same amoun t of infection. Fr1 marker data can be used to make infe rences about resistance to fusiform rust disease in families from parent 17. I had thre e screens (ten-gall, one-gall and field) to contrast and compare the relative resistance levels of two genotyped families 0 and 1. Similar percentages of disease incidence in te n-gall, one-gall and fiel d screens at both the ramet (36%, 31% and 26% respectively) and the clone (62%, 49% and 51% respectively) levels, gave us confidence to compare them without further adjust ment (Table 3-1). All megagametophytes from families 0 and 1 were genotyped and classified as either Fr1 or Fr1 The marker J7_485A cosegregated with presence/absence of galls in families 0 and 1 in the greenhouse and field scre ens (Table 3-3). In families 0 and 1 most of clones behaved as expected. In the ten-gall screen, twenty four out of twenty nine clones that were Fr1 / Fr1 were disease free whereas twenty six clones were disease free out of thirty clones that were Fr1 / Fr1 In both the one-gall and th e field screens, all the clones from family 0 and 1 that were Fr1 / Fr1 did not have any disease symptoms as expected. Twenty two out of tw enty five clones that were Fr1 / Fr1 were diseased in the one gall and twelve out of sixt een that were genetically susceptible were galled in the field test. As I stated before ten-gall inoc ulum collected from Madison County, FL was avirulent to Fr1

PAGE 55

43 Table 3-3 Segregation of marker J7_485A linked to Fr1 gene in families 0 and 1 across ten-gall, one-gall and field screens (658 ramets combined) with disease phenotype. Parent number 17 is hete rozygous for pathotype-specific resistance gene Fr1 Family 0 is full sib test crosses between parent 17 ( Fr1 / Fr1 ) and 18 ( Fr1 / Fr1 ), whereas family 1 is test cross between 17 ( Fr1 / Fr1 ) and 19 ( Fr1 / Fr1 ). Gray cells highlight the clones that were not in the expected class. Inoculum Genotype Fr1 (+) or fr1 (-) Gall+-+-+-+-+-+Family 02131130141000971 Family 12131320151230953 Overall observed426245029223018124 Overall expected030290029250018160 Ten gallOne gall +-+Field +Inoculum Genotype Fr1 (+) or fr1 (-) Gall+-+-+-+-+-+Family 02131130141000971 Family 12131320151230953 Overall observed426245029223018124 Overall expected030290029250018160 Ten gallOne gall +-+Field +However neither the one gall inoculum (N assau, FL), nor the naturally existing inoculum in the field trial (Randolph, GA) was tested for avirulence. Yet, 0% of the Fr1 /clones from genotyped families 0 and 1 were dis eased in either the one-gall trial or the field experiment, and 83% of the Fr1 / Fr1 clones from these families were diseased in the one-gall and field screens. This suggests th at the inocula used in those screens were avirulent to Fr1 too. Furthermore, in an unpublishe d study varying the numbers of galls collected from counties; Santa Rosa, Madis on, Alachua and Bradford, Florida virulence against Fr1 was found to be very low in Flor ida (personal communication Dr Henry Amerson) (Figure 3-1). There were a few cases where clones did not perform as anticipated. Four out of thirty clones in the ten-gall screen developed galls even though they were genotyped as Fr1 / Fr1 (Table 3-3). To investigate potential mi slabeling problems needle samples from all the ramets belonging to these four cl ones were re-genotyped with markers AJ4 420 and J7 470 to ensure their identity. Based on the marker information, all plants were marker (-) for AJ4 420 and (+) for J7 470 confirming the genotypes previously assessed.

PAGE 56

44 One possible explanation for th e lack of marker-trait co-segregation in these cases was genetic recombination between the marker and Fr1 To evaluate the likelihood of this occurrence, I calculated the probability of r ecombination for both full sib families (0 and 1) and low values (9.3x10-6, 2.1x10-4 respectiv ely) suggested that a recombination event between the marker and Fr1 gene was extremely unlikely. On the other hand five out of twenty nine clones in the ten-gall screen, thr ee out of twenty five clones in the one-gall screen and four out of sixteen clones in the field did not show dis ease although they were genotyped as Fr1 / Fr1 (Table 3-3). These were potentially examples of physiological escapes that are described in more detail in the next paragraph. The Genetic Basis for Escape Rate Clonal replication provided multiple observations of individual host genotypes and helped identify escapes (disease free ramets from a susceptible clone, Figure 3-2). Because most clones that were susceptible ha d at least one ramet was diseased (83 % of Fr1 / Fr1 clones fit this category; Table 3-3), I used all of the families from greenhouse and field screens in the analysis. In an ideal ized experiment in which the escape rate (ER) was zero, all ramets from a given clone would be either gall-f ree or galled. If a histogram is plotted illustrati ng the percentage of galled rame ts for a given clone in an experiment without any escapes I would have tw o bars; one at 0 percen t (resistant); and a second one at 100 percent (susceptible). To ev aluate the distribution of percentage of galled ramets per clone in each screen, I plot ted histograms for the ten-gall, one-gall and field tests (Figure 3-4). The existence of th e bars at 20%-80% drew attention to the significance of escapes in these screens. This suggested a role for one or more environmental factors that prevented norma l expression of the disease phenotype.

PAGE 57

45 0 50 100 150 200 250 300 350 400 450 500 550 0%20%40%60%80%100%Percentage of galled ramets per clone in Ten-gall screenNumber of clones 0 50 100 150 200 250 300 350 400 0%20%40%60%80%100% Percentage of galled ramets per clone in One-gall screenNumber of clones 0 50 100 150 200 250 300 350 400 450 500 0%20%40%60%80%100%Percentage of galled ramets per clone in the FieldNumber of clones Number of clones Number of clones Number of clonesA C B 0 50 100 150 200 250 300 350 400 450 500 550 0%20%40%60%80%100%Percentage of galled ramets per clone in Ten-gall screenNumber of clones 0 50 100 150 200 250 300 350 400 0%20%40%60%80%100% Percentage of galled ramets per clone in One-gall screenNumber of clones 0 50 100 150 200 250 300 350 400 450 500 0%20%40%60%80%100%Percentage of galled ramets per clone in the FieldNumber of clones 0 50 100 150 200 250 300 350 400 450 500 550 0%20%40%60%80%100%Percentage of galled ramets per clone in Ten-gall screenNumber of clones 0 50 100 150 200 250 300 350 400 0%20%40%60%80%100% Percentage of galled ramets per clone in One-gall screenNumber of clones 0 50 100 150 200 250 300 350 400 450 500 0%20%40%60%80%100%Percentage of galled ramets per clone in the FieldNumber of clones Number of clones Number of clones Number of clonesA C B Figure 3-4 Distribution of percen tage of galled ramets by clone in the ten-gall (A), onegall (B) and field (C) sc reens. There were a to tal of 1471 genotypes (i.e. clones) in all the experiments and each clone was replicated 1-5 times in each experiment.

PAGE 58

46 0 20 40 60 80 100 010203040 ColumnRow Figure 3-5 Random distribution of fusiform rust disease resistance performance of ramets from clones that had at least one dise ased ramet in Randolph, GA field trial. Ramets that formed galls were illust rated as full circles whereas healthy individuals were presented as empty ci rcles. Column and row refers to the exact location of each ramet as they were planted in the field. A lack of circle indicates the position of a rame t in a disease-free genotype.

PAGE 59

47 ER might be an environmentally driven event or it might be under control of genetic factors. Replicati on of the clones as ramets and placing them randomly in different blocks enabled us to calculat e escape rate for each clone and compute heritability for this trait. Cl ones with no galled ramets were not included in the analysis since genetic and physiological resistance cannot be distinguished in these cases. According to our calculations, ER was nearly as heritable as score (disease incidence) in both greenhouse screens (Table 3-2). This was co nsistent with the explanation that escape rate in the greenhouse was controlled in part by genes. In contrast, when I ran the field data for the same trait, the heritability wa s zero. Thus, ER was only heritable when the cuttings were in the greenhouse and it was non-ge netically controlled in the field. I ruled out the possibility that infection occurred in a specific pattern (i.e. only on the north side) and show that infection was spread uniform in the field; I graphed the distribution of ramets from susceptible clones in the field (Figure 3-5). I also te sted distribution of resistant and susceptible clones in field to find out if there was any non-random pattern to their placement in field area. Asymptotic Z-test results were not significant implying that the resistant and susceptible clones and rame ts were distributed in field in a random fashion. Discussion In this study I utilized a la rge, structured population of l oblolly pine that had been phenotyped in the greenhouse and in the field fo r resistance to fusiform rust disease, which is an endemic pathosystem in whic h specific resistance has evolved. Agreement Among Greenhouse and Field Screens Greenhouse disease screens are performe d to predict resi stance classes the genotypes will fall into in the field. In the greenhouse I can control nearly all conditions

PAGE 60

48 whereas our influence on conditions in the fiel d is limited to experime ntal design. One of the most important parameter in disease screen is the inoculum; C quercuum which has varying pathogenicity in geogr aphically different field site s (Kuhlman, 1990; Powers and Langdon, 1977; Powers Jr, 1985; Snow et al., 1975; Snow and Kais, 1970; Walkinshaw and Bey, 1981). An implication of genetic vari ation at the pathogen side is unpredictable performance of elite crosses across field sites (McKeand et al., 2003). Furthermore, it was discovered recently that in the field, a single gall is usually induced by a single C. quercuum basidiospore (Kubisiak et al., 2004), whereas concentr ated basidiospore spray system (CBS) inoculation allows multiple ha plotypes to infect and form a single gall (Kubisiak et al., 2005). Thus, the host can be screened for resistance against different genotypes of virulent pathoge n in a single experiment a nd resistant genotypes would perform more consistently across the sites. In this study I not only conservatively score a clone as susceptible if it had at least one ra met diseased but I also screened both in the greenhouse and the field. This cautious approach aided us in identif ication of resistant individuals with complete pene trance of disease resistance trait. Thus, concentrated basidiospore spray system I used to inocul ate the clonally propagated material in RSC and transformation of continuous data (clonal m eans) to binary scale (0< ramets diseased: susceptible, 0=ramets diseased: resistan t) conservatively predicted resistance/ susceptibility of the genotypes I was testing. I found that greenhouse and field data show ed high genetic correlation, presumably due to genetic similarity of th e inocula utilized in the gree nhouse screen and the inoculum in the natural ecosystem at the field site. Th ese results support the earlier reports where gall score in the field yielded similar heritabi lities as greenhouse scre ens (de Souza et al.,

PAGE 61

49 1990; Miller, 1983). Moreover the genetic co rrelation between gr eenhouse screens and the field trial was very high suggesting greenhou se screens can be used to select elite parents, to breed for fusiform rust resistance in the field. Fusiform rust rankings from the Randolp, GA field site was obtai ned from 2nd year data whic h would be considered as preliminary for this disease. Multiple field site data would be avai lable in near future and then I can compare the field fusiform ru st rankings with the one s I calculated from the greenhouse screens. Marker-trait Cosegregation for Fu siform Rust Disease Resistance Markers segregating with resistance genes have been used for selection purposes over the last decade (Francia et al., 2005). This kind of information recently became available for the fusiform rust-loblolly pi ne pathosystem (Wilcox et al., 1996). Host genotyping with markers linked to the know n pathotype-specific resistance gene Fr1 revealed marker-trait cosegregation in both greenhouse and field scr eens. I used this RAPD marker data to predict performances of clones from the two families which were segregating for Fr1 resistance gene. The ten-gall i noculum used in the greenhouse experiment was tested for virulence towards Fr1 before this screen and found to be avirulent (Avr) since after the challenge with this inoculum the clones with Fr1 /(resistant) genotypes were gall-free and Fr1 / Fr1 (susceptible) genotypes were mostly diseased (personal communicati on Dr Henry Amerson). In the ten-gall screen clones that were genotyped for Fr1 performed as expected from their genotypes, confirming the previous results. Cosegregation data in the one-gall screen showed that Fr1 /genotypes were disease free and Fr1 / Fr1 genotypes were mostly diseased, these results suggested that the one-gall inoculum was also avirulent to Fr1 resistance gene. I did not have control over inoculum in the field as I did in the greenhouse screens. However, the data

PAGE 62

50 collected in Randolph, GA on families 0 and 1 indicated that the marker was holding up quite well. Thus, all the inocula I used for these screens were avirulent to Fr1 I identified some clones that were genotyped as Fr1 /but expressed a susceptible phenotype. I initially considered that these may have been cases in which Fr1 / Fr1 genotypes were mislabeled as Fr1 /-. I repeated the genotypi ng reactions on diploid tissues of diseased cuttings and verified th at all of ramets for each of the unexpected classes of genotypes (Table 3-3, grey boxes) ga ve rise to the same genotypic classes that were assigned to them based on megagametophyte genotyping. Hence I favor the explanation that these exceptions to marker-t rait cosegregation are due to a low level of virulence in the inoculum, i.e., a low fre quency of basidiospores with virulence to Fr1 Consistent with this view is the observation that all four cases of diseased Fr1 genotypes in the ten-gall screen were due to single di seased ramets (data not shown). Moreover tengall inoculum was a mixture of ten galls from a high hazard site; also the inoculation load in greenhouse was much higher th an in the field. Diseased Fr1 /clones could be the result of a recombination event between th e marker J7_485A and the actual resistance gene Fr1 ; however, our calculations show that the likelihood of havi ng a recombination between the marker and the resistance gene in the families 0 and 1 was very low, thus unlikely. Thus, a much more likely explanati on for these cases of disease in the presence of the DNA marker was that these four genotypes were infected with virulent C. quercuum genotypes that were present at rela tively low frequency in the ten-gall inoculum. Penetrance of the Fusiform Rust Disease Phenotype Clonal propagation enabled us to quantify the penetrance of the fusiform rust disease phenotype in genotyped and non-ge notyped families within the structured

PAGE 63

51 population. Although the disease ph enotype was expressed at a similar frequency in both greenhouse and field screens as revealed by similar proportions of galled ramets and clones the penetrance of the disease phe notype was dramatically reduced in both greenhouse screens relative to th e field. The basis for this co nclusion is that escape rate (the lack of disease symptom development in a susceptible host genotype) was heritable and similarly so in magnitude across both greenhouse screens, but not heritable in the field. The biological explanation for the re duced penetrance of the disease phenotype in the greenhouse could be driven by pathogenic variation; there may be a low level of avirulence in the ten-gall and one-gall in ocula that correspond to unmonitored host resistance genes that are segregating in the structured population. If this explanation is correct, then the ten-gall a nd one-gall inocula must harbor similarly low levels of avirulent pathotypes such that equivalent heritability estimates are obtained, and the heritability of escape rate (ER) is being driven by segregation of resistance genes in the structured population. Alte rnatively, the biological explanation for the reduced penetrance of the disease phenotype in the greenhouse could be driven by host physiological genetics; there may be inconsis tent growth (shoot flush) rhythms among genotypes that lead to a lack of infection in some genotypes at the time of inoculation. If this model is correct, then the heritability of ER is being driven by segregation of genes that directly or indirectly regulate production of shoot tissu es that are pot ential infection courts. Distinguishing between these comp eting models should be feasible when sufficient marker coverage allows associ ation testing between candidate genes and disease phenotypes (Bro wn et al., 2004).

PAGE 64

52 Pathogen Infection in an Ec ologically Relevant Setting Certainly there were many differences in both host and the pathogen dynamics in the greenhouse compared to the field. Because the artificial inoculat ion with rust spores in a greenhouse screening trial normally occurs within a na rrow window of time (Knighten, 1988), all ramets may not have succ ulent shoots that are susceptible to the pathogen at that time. This phenomenon might st em from the fact rooted cuttings harden off faster than seedlings, preventing an otherwise successful infection. C. quercuum prefers young actively growing plant tissue to infect (Gri ggs and Schmidt, 1977), so a genetically susceptible ramet might not be convenient for infection if it already had hardened off. In contrast, fiel d-grown trees may be exposed to inoculum periodically over a much longer time span of several week s after spring (Schmidt, 1998), so it is much more likely for the pathogen to find the host in a succulent state. Random distribution of more favorable micro sites with more water, fertilizer and sunlight likely affected the escapes in the field. This approach reinforced the conclusions of pr evious reports (Foster and Anderson, 1989; Frampton et al., 2000; Stel zer et al., 1998) where rooted cuttings became physiologically equivalent to the seed lings after several years in field and had higher resistance to fusiform rust then the seedlings. It is very likely that these changes in both host and the pathogen lead to loss of heritability of escape rate in the field. I also investigated the po ssibility of a non-random disp ersion of basidiospores on field. However, I did not find any eviden ce to support uneven disp ersion of inoculum over the field which would result in spatial pattern of susceptibl e ramets suggesting a driver of zero heritability. The diseased plan ts were scattered all over the field site in a random fashion. Thus, the lack of ER herita bility in the field was not a consequence of

PAGE 65

53 spatially nonrandom infection. I surmise that extended periods of inoculation may occur under natural conditions, obscuring genetic in fluences of host shoot phenology on ER. In this study I examined the architecture of fusiform rust disease resistance in a large structured population in which I consid ered resistance as a binomial trait (i.e., resistant or susceptible) for each host genotype. This is in contrast to other studies in which disease resistance was scored on a c ontinuous scale for each genotype, based upon the proportion of ramets that exhibited diseas e symptoms (Frampton et al., 2002; Isik et al., 2004; Kayihan et al., 2005). Both appr oaches have value, based on objectives; evaluating resistance at clonal level might aid answering biol ogically important questions about disease inheritance wher eas the clonal mean for diseas e incidence approach might reveal quantitative genetically im portant questions such as the amount of epistasis. Both approaches have reduced amount of error with the use of clonal propagation. Moreover, the existence of these clonally replicated field trials presen ts an opportunity to monitor potential shifts in pathogen vi rulence that may occur in the C. quercuum population. Such shifts may occur in part due to increased pl anting of resistant genot ypes in the field, and create potentially greater risks if planting stock is clonal.

PAGE 66

54 CHAPTER 4 TRANSCRIPT PROFILING REVEALS POTENTIAL MECHANISMS OF FUSIFORM RUST DISEASE DEPENDENT SHIFTS IN PINE STEM DEVELOPMENT Introduction Loblolly pine is one of the most econo mically important tree species in the southeastern United States si nce loblolly pine plantations cover nearly 13.4 hectares (Schultz, 1999) in this region and over 1 bi llion seedlings are planted annually (McKeand et al., 2003). These plantations have b een threatened by the endemic fungus Cronartium quercuum Berk. Miyable ex Shirai f. sp. fu siforme (Cqf) (Burdsall and Snow, 1977) which incites fusiform rust disease. Fusiform rust is one of the most destructive fungal diseases in the South causing damage in m illions of dollars every year (Cubbage et al. 2000). Cqf is a biotrophic fungus that induces ga ll formation on susceptible trees. The pathogen causes a number of abnormal changes in the stem the galls themselves have an organized cellular structure distinct from a normal stem when viewed using light microscopy. The Cqf hyphae are intimately a ssociated with cortical cells phloem and xylem ray cells, and with cambial cells, with the hyperplasia (swelling) of the stem apparently due to an increase in the number of xylem ray cells and vertical resin ducts in the diseased stem relative to the healthy tissue (Jackson and Parker, 1958). There is evidence that galls disrupt water transport in diseased trees based on nuclear magnetic resonance imaging of galled and healthy stems (MacFall, 1994) and expression of desiccation-associated genes in galled vs. h ealthy stems (Myburg et al., in press). Galls

PAGE 67

55 weaken the structural integrity of stems such that diseased trees are more susceptible to breakage by wind, resulting in reduced stocki ng in stands (Cubbage et al., 2000). Thus the galls alter both structural and fu nctional features of pine stems. The interactions between the host and th e pathogen during gall formation are far from being completely understood in the lobl olly pine-fusiform rust system. However recently, transcript enrichment techniques (Warren and Covert, 2004) and microarray analysis have begun to identify genes from both pine and Cqf that are differentially expressed at infection, gall in itiation and gall expansion stag es (Myburg et al., in press). The Myburg et al. (in press) study is of particular relevance to this chapter, in that I have re-analyzed the data presented in Myburg et al. to extract information on the expression profiles of differentially regulated genes. The Myburg et al. paper presented the overall study design and the analysis focused on c ontrasts between selected time points to identify genes with potential roles in specific stages of gall development (i.e., infection, initiation and expansion). In this study, I focused on the actual expression profiles of individual genes, which I defi ne as the observed change in transcript abundance across time intervals. This analysis allowed me to identify genes whose profiles differed by treatment (pathogen vs. control), genotype (r esistant vs. susceptible) and disease state (diseased vs. healthy). Materials and Methods Plant Material, Genotyping and Harvesting The genetic materials, fingerprinting and sample collection methods are described in a previous study (Myburg et al. in press) Briefly, seedlings from the cross (10-5 4666-4 ) segregate for Fr1 because it is a testcross be tween genotype 10-5, which is heterozygous for the dominant resistance gene Fr1 (Wilcox et al., 1996), and 4666-4

PAGE 68

56 ( Fr1 / Fr1 ). The megagametophytes were harvested from each germinating seedlings and the haploid tissue was screened for RA PD markers J7_470 (Wilcox et al., 1996) and AJ4_420 that define a ~1 cM interval containing Fr1 The RAPD markers identified 350 resistant ( Fr1 / Fr1 ; +J7_470, -AJ4_420) and 350 susceptible ( Fr1 / Fr1 ; -J7_470, +AJ4_420) individuals. Among the 350 individuals in each resistan ce class 210 individuals (15 seedlings x 2 biological reps x 7 time points) were challenged with Cqf and 140 individuals (10 seedlings x 2 biological reps x 7 time points) were inoculat ed with distilled water (as control). Twenty additional seedlings were wa ter inoculated and included in the study as index plants. These index plants were marked with two black ink spots, one immediately below the apical bud and the ot her approximately 1.5 cm belo w the first spot where the potential gall formation with take place. Us ing the references from the index plants, tissue from this region was harvested before the onset of visible disease symptom. The first harvest time point was 90 min after inoc ulation followed by additional harvests at 6hrs, 24hrs, 7 days, 28 days, 56 days and 112 days post inoculation. Fungal Material and Inoculation A single aecisospore isolate of C. quercuum that was avirulent to Fr1 (SC 20-21, obtained from E.G. Kuhlman, US DA-FS, retired) was used in the inoculations that were performed at the Resistance Screening Center Asheville, NC following their standard inoculation protocol except inoculum was increased by 400% to minimize escapes. Microarray Experimental design, microarray prepar ation, target synthesis, microarray hybridization and scanning were descri bed in Myburg et al. (in press).

PAGE 69

57 Statistical Analysis The experiment was implemented in a balanced incomplete block design (Kerr and Churchill, 2001) and analyzed using a mixe d model approach (Wolfinger, 2001). In order to compare the gene expression on several arrays treated with RNA from different treatment X genotype combinations that were dyed with 2 different dyes I applied a normalization data to the entire dataset. The normalization model was chosen with respect to the significance of the effects that extracted from the full model where every effect and the combinations of effects we re tested. I assumed dye, genotype, time and treatment did not change the overall level of gene expression in a single biological sample. The resulting normalization model fo r the log2 transformed data (yijklm) was: ijklm lm jl m l k j i ijklme T M M D M T G D a y * where: is the population mean. ai is the random variable array ~NID(0, 2 a), i=1 Dj is the fixed resolvable dye, j=1 Gk is the fixed resolvable genotype, k=1 Tl is the fixed resolv able treatment, l=1 Mm is the fixed resolvable time, m=1 D*Mjm is the fixed resolvable dye by time T*Mlm is the fixed resolvable treatment by time eijklm is the random variable error within the experiment ~NID(0, 2 e). Before residual values derived from this model were incorporated into the genespecific model I deleted the two early time point s (1.5 hrs and 6hrs) since any of the three

PAGE 70

58 time points within a 24 hr time period should re present the stage of stem development at the time of the inoculation. The multiple early sample times were initially chosen to try and capture gene expression shifts associat ed with a rapid hype rsensitive response; however this was unsuccessful (Myburg et al., in press). I chose to re tain the 24 hr time point since the inoculation procedure involves incubation of trees in a humid chamber for 6-12 hr, whereas the 24 hr (hereafter referred to as day), 7 day, 56 day and 112 day samplings were all carried out in the greenhouse. The 1 day, 7 day, 56 day and 112 day data were then analyzed using the gene model: ijklm klm kl lm kl m l k ij i ijklme M T G T G T M M G M T G a s a y * * ) ( where: is the population mean. ai is the random variable array ~NID(0, 2 a), i=1. s(a)ij is the random variable s pot number(array) ~NID(0, 2 s(a), j=1. Gk is the fixed resolvable genotype, k=1. Tl is the fixed resolvable treatment, l=1. Mm is the fixed resolvable time, m=1. G*Mkm is the fixed resolvable genotype by time. T*Mlm is the fixed resolvable treatment by time. T*Mklm is the fixed resolvable genotype by treatment by time eijklm is the random variable error within the experiment ~NID(0, 2 e). I used PROC MIXED in SAS (SAS Institut e Inc. SAS/STAT Software version 9, SAS Institute, Cary, NC) to run both the array and gene level models (Wolfinger et al., 2001) (Figure 4-1). I identified genes that we re significant for genot ype, treatment, time,

PAGE 71

59 genotype by time, genotype by treatment, time by treatment and genotype by time by treatment using the ANOVA F-test combined w ith Bonferroni multiple testing correction (p=0.01). Using this conserva tive criterion for significance, a total of 861 genes were significant for one or more of these effects. ANOVA F-test + Bonferroni multiple testing correction (p=0.01) 3,705 genes on 56 arrays 861 significant genes with 1 or more significant interactions All of 27 possible gene*treatment profiles are occupied (Figure 4-2) 218 genes left Same profile all treatments 72 R vs. S 29 C vs. I 20 H vs. D 97 4 treatments in a 2x2 factorial arrangement (C=water control, I=pathogen inoculated) x (R=Fr1/fr1, S=fr1/fr1) 4 time points (1day, 7days, 56day, and 112days) Array-level adjustment for the entire experiment (Wolfingeret al., 2001) Gene*treatment interactions each assigned a profileacross the three time intervals based on increased (>+3SD), unchanged (+3SD>x>-3SD) or decreased (<-3SD) transcript abundance across adjacent time points Identification of genes whose profiles vary by treatment No. of genes: ANOVA F-test + Bonferroni multiple testing correction (p=0.01) 3,705 genes on 56 arrays 861 significant genes with 1 or more significant interactions All of 27 possible gene*treatment profiles are occupied (Figure 4-2) 218 genes left Same profile all treatments 72 R vs. S 29 C vs. I 20 H vs. D 97 Same profile all treatments 72 R vs. S 29 C vs. I 20 H vs. D 97 4 treatments in a 2x2 factorial arrangement (C=water control, I=pathogen inoculated) x (R=Fr1/fr1, S=fr1/fr1) 4 time points (1day, 7days, 56day, and 112days) Array-level adjustment for the entire experiment (Wolfingeret al., 2001) Gene*treatment interactions each assigned a profileacross the three time intervals based on increased (>+3SD), unchanged (+3SD>x>-3SD) or decreased (<-3SD) transcript abundance across adjacent time points Identification of genes whose profiles vary by treatment No. of genes: Figure 4-1 Flow chart illustra ting the procedure to identify significant and biologically interesting gene expression profiles. ANOVA was performed for each of the 3705 genes. After experimentwise correc tion for multiple testing, significant gene*treatment interactions were assi gned profiles comprised of three time intervals (1d to 7d; 7d to 56d; 56d to 112d) based on shifts in the LS mean for transcript abundance at each adjacent time point. Biologically interesting profile contrasts are explored; SI vs. rest is equivalent to diseased vs. healthy.

PAGE 72

60 I established profiles of tr anscript abundance that these 861 genes followed on the four time points (i.e. three intervals) (Figure 4-1). To establish these profiles I began with the least square mean for each significant ge ne in each treatment and time point. Using the pdiff option I identified mean differences and standard deviations associated with the mean differences, for each consecutive time point within each treatment. At any given time increment (e.g between 1 day and 7 da ys), expression of a gene can increase, decrease or not change. A d ecision rule was applied in which expression for a gene was declared not changed if the absolute mean difference between the adjacent time points was within 3 standard deviations. However if the absolute mean difference of gene expression was greater than 3 standard deviatio ns within a time increment it was declared increased or decreased. I simply joined the three adjacent time intervals for a gene in order to assign it a gene profile. A gene pr ofile is thus comprised of three integers representing the change in gene expression across the three adjacent time intervals in the experiment. For example, a gene that increase s in expression at each interval has a profile 1 1 1, whereas a gene that decreases during the first in terval and does not change thereafter has a profile -1 0 0. Results Among 3705 genes that were evaluated for gene expression, 861 genes were significant for time, genotype, treatment time*treatment, treatment*genotype, time*genotype, or time*genotype*treatment w ith Bonferroni corrections for multiple testing (p=0.01) (Figure 4-1). Since there were only 3 time in tervals and 3 possible trends within each interval, there can be only 27 distinct profile s. The diversity of profile types among significant genes was sufficiently high to occupy all 27 possible profile groups. I

PAGE 73

61 -3 -2 -1 0 1 2 log2 1 2 3 4Time point A(N)=52 -3 -2 -1 0 1 2 log2 1 2 3 4Time point B(N)=176 -3 -2 -1 0 1 2 log2 1 2 3 4Time point C(N)=40 -3 -2 -1 0 1 2 log2 1 2 3 4Time point D(N)=121 -3 -2 -1 0 1 2 log2 1 2 3 4Time point E(N)=606 -3 -2 -1 0 1 2 log2 1 2 3 4Time point F(N)=68 -3 -2 -1 0 1 2 log2 1 2 3 4Time point G(N)=66 -3 -2 -1 0 1 2 log2 1 2 3 4Time point H(N)=181 -3 -2 -1 0 1 2 log2 1 2 3 4Time point I(N)=12 -3 -2 -1 0 1 2 log2 1 2 3 4Time point J(N)=50 -3 -2 -1 0 1 2 log2 1 2 3 4Time point K(N)=270 -3 -2 -1 0 1 2 log2 1 2 3 4Time point L(N)=57 -3 -2 -1 0 1 2 log2 1 2 3 4Time point M(N)=38 -3 -2 -1 0 1 2 log2 1 2 3 4Time point N(N)=536 -3 -2 -1 0 1 2 log2 1 2 3 4Time point O(N)=152 -3 -2 -1 0 1 2 log2 1 2 3 4Time point P(N)=4 -3 -2 -1 0 1 2 log2 1 2 3 4Time point Q(N)=156 -3 -2 -1 0 1 2 log2 1 2 3 4Time point R(N)=13 -3 -2 -1 0 1 2 log2 1 2 3 4Time point S(N)=29 -3 -2 -1 0 1 2 log2 1 2 3 4Time point T(N)=182 -3 -2 -1 0 1 2 log2 1 2 3 4Time point U(N)=11 -3 -2 -1 0 1 2 log2 1 2 3 4Time point V(N)=80 -3 -2 -1 0 1 2 log2 1 2 3 4Time point W(N)=5 -3 -2 -1 0 1 2 log2 1 2 3 4Time point X(N)=399 -3 -2 -1 0 1 2 log2 1 2 3 4Time point Y(N)=7 -3 -2 -1 0 1 2 log2 1 2 3 4Time point Z(N)=132 -3 -2 -1 0 1 2 log2 1 2 3 4Time point ZAA(N)=5 -3 -2 -1 0 1 2 log2 1 2 3 4Time point A(N)=52 -3 -2 -1 0 1 2 log2 1 2 3 4Time point B(N)=176 -3 -2 -1 0 1 2 log2 1 2 3 4Time point C(N)=40 -3 -2 -1 0 1 2 log2 1 2 3 4Time point D(N)=121 -3 -2 -1 0 1 2 log2 1 2 3 4Time point E(N)=606 -3 -2 -1 0 1 2 log2 1 2 3 4Time point F(N)=68 -3 -2 -1 0 1 2 log2 1 2 3 4Time point G(N)=66 -3 -2 -1 0 1 2 log2 1 2 3 4Time point H(N)=181 -3 -2 -1 0 1 2 log2 1 2 3 4Time point I(N)=12 -3 -2 -1 0 1 2 log2 1 2 3 4Time point J(N)=50 -3 -2 -1 0 1 2 log2 1 2 3 4Time point K(N)=270 -3 -2 -1 0 1 2 log2 1 2 3 4Time point L(N)=57 -3 -2 -1 0 1 2 log2 1 2 3 4Time point M(N)=38 -3 -2 -1 0 1 2 log2 1 2 3 4Time point N(N)=536 -3 -2 -1 0 1 2 log2 1 2 3 4Time point O(N)=152 -3 -2 -1 0 1 2 log2 1 2 3 4Time point P(N)=4 -3 -2 -1 0 1 2 log2 1 2 3 4Time point Q(N)=156 -3 -2 -1 0 1 2 log2 1 2 3 4Time point R(N)=13 -3 -2 -1 0 1 2 log2 1 2 3 4Time point S(N)=29 -3 -2 -1 0 1 2 log2 1 2 3 4Time point T(N)=182 -3 -2 -1 0 1 2 log2 1 2 3 4Time point U(N)=11 -3 -2 -1 0 1 2 log2 1 2 3 4Time point V(N)=80 -3 -2 -1 0 1 2 log2 1 2 3 4Time point W(N)=5 -3 -2 -1 0 1 2 log2 1 2 3 4Time point X(N)=399 -3 -2 -1 0 1 2 log2 1 2 3 4Time point Y(N)=7 -3 -2 -1 0 1 2 log2 1 2 3 4Time point Z(N)=132 -3 -2 -1 0 1 2 log2 1 2 3 4Time point ZAA(N)=5 Figure 4-2. Analyses of mean gene expre ssion data support 27 distinct profile groups, A through AA. The genes that were significant for time, genotype, treatment, time*treatment, treatment*genotype, time*genotype, or time*genotype*treatment (p=0.01) were categorized into 1 of 3 possible profiles (i.e. up, down or unchanged) base d on variation 3 standard deviations from the mean. Mean gene expressi on (y-axis) was plotted across the time intervals (x-axis) for each profile gro up. Time points are indicated as 1=1 day, 2=7 days, 56 days=3 and 112 days=4. N= the number of gene or gene interactions that fall into a particular profile group.

PAGE 74

62 described these profile groups and counted th e genetreatment combinations falling into these groups (Figure 4-2). The frequency at which non-regulated gene and gene interaction profiles were detected (N; N=536) however the number of genes that were unregulated in all four treatments was mu ch lower (N=5 genes; data not shown). The criteria I used to distinguish profiles allowed me to investig ate the trends of a given gene expressed in resistant-control ( RC), resistant-inoculated (RI), susceptiblecontrol (rC), susceptible-inoculated (rI) tr eatment conditions. There were 72 genes that did not respond differently to the genotype-tre atment combinations; in other words, the profile of gene expression was not affected by genotype or by funga l inoculation (Figure 4-3). Sum of Trajectories 0 10 20 30 40 50 123 Time Interval A B cellular development metabolism protein synthesis and stability signal transduction unknown or no hits cellular development metabolism protein synthesis and stability signal transduction unknown or no hitsSum of Trajectories 0 10 20 30 40 50 123 Time Interval Sum of Trajectories 0 10 20 30 40 50 123 Time Interval A B cellular development metabolism protein synthesis and stability signal transduction unknown or no hits cellular development metabolism protein synthesis and stability signal transduction unknown or no hits Figure 4-3. Genes with the same expression profile in all treatment combinations were predominantly induced during the first time interval. (A) For each time interval, profiles were summed across the significant genes an d plotted. Time interval 1=interval from time point 1 (1d) to time point 2 (7d), interval 2=interval from time point 2 (7d) to time point 3 (56d) and interval 3=interval from time point 3 (56d) to time point 4 (112d). (B) Chart of the genes significantly up-regulated in the first time interval but unchanged with respect to treatment or genotype after cate gorization into f unctional groups.

PAGE 75

63 As stated before each gene was assigned a se ries of numbers (1, 0 or -1) to describe the trend within a treatment. The sum of thes e numbers (across genes) for a given time interval illustrates the profile of overall expression for those genes (Figure 4-3A). According to this analysis the genes that had the same profiles in all treatment combinations were predominantly up-regulated within the first time interval (1 day-7 days). For the rest of the time intervals th e expressions of the 72 genes were nearly equally up and down regulated. There were 41 ge nes with a profile in which expression increased in interval 1 (Figure 4-3 B); thes e genes were in category E for all genetreatment combinations (Figure 4-2). I extended this analysis to identify pot entially interesting genes whose profiles changed according to treatment (Figure 4-4) Genes significantly regulated between control vs. inoculated treatment classes di ffered mainly in the 3rd interval (56-112 days) (Figure 4-4A). In contrast, genes re gulated between resistant vs. susceptible genotypic classes differed mainly in the 1st time interval (1-7 days) (Figure 4-4B). In the comparison of healthy (resistant-inoculated, resistant-control, susceptible-control) vs. diseased (susceptible-inoculated) classes (Figure 4-4C) the 2nd time interval (7-56 days) was the interval during which mo st genes were regulated. Among these biologically interesting comparis ons only the healthy vs. diseas ed contrast revealed genes of fungal origin. There were 13 known fungal ge nes regulated in this category most of these occurring in the second time interval (7days-56 days) with possible additional fungal genes classified as unknow n origin (Figure 4-4C).

PAGE 76

64 0 10 20 30 40 50Number of genes 123121323123 Time interval 0.0 2.5 5.0 7.5 10.0 123121323123 Time intervalNumber of genes 0.0 2.5 5.0 7.5 10.0 12.5 123121323123 Time intervalNumber of genesABC 0 10 20 30 40 50Number of genes 123121323123 Time interval 0.0 2.5 5.0 7.5 10.0 123121323123 Time intervalNumber of genes 0.0 2.5 5.0 7.5 10.0 12.5 123121323123 Time intervalNumber of genesABC Figure 4-4. Profile groups can be categorized into biologically interesting clusters with distinct changes in gene expression patterns. Ge nes significantly regulated across control and inoculated treatme nt classes (A), across resistant and susceptible genotypic classes (B) and across healthy (resistant-inoculated, resistant-control, susceptible-control) and diseased (susceptible-inoculated) classes (C) are shown. Time intervals 1, 2, and 3 are as defined in Figure 2. 12 = intervals 1 and 2 combined, 23= 2 and 3 combined, 13= 1 and 3 combined, and 123= 1, 2 and 3 combined. The genes that originated from host are represented by a black bar, genes with no known origin are represented by a gray bar and fungal genes are repres ented by a white bar. I identified genes that were differentially regulated in diseased vs. healthy tissues, and found there were dramatic shifts in all th ree intervals. The three intervals had been previously characterized as reflecting distinct phases of disease deve lopment, specifically the 1st time interval = infection, 2nd time interval = gall initiation and 3rd time interval = gall expansion (Myburg et al., in press). I contra sted the profiles in diseased vs. healthy tissues in order to identify the direction of regulation imposed by the pathogen (Figure 45). In the infection interval, most of the genes that are differentially regulated were down-regulated in diseased seedlings. During gall initiation the majority of the genes were up-regulated and 11 of the 50 genes th at were in this class were fungal genes whereas 27 of them were pine genes and th e rest (12 genes) were unknown. The gall expansion phase was dominated by pine genes, with half of the genes in this class upregulated and the other half down-regul ated under the influence of Cqf.

PAGE 77

65 NXCI_082_G01calcium ion binding NXCI_153_A02_FCLAVATA1 receptor kinase NXLV_022_H08_Fras-related protein RAB8-5 NXNV_096_C09putative asparaginesynthetase NXNV_125_E12_FASP1 (ASPARTATE AMINOTRANSFERASE 1) NXNV_147_G03_Fcysteinproteinase(by similarity) NXPV_062_E04_FAUX1 (AUXIN RESISTANT 1) NXLV_100_F02| auxin:hydrogen symporter NXSI 059 G09Adenosine kinase NXSI 102 F11putative UMP/CMP kinasea NXSI_067_F10_Fputative nitrilase 2 NXNV015H07Calcium-binding EF-hand NXNV_073_G04Putative cellulose synthasecatalytic subunit NXNV_132_G11_FCellulose synthase NXLV100_F02_FAuxin:hydrogen symporter NXLV_012_A05_FAnnexin NXLV_049_G11_FLMW heat shock protein NXNV_135_E01_FDirigent-like protein pDIR4 NXSI_099_H06_F(+)-alpha-pinenesynthase NXSI_104_B11FED A; electron transporter/ iron ion binding NXCI_018_A08pectatelyase NXCI_027_G06Inositol-3-phosphate synthase NXCI_067_H06S-adenosyl-methionine-sterol-C-methyltransferase NXCI_075_D09Epoxidehydrolase NXLV_079_G07_FLeucine-rich repeat transmembraneprotein kinase1 NXNV_122_C07_FCalcium-dependent protein kinase NXSI_063_D01Flavanone3-hydroxylase NXSI_076_E08Putative mitochondrial dicarboxylatecarrier protein NXSI_101_B01Putative glycine-rich protein NXSI_103_D11_FPutative sinapylalcohol dehydrogenase pi134-1Acyltransferase/ carboxylic ester hydrolase/ lipase NXNV 129 F06PIN1-like auxin transport protein NXNV_163_F07_FAcyl-CoAbinding NXPV_068_E06_FXyloglucanendotransglycosylase NXRV064_C07_FPutative xyloglucanendotransglycosylase NXRV079_D01_FXyloglucanendotransglycosylase NXSI_103_E12_FXyloglucanendotransglycosylaseXET1 NXLV103_E01_FFlavanone3-hydroxylase 2 NXSI 008 G11Late embryogenesis abundant protein pi295Type II proteinaseinhibitor family protein infection 24 genes (21 host, 3 unknown) gall elongation 20 genes (15 host, 2 fungal, 3 unknown) gall initiation 50 genes (27 host, 11 fungal, 12 unknown) NXCI_082_G01calcium ion binding NXCI_153_A02_FCLAVATA1 receptor kinase NXLV_022_H08_Fras-related protein RAB8-5 NXNV_096_C09putative asparaginesynthetase NXNV_125_E12_FASP1 (ASPARTATE AMINOTRANSFERASE 1) NXNV_147_G03_Fcysteinproteinase(by similarity) NXPV_062_E04_FAUX1 (AUXIN RESISTANT 1) NXLV_100_F02| auxin:hydrogen symporter NXSI 059 G09Adenosine kinase NXSI 102 F11putative UMP/CMP kinasea NXSI_067_F10_Fputative nitrilase 2 NXNV015H07Calcium-binding EF-hand NXNV_073_G04Putative cellulose synthasecatalytic subunit NXNV_132_G11_FCellulose synthase NXLV100_F02_FAuxin:hydrogen symporter NXLV_012_A05_FAnnexin NXLV_049_G11_FLMW heat shock protein NXNV_135_E01_FDirigent-like protein pDIR4 NXSI_099_H06_F(+)-alpha-pinenesynthase NXSI_104_B11FED A; electron transporter/ iron ion binding NXCI_018_A08pectatelyase NXCI_027_G06Inositol-3-phosphate synthase NXCI_067_H06S-adenosyl-methionine-sterol-C-methyltransferase NXCI_075_D09Epoxidehydrolase NXLV_079_G07_FLeucine-rich repeat transmembraneprotein kinase1 NXNV_122_C07_FCalcium-dependent protein kinase NXSI_063_D01Flavanone3-hydroxylase NXSI_076_E08Putative mitochondrial dicarboxylatecarrier protein NXSI_101_B01Putative glycine-rich protein NXSI_103_D11_FPutative sinapylalcohol dehydrogenase pi134-1Acyltransferase/ carboxylic ester hydrolase/ lipase NXNV 129 F06PIN1-like auxin transport protein NXNV_163_F07_FAcyl-CoAbinding NXPV_068_E06_FXyloglucanendotransglycosylase NXRV064_C07_FPutative xyloglucanendotransglycosylase NXRV079_D01_FXyloglucanendotransglycosylase NXSI_103_E12_FXyloglucanendotransglycosylaseXET1 NXLV103_E01_FFlavanone3-hydroxylase 2 NXSI 008 G11Late embryogenesis abundant protein pi295Type II proteinaseinhibitor family protein infection 24 genes (21 host, 3 unknown) gall elongation 20 genes (15 host, 2 fungal, 3 unknown) gall initiation 50 genes (27 host, 11 fungal, 12 unknown) Figure 4-5 Clustergram of gene profile differe nces (or contrasts) between diseased and healthy treatments. Genes whose expre ssion in diseased tissues are higher than in healthy tissues for a given interval are shown in red; if lower, green; if identical, black. Examples of genes in se lected clusters are shown on the right. The genes regulated during the infection phase are similar to genes that encode proteins involved in auxin transport (NXLV_100_F02 = auxin:hydrogen symporter;

PAGE 78

66 NXPV_062_E04 F = AUX1 auxin transporter) and auxin biosynthesis (NXSI_067_F10F = nitrilase 2, an IAA biosynthetic gene). Th e infection phase also appears to involve calcium fluxes as indicated by regulation of genes involved in calcium signaling (NXCI_082_G01 = calcium ion binding protei n). During gall i nitiation cell wall modification enzymes appear to be induced under the influence of Cqf (NXNV_132_G11F = cellulose syntha se; NXCI_018_A08 = pectate lyase; NXSI_101_B01 = putative glycine-rich prot ein; NXSI_103_D11F = putative sinapyl alcohol dehydrogenase). During the gall expans ion phase, four distinct members of the xyloglucan endotransglycosylase family were down-regulated, potentially in association with additional cell wall arch itecture modifications duri ng gall growth (NXSI_103_E12F; NXPV_068_E06F; NXRV064_C07_F; NXRV079_D 01F = xyloglucan endotransglycosylase). Discussion Cronartium quercuum is a biotrophic, macrocyclic heteroecious fungus (Burdsall and Snow, 1977) that incites abnormal changes in the fusiform rust susceptible pine stem such as swollen phloem cells, and increase in the number of resin ducts and ray cells (Jackson and Parker, 1958; Gray and Amerson, 1983, Jewell et al., 1962; Miller et al., 1976). To identify genes and processes that may underlie the development of disease symptoms, I used a microarray dataset derive d from a time course analysis of fusiform rust disease development. The genes on the microarray included ESTs and cDNAs obtained from subtraction libraries and fr om genes that are expressed during the interaction between the host and the pathogen, an approach that has been successful in identifying genes that may condition dis ease phenotypes (Birch and Kamoun, 2000; Wan et al., 2002). These powerful tools became ava ilable for the fusiform rust-loblolly pine

PAGE 79

67 pathosystem recently (Warren and Cover, 2004; Myburg et al., in press). In this study I investigated fungal and pine gene expressi on on the pine stems obtained from resistant and susceptible seedlings that were inoculated with Cqf or water using an experimental design that involved 4 time points. I captured a diverse array of expr ession profiles across the genes that were significant for the main ef fects or interactions of the main effects I was testing. Transcription Profiling Reveals Differential Gene Expression Diverse patterns of gene expression were observed in this study, such that all possible combinations of 27 profiles were occupied by at least one gene or gene interaction. The chip-level model that I us ed, effectively corrected for statistically significant treatment and interaction effects at the chip level prior to the gene-level analysis. This approach assumes that large-sc ale unidirectional shifts in gene expression in any particular treatment or treatment comb ination are based on technical artifacts, not biologically meaningful effect s, in the microarray experiment. A chip-level adjustment model with fewer terms may be more biologically appropriate for this study, in that it may identify more genes whose expression is altered; on the other hand, such an approach may lead to more false positives. The decision to use 3 SD as a criterion for significance was based on striki ng a balance between identify ing potentially interesting biological mechanisms, while still being suffi ciently conservative to exclude most false positives. An indication that this criteri on was reasonably conservative was the observation that although all of the gene a nd gene interactions were highly significant experimentwise (after Bonfe rroni correction), 648 profiles were declared biologically non-significant based on the 3 SD criterion.

PAGE 80

68 There were 72 genes that showed identic al profiles regardless of treatment (genotype or pathogen). The majority of the ge nes in this category belonged to group E which is characterized by an increase in inte rval 1 and stable expression thereafter. The 1 day to 1 week period of pine seedling growth dur ing this study is predicted to be a time of active primary stem growth and development. The annotation of the genes in group E for all treatments revealed that they bel onged to functional groups such as metabolism, protein stability, signal transduction and cellular development, which is consistent with the kinds of functions expected in an activel y growing seedling. Notably, these genes are not affected by genotype or by pathogen challe nge, so they are more relevant to stem development in loblolly pine than to diseas e development per se. Such genes presumably reflect the juvenile developmental state of th e stem when the seedlings were harvested at the earliest time point in the experiment. When the transcriptomes of control and pathogen inoculated seedlings are compared, one might expect that the differen ce in gene expression w ould be at the first time interval since that would be the period during which pathogen spores are germinating and contributing to the transcri ptome (this might occur on both resistant and susceptible seedlings). Interestin gly, this analysis s uggests otherwise, in that the effect of Cqf inoculation is on the regulation of genes much later than the first time interval. In fact, the effects of Cqf inoculation on ge ne expression profiles gradually increased through the 3rd time interval. Thus, both resist ant and susceptible plants gave the same response to a fungal infection a ttempt that resulted in differe nces between the control and the inoculated plants. It is intriguing to speculate that some kind of pathogen-induced systemic response might be responsible for this observation. However, known examples

PAGE 81

69 of long-term pathogen-induced responses (s uch as systemic acquired resistance and induced systemic resistance) are typically inci ted by an incompatible interaction (resistant host) distinct from the compatible interacti on (susceptible host), wh ich is not consistent with my findings. In addition, the number of genes in this group was relatively small (N=20), so further gene expr ession studies should be performed to confirm and resolve this phenomenon. The comparison of profiles in resistan t vs. susceptible genotypes revealed differential gene expression, regardless of whet her the plants were inoculated with Cqf or the water control. This is interesting in and of itself, since it suggests that allelic differences at the Fr1 locus (and loci linke d within 1 cM of Fr1 ) can be detected at the level of gene expression. The level of ge netic resolution in this study was reasonably high, in that flanking markers were used for genotyping (thus only double recombinants within a 1 cM interval would be misclassified) and a relatively large number of seedlings was bulked for each time point (50 Fr1 / Fr1 and 50 Fr1 / Fr1 seedlings for each time point). Therefore the gene expression comparis ons at each time point are likely to reflect allelic differences at or near Fr1 but mixtures of both allele s at unlinked loci. One would expect that regulation could occu r in cisor in trans-. If cis-regulation is being observed here, then markers within the regulated genes coul d be used to create a fine map of loci in the Fr1 interval. If trans-regulation predominates (e.g., Kirst et al. 2005), then allelic configuration at or near Fr1 may induce downstream signaling mechanisms that are manifest on the microarrays. Influences of Fusiform Rust Dise ase Development on Gene Profiles The healthy vs. diseased comparison re vealed over a hundred genes that were differentially regulated across all time intervals. In this section, I elaborate on the types of

PAGE 82

70 putative gene functions and physiological m echanisms that may be involved in the development of the fusiform rust disease state. In the first interval, a dramatic (down -) regulation of host re gulatory genes and auxin biosynthesis and transport genes, in th e developing disease stat e. Nitrilase 2 is the enzyme that catalyzes the conversion of i ndole-3-acetonitrile (IAN) to indole-3-acetic acid (IAA) which is active auxin (Woodwar d and Bartel, 2005). The auxin:hydrogen symporter is an efflux auxin carrier whereas AUX1 is an influx auxin carrier. Therefore, in diseased seedlings it is feasible that auxin transport as well as bi osynthesis is impaired (or its transport is modified under the influen ce of the pathogen) with in infected cells. Ca+2 is a well-known second messenger acti ng downstream of many stimuli including hormone signaling, and specifica lly in auxin signaling sin ce calmodulin binding proteins are encoded by members of the auxin response gene family (SAURs; Yang and Poovaiah, 2000). Interestingly, during gall expansion ther e is a similar, coordi nated regulation of a putative auxin transport protein-encoding gene (Pin1-like) and a gene encoding a protein containing a putative calcium-binding EF-h and. This provides another potential connection between auxin and calcium signaling in fusiform rust disease development. A potentially fruitful area for future research would be to quantify hormones in developing fusiform rust galls, since there is also evid ence for a potential role of gibberellins in later stages of gall development (Myburg et al., in press). In the second interval, tr anscripts encoded by the fungal pathogen were detected. The detection of fungal transcripts presumably reflects the establishment of a compatible interaction with the host (Heath, 1997) whereby the pa thogen has begun to disperse within the host tissue (Walkinshaw, 1978) to manipulate the host sink to its needs. Since

PAGE 83

71 Cronartium spp. are ecologically and economically important pathogens on pines (for example, the white pine blister rust is incited by Cronartium ribicola J. C. Fisch; Jurgens et al., 2003; Hudgins et al., 2005) there woul d be value in using transcript profiling approaches to better understand gene expressi on shifts in the pat hogen component of the disease interaction in the future. In the second interval I also observed disease-altered prof iles of host genes potentially involved in modifying cell wall ar chitecture. Cellulose synthase and pectate lyase were shown to be up-regulated in xylem (compared to leaves) (Paux et al., 2004), and in Populus reaction wood (AnderssonGunneras et al., 2006). Sinapyl alcohol dehydrogenase is involved in the monolignol precursor pathway leading to lignin biosynthesis (Anterola et al., 2002) and glycin e-rich proteins are thought to be involved in structural integrity and i nducible reinforcement of plant cell walls (Ringli et al., 2001). Thus, host genes related to cell wall synthesis may be induced as a suite of genes required for rapid cell wall biosynthesis associ ated with the initiation of a gall. In the third interval I observed regulati on of a multigene family whose products are involved in growth and cellular architecture (l oosening and/or tightening of the cell wall). Cellulose microfibrils are typically stabil ized by xyloglucan moieties, which can be cleaved via xyloglucan endotransgl ycosylase (XET) so that cell can alter its shape (Fry et al., 1992). One of the family members is indu ced in the gall expans ion phase, whereas four members are repressed, suggesting the family members have non-redundant roles in cell wall modification. It is interesting to note the rapid yet organized manner in which specific cell types in fusiform rust galls deve lop ray cells increase in size and number, whereas resin ducts increase in number onl y (Jackson and Parker, 1958) whereas the

PAGE 84

72 attachments among parenchymal cells appear to loosen (Walkinshaw, 1978). Given the complex dynamics of cell architecture changes, it is tempting to speculate that some of the XETs may play roles in specific cell type s and thus be performing distinct functions accordingly. These results validate the stages of disease development proposed by Myburg et al. (in press) that seedlings can be analyzed according to the discrete disease phases of infection (1-7days), gall initiation (756 days) and gall expansion (56-112 days).

PAGE 85

73 CHAPTER 5 CONCLUSION Throughout this study I investigated two fungal diseases that are threatening loblolly pine plantations in s outheastern US. The first disease is fusiform rust which is incited by biotrophic fungus Cronartium quercuum (Burdsall and Snow, 1977), the second one is pitch canker which is incited by the nect rorophic fungus Fusarium circinatum (Nirenberg and ODonnell, 1998). Biotr ophic and necrotrophic fungi have different life cycles and infection mechan isms. Thus, a resistant host would respond differently to these two fungal diseases. A fu siform rust resistant host would recognize C. querc uum induce HR whereas F. circinatum the pathogen can survive HR based defense by detoxification which is a common protection for necrotrophic fungi (Mayer, Staples, and Gil-ad 2001). A nother major difference between C. quercuum and F. circinatum is the fact that F. circinatum requires wound to enter the host (Kuhlman 1987) while C. quercuum does not. Another host defense after the infection can be delimiting fungal growth by forming cell appositions to stop the disease progression. It was demonstrated that to C. quercuum triggers necrosis in resistant loblolly pine and forms wall appositions that are part ially composed of callose to prevent infection (Gray and Amerson 1983). It was also reported that a dist inct lesion formation is a type of resistance which limits F. circinatum growth in loblolly pine (Barrows-Broaddus and Dwinell 1983). I propose C. quercuum and F. circinatum have very different life cycles and infection styles as a result th e way host responds to them shou ld be distinct, too. Later in

PAGE 86

74 the study I showed that in loblol ly pine resistance to fusifo rm rust is not correlated to resistance /susceptibility to pitch canker and vice versa. I also investigated Fr1 marker co-segregation with the disease phenotype in two families that were a result of a cross between Fr1/fr1 and fr1/fr1 parents. The offspring from the crosses were genotyped for Fr1 locus. I showed that Fr1 marker information is predictive of resistance in the both greenhouse and the field as long as the inoculum is Avr to Fr1 locus. Using clonally propagated mate rial I increased the precision and also identified the ramets that escape. I showed that escape is a heritable trait in the greenhouse whereas it becomes inheritable in the field due to inoc ulum source that is presented for longer periods of time compar ed to a few minutes in the artificial inoculations. In the last part of this study I switched gears to investigate gene expression in resistant ( Fr1/fr1 ) and susceptible ( fr1/fr1 ) individuals that were inoculated with water or C. quercuum in a time frame of 112 days. I showed th at there are genes that have unique expression profiles across contro l vs. inoculated, resistant vs susceptible and healthy vs. diseased. These results would be helpful to parse the disease development in susceptible loblolly pine seedlings. Neither fusiform rust nor pitch canker disease resistance is far from being completely understood. Although this study brings us closer to the answers scientists are seeking, more research on biology of these fungal diseases is needed.

PAGE 87

75 APPENDIX A SAS SCRIPTS FOR MICROARRAY ANALYSIS /* Fr1 2 is the dataset that contains treatme nt genotype time_point gene_name log2 expression here I am dropping all the contro l genes since they are causing extra noise*/ data Fr1 2; set tmp1. Fr1 1; if gene='Blank water control' or gene=' Blank' or gene='BLAN K' or gene='Water Control' or gene=' Control BA R' or gene=' Cont rol GFP' or gene=' Control Globulin' or gene='SP3' or gene='SP2' or gene='SP1' or gene='Sp3' or ge ne='Sp1' or ge ne='Sp2' or gene='Spike Control Sp1' or gene='Spike Control Sp2' or gene='Spike Control Sp3' or gene='Spike Control Sp4' then delete; run; /*after control_genes are deleted, run the array-level analysis to get rid of major effects like dye*/ proc sort data= Fr1 2; by dye gene time; run; proc mixed data= Fr1 2; class array dye genotype treat time; model log2i= dye genotype treat time dye*time treat*time/OutPred=no_contr; random array; run; /* after this delete time poi nts=all, 6 hours, 90 mins*/ proc sort data=tmp1.Frno_cont r; by genotype time; run; data Frno_contr_r_i_6_90; set tmp1.Frno_contr; if genotype='r' and treat='I' then delete; if time='all' or time='6' or time='90' then delete; run; /*now I can run the se cond level analysis where gene by gene*/

PAGE 88

76 proc sort data=Frno_contr_r_i_6_90; by gene; proc mixed; by gene; class array treat time2 dye spot_number genotype; model resid=time2 treat genotype time 2*treat treat*genotype time2*genotype time2*genotype*treat; random array spot_number(array); ods out put Tests3=pvalno_c_r_i_6_90; run; proc sort data=pvalno_c _r_i_6_90; by probf; run; data Gene_no_c_r_i_6_90; set pvalno_c_r_i_6_90; pr oc sort; by ProbF; where ProbF<3.85683E-07 ; run; proc sort data=Gene_no_c_r_i_6_90; by gene; run; /* get the significant genes only one copy of gene should be in the list*/ data once_sig_gene_no_690ri; set Gene_no_c_r_i_6_90; by gene; if first.gene; run; /*here I merge the significant gene list with the residuals from the array level analysis to get lsmeans and stdev associated with them*/ proc sort data=Frno_contr_r_i_6_90; by gene;run; proc sort data=once_sig_gene_no_690ri; by gene;run; data res_no_690ri; merge Frno_contr_r_i_6_90 once_sig_gene_no_690ri; by gene; run; proc sort data=res_no_690ri; by ProbF; data clean_res_no_690ri; set res_no_690ri; if ProbF<0 then delete; run; proc sort data=clean_res_no_690ri; by gene; run; /*it is to get lsmeans and stdev*/ proc mixed; by gene; class array treat time2 dye spot_number genotype; model resid=time2 treat genotype treat*time2*genotype;

PAGE 89

77 random array spot_number(array); lsmeans treat*time2*genot ype; ods output LSMeans=wo_all_G_T_T_m eans diffs=wo_all_pdiff; run; proc sort data=Wo_all_pdiff; by time2 _time2; data clean_Wo_all_pdiff; set Wo_all_pdiff; if time2=1 and _time2=1 then delete; if time2=2 and _time2=2 then delete; if time2=3 and _time2=3 then delete; if time2=4 and _time2=4 then delete; if time2=5 and _time2=5 then delete; if time2=6 and _time2=6 then delete; if time2=1 and _time2=3 then delete; if time2=1 and _time2=4 then delete; if time2=1 and _time2=5 then delete; if time2=1 and _time2=6 then delete; if time2=2 and _time2=4 then delete; if time2=2 and _time2=5 then delete; if time2=2 and _time2=6 then delete; if time2=3 and _time2=5 then delete; if time2=3 and _time2=6 then delete; if time2=4 and _time2=6 then delete; if time2=2 and _time2=1 then delete; if time2=3 and _time2=1 then delete; if time2=4 and _time2=1 then delete; if time2=5 and _time2=1 then delete; if time2=6 and _time2=1 then delete; if time2=3 and _time2=2 then delete; if time2=4 and _time2=2 then delete; if time2=5 and _time2=2 then delete; if time2=6 and _time2=2 then delete; if time2=4 and _time2=3 then delete; if time2=5 and _time2=3 then delete; if time2=6 and _time2=3 then delete; if time2=5 and _time2=4 then delete; if time2=6 and _time2=4 then delete; if time2=6 and _time2=5 then delete; if treat='C' and _treat='I' then delete; if genotype='R' and _genotype='r' then delete; if treat='I' and _treat='C' then delete; if genotype='r' and _genotype='R' then delete; run;

PAGE 90

78 APPENDIX B ASREML SCRIPT FOR ASYMTOTIC Z-TEST Test E fusiform rust only susceptible clones clone 440 !A family 61 !A female 32 !P male 32 !P rep 4 !A inc 110 !A row 55 !A col 20 !A ncol 40 !A nrow 110 !A score /gck/parped.txt !ALPHA /gck/VINF.prn score ~ mu !r rep female and(male) family clone

PAGE 91

79 APPENDIX C HEALTHY VS. DISEASED GENE LIST CloneID GenotypeTreatmentFirst time increment Second time increment Third time increment J4 R C 0 0 0 J4 R I 0 0 0 J4 S C 0 0 0 J4 S I 0 -1 0 07 E10 R C -1 0 0 07 E10 R I -1 0 0 07 E10 S C -1 0 0 07 E10 S I -1 0 1 37 G12 R C -1 0 1 37 G12 R I -1 0 1 37 G12 S C -1 0 1 37 G12 S I 0 0 1 G1 R C 0 0 0 G1 R I 0 0 0 G1 S C 0 0 0 G1 S I 0 -1 0 G12 R C 0 0 0 G12 R I 0 0 0 G12 S C 0 0 0 G12 S I 0 -1 0 G16 R C 0 0 0 G16 R I 0 0 0 G16 S C 0 0 0 G16 S I 1 -1 -1 G17 R C 0 0 0 G17 R I 0 0 0 G17 S C 0 0 0 G17 S I 0 -1 0 G27 R C 0 0 0 G27 R I 0 0 0 G27 S C 0 0 0 G27 S I 0 -1 0 G30 R C 0 0 0 G30 R I 0 0 0

PAGE 92

80 G30 S C 0 0 0 G30 S I 0 -1 0 G34 R C 0 0 0 G34 R I 0 0 0 G34 S C 0 0 0 G34 S I 0 -1 0 G35 R C 0 0 0 G35 R I 0 0 0 G35 S C 0 0 0 G35 S I 0 -1 0 G39 R C 0 0 0 G39 R I 0 0 0 G39 S C 0 0 0 G39 S I 0 -1 0 G41 R C 0 0 0 G41 R I 0 0 0 G41 S C 0 0 0 G41 S I 1 -1 0 G45 R C 0 0 0 G45 R I 0 0 0 G45 S C 0 0 0 G45 S I 1 -1 -1 G51 R C 0 0 0 G51 R I 0 0 0 G51 S C 0 0 0 G51 S I 0 -1 0 G56 R C 0 0 0 G56 R I 0 0 0 G56 S C 0 0 0 G56 S I 0 -1 0 G8 R C 0 0 0 G8 R I 0 0 0 G8 S C 0 0 0 G8 S I 0 -1 -1 H8 R C 0 -1 0 H8 R I 0 -1 0 H8 S C 0 -1 0 H8 S I 0 0 0 NXCI_002_E02 R C 0 -1 0 NXCI_002_E02 R I 0 -1 0 NXCI_002_E02 S C 0 -1 0 NXCI_002_E02 S I 0 0 0 NXCI_004_G05 R C -1 0 0 NXCI_004_G05 R I -1 0 0

PAGE 93

81 NXCI_004_G05 S C -1 0 0 NXCI_004_G05 S I -1 -1 0 NXCI_018_A08 R C -1 0 0 NXCI_018_A08 R I -1 0 0 NXCI_018_A08 S C -1 0 0 NXCI_018_A08 S I -1 -1 0 NXCI_027_G06 R C -1 1 0 NXCI_027_G06 R I -1 1 0 NXCI_027_G06 S C -1 1 0 NXCI_027_G06 S I -1 0 0 NXCI_034_F04 R C 0 0 0 NXCI_034_F04 R I 0 0 0 NXCI_034_F04 S C 0 0 0 NXCI_034_F04 S I 0 -1 -1 NXCI_042_D04_F R C -1 0 0 NXCI_042_D04_F R I -1 0 0 NXCI_042_D04_F S C -1 0 0 NXCI_042_D04_F S I -1 -1 0 NXCI_057_B05 R C -1 0 0 NXCI_057_B05 R I -1 0 0 NXCI_057_B05 S C -1 0 0 NXCI_057_B05 S I 0 -1 0 NXCI_067_H06 R C -1 0 0 NXCI_067_H06 R I -1 0 0 NXCI_067_H06 S C -1 0 0 NXCI_067_H06 S I -1 -1 0 NXCI_070_D01 R C -1 0 0 NXCI_070_D01 R I -1 0 0 NXCI_070_D01 S C -1 0 0 NXCI_070_D01 S I -1 -1 0 NXCI_075_D09 R C 0 0 0 NXCI_075_D09 R I 0 0 0 NXCI_075_D09 S C 0 0 0 NXCI_075_D09 S I 0 -1 0 NXCI_075_E11 R C -1 -1 1 NXCI_075_E11 R I -1 -1 1 NXCI_075_E11 S C -1 -1 1 NXCI_075_E11 S I -1 -1 0 NXCI_082_E07_F R C 0 -1 1 NXCI_082_E07_F R I 0 -1 1 NXCI_082_E07_F S C 0 -1 1 NXCI_082_E07_F S I 0 -1 0 NXCI_082_G01 R C -1 0 0 NXCI_082_G01 R I -1 0 0

PAGE 94

82 NXCI_082_G01 S C -1 0 0 NXCI_082_G01 S I 0 0 0 NXCI_099_A12 R C -1 0 0 NXCI_099_A12 R I -1 0 0 NXCI_099_A12 S C -1 0 0 NXCI_099_A12 S I 0 0 0 NXCI_111_C10 R C -1 0 0 NXCI_111_C10 R I -1 0 0 NXCI_111_C10 S C -1 0 0 NXCI_111_C10 S I 0 -1 0 NXCI_150_F06_F R C -1 1 -1 NXCI_150_F06_F R I -1 1 -1 NXCI_150_F06_F S C -1 1 -1 NXCI_150_F06_F S I -1 0 0 NXCI_153_A02_F R C -1 -1 0 NXCI_153_A02_F R I -1 -1 0 NXCI_153_A02_F S C -1 -1 0 NXCI_153_A02_F S I 0 -1 0 NXCI_164_A06_F R C -1 0 0 NXCI_164_A06_F R I -1 0 0 NXCI_164_A06_F S C -1 0 0 NXCI_164_A06_F S I 0 -1 0 NXLV100_F02_F R C -1 0 0 NXLV100_F02_F R I -1 0 0 NXLV100_F02_F S C -1 0 0 NXLV100_F02_F S I 0 0 0 NXLV103_E01_F R C -1 1 0 NXLV103_E01_F R I -1 1 0 NXLV103_E01_F S C -1 1 0 NXLV103_E01_F S I -1 0 -1 NXLV105_B02_F R C -1 0 0 NXLV105_B02_F R I -1 0 0 NXLV105_B02_F S C -1 0 0 NXLV105_B02_F S I -1 1 0 NXLV129_C12_F R C -1 0 0 NXLV129_C12_F R I -1 0 0 NXLV129_C12_F S C -1 0 0 NXLV129_C12_F S I -1 -1 0 NXLV_012_A05_F R C 0 -1 0 NXLV_012_A05_F R I 0 -1 0 NXLV_012_A05_F S C 0 -1 0 NXLV_012_A05_F S I 0 0 0 NXLV_022_H08_F R C -1 0 0 NXLV_022_H08_F R I -1 0 0

PAGE 95

83 NXLV_022_H08_F S C -1 0 0 NXLV_022_H08_F S I 0 0 -1 NXLV_023_D12_F R C -1 -1 0 NXLV_023_D12_F R I -1 -1 0 NXLV_023_D12_F S C -1 -1 0 NXLV_023_D12_F S I 0 0 0 NXLV_049_G11_F R C -1 0 -1 NXLV_049_G11_F R I -1 0 -1 NXLV_049_G11_F S C -1 0 -1 NXLV_049_G11_F S I -1 0 0 NXLV_079_G07_F R C -1 0 0 NXLV_079_G07_F R I -1 0 0 NXLV_079_G07_F S C -1 0 0 NXLV_079_G07_F S I -1 1 0 NXNV 129 F06 R C 1 0 0 NXNV 129 F06 R I 1 0 0 NXNV 129 F06 S C 1 0 0 NXNV 129 F06 S I 1 0 1 NXNV015H07 R C -1 1 -1 NXNV015H07 R I -1 1 -1 NXNV015H07 S C -1 1 -1 NXNV015H07 S I -1 1 0 NXNV027F10 R C 0 1 0 NXNV027F10 R I 0 1 0 NXNV027F10 S C 0 1 0 NXNV027F10 S I 0 0 0 NXNV_073_G04 R C -1 1 0 NXNV_073_G04 R I -1 1 0 NXNV_073_G04 S C -1 1 0 NXNV_073_G04 S I 0 0 0 NXNV_096_C09 R C 0 0 0 NXNV_096_C09 R I 0 0 0 NXNV_096_C09 S C 0 0 0 NXNV_096_C09 S I 0 0 -1 NXNV_118_E06 R C -1 0 0 NXNV_118_E06 R I -1 0 0 NXNV_118_E06 S C -1 0 0 NXNV_118_E06 S I 0 -1 0 NXNV_122_C07_F R C -1 0 0 NXNV_122_C07_F R I -1 0 0 NXNV_122_C07_F S C -1 0 0 NXNV_122_C07_F S I 0 0 0 NXNV_125_E12_F R C -1 0 0 NXNV_125_E12_F R I -1 0 0

PAGE 96

84 NXNV_125_E12_F S C -1 0 0 NXNV_125_E12_F S I -1 -1 0 NXNV_132_G11_F R C -1 1 0 NXNV_132_G11_F R I -1 1 0 NXNV_132_G11_F S C -1 1 0 NXNV_132_G11_F S I -1 0 0 NXNV_135_E01_F R C -1 -1 0 NXNV_135_E01_F R I -1 -1 0 NXNV_135_E01_F S C -1 -1 0 NXNV_135_E01_F S I 0 -1 0 NXNV_147_G03_F R C -1 1 0 NXNV_147_G03_F R I -1 1 0 NXNV_147_G03_F S C -1 1 0 NXNV_147_G03_F S I 0 0 0 NXNV_159_G03 R C 0 -1 0 NXNV_159_G03 R I 0 -1 0 NXNV_159_G03 S C 0 -1 0 NXNV_159_G03 S I 0 0 0 NXNV_163_F07_F R C -1 0 0 NXNV_163_F07_F R I -1 0 0 NXNV_163_F07_F S C -1 0 0 NXNV_163_F07_F S I 0 0 0 NXNV_173_B11_F R C -1 1 0 NXNV_173_B11_F R I -1 1 0 NXNV_173_B11_F S C -1 1 0 NXNV_173_B11_F S I 0 0 0 NXPV_062_E04_F R C -1 0 0 NXPV_062_E04_F R I -1 0 0 NXPV_062_E04_F S C -1 0 0 NXPV_062_E04_F S I 0 0 0 NXPV_068_E06_F R C -1 1 -1 NXPV_068_E06_F R I -1 1 -1 NXPV_068_E06_F S C -1 1 -1 NXPV_068_E06_F S I -1 1 0 NXPV_076_C12_F R C -1 1 -1 NXPV_076_C12_F R I -1 1 -1 NXPV_076_C12_F S C -1 1 -1 NXPV_076_C12_F S I -1 1 0 NXRV064_C07_F R C -1 0 -1 NXRV064_C07_F R I -1 0 -1 NXRV064_C07_F S C -1 0 -1 NXRV064_C07_F S I -1 0 0 NXRV079_D01_F R C -1 1 -1 NXRV079_D01_F R I -1 1 -1

PAGE 97

85 NXRV079_D01_F S C -1 1 -1 NXRV079_D01_F S I -1 1 0 NXRV118_B08_F R C -1 0 0 NXRV118_B08_F R I -1 0 0 NXRV118_B08_F S C -1 0 0 NXRV118_B08_F S I -1 -1 0 NXSI 008 G11 R C 0 -1 0 NXSI 008 G11 R I 0 -1 0 NXSI 008 G11 S C 0 -1 0 NXSI 008 G11 S I 1 -1 -1 NXSI 059 G09 R C 0 -1 0 NXSI 059 G09 R I 0 -1 0 NXSI 059 G09 S C 0 -1 0 NXSI 059 G09 S I 1 -1 0 NXSI 102 F11 R C -1 -1 0 NXSI 102 F11 R I -1 -1 0 NXSI 102 F11 S C -1 -1 0 NXSI 102 F11 S I 0 -1 0 NXSI_013_C04 R C 0 -1 0 NXSI_013_C04 R I 0 -1 0 NXSI_013_C04 S C 0 -1 0 NXSI_013_C04 S I 1 -1 0 NXSI_027_G10 R C -1 0 0 NXSI_027_G10 R I -1 0 0 NXSI_027_G10 S C -1 0 0 NXSI_027_G10 S I -1 -1 0 NXSI_040_C01 R C -1 0 0 NXSI_040_C01 R I -1 0 0 NXSI_040_C01 S C -1 0 0 NXSI_040_C01 S I -1 -1 0 NXSI_041_B01 R C 0 -1 0 NXSI_041_B01 R I 0 -1 0 NXSI_041_B01 S C 0 -1 0 NXSI_041_B01 S I 0 0 0 NXSI_055_H08 R C -1 0 0 NXSI_055_H08 R I -1 0 0 NXSI_055_H08 S C -1 0 0 NXSI_055_H08 S I -1 1 0 NXSI_060_E02 R C -1 0 0 NXSI_060_E02 R I -1 0 0 NXSI_060_E02 S C -1 0 0 NXSI_060_E02 S I 0 -1 0 NXSI_063_D01 R C -1 1 0 NXSI_063_D01 R I -1 1 0

PAGE 98

86 NXSI_063_D01 S C -1 1 0 NXSI_063_D01 S I -1 0 0 NXSI_064_A03 R C 1 -1 0 NXSI_064_A03 R I 1 -1 0 NXSI_064_A03 S C 1 -1 0 NXSI_064_A03 S I 1 -1 -1 NXSI_067_F10_F R C -1 0 0 NXSI_067_F10_F R I -1 0 0 NXSI_067_F10_F S C -1 0 0 NXSI_067_F10_F S I 0 0 0 NXSI_069_F12_F R C -1 1 -1 NXSI_069_F12_F R I -1 1 -1 NXSI_069_F12_F S C -1 1 -1 NXSI_069_F12_F S I -1 0 0 NXSI_076_E08 R C -1 1 0 NXSI_076_E08 R I -1 1 0 NXSI_076_E08 S C -1 1 0 NXSI_076_E08 S I -1 0 0 NXSI_092_E10 R C -1 -1 1 NXSI_092_E10 R I -1 -1 1 NXSI_092_E10 S C -1 -1 1 NXSI_092_E10 S I -1 -1 0 NXSI_098_C01 R C 1 -1 0 NXSI_098_C01 R I 1 -1 0 NXSI_098_C01 S C 1 -1 0 NXSI_098_C01 S I 1 -1 -1 NXSI_099_H06_F R C -1 0 0 NXSI_099_H06_F R I -1 0 0 NXSI_099_H06_F S C -1 0 0 NXSI_099_H06_F S I -1 1 0 NXSI_101_B01 R C -1 0 0 NXSI_101_B01 R I -1 0 0 NXSI_101_B01 S C -1 0 0 NXSI_101_B01 S I -1 -1 0 NXSI_103_D11_F R C -1 1 0 NXSI_103_D11_F R I -1 1 0 NXSI_103_D11_F S C -1 1 0 NXSI_103_D11_F S I -1 0 0 NXSI_103_E12_F R C -1 1 -1 NXSI_103_E12_F R I -1 1 -1 NXSI_103_E12_F S C -1 1 -1 NXSI_103_E12_F S I -1 1 0 NXSI_104_B11 R C 0 0 1 NXSI_104_B11 R I 0 0 1

PAGE 99

87 NXSI_104_B11 S C 0 0 1 NXSI_104_B11 S I 0 1 1 NXSI_114_A04 R C 0 -1 1 NXSI_114_A04 R I 0 -1 1 NXSI_114_A04 S C 0 -1 1 NXSI_114_A04 S I 0 -1 0 NXSI_115_A12_F R C -1 0 0 NXSI_115_A12_F R I -1 0 0 NXSI_115_A12_F S C -1 0 0 NXSI_115_A12_F S I 0 0 0 NXSI_133_G11 R C 0 -1 0 NXSI_133_G11 R I 0 -1 0 NXSI_133_G11 S C 0 -1 0 NXSI_133_G11 S I 0 0 0 pi134-1 R C 0 1 0 pi134-1 R I 0 1 0 pi134-1 S C 0 1 0 pi134-1 S I 0 0 0 pi150-2 R C 0 -1 1 pi150-2 R I 0 -1 1 pi150-2 S C 0 -1 1 pi150-2 S I 0 -1 0 pi226 R C 0 0 0 pi226 R I 0 0 0 pi226 S C 0 0 0 pi226 S I 0 -1 0 pi243 R C 0 0 0 pi243 R I 0 0 0 pi243 S C 0 0 0 pi243 S I 1 -1 -1 pi266 R C 0 -1 1 pi266 R I 0 -1 1 pi266 S C 0 -1 1 pi266 S I 0 0 1 pi295 R C 0 0 0 pi295 R I 0 0 0 pi295 S C 0 0 0 pi295 S I 1 0 -1 pi310 R C 1 0 0 pi310 R I 1 0 0 pi310 S C 1 0 0 pi310 S I 1 -1 0 pi64-9 R C 0 0 0 pi64-9 R I 0 0 0

PAGE 100

88 pi64-9 S C 0 0 0 pi64-9 S I 0 -1 -1

PAGE 101

89 APPENDIX D GENES THAT ARE REGULATED ACROSS TIME 01 D11 01 F09 01 F11 02 A06 02 B01 02 B03 02 C04 02 D01 02 G09 03 B07 03 D10 03 E05 03 F07 03 G03 03 H06 04 A02 04 D09 04 E10 04 F05 04 H02 06 A10 06 B06 06 C03 06 C04 06 F05 06 G07 06 H04 06 H05 07 D02 07 E10 08 A10 08 B05 08 H06 12 D06 12 E03 12 E05 13 C12 13 F12 13 H06 14 G06

PAGE 102

90 15 C08 15 G05 16 C01 21 B10 21 E01 21 E10 22 H10 23 A08 23 E11 23 F03 23 G12 24 F06 25 E06 27 B08 27 G04 27 G09 27 G10 30 C04 30 G05 30 H08 33 H11 34 E09 37 D08 37 D12 37 E04 37 E10 37 F05 37 G12 38 F06 38 G04 40 A03 40 D05 40 E09 40 F04 Eli3 G1 G10 G11 G12 G13 G16 G17 G18 G19 G2 G26

PAGE 103

91 G27 G3 G30 G31 G32 G33 G34 G35 G36 G37 G38 G39 G41 G45 G46 G51 G55 G56 G57 G59 G6 G60 G63 G65 G66 G67 G68 G7 G72 G73 G75 G8 G9 H16 H17 H21 H24 H25 H27 H39 H5 H6 H7 H8 J10

PAGE 104

92 J11 J16 J18 J19 J2 J4 NXCI 048 F02 NXCI 056 A03 NXCI 069 B02 NXCI 147 C04 NXCI_002_A12 NXCI_002_D01_F NXCI_002_E02 NXCI_002_E07 NXCI_002_G10_F NXCI_002_G11_F NXCI_002_H01_F NXCI_002_H03_F NXCI_002_H04 NXCI_004_G05 NXCI_005_G03 NXCI_007_G08_F NXCI_008_C01 NXCI_008_D06_F NXCI_008_D12_F NXCI_008_F09_F NXCI_008_G03_F NXCI_008_H07_F NXCI_008_H10 NXCI_009_A10 NXCI_009_B05_F NXCI_009_B08_F NXCI_018_A08 NXCI_018_D03 NXCI_018_H04 NXCI_019_E11 NXCI_020_A02 NXCI_020_G08_F NXCI_021_D03 NXCI_022_E07 NXCI_022_G01_F NXCI_023_D01 NXCI_023_F12 NXCI_025_F02 NXCI_025_G06 NXCI_026_C06

PAGE 105

93 NXCI_026_D09_F NXCI_026_F10 NXCI_027_E04 NXCI_027_F06_F NXCI_027_F08_F NXCI_027_G06 NXCI_028_B02_F NXCI_029_G10_F NXCI_029_H10_F NXCI_031_A08_F NXCI_031_C04_F NXCI_031_E05 NXCI_033_C02 NXCI_033_F03 NXCI_034_B01 NXCI_034_F04 NXCI_037_B03_F NXCI_040_B11 NXCI_041_E04 NXCI_042_D04_F NXCI_042_D08 NXCI_043_A11 NXCI_043_F09_F NXCI_044_A12 NXCI_044_F11_F NXCI_045_C01 NXCI_045_G05_F NXCI_046_E05 NXCI_047_A08_F NXCI_047_C05 NXCI_048_H04 NXCI_050_F08 NXCI_055_C01 NXCI_055_C06 NXCI_057_B05 NXCI_057_E03 NXCI_057_E05 NXCI_058_C02 NXCI_058_H01_F NXCI_060_A12_F NXCI_061_B09 NXCI_061_F02_F NXCI_062_H01_F NXCI_064_E04 NXCI_066_A11 NXCI_066_F01

PAGE 106

94 NXCI_066_G08 NXCI_066_H04 NXCI_067_A10 NXCI_067_H06 NXCI_069_A02 NXCI_069_H11 NXCI_070_B10 NXCI_070_D01 NXCI_070_E11 NXCI_070_G08 NXCI_071_B03 NXCI_071_C01 NXCI_071_F03 NXCI_075_B02_F NXCI_075_C07 NXCI_075_D09 NXCI_075_E11 NXCI_076_A09 NXCI_076_A10 NXCI_076_E05_F NXCI_076_F07 NXCI_076_F09 NXCI_076_G08_F NXCI_082_E07 NXCI_082_E07_F NXCI_082_G01 NXCI_082_G01_F NXCI_083_A06 NXCI_083_D09_F NXCI_084_A07_F NXCI_084_G02 NXCI_085_E04 NXCI_085_E12 NXCI_085_H12 NXCI_086_A09 NXCI_087_F06 NXCI_087_F07 NXCI_093_E01 NXCI_093_H05 NXCI_094_C09 NXCI_094_C11 NXCI_094_E12 NXCI_094_G11 NXCI_095_C01 NXCI_095_D10 NXCI_096_A07

PAGE 107

95 NXCI_096_C05 NXCI_097_A07 NXCI_098_D10 NXCI_098_F10 NXCI_099_A12 NXCI_101_B10 NXCI_102_F06 NXCI_103_A12 NXCI_106_C10 NXCI_106_F03_F NXCI_107_G09_F NXCI_107_H04_F NXCI_108_E05 NXCI_109_F09 NXCI_111_C10 NXCI_114_B08 NXCI_115_C04 NXCI_116_C11_F NXCI_116_D01 NXCI_118_F05 NXCI_122_H05 NXCI_124_C07 NXCI_125_D10_F NXCI_127_D04_F NXCI_128_G07_F NXCI_130_C09 NXCI_131_H09 NXCI_132_B08_F NXCI_132_B11 NXCI_132_E06 NXCI_132_G02_F NXCI_132_H04 NXCI_134_B04_F NXCI_134_H12 NXCI_136_A08 NXCI_144_F06 NXCI_149_E03 NXCI_149_F01 NXCI_149_F01_F NXCI_149_H12 NXCI_150_A07 NXCI_150_F06_F NXCI_151_G03_F NXCI_153_A02_F NXCI_153_F03_F NXCI_153_G06

PAGE 108

96 NXCI_153_G06_F NXCI_153_H08_F NXCI_155_E05_F NXCI_155_H03 NXCI_155_H06_F NXCI_157_D11_F NXCI_162_A05_F NXCI_164_A06_F NXCI_164_B04_F NXCI_164_C11_F NXLV082_A10_F NXLV082_F03_F NXLV085_B05_F NXLV088_B11_F NXLV090_G07_F NXLV098_B10_F NXLV098_E06_F NXLV100_B02_F NXLV100_F02_F NXLV101_A05_F NXLV101_G02_F NXLV103_E01_F NXLV105_B02_F NXLV105_E07_F NXLV106_G06_F NXLV106_G10_F NXLV109_B02_F NXLV111_E05_F NXLV112_B02_F NXLV112_H10_F NXLV118_B08_F NXLV127_E02_F NXLV129_C12_F NXLV133_D07_F NXLV_007_D11_F NXLV_009_A11_F NXLV_010_H05_F NXLV_012_A05_F NXLV_014_C04_F NXLV_020_A08_F NXLV_020_E04_F NXLV_022_E02_F NXLV_022_H08_F NXLV_023_D12_F NXLV_024_A02_F NXLV_024_G06_F

PAGE 109

97 NXLV_029_D05_F NXLV_030_A06_F NXLV_031_D02_F NXLV_037_E01_F NXLV_039_H10_F NXLV_041_H10_F NXLV_049_G11_F NXLV_065_A09_F NXLV_070_F01_F NXLV_077_B11_F NXLV_077_H10_F NXLV_079_G07_F NXLV_080_H03_F NXNV 074 G09 NXNV 100 H01 NXNV 118 C02 NXNV 129 F06 NXNV002C02 NXNV003F06 NXNV005A03 NXNV005D01 NXNV005H04 NXNV006A11 NXNV015F05 NXNV015H07 NXNV018A10 NXNV019E06 NXNV021C03 NXNV027B07 NXNV027F10 NXNV028A02 NXNV047B02 NXNV047B05 NXNV047E10 NXNV_001_B08 NXNV_005_B04 NXNV_007_G05 NXNV_007_G06 NXNV_008_F05 NXNV_009_C04 NXNV_010_H01 NXNV_015_H07 NXNV_018_E08 NXNV_044_C04 NXNV_044_D09 NXNV_044_G02

PAGE 110

98 NXNV_044_G05 NXNV_045_E12 NXNV_046_A05 NXNV_046_B04 NXNV_046_D01 NXNV_046_F03 NXNV_048_F06 NXNV_060_F07 NXNV_060_H10 NXNV_066_B07 NXNV_067_A11 NXNV_067_B01 NXNV_070_F06 NXNV_071_H03 NXNV_072_F02_F NXNV_072_G08_F NXNV_073_G04 NXNV_074_G01 NXNV_074_G06 NXNV_074_H01_F NXNV_078_B01_F NXNV_079_G02 NXNV_081_D10 NXNV_083_A10 NXNV_083_D09 NXNV_083_E04 NXNV_083_E11_F NXNV_086_B04 NXNV_086_C07 NXNV_089_A02 NXNV_089_B08 NXNV_089_B08_F NXNV_089_C07_F NXNV_089_E08_F NXNV_089_G04 NXNV_091_F02 NXNV_095_B07_F NXNV_095_C08_F NXNV_095_F08_F NXNV_096_A01_F NXNV_096_A04 NXNV_096_C09 NXNV_096_E01 NXNV_098_D05 NXNV_105_D01_F NXNV_106_A05

PAGE 111

99 NXNV_106_A11_F NXNV_106_F12_F NXNV_108_E11_F NXNV_108_G08_F NXNV_117_G05_F NXNV_118_E06 NXNV_120_E10 NXNV_120_F03_F NXNV_120_F05 NXNV_120_G04_F NXNV_120_H02 NXNV_122_C07_F NXNV_122_E06_F NXNV_123_B06 NXNV_124_H02_F NXNV_125_E12_F NXNV_127_E04 NXNV_127_F09 NXNV_128_D06_F NXNV_129_G07 NXNV_130_G07 NXNV_131_F05 NXNV_131_F08 NXNV_132_B03 NXNV_132_G06_F NXNV_132_G11_F NXNV_133_A09_F NXNV_133_G06 NXNV_133_H03_F NXNV_134_H10 NXNV_135_E01_F NXNV_136_C05_F NXNV_136_C12_F NXNV_139_B12 NXNV_139_C09 NXNV_139_C11_F NXNV_144_C01_F NXNV_145_F12 NXNV_147_G03_F NXNV_147_H08 NXNV_148_E12_F NXNV_148_G06_F NXNV_149_B07 NXNV_149_E07 NXNV_150_D05_F NXNV_151_G10

PAGE 112

100 NXNV_153_F09 NXNV_158_B06_F NXNV_158_D09 NXNV_159_C03_F NXNV_159_G03 NXNV_160_C09 NXNV_160_C10 NXNV_162_D12_F NXNV_162_H07 NXNV_162_H07_F NXNV_163_F07_F NXNV_163_G07 NXNV_163_G09 NXNV_163_G10_F NXNV_164_H08 NXNV_173_B11_F NXNV_173_C05_F NXNV_173_E07 NXNV_181_A11_F NXNV_181_B11_F NXNV_181_F08_F NXNV_181_H08 NXNV_185_D06 NXNV_185_F02 NXNV_186_A10 NXNV_186_C11_F NXNV_186_G05 NXPV_007_H07_F NXPV_008_H09_F NXPV_010_A07_F NXPV_010_B09_F NXPV_010_C08 NXPV_010_E07_F NXPV_010_H01_F NXPV_011_F10_F NXPV_012_H12_F NXPV_013_A04_F NXPV_013_C08_F NXPV_020_G09_F NXPV_021_F10_F NXPV_021_G10_F NXPV_025_E07_F NXPV_028_D12_F NXPV_028_H06_F NXPV_035_A07_F NXPV_037_C02_F

PAGE 113

101 NXPV_041_A04_F NXPV_041_B08_F NXPV_042_C11_F NXPV_044_F01_F NXPV_048_C09_F NXPV_049_C03_F NXPV_052_D04_F NXPV_055_C02_F NXPV_056_F09_F NXPV_062_E04_F NXPV_062_E05_F NXPV_063_A12_F NXPV_066_B11_F NXPV_068_E06_F NXPV_075_B11_F NXPV_076_A06_F NXPV_076_C12_F NXPV_077_B05_F NXPV_078_G08_F NXPV_079_D06_F NXPV_088_C08_F NXPV_094_G04_F NXPV_096_D02_F NXPV_097_B11_F NXPV_108_B07_F NXPV_123_C07_F NXPV_128_F03_F NXPV_133_B10_F NXRV055_C03_F NXRV060_D09_F NXRV060_D10_F NXRV061_G09_F NXRV061_H10_F NXRV062_H08_F NXRV064_C07_F NXRV064_G04_F NXRV066_G02_F NXRV066_H10_F NXRV072_A01_F NXRV075_C07_F NXRV077_A04_F NXRV078_H08_F NXRV079_D01_F NXRV084_E09_F NXRV087_B06_F NXRV097_A07_F

PAGE 114

102 NXRV100_H07_F NXRV101_H07_F NXRV107_G02_F NXRV112_F01_F NXRV112_F11_F NXRV114_B02_F NXRV117_A05_F NXRV118_B08_F NXRV118_E03_F NXRV120_F06_F NXRV125_B10_F NXRV126_E10_F NXRV128_D08_F NXRV130_E04_F NXRV132_G06_F NXRV_003_A04_F NXRV_003_H02_F NXRV_011_E07_F NXRV_016_F09_F NXRV_017_A01_F NXRV_022_D06_F NXRV_025_D11_F NXRV_025_E11_F NXRV_025_E12_F NXRV_037_H06_F NXS1_121_C05 NXSI 044 F04 NXSI 060 H06 NXSI 124 B08 NXSI 008 G11 NXSI 028 D09 NXSI 059 G09 NXSI 102 F11 NXSI 117 B05 NXSI 123 D02 NXSI_007_A07_F NXSI_012_H11 NXSI_013_C04 NXSI_021_B12 NXSI_021_D01 NXSI_021_E09 NXSI_025_H02 NXSI_027_G10 NXSI_028_G05 NXSI_029_F11 NXSI_030_C06

PAGE 115

103 NXSI_031_G08 NXSI_036_C05 NXSI_039_D01 NXSI_040_C01 NXSI_040_D02 NXSI_040_D03_F NXSI_041_B01 NXSI_041_H12 NXSI_042_G07_F NXSI_042_H05 NXSI_043_C03 NXSI_043_H03 NXSI_044_C10 NXSI_045_A04 NXSI_045_B09 NXSI_045_G03 NXSI_046_B05 NXSI_047_A11_F NXSI_049_A04_F NXSI_050_C07_F NXSI_051_G07 NXSI_053_B02 NXSI_053_G04 NXSI_053_G05 NXSI_054_A01 NXSI_054_A09 NXSI_055_B06 NXSI_055_F11 NXSI_055_H08 NXSI_056_F07 NXSI_058_B04 NXSI_058_G02 NXSI_060_B07 NXSI_060_E02 NXSI_061_F04_F NXSI_062_E07 NXSI_063_D01 NXSI_063_E04 NXSI_064_A03 NXSI_064_H06 NXSI_065_C08 NXSI_067_C11 NXSI_067_F10_F NXSI_067_H09_F NXSI_068_G09 NXSI_069_F12_F

PAGE 116

104 NXSI_073_F05 NXSI_075_B04_F NXSI_076_E08 NXSI_077_F09 NXSI_079_D06 NXSI_079_D09_F NXSI_081_D01 NXSI_082_H01 NXSI_083_G10_F NXSI_088_C05_F NXSI_089_E04 NXSI_089_H07 NXSI_090_C05 NXSI_092_E10 NXSI_092_H03_F NXSI_096_G02_F NXSI_097_H07 NXSI_098_A04 NXSI_098_C01 NXSI_099_F06 NXSI_099_F10 NXSI_099_G06 NXSI_099_H06_F NXSI_100_A04 NXSI_100_C11 NXSI_100_D07 NXSI_100_F02 NXSI_100_F12 NXSI_101_B01 NXSI_101_E11 NXSI_101_H03_F NXSI_102_D03 NXSI_102_F12 NXSI_102_H05 NXSI_103_A08 NXSI_103_A10 NXSI_103_B01 NXSI_103_C04 NXSI_103_D11_F NXSI_103_E12_F NXSI_103_F08 NXSI_103_H03 NXSI_104_B11 NXSI_104_E11 NXSI_104_H10 NXSI_105_G10

PAGE 117

105 NXSI_107_C09 NXSI_108_D12 NXSI_108_H05 NXSI_112_B07 NXSI_112_D01 NXSI_112_D08 NXSI_112_G05_F NXSI_113_B09 NXSI_113_C10_F NXSI_113_D07 NXSI_113_E06_F NXSI_113_G11_F NXSI_113_H02 NXSI_114_A04 NXSI_114_D12 NXSI_114_G07 NXSI_116_A11 NXSI_116_B04 NXSI_116_F02 NXSI_117_B05 NXSI_117_C06_F NXSI_118_A03 NXSI_118_B03 NXSI_118_F05_F NXSI_119_D08 NXSI_119_F11 NXSI_120_D02_F NXSI_121_A05 NXSI_121_F04_F NXSI_122_H10 NXSI_124_C04 NXSI_127_E02 NXSI_131_C03 NXSI_132_F03 NXSI_132_H01 NXSI_115_A12_F NXSI_133_B03_F NXSI_133_B05 NXSI_133_G06 NXSI_133_G11 NXSI_134_E02 NXSI_136_C07_F NXSI_137_D09 NXSI_137_E06 NXSI_142_F05 NXSI_143_G11

PAGE 118

106 NXSI_144_H01 NXSI_145_D04 NXSI_145_E11_F PC 03 F07 PC 04 B12 PC 05 A11 PC 07 F10 PC 18 B08 PC 19 E01 ST 32 C09 ST_02_E09 ST_06_F05 ST_15_G05 ST_19_A09 ST_20_B02 ST_21_E01 ST_22_F09 ST_23_F07 ST_24_H10 ST_25_C07 ST_29_A08 ST_35_A01 ST_35_B03 ST_35_D08 ST_36_C08 ST_37_B11 ST_40_A03 pi107-2 pi111-4 pi113-1 pi115-1 pi118-1 pi121-1 pi129-1 pi134-1 pi143-1 pi148-1 NXSI_136_H09_F pi150-2 pi152-5 pi167-5 pi191-2 pi193-3 pi194-1 pi196-1 pi201-2

PAGE 119

107 pi226 pi235 pi240 pi243 pi255 pi261 pi263 pi266 pi267 pi270 pi271 pi273 pi274 pi275 pi278 pi284 pi287 pi288 pi293 pi295 pi305 pi306 pi310 pi311b pi315 pi46-1 pi54-5 pi59-1 pi64-9 pi70-2 pi73-1 pi76-1 pi78-1 pi79-2 pi90-2 pi97-3 pic56-12 pidd1

PAGE 120

108 APPENDIX E MICROARRAY PROCEDURE Indirect Incorporation of Cy Dyes SuperScript Indirect cDNA Labeling from total RNA First-Strand cDNA Synthesis us ing Invitrogen's SuperScrip t Indirect cDNA Labeling Kit (Catalog numbers L1014-01 and L1014-02). What follows is the protocol I use for c DNA synthesis from total RNA and indirect cDNA labeling prior to microarray hybridizations. The protocol be low is similar to that in the instruction manual provided with this ki t but has been modified previously by Dr. Rob Alba. This is a copy of his proce dure and I have added a couple of my own modifications. His originals can be found at http://ted.bti.cornell.edu/array/interface/protocol/protocol.html First-Strand cDNA Synthesis Rxn. Mix and briefly spin each kit component before use. Prepare rxns as follows: Xul DEPC-H2O Xul Rnase free Dnase treated total RNA (15 to 20 ug/rxn) 2ul Anchored Oligo(dT)20 Primer (2.5ug/ul) 1ul Random Hexamer Total Volume = 18ul Incubate tubes at 70oC for 5 min, and then place on ice for at least 1 min.

PAGE 121

109 Add the following to each rxn tube on ice: 6ul 5X First-Strand buffer 1.5ul 0.1 M DTT 1.5ul 10-mM dNTP mix 1ul RNaseOUT (40U/ul) 2ul SuperScript III RT (400U/ul) Total Volume = 30ul Mix gently and spin briefly. In cubate tube at 46oC Overnight Add 15ul of 1N NaOH to each r xn tube and mix thoroughly. Incubate tube at 70o C for 10 min. Add 15ul of 1N HCl; mix gently. Add 20ul 3M NaOAc (pH 5.2); mix gently. Purifying First-Strand cDNA. Add 500ul of Loading Buffer to the c DNA (from Step 9) and mix well. Place a SNAP Column on a collection tube and load your cDNA on the column. Spin at 14,000g at room temp for 1 min; discard the flow-through. Place the SNAP Column onto the same collection tube and add 500ul of Wash Buffer. Spin at 14,000g at room temp for 1 min; discard the flow-through. Repeat Steps 4 and 5 twice more, fo r a total of thr ee 500ul washes. Spin one more time at 14,000g at room te mp for 1 min; discard the flow-through. Place the SNAP Column onto a new 1.5-ml tube. Add 50ul of DEPC-treated wate r to the SNAP Column and incubate at room temp for 1 min. Elute the cDNA via spin at 14,000g at room temp for 1 min.

PAGE 122

110 Repeat Step 9, using the same 1.5-ml tube. Add 10ul of 3M sodium acetate (pH 5.2) to the eluent from steps 9 and 10. Add 4ul of glycogen (20mg/ml) to the tube and mix. Add 250ul of ice-cold 100% EtOH, and in cubate the tube -8 0oC for 30 min. Spin the tube at 14,000g at 4oC for 20 mi n. Carefully remove the supernatant. Add 500ul of ice-cold 70% EtOH and spin th e tube at 14,000g for 2 min. Carefully remove the supernatant. Air dry the sample for 5-10 min; ensure that al l EtOH is removed. I can let it sit for a few weeks in fridge. Warm the 2X Coupling Buffer at 37oC for 5 min and re-suspend the cDNA sample in 5ul of warm 2X Coupling Buffer. Heat the cDNA/Coupling buffer at 50oC for 10 min and vortex well. Ensure that your cDNA pellet is fully re-suspended in the 2X Coupling Buffer. Labeling with Fluorescent Dye. When prep aring the rxn, be careful to minimize exposure of the dye solution to light. Also DMSO is hygroscopic and will absorb moisture from the air, which will react with the NHS ester of the dye and significantly reduce the coupling rxn effici ency. Keep the DMSO supplie d in the kit in an amber screw-capped vial at -20oC, and let the vial warm to room temperature before opening to prevent condensation. Use only the DMSO provided with this kit. Open one packet of Cy3or Cy5dye and a dd 45ul of DMSO directly to the dye vial. Add 5ul of the DMSO/dye solution to the tube from Step 17. Mix well and incubate the tube at ro om temp in the dark for 1 hr. Add 20ul of 3M Sodium Acetate (pH 5.2) to the dye-coupled cDNA solution.

PAGE 123

111 Add 500ul of Loading Buffer to the c DNA solution. Mix well by vortexing. Place a SNAP Column onto a clear collection tube and load the cDNA/buffer solution. Spin at 14,000g at room temp for 1 min; discard the flow-through. Place the SNAP Column on the same collecti on tube; add 500ul of Wash Buffer to column. Spin at 14,000g at room temp for 1 min; discard the flow-through. Repeat Steps 8-9 three times, for a total of four 500ul washes. Spin one more time at 14,000g at room te mp for 1 min; discard the flow-through. Place the SNAP Column onto a new amber collection tube. Add 63ul of DEPC-water to the SNAP Column and incubate at room temp for 1 min. Spin at 14,000g at room temp for 1 min a nd collect the flow-thr ough. The flow-through should contain 60ul of your purified dye-coupled cDNA. Assessing the Labeling Procedure. Use UV/ VIS spectroscopy to assess the labeling procedure prior to microarray hybridization. Th is technique is described briefly in the Appendix of the Instruction Manual for th e Invitrogen cDNA labeling kit (page 8). Additional information for assessing your labe ling rxns can be obtained from the TIGR website (TIGR Standard Operating Procedur e #M004). Optimal labeling reactions have the following characteristics: A260 > 0.5; A 450 (for Cy3 labeling) < 0.2; A520 (for Cy5 labeling) < 0.2; pmols cDNA > 1000; pmols Cy > 100; nucleotides per Cy molecule < 50. The Instruction Manual for Corning's Pront o!Plus kit suggests that Frequency of Incorporation (FOI) and the FOI/Yield Ratio be determined as well. Preferably, FOI 20 to 50, and FOI/Yield Ratio 4 to 8.5%

PAGE 124

112 Hybridization What follows is the protocol I use for pr e-hybridization, hybridi zation, washes, and subsequent scanning of our Pine microarrays. These protocols were derived from similar ones developed by Dr. Rob Alba at the Boyce Thompson Institute at Cornell University http://ted.bti.cornell.edu/ The pre-hybridization and wash protocols derive from the lab of John Quackenbush at TIGR (http://atarrays.tigr.org/PDF/Probehyb.pdf ), with a few modifications. Reagents/Materials Required Pre-hybridization ("block" ) solution, (5X SSC, 0.1% SDS, 1% BSA) Filter pre-hyb solution using a 0.2um Corning filter unit Wash Solution #1 (1X SSC, 0.2% SDS; pre-heat to 43oC) Filter wash solution using a 0.2um Corning filter unit Wash Solution #2 (0.1X SSC, 0.2% SDS; room temperature) Filter wash solution using a 0.2um Corning filter unit Wash Solution #3 (0.1X SSC; room temperature) Filter wash solution using a 0.2um Corning filter unit Isopropyl alcohol 0.1% SDS Milli-Q H20 Coplin staining jars LifterSlipsTM(Erie Scientific Co.; catalog #22X50I-2-4711) 50mL FalconTM tubes (place the cap from a 14mL FalconTM culture tube at the bottom of each 50mL FalconTM tube to elevate the arrays during centrifugation) Genomic Solutions Hybridizati on Chambers (catalog # JHYB200003)

PAGE 125

113 Hybridization Solution: 50% formamide, 5X SSC, 5X Denhardt's solution, 0.5% SDS, 5mM potassium phosphate monobasic. Filter hyb solution using a 0.2um Corning filter unit. Pre-hybridization/Blocking the Arrays Pre-heat (to 43oC) mL of the filtered prehybridization solution in a Coplin jar. Incubate arrays in warm prehybridization solution for 45min. Rinse the "blocked" arrays via 5 dips in Mill i-Q H20, 5 more dips in fresh Milli-Q H20, and three dips in 100% isopropyl alcohol. Quickly dry the rinsed arrays by centrif ugation (1min; 1000 rpm) in a 50mL FalconTM tube. Do not let arrays dry out prior to centrifugation. Excessive centrifugal force will crack array slides. Inspect arrays carefully after centrifugation; if they are not clean, repeat steps 3 and 4. Use arrays immediately af ter pre-hybridization. Preparation of Cy-Labeled cDNA Targets Conduct a spectrophotometric assay to asses the success of each la beling reaction using the procedure described in TIGR's St andard Operating Procedure #M004. Refer to: http://pga.tigr.org/sop/M004_1a.pdf). Using the formulae presented in TIGR's SOP #M004, calculate the total pmol of synthesized cDNA, the total pmol of incorpor ated Cy dye for each labeling reaction, and the nucleotide/dye ratio for each reaction. Optimal labeling reactions ge nerate >2000 pmols of cDNA, >150 pmol of Cy dye, and a nucleotide/dye ratio that is <50.

PAGE 126

114 For each Cy-labeled cDNA sample that will be used in a two-color (competitive) hybridization, calculate the volume of Cy5-labeled cDNA and Cy3-labeled cDNA that is equivalent to 50 pmol of incorporated Cy5 and 50 pmol of incorporated Cy3. Combine the volumes calculated in step 3 in a single microfuge tube and dry the combined cDNA targets in a roto-evaporator (45oC). Resuspend the combined and dried cDNA targets in 70ul of Corning Universal Hybridization Solution. Incubate re-suspended cDNA targets at 95oC for 5min Centrifuge at max speed for 1 minute (room temperature). Clean "blocked" slides by blowing compresse d air passed through a 0.2 m filter disk and then place array in a Genomic Solutions (or similar, e.g. Corning) Hybridization Chamber; fill humidity wells as per the inst ructions that come with the hybridization chambers (I use 20 ul) Clean LifterSlipsTM by blowing compressed air passed through a 0.2 m filter disk Carefully cover the array with a clean dry LifterSlipTM. Carefully pipette 65ul of re-suspended cDNA ta rgets to the edge of the LifterSlipTM, allowing the solution to cover the array by capillary action. Seal the array chamber (without moving the LifterSlipTM) as per the instructions that come with the hybridization chambers. Incubate the sealed chamber contai ning array at 43oC for 12 hours. Conduct the hybridizat ion in the dark. Washing Arrays afte r Hybridization Fill two foil-covered Coplin jars wi th pre-heated Wash Solution #1.

PAGE 127

115 Remove the LifterSlipTM from the array surface by dipping a rrays in the first Coplin jar containing Wash Solution #1; th e LifterSlipTM should slide off the array easily. Place the array in the second Coplin jar c ontaining pre-heated Wash Solution #1. Incubate arrays in Wash Solu tion #1 for 10 minutes at 43oC. Transfer arrays to a new foil-covered Coplin jar containing Wash Solution #2. Incubate arrays in Wash Solution #2 for 10 minutes at room temperature Agitate arrays gently during wash step. Transfer arrays to a new foil-covered Coplin jar containing Wash Solution #3. Incubate arrays in Wash Solution #3 for 10 minutes at room temperature Agitate arrays gently during wash step. Immediately dry the washed arrays via ge ntle centrifugation, as described above. Do not let the arrays dry out pr ior to the centrifugation step. 1 minute at 1000-1500 rpm Place arrays in foil-covered slide tray until scanning. Scanning Arrays We scan our arrays immedi ately after they are washed/dried using a two-channel confocal microarray sca nner (ScanArray5000; GSI Lumonics, MA) and the associated ScanArray software (v3.1, Packard BioChip Technologies, MA). After laser focusing and balancing of the two cha nnels, scans are conducte d at a resolution of 10 um with the laser power typically set between 90-100% of maximum and the photomultiplier tube typically set at 65-75% of maximum. Excita tion/emission settings are 543 um/570 um and 633 um/670 um for the Cy3 and Cy5 fluors, respectively. Raw fluorescence image data is saved as .tif files.

PAGE 128

116 LIST OF REFERENCE Al-Rabab'ah MA, Williams CG. 2002. Population dynamics of Pinus taeda L. based on nuclear microsatellites. Forest Ecology and Management 163:263-271. Anderson RL, McCartney TC, Cost CD, Devine H, Botkin M. 1988. Fusiform rust hazard haps for loblolly and slash pine. USDA Forest Service Southeastern Forest Experiment Station Research Note SE-351:1-7. Andersson-Gunner S, Mellerowicz EJ, Love J, Segerman B, Ohmiya Y, Coutinho PM, Nilsson P, Henrissat B, Moritz T, S undberg B. 2006. Biosynthesis of cellulose enriched tension wood in Populus: global an alysis of transcripts and metabolites identifies biochemical and developm ental regulators in secondary wall biosynthesis. Plant Journal 45:144-165. Anterola AM, Jeon JH, Davin LB, Lewis NG. 2002. Transcriptional control of monolignol biosynthesis in Pinus taeda Factors affecting monolignol ratios and carbon allocation in phenylpropanoid metabo lism. Journal of Biological Chemistry 277:18272-18280. Baker JB, Langdon OG (1990) Pinus taeda L., Loblolly Pine. In: Burns, R.M. Honkala, B.H (eds) Silvics of North America. Handbook 654 ed. Washington, D.C, US Department of Agriculture, 497-512 Baltunis BS, Huber DA, White TL, Goldfarb B, Stelzer HE. 2005. Genetic effects of rooting loblolly pine stem cuttings from a partial dial lel mating design. Canadian Journal of Forest Research-Revue Cana dienne De Recherche Forestiere 35:10981108. Barrows-Broaddus J, and Dwinell LD. 1983. Histopathology of Fusarium moniliforme var. subglutinans in four species of southern pines pitch canker, Pinus virginiana virginia pine, Pinus elliotii var. elliotii slash pine, Pinus taeda loblolly pine, Pinus serotina pond pine. Phytopathology. June:882-889. Birch PRJ, Kamoun S. 2000. Studying interactio n transcriptomes: coordinated analyses of gene expression during plant-microorga nism interactions. New Technologies for Life Sciences: A Tr ends Guide:77:72. Brown GR, Gill GP, Kuntz RJ, Langley CH, Neale DB. 2004. Nucleotide diversity and linkage disequilibrium in loblolly pine. Proceedings of the National Academy of Sciences of the United Stat es of America 101:15255-15260. Bubeck DM, Goodman MM, Beavis WD, Gr ant D. 1993. Quantitative trait loci controlling resistance to gray leaf-spot in maize. Crop Science 33:838-847

PAGE 129

117 Burdsall HH, Snow GA. 1977. Taxonomy of Cronartium qercuum and Cronartium fusiforme Mycologia 69:503-508. Carey WA, Kelley WD. 1993. Seedling production trends and fusiform rust control practices at southern nurs eries, 1981-1991. Southern Jour nal of Applied Forestry 17:207-211 Cubbage FW, Pye JM, Holmes TP, Wagner JE. 2000. An economic evaluation of fusiform rust protection research. Southern Journal of Applied Forestry. 24:77-85. de Souza SM, White TL, Schmidt RA, Young CH, Anderson RL. 1990. Evaluating fusiform rust symptoms on greenhouse-grown slash pine seedlings to predict field resistance. Plant Disease 74:969-974. Dwinell LD, Barrows-Broaddus JB. 1981. Pitc h Canker in Seed Orchards. Proc 16th Southern Forest Tree Improvement Conf erence, Blacksburg, Virginia 234-240 Dwinell LD, Barrows-Broaddus JB, Kuhlman EG. 1985 Pitch canker: a disease complex of southern pines. Plant Disease 69:270-276 Dwinell LD, Barrowsbroaddus JB. 1982. Pitch Ca nker in Southern Pine Seed Orchards. Phytopathology 72:979-979. Falconer DS, Mackay TFC. 1996. Introduction to quantitative genetics, 4th ed. New York: Longman and Co. 464. Food and Agricultural Organi zation 2006. Global Forest Resources Assessment 2005. In: Organization FaA, editor.: FAO. Farnir F, Grisart B, Coppieters W, Riquet J, Berzi P, Cambisano N, Karim L, Mni M, Moisio S, Simon P, Wagenaar D, Vilkki J, Georges M. 2002. Simultaneous mining of linkage and linkage disequilibrium to fi ne map quantitative tr ait loci in outbred half-sib pedigrees: Revisiting the location of a quantitative trait locus with major effect on milk production on bovine chromosome 14. Genetics 161:275-287 Flint-Garcia SA, Thornsberry JM, Buckler ES 2003. Structure of linkage disequilibrium in plants. Annual Review of Plant Biology 54:357-374. Foster GS, Anderson RL. 1989. Indirect selectio n and clonal propagation of loblolly pine seedlings enhance resistance to fusiform rust. Canadian Journal of Forest ResearchRevue Canadienne De Recherche Forestiere 19:534-537. Frampton J, Isik F, Goldfarb B. 2002. Effects of nursery characteristic s on field survival and growth of loblolly pine rooted cutti ngs. Southern Journal of Applied Forestry 26:207-213. Frampton J, Li B, Goldfarb B. 2000. Early fiel d growth of loblolly pine rooted cuttings and seedlings. Southern Journal of Applied Forestry. 24:98-105.

PAGE 130

118 Francia E, Tacconi G, Crosatti C, Barabasc hi D, Bulgarelli D, Dall'Aglio E, Vale G. 2005. Marker assisted selection in crop plants. Plant Cell Tissue and Organ Culture 82:317-342. Fry SC, Smith RC, Renwick KF, Martin DJ, Hodge SK, Matthews KJ. 1992. Xyloglucan endotransglycosylase, a new wall-loos ening enzyme activity from plants. Biochemistry. J 282:821828. Gilmour AR, Cullis BR, Welham SJ, Thompson R. 2004. ASREML reference manual, Orange Agriculture Institute. New South Wales, Australia. Glazebrook J. 2005. Contrasting mechanisms of defense against biotrophic and necrotrophic pathogens. Annual Reviews of Phytopathology 43:205227. Gray DJ, Amerson HV. 1983. Invitr o resistance of embryos of Pinus taeda to Cronartium quercuum f sp fusiforme ultrastructure and hist ology. Phytopathology 73:14921499. Griggs MM, Schmidt RA. 1977. Increase and spr ead of fusiform rust. In: Management of Fusiform Rust in Southern Pines Sympos ium Proceedings. University of Florida, Gainesville. 32-38. Hammond-Kosack KE, Jones JDG. 1997. Plant disease resistance genes. Annual Review of Plant Physiology and Plan t Molecular Biology 48:575-607. Heath MC. 1997. Signalling betw een pathogenic rust fungi a nd resistant or susceptible host plants. Annals of Botany 80:713-720. Henderson CR. 1973. Sire evaluation and ge netic trends. In: Proceedings Animal Breeding and Genetics Symposium in Honor of J. Lush, 3, Champaign, Illinois: American Society of Animal Science. p 10-41. Huber DA, White TL, Hodge GR. 1992. The effi ciency of half-sib, half-diallel and circular designs in the estimation of genetic parameters in forestry: a simulation. Forest Science 38:757-776 Huber DA. 1993. Optimal mating designs a nd optimal techniques for analysis of quantitative traits in forest genetics. In: Forest Resources and Conservation,. Gainesville, FL: University of Florida. Hudgins JW, McDonald GI, Zambino PJ Klopfenstein NB, Franceschi VR. 2005. Anatomical and cellular responses of Pinus monticola stem tissues to invasion by Cronartium ribicola Forest Pathology 35:423-443. Isik F, Li BL, Frampton J, Goldfarb B. 2004. Efficiency of seedlings and rooted cuttings for testing and selection in Pinus taeda Forest Science 50:44-53.

PAGE 131

119 Jackson LWR, Parker JN. 1958. Anatomy of fusiform rust galls on loblolly pine. Phytopathology 48:637-640. Jannink JL, Bink M, Jansen RC. 2001. Using co mplex plant pedigrees to map valuable genes. Trends in Plant Science 6:337-342. Jewell FF, Mallett SL, True RP. 1962. Histology of Cronartium fusiforme in Slash Pine Seedlings. Phytopathology 52:850-&. Jurgens JA, Blanchette RA, Zambino PJ, Davi d A. 2003. Histology of white pine blister rust in needles of resistant and susceptible eastern white pine. Plant Disease 87:1026-1030. Kayihan GC, Huber DA, Morse AM, White TL, Davis JM. 2005. Genetic dissection of fusiform rust and pitch canker disease tr aits in loblolly pine. Theoretical and Applied Genetics 110:948-958. Kerr MK, Churchill GA. 2001. Experimental design for gene expr ession microarrays. Biostatistics 2:183-201. Kirst M, Basten CJ, Myburg AA, Zeng ZB, Sederoff RR. 2005 Genetic architecture of transcript-level variation in differenti ating xylem of a eucal yptus hybrid. Genetics 169 (4): 2295-2303 APR 2005 Knighten JL. 1988. Resistance screening cente r procedures manual: a step by step guide used in the operational scr eening of southern pines for resistance to fusiform rust. Region 8, USDA Forest Service, Forest Pest Management, 62 Kubisiak TL, Amerson HV, Nelson CD. 2005. Ge netic interaction of the fusiform rust fungus with resistance gene Fr1 in loblolly pine. Phytopathology 95:376-380. Kubisiak TL, Roberds JH, Spaine PC, Doudr ick RL. 2004. Microsat ellite DNA suggests regional structure in th e fusiform rust fungus Cronartium quercuum f. sp fusiforme. Heredity 92:41-50. Kuhlman EG. 1992. Interaction of virule nt single-gall rust isolates of Cronartium quercuum f sp fusiforme and resistant families of loblolly pine. Forest Science 38:641-651 Kuhlman EG. 1990. Frequency of single-gall isolates of Cronartium quercuum f. sp. fusiforme with virulence toward three resistant loblolly pine families. Phytopathology 80:614-617. Kuhlman EG. 1987. Effects of inoculation treatment with Fusarium moniliforme var. subglutinans on dieback of loblolly and slash pine seedlings. Plant Disease 71:161162.

PAGE 132

120 Kuhlman EG, Amerson HV, Jordan AP, Pepper WD. 1997. Inoculum density and expression of major gene resistance to fusi form rust disease in loblolly pine. Plant Disease 81:597-600. Kuhlman EG, Cade S. 1985. Pitch canker di sease of loblolly and pond pines in North Carolina plantations. Plant Disease 69:175-176 Kuhlman EG, Dianis SD, Smith TK. 1982. Ep idemiology of pitch canker disease in a loblolly pine seed orchard in North Carolina Fusarium moniliforme var. subglutinans, Pinus taeda Phytopathology 72:1212-1216 Kuhlman EG, Powers HR. 1988. Resistance respons es in half-sib loblolly pine progenies after inoculation with Cronartium quercuum f.sp f usiforme Phytopathology 78:484-487. Kus JV, Zaton K, Sarkar R, Cameron RK. 2002. Age related resistance in Arabidopsis is a developmentally regulated defense response to Pseudomonas syringae Plant Cell 14:479-490 Lev-Yadun S, Sederoff R. 2000. Pines as m odel gymnosperms to study evolution, wood formation, and perennial growth. Journa l of Plant Growth Regulation 19:290-305. Lewis DH. 1973. Concepts in fungal nutrition and the origin of biotrophy. Biological Reviews 48:261-278. MacFall JS, Spaine P, Doudrick R, Johnson GA. 1994. Alterations in growth and watertransport processes in fusiform rust galls of pine, determined by magneticresonance microscopy. Phytopathology 84:288-293. Mayer AM, Staples RC and Gilad NL. 2001. Mechanism of survival of necrotrophic fungal plant pathogens in hosts expr essing the hypersensitive response. Phytochemistry 58:33-41. McKeand S, Mullin T, Byram T, White T. 2003. Deployment of genetically improved loblolly and slash pines in the s outh. Journal of Forestry 101:32-37. McKeand SE, Li B, Amerson HV. 1999. Genetic variation in fusiform rust resistance in loblolly pine across a wide geographi c range. Silvae Genetica 48:255-260. McKeand SE, Mullin T, Byram T. 2003. Deployment of geneticaly improved loblolly and slash pines in the South. Journal of Forestry 101:32-37. McKeever DB, Howard JL. 1996. Value of timbe r and agricultural products in the United States, 1991. Forest Products Journal 46:45-50. McLeod AR, Long SP. 1999. Free-air carbon di oxide enrichment (FACE) in global change research: A review. Advances in Ecological Research, Vol 28 28:1-56.

PAGE 133

121 Meuwissen THE, Karlsen A, Lien S, Olsa ker I, Goddard ME. 2002. Fine mapping of a quantitative trait locus for twinning rate using combined linkage and linkage disequilibrium mapping. Genetics 161:373-379 Miller T, Cowling EB, Powers HR, Blal ock TE. 1976. Types of resistance and compatibility in slash pine seedlings infected by Cronartium fusiforme Phytopathology 66:1229-1235. Miller T. 1983. Fusiform rust resistance in lobl olly pine: artificial i noculation versus field performance. Plant Disease 67:33-34. Morse A, Nelson CD, Covert SF, Holliday AG, Smith KE, Davis JM. 2004. Pine genes regulated by the necrotrophic pathogen Fusarium circinatum Theoretical and Applied Genetics t 19 (5) 922-932 Myburg H, Morse AM, Kubisiak TL, Osborne JA, Garcia SA, Nelson CD, Davis JM, Covert SF, van Zyl LM. 2006 Differential gene expression in loblolly pine ( Pinus taeda L.) challenged with the fusiform rust fungus, Cronartium quercuum f.sp. fusiforme Physiological and Molecular Plant Pathology Nirenberg HI, O'Donnell K. 1998. New Fusarium species and combinations within the Gibberella fujikuroi species complex. Mycologia 90:434-458 Oak SW, Blakeslee GM, Rockwood DL. 1987. Pitch canker resistant slash pine identified by greenhouse screening. Pro ceedings of 19th Southern Forest Tree Improvement Conference, College Station, Texas 132-139 Oliver RP, Ipcho SVS. 2004. Arabidops is pathology breathes new life into the necrotrophs-vs.-biotrophs classification of fungal pathogens. Molecular Plant Pathology 5:347-352. Patterson H, Thompson R. 1971. Recovery of inter-block information when block sizes are unequal. Biometrika 58:545-&. Paux E, Tamasloukht MB, Ladouce N, Sivadon P, Grima-Pettenati J. 2004. Identification of genes preferentially expressed duri ng wood formation in Eucalyptus. Plant Molecular Biology 55:263-280. Powers HR, Langdon OG. 1977. Pond pine clones vary in resistance to Cronartium fusiforme Plant Disease Reporter 61:916-918. Powers HR,Jr, Zobel BJ. 1978 Progeny of specific loblolly pine ( Pinus taeda ) clones vary in fusiform rust ( Cronartium fusiforme ) resistance according to seed orchard of origin. For Sci 24:227-230 Powers HR. 1980. Pathogenic variation among single aeciospore isolates of Cronartium quercuum f sp fusiforme Forest Science 26:280-282.

PAGE 134

122 Powers Jr HR. 1985. Response of sixteen lobl olly pine families to four isolates of Cronartium quercuum f. sp. fusiforme IUFRO Proc., Rusts of Hard Pines Working Party Conf. J. Barrows-Broaddus and HR Powers, eds. Georgia Center for Continuing Education, University of Georgia, Athens:89-96. Rafalski JA. 2002. Novel genetic mapping tools in plants: SNPs and LD-based approaches. Plant Science 162:329-333. Ringli C, Keller B, Ryser U. 2001. Glycine-ri ch proteins as stru ctural components of plant cell walls. Cellular and Molecula r Life Sciences (CMLS) 58:1430-1441. Schmidt RA, Gramacho KP, Miller T, Young CH. 2000. Components of partial resistance in the slash pine-fusiform rust pathosystem. Phytopathology 90:10051010 Schmidt RA. 1998. Fusiform rust disease of southern pines: Biology, ecology and management In. Agriculture Station Ins titute of Food Agriculture Science and Technology Bulletin, University of Florida. Schmidtling RC, Carroll E, LaFarge T. 1999. A llozyme diversity of selected and natural loblolly pine populations Silvae Genetica 48:35. Schultz RP. 1999. Loblolly the pine for the twenty-first century. New Forests 17:71-88. Snow GA, Dinus RJ, Kais AG. 1975. Variation in pathogenicity of diverse sources of Cronartium fusiforme on selected slash pine fami lies. Phytopathology 65:170-175. Snow GA, Kais AG. 1970. Pathogeni c variability in isolates of Cronartium fusiforme from five southern stat es. Phytopathology 60:1730-1731. Stelzer HE, Doudrick RL, Kubisiak TL, Nelson CD. 1997. Deviation of host and pathogen genotypes in the fusiform rust pathosystem on slash pine using a complimentary genetic model and diallel da ta. Proceedings of 24th Southern Forest Tree Improvement Conference. Orlando, FL:320-330. Stelzer HE, Foster GS, Shaw V, McRae JB 1998. Ten year growth comparison between rooted cuttings and seedlings of loblo lly pine. Canadian Journal of Forest Research-Revue Canadienne De Recherche Forestiere 28:69-73. Storer AJ, Wood DL, Gordon TR. 2002. The ep idemiology of pitch canker of Monterey pine in California. Forest Science 48:694-700. Thomma BPHJ, Penninckx IAMA, Broekaert WF, Cammue BPA. 2001. The complexity of disease signaling in Arabidopsis. Current Opinion in Immunology 13:63-68 Walkinshaw CH, Bey CF. 1981. Reaction of field -resistant slash pines to selected isolates of Cronartium quercuum f. sp. fusiforme Phytopathology 71:1090-1092.

PAGE 135

123 Wan J, Dunning MF, Bent AF. 2002. Pr obing plant-pathogen interactions and downstream defense signaling using DNA microarrays. Functi onal & Integrative Genomics 2:259-273. Wang GL, Mackill DJ, Bonman JM, Mccouch SR, Champoux MC, Nelson RJ. 1994. RFLP mapping of genes conf erring complete and partial resistance to blast in a durably resistant rice cultivar. Genetics 136:1421-1434 Warren JA, Covert SF. 2004. Differential expression of pine and Cronartium quercuum f. sp fusiforme genes in fusiform rust galls. Applied and Environmental Microbiology 70:441-451. Wilcox PL, Amerson HV, Kuhlman EG, Liu BH, OMalley DM, Sederoff RR. 1996. Detection of a major gene for resistance to fusiform rust disease in loblolly pine by genomic mapping. Proceedings of the Nati onal Academy of Sciences of the United States of America 93:3859-3864. Wolfinger RD, Gibson G, Wolfinger ED, Benne tt L, Hamadeh H, Bushel P, Afshari C, Paules RS. 2001. Assessing Gene Significan ce from cDNA Microarray Expression Data via Mixed Models. Journal of Computational Biology 8:625-637. Woodward AW, Bartel B. 2005. Au xin: Regulation, Action, a nd Interaction. Annals of Botany 95:707-735. Wu RL, Ma CX, Casella G. 2002. Joint linka ge and linkage dise quilibrium mapping of quantitative trait loci in natu ral populations. Genetics 160:779-792 Yamada Y (1962) Genotype by environment inte raction and genetic correlation of the same trait under different environments. Japanese Journal of Genetics 37:498-509 Yang T, Poovaiah BW. 2000. Molecular and Biochemical Evidence for the Involvement of Calcium/Calmodulin in Auxin Action. Journal of Biological Chemistry 275:3137-3143

PAGE 136

124 BIOGRAPHICAL SKETCH Gogce Ceren Kayihan was born in Ankara, Turkey in 1975. She studied in Amasya until the end of her high school education. In 1993, she entered Biological Sciences Department in Middle East Technical Un iversity. Upon her graduation in 1997 she started working on her masters thesis in the same department with her advisor, Dr. Zeki Kaya. Her love of plants and his research inte rest in conifers ended her up in a project where she studied genetic variation of Cedr us libani populations in Turkey. In 2001 she was accepted to the School of Forest Resour ces and Conservation at the University of Florida as a PhD student and started to work with Dr. Timothy White and Dr. John Davis.


Permanent Link: http://ufdc.ufl.edu/UFE0015719/00001

Material Information

Title: Genetic Architecture of Fungal Disease Traits in Loblolly Pine
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0015719:00001

Permanent Link: http://ufdc.ufl.edu/UFE0015719/00001

Material Information

Title: Genetic Architecture of Fungal Disease Traits in Loblolly Pine
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0015719:00001


This item has the following downloads:


Full Text












GENETIC ARCHITECTURE OF FUNGAL DISEASE TRAITS IN LOBLOLLY PINE


By

GOGCE CEREN KAYIHAN













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


2006

































Copyright 2006

by

Gogce Ceren Kayihan

































Dedicated to my family and my cats















ACKNOWLEDGMENTS

This study could not have become a reality without the hard work of many talented

and smart people who funded, designed the project and physically worked in the field, so

I would like to take this opportunity to thank them all. I would not be able to do any of

this work without the help from my advisors Dr. Timothy L. White, Dr. John M. Davis,

Dr. Dudley A. Huber, Dr. Dana Nelson and Dr. Rongling Wu who were extremely

patient, giving and kind. I would like to thank Davis lab members who worked with me

day/night and always kept their sense of humor to get all of us through very hot days in

the greenhouse.

My family, the Kayihans although thousands of miles away, had never left me

alone and endured all my frustration and depression with me, I can't imagine going

through this without their support. I would also like to mention my lovely cats Kouketsu

and Suki who were my only family here for the last 5 years and Shinsetsu, Cakal, late

Takashi and Tsubasa who were always in my heart.

My friends, who laughed with me, cooked for me, watched weird TV shows with

me and made the PhD process much lighter on my shoulders. I also want to take this

opportunity to thank my therapists who helped me break a lot of waves.

Last but certainly not the least I would like to thank Stefan Crynen for all the

support he has given during the last very stressful months.
















TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TABLES .................................................... ........ .. .............. viii

LIST OF FIGURES ......... ......................... ...... ........ ............ ix

ABSTRACT ........ .............. ............. ...... ...................... xi

CHAPTER

1 IN T R O D U C T IO N ............................................................................. .............. ...

2 GENETIC DISSECTION OF FUSIFORM RUST AND PITCH CANKER
DISEASE TRAITS IN LOBLOLLY PINE ............................................ ................6

Introduction ............... ...... ... ........... ......... ........................ 6
M materials and M methods .............. ....... ... .............................. ... ............... 8
Genetic Material, Plant Propagation, and Experimental Design .........................8
Pitch Canker: Inoculations and Data Collection ................................................9
Fusiform Rust: Inoculations and Data Collection ...............................................11
Estimation of Genetic Parameters ......................................................... 12
G enetic C orrelations ............................................................... .........................14
R e su lts................. ............ ...... .. ... ..... ........................... ................ 1 5
Pitch Canker Disease Resistance is Heritable ...................................................15
Two Distinct Inoculation Procedures Reveal Similar Heritabilities for Lesion
L en gth .................. ........ ...... .. .. ...... .................... ....................17
Disease Traits Associated with Fusiform Rust are Independently Inherited ......17
Host Genes Underlying Resistance to Pitch Canker and Fusiform Rust are
Independent ......................... ... ...................... ....... .... ......... 22
Efficiency of Using Multiple Ramets per Genotype ............... .... ....... .....22
D iscu ssion .................. .... .......... ... .... ............ .............. ................. 24
Genetic variation for pitch canker resistance ...................................................24
Gall score and gall length are the most heritable fusiform rust traits..................26
Resistance to pitch canker and fusiform rust are under the control of two
different m echanism s ............... .... ..................................... ..................... 28
Phenotyping for disease trait dissection in loblolly pine................................29









3 FUSIFORM RUST RESISTANCE COSEGREGATES WITH AN FRI-LINKED
MARKER AND REVEALS VARIABLE PENETRANCE OF THE DISEASE
PH EN O TYPE ................ ......................... ......... ....................30

In tro d u ctio n .................................................. ................. ................ 3 0
M materials and M methods ....................................................................... ..................32
Genetic M material ...................................... ........ ...... .......... ....32
Genotyping Fam ilies 0 and 1 for Fr ...................................... ............... 32
G reenhou se screen ............. ........................................................ .. .... .... ... 33
F ie ld ............................................................................................................... 3 5
D ata A nalysis................................................ 35
G enetic Correlations .................. ..................................... ........ .... .. .38
A sym ptotic Z -test ................................................................ ...... .... .... 39
R results .......................... ..... .............. ............................40
Inheritance of Fusiform Rust Resistance in the Greenhouse and Field...............40
V alidation of Fr m arker ................................................................ ............... 41
The Genetic Basis for "Escape Rate"................... .......................................... 44
Discussion .............................. ... ............ ........ .............. 47
Agreement Among Greenhouse and Field Screens...........................................47
Marker-trait Cosegregation for Fusiform Rust Disease Resistance ....................49
Penetrance of the Fusiform Rust Disease Phenotype............... ...................50
Pathogen Infection in an Ecologically Relevant Setting ...............................52

4 TRANSCRIPT PROFILING REVEALS POTENTIAL MECHANISMS OF
FUSIFORM RUST DISEASE DEPENDENT SHIFTS IN PINE STEM
D E V E L O PM E N T ........ .... ......... .. ........ ............................ ....................54

Introduction ............... ........... ........................ ............................54
M materials and M methods ................ .... ................................................ ............... 55
Plant Material, Genotyping and Harvesting ..................................................55
Fungal M material and Inoculation ........................................ ...... ............... 56
M ic ro array ...................................................... ................ 5 6
Statistical A analysis .......................... ............ ............... .......... 57
R e su lts ...........................................................................................6 0
D iscu ssion ....................................................... ......................... 66
Transcription Profiling Reveals Differential Gene Expression.........................67
Influences of Fusiform Rust Disease Development on Gene Profiles.........69

5 C O N C L U SIO N ......... ...................................................................... ......... .. ..... .. 73

APPENDIX

A SAS SCRIPTS FOR MICROARRAY ANALYSIS ..............................................75

B ASREML SCRIPT FOR ASYMTOTIC Z-TEST................................. ...............78

C HEALTHY VS. DISEASED GENE LIST.........................................................79









D GENES THAT ARE REGULATED ACROSS TIME ...........................................89

E M ICROARRAY PROCEDURE ........................................ ......................... 108

Indirect Incorporation of Cy Dyes.......................... ........................... 108
Hybridization ........... .. ............................................. .... ......... 112

LIST OF REFEREN CE S ......... .................................. ........................ ............... 116















LIST OF TABLES


Table page

2-1 Summary of the four inoculation experiments reported in this study ................... 10

3-1 Summary of the greenhouse and field screens reported in this study. The 63
families and most of the clones screened were the same across the ten gall, one
gall and fi eld screens. ...................... .. ...... .................... .... .... ........... 33

3-2 Summary of score (disease incidence) and escape rate datasets along with
narrow sense heritabilities (h2) and broad sense heritability (H2) for escape rate
and score in ten-gall, one-gall and field fusiform rust screens..............................40

3-3 Segregation of marker J7_485A linked to Fr] gene in families 0 and 1 across
ten-gall, one-gall and field screens (658 ramets combined) with disease
phenotype. Parent number 17 is heterozygous for pathotype-specific resistance
gene F r .............................................................................43















LIST OF FIGURES


Figure page

2-1 A circular mating design was used to generate the plant material. Thirty-two
parents were crossed following a circular design, and the resulting progeny was
used as the m material screened for this study. ........................................ ..................9

2-2 Frequency distributions and genetic correlation for pitch canker lesion length.......16

2-3 Heritability estimates for pitch canker lesion lengths. ...........................................18

2-4 Frequency distributions and genetic correlations for fusiform rust disease traits....19

2-5 Heritability estimates and family rank scatter plots for fusiform rust disease
tr a its ............................................................................. 2 1

2-6 No genetic correlation between pitch canker and fusiform rust resistance. Family
rank-rank scatter plot based on predicted family means for pitch canker and
fusiform rust, fitted with a least squares regression line..........................................23

2-7 Efficiency is inversely proportional to the number of ramets per genotype ...........23

3-1 The inoculum sources used in the ten-gall and one-gall greenhouse trials and
field screens mapped in Florida and Georgia along with the other areas that were
assessed for virulence ............................................... .......... .. .......... 34

3-2 Diagrams illustrating genotype (clone) based phenotyping for disease resistance,
susceptibility and escape rate. ............................................................................36

3-3 Scatter plot of ranks based on BLUP-predicted family genetic values for ten-gall
and field were plotted against each other (a rank of' 1' is the most resistant and
'63' the m ost susceptible). ...................................... .................. ............. 41

3-4 Distribution of percentage of galled ramets by clone in the ten-gall (A), one-gall
(B) and field (C) screens. There were a total of 1471 genotypes (i.e. clones) in
all the experiments and each clone was replicated 1-5 times in each experiment...45

3-5 Random distribution of fusiform rust disease resistance performance of ramets
from clones that had at least one diseased ramet in Randolph, GA field trial.........46









4-1 Flow chart illustrating the procedure to identify significant and biologically
interesting gene expression profiles. ANOVA was performed for each of the
3 7 0 5 g e n e s ...................................... ................................ ................ 5 9

4-2 Analyses of mean gene expression data support 27 distinct profile groups, "A"
through "A A ." ........................................................................6 1

4-3 Genes with the same expression profile in all treatment combinations were
predominantly induced during the first time interval.............................................62

4-4 Profile groups can be categorized into biologically interesting clusters with
distinct changes in gene expression patterns ........................................................64

4-5 Clustergram of gene profile differences (or contrasts) between diseased and
healthy treatm ents.. .................... .................... ...... ................. 65















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

GENETIC ARCHITECTURE OF FUNGAL DISEASE TRAITS IN LOBLOLLY PINE
By

Gogce Ceren Kayihan

August 2006

Chair: Timothy L. White
Cochair: John M. Davis
Major Department: Forest Resources and Conservation

In the southeastern United States, loblolly pine (Pinus taeda L) is the most

common tree species covering nearly 13.4 hectares in southern United States with over 1

billion seedlings produced every year. This popular pine species bring $30 billion and

110,000 jobs to the region. However, two endemic fungal diseases are threatening this

productive view: fusiform rust incited by Cronartium quercuum Berk. Miyable ex Shirai

f. sp.fusiforme and pitch canker incited by Fusarium circinatum Nirenberg et O'Donnell.

Loblolly pine is not totally susceptible to these diseases and it has been shown by many

researchers, using natural and artificial inoculations, that loblolly pine families show

genetic variation in resistance to both fusiform rust and pitch canker diseases. Precision

was acquired by a combination of clonal propagation, which allows repeat observations

of the same genotypes and the use of a mixed linear model (GAREML) to adjust for

environmental effects. In the first part of this study, I identified traits, clones, families,

and parents that guide a genetic approach to dissecting disease traits in loblolly pine. I









verified that pitch canker and fusiform rust traits are heritable and identified the disease

traits that are genetically distinct from one another. Second, I used DNA marker

information that was developed in previous mapping studies to distinguish host

genotypes that carry/lack the pathotype-specific Frl allele. I tested the hypothesis that the

Frl allele is predictive of resistance in greenhouse and field experiments. Because these

studies involved clonally propagated materials, I also quantified the extent to which

genetic and non-genetic factors influence disease expression levels and escape rate in

greenhouse and field trials. Finally, I used gene expression data obtained from a very

complex design of microarray experiments using diseased and healthy loblolly pine

clones from a family that is segregating for Fr], to identify genes that are differentially

regulated in diseased and healthy individuals. I contrasted gene expression in diseased

and healthy individuals over a time frame of 4 months. Together, these studies revealed

the genetic architecture of fusiform rust disease resistance in scales ranging from the

population level to the molecular level.














CHAPTER 1
INTRODUCTION

Forests cover one-third of the earth's terrestrial surface and provide social,

economical and environmental benefits (FAO, 2006). Pine is a dominant plant species in

Europe, Asia and America and has been used both as source for forest products and as a

model organism to study wood formation (Lev-Yadun and Sederoff, 2000). In the

southeastern United States, loblolly pine (Pinus taeda L) is the most common tree species

covering nearly 13.4 hectares in southern United States (Schultz, 1999) with over 1

billion seedlings planted every year (McKeand et al., 2003). This popular pine species

brings $30 billion and 110,000 jobs to the region (Schultz, 1999). In addition, loblolly

pine plantation and natural forests offer habitat for many diverse species, control erosion,

improve water quality, provide recreation and sustain rural communities.

As management of loblolly pine plantations intensive to maximize product yield,

new problems started to emerge in production of healthy loblolly pine. Among these

problems, two endemic fungal diseases attracted the attention of many researchers and

breeders: fusiform rust (incited by Cronartium quercuum Berk. Miyable ex Shirai f sp.

fusiforme) (Burdsall and Snow, 1977) and pitch canker (incited by Fusarium circinatum

Nirenberg et O'Donnell) (Nirenberg and ODonnell, 1998). Loss of millions of dollars

(Cubbage et al., 2000) pushed breeders and researchers to investigate fusiform rust, one

of the most economically destructive diseases of the southeastern United States. It is

incited by Cronartium quercuum, a biotrophic pathogen that alternates its life cycle with

pine and oak as hosts.









Fusiform shaped galls on pine hosts are the major symptom of fusiform rust. As

disease progresses through the years these infections may take the form of sunken

cankers. Galls on stems decrease the wood quality and sometimes kill the plant (Schmidt,

1998). Both specific resistance, i.e., "gene-for-gene" interactions (Powers, 1980; Stelzer

et al., 1997; Wilcox et al., 1996), and partial resistance in the form of short galls (Schmidt

et al., 2000) have been demonstrated for the C. quercuum- pine pathosystem. Loblolly

pine is not totally susceptible to this disease and it has been shown by many researchers,

using natural and artificial inoculations, that loblolly pine families show genetic variation

in resistance to fusiform rust (Kuhlman and Powers, 1988; McKeand et al., 1999).

The pitch canker disease is not as economically important as fusiform rust in

southeastern United States although it can damage southern pine plantations sporadically

in the USA and it is an important problem for Pinus radiata in California (Storer et al.,

2002). In the southeastern United States seedling production can be severely hampered

by this disease (Dwinell et al., 1985). The pitch canker inciting agent, F. circinatum, is a

necrotrophic fungus that survives on dead tissues. A successful infection results in

symptoms like resinous lesions on stems and branches that cause seedling mortality and

decreased growth rates and crown die-back of plantation trees (Dwinell et al. 1985).

Although genetic variation among loblolly pine families (Kuhlman et al., 1982) and

clones (Dwinell and Barrowsbroaddus, 1982) has been detected, the genetic architecture

of resistance has not been thoroughly investigated.

Knowing that family level genetic variation for both diseases exists in the same

species, namely, loblolly pine, provided an opportunity to investigate and contrast the

nature and architecture of resistance to the diseases incited by the biotrophic and









necrotrophic fungi. The main difference between the two types of fungi is that

necrotrophs survive on the dead plant cells and biotrophs feed on living plant cells

(Lewis, 1973). Thus the damage they cause is significantly different and the host defense

mechanisms against them may also vary. For example, biotrophic fungi are typically

associated with gene-for gene systems (Glazebrook, 2005; Hammond-Kosack and Jones,

1997) and necrotrophs are often linked to quantitative disease resistance genes (Oliver

and Ipcho, 2004). Thus, I hypothesized that resistance and responses on the pathogen and

host sides would differ for the two distinct pathosystems.

Complex traits such as disease resistance can be dissected by two core activities,

genotyping and phenotyping. These two different sets of data can be analyzed in two

ways: by linkage, which uses QTL or linkage analysis approach; or by association, which

uses linkage disequilibrium to make genotype-phenotype association (Jannink et al.,

2001). Association genetics is gaining favor as an approach to identify genes that underlie

complex traits (Rafalski, 2002). The association approach relies on linkage

disequilibrium between marker loci and target trait loci, and because many unrelated

individuals are examined in a single association experiment, it is possible to evaluate the

effects of many alleles across a broad sample of the population (Flint-Garcia et al., 2003).

In contrast QTL approaches evaluate pairs of segregating alleles typically within single

families (Jannink et al., 2001). Loblolly pine is an ideal organism for association genetics

because loblolly pine has natural and outcrossing populations distributed across large

areas that have high gene-flow and little population substructure (Al-Rabab'ah and

Williams, 2002; Brown et al., 2004b; Schmidtling et al., 1999). Also with loblolly pine it

is possible to create large experimental populations and clonally propagate them to









detect, verify and evaluate phenotypes and genotypes. Loblolly pine also has desirable

levels of nucleotide diversity (0=0.005; Brown et al., 2004) and limited linkage

disequilibrium (<2500 bp on average, Brown et al., 2004).

As a first step toward dissecting complex disease traits in loblolly pine, I have

undertaken this study to evaluate a variety of disease phenotypes in a population of 32

unrelated parents mated to form approximately 63 full-sib families that were clonally

propagated to form hundreds or thousands of clones depending upon the experiment.

Although an ideal association population would contain hundreds of unrelated individuals

(McLeod and Long, 1999), this population is an excellent starting point to evaluate the

heritabilities and genetic relationships among the two sets of disease traits that can then

be dissected by association or QTL mapping.

An important aspect of experimental material is its clonal propagation to form a

hierarchy of genetic relationships (parent, full-sib family, and clone) that facilitate the

dissection of genetic architecture of diseases and their genetic relations. Since micro-

environmental variation for a given genotype can be calculated using its vegetative

propagules, I can obtain a more precise estimate of genetic components of disease

resistance. Given that I am using the same genotypes to predict breeding values for both

pitch canker and fusiform rust disease resistance, the values can be compared to look for

correlations that will be informative when I try to understand the underlying genetic

architecture.

The clonally propagated material of clones screened in the greenhouse conditions

was also planted in several field sites. The first year fusiform rust disease incidence data

from the naturally inoculated field site can be compared with the greenhouse screen to









confirm resistant and susceptible genotypes, since the ultimate goal of resistance

screening is to identify the resistant genotypes that will be disease free in the plantations.

Genomic mapping has identified the region containing FrI (fusiform resistance-1)

conferring pathotype-specific resistance to fusiform rust (Wilcox et al., 1996). RAPD

marker J7_485A was linked to the FrI locus in progeny of a single loblolly pine parent.

Thus, the progeny that have this marker were resistant whereas the ones without the

marker were susceptible to fusiform rust incited by C. quercuum with the avirulence gene

(Avrl). This genetic marker was consistently predictive of fusiform rust resistant trees in

greenhouse (Kubisiak et al., 2005; Kuhlman et al., 1997) and field screens (Wilcox et al.,

1996). Two families among 63 families that were screened for fusiform rust resistance in

the greenhouse and the field were genotyped for Frl. Thus, clones belonging to the two

genotyped families can be used to verify the resistance prediction power of the genetic

marker.

The genotypic information on progeny of the families that are segregating for the

RAPD marker J7_485A can also be useful in molecular genetics studies. Microarray

technology which became available with the last decade (ref) can be used to identify

genes regulated in response to inoculation with C. quercuum. With the genetic marker

information genetically resistant and susceptible individuals can be isolated to be

challenged by FrI avirulent strains of C. quercuum. The host responses, disease

development and the interactions between the host and the pathogen can be revealed at

the molecular genetics level.














CHAPTER 2
GENETIC DISSECTION OF FUSIFORM RUST AND PITCH CANKER DISEASE
TRAITS IN LOBLOLLY PINE

Introduction

Pinus species are both economically and ecologically important. Pines grown in the

southeastern United States generate nearly half of the nation 's pulpwood, with an annual

harvest value of approximately $19 billion (McKeever and Howard, 1996). Loblolly pine

(Pinus taeda L.) is the most widely planted Pinus species in this region, averaging 74%

of the annual seedling production (Carey and Kelley, 1993). In addition to plantations,

loblolly pine is the predominant species on 11.7 million ha of native forest (Baker and

Langdon, 1990), where it impacts the welfare of nearly 400 species of vertebrates

(Schultz, 1999).

Loblolly pine is a host for two endemic pathogens, Cronartium quercuum Berk.

Miyable ex Shirai f sp. fusiforme (Burdsall and Snow, 1977), the inciting agent of

fusiform rust disease, and Fusarium circinatum Nirenberg et O 'Donnell (Nirenberg and

O 'Donnell, 1998), the inciting agent of pitch canker disease. Fusiform rust is one of the

most destructive fungal diseases in the southeastern United States, causing damage

ranging from $25-$135 million per year (Cubbage et al., 2000). The major symptom of

fusiform rust disease is the formation of stem galls that lead to decreases in survival,

wood quality, and growth. Genetic variation in resistance at the family level has been

demonstrated for fusiform rust (Kuhlman and Powers, 1988; McKeand et al., 1999).

Based on controlled inoculation studies carried out on specific loblolly and slash pine









Pinus elliottii Engelm. var. elliottii) families, specific resistance-i.e., "gene-for-gene"

interactions-has evolved (Powers, 1980; Stelzer et al., 1997; Wilcox et al., 1996), as

well as partial resistance in the form of short galls (Schmidt et al., 2000), which may be

genetically distinct from specific resistance.

Pitch canker is also a significant, although more episodic, disease problem (Dwinell

et al., 1985). Symptoms of pitch canker disease include resinous lesions on stems and

branches that cause seedling mortality, decreased growth rates, and crown dieback

(Dwinell et al., 1985). A considerable amount of genetic variation for pitch canker

resistance has been detected in loblolly pine families (Kuhlman et al., 1982) and clones

(Dwinell and Barrows-Broaddus, 1981); however, the genetic architecture of resistance is

not well understood.

Our goal in this work was to obtain precise estimates of pitch canker and fusiform

rust disease phenotypes expressed in loblolly pine. Precision was acquired by a

combination of clonal propagation, which allows repeat observations of the same

genotypes, and is now feasible in loblolly pine (Frampton et al., 2000), testing of over

one thousand pedigreed genotypes, and the use of a mixed linear model (GAREML) to

adjust for environmental effects (Huber, 1993). In this study, I identified traits, clones,

families, and parents that guide a genetic approach to dissecting disease traits in loblolly

pine. I verified that pitch canker and fusiform rust traits are heritable and identified the

disease traits that are genetically distinct from one another. This work creates the baseline

knowledge required for identifying genes that condition phenotypes of interest, either

through genetic linkage analysis within defined pedigrees, or by association in

populations of unrelated genotypes (Flint-Garcia et al., 2003; Jannink et al., 2001).









Materials and Methods

Genetic Material, Plant Propagation, and Experimental Design

The 63 loblolly pine families screened in this study were obtained from a circular

mating design with some off-diagonal crossing. Members of the Cooperative Forest

Genetics Research Program at the University of Florida and the North Carolina State

University-Industry Cooperative Tree Improvement Program (Figure 2-1) provided the

32 parents and generated the full-sib families and clones screened in this study. Forty-

nine seeds from each full-sib family were germinated and grown into hedges for clonal

propagation. Maintenance of hedges and propagation of cuttings is reported in Baltunis et

al. (2005). In brief, cuttings were set in July 2001, assessed for rooting after 9 weeks, and

clones with the highest rooting ability selected for this experiment. The number of clones

within families and the number of ramets (i.e., rooted cuttings) for each clone was not

equal, since families did not produce the same number of clones, and clones had different

rates of rooting. Cuttings assigned to a greenhouse screen were chosen at random from

the ramet pool of each available clone (Table 2-1). The screens were grouped according

to the disease (fusiform rust or pitch canker). The fusiform rust screens were conducted

using two types of inoculum (a one-gall mix or a ten-gall mix), whereas both pitch canker

screens used a single inoculum. The experimental design was a randomized complete

block with single-tree plots arranged in an alpha lattice with an incomplete block size of

20. The clones were replicated with a maximum number of five ramets per experiment.

Ramets were pruned twice to stimulate synchronous elongation of multiple succulent










ID 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
1 16 17 18 19
2 6 18 7 20 18 19 22
3 34 23 19 19
4 69 37 15
5 58 19 18 20 21
6 15 10 19 19 18
7 45 11 64 21 191
8 50 56 18 19 21
9 40 26 20
10 25 49 21
11 39 44 18 21
12 28 21 19 21
13 35 41 19 19 21
14 46 65 18
15 7 68 20 22
16 70 38 22 21
17 16 36 42 30 31
18 27 60 0 21 22
19 1 22 22 15

21 13 29 19
22 8 19 18 20
23 48 67 20 18 20
24 51 55 17
25 66 43 62 21 19
26 9 18 19
27 54 31 32 20
28 33 52 21 14 22
29 3 2
30 12 53
31 19
32 61 4
Figure 2-1 A circular mating design was used to generate the plant material. Thirty-two
parents were crossed following a circular design, and the resulting progeny
was used as the material screened for this study. The numbers in the cells
above the diagonal are the number of clones used from a given cross, and the
numbers below the diagonal are the family identification numbers

shoots for inoculation. The initial pruning occurred in March 2002, 8 months after

setting, by cutting back the shoots from 10-15 cm to 3-4 cm each. The second pruning

occurred 6 weeks prior to inoculations for both pitch canker and fusiform rust screens;

shoots were succulent and 5-8 cm in length at the time of inoculation. After pruning, all

trees were fertilized weekly with Miracle-Gro 15-30-15 until inoculation.

Pitch Canker: Inoculations and Data Collection

The larger of the two pitch canker screens was conducted at the USDA Forest Service

Resistance Screening Center in Bent Creek, North Carolina, and is referred to as









Table 2-1 Summary of the four inoculation experiments reported in this study
Testa #Families # Clones Rangeb # Observationsc
Ramets
RSC pitch canker 63 1065 7-31 4483 7664
UF pitch canker 60 362 1-24 1316 3119
Ten Gall fusiform rust 63 1360 17-31 5473 11,395
One Gall fusiform rust 63 698 2- 30 2743 5195
aRSC USDA Forest Service Resistance Screening Center, UF University of Florida
bNumber of clones within families
cNumber of observations exceeds the number of ramets because multiple shoots were
assessed on a given ramet


the "RSC" screen in this manuscript. New growth (5-7 cm) was inoculated following

the standard RSC protocol (Oak et al., 1987) with F. circinatum isolate S45 (Forest

Pathology laboratory collection, University of Florida) at a density of 92,500 spores/ml.

In brief, prior to spray inoculation, shoot tips were excised from two shoots on each

ramet. After inoculation, plants were placed in a high-humidity chamber for 24 h, then

transferred to a greenhouse and maintained at an average temperature of 200C for

3 months.

The smaller of the two pitch canker screens was conducted at the University of

Florida and is referred to here as the "UF' screen. Plants were pruned 6 weeks before

inoculation with the same S45 isolate. One shoot tip per ramet was excised, and 1 pl of a

500-spores/pl solution was applied to the wound with a micropipette. All plants were

incubated under high humidity for 24 h. The test was kept in the greenhouse for 36 days

at an average temperature of 300C.

Disease symptoms were measured at 90 days (RSC) and 36 days (UF). Shoot

length and lesion length were measured (in millimeters) on one shoot chosen at random

from each ramet at the RSC and on the single shoot inoculated per ramet at UF.









Both the RSC and UF pitch canker raw data sets included only one lesion-length

and shoot-length measurement for each ramet. Prior to analysis, the data were

standardized by experiment, using the phenotypic standard deviation calculated from the

appropriate linear model for the screen.

Fusiform Rust: Inoculations and Data Collection

Plants were pruned twice before inoculation to stimulate elongation of multiple

shoots per ramet. Both rust screens were inoculated at the RSC, following standard

protocols (Knighten, 1988). The ten-gall test was inoculated at a density of

52,000 spores/ml with aeciospores pooled from a ten-gall collection obtained from a 6-

year-old loblolly pine plantation near Lee, Florida (designated L-10-1-99, provided by

Dr. Henry Amerson, NC State University) The one-gall test was inoculated at a density of

50,700 spores/ml with aeciospores isolated from a single gall obtained from a branch of

slash pine family 84-57 near Callahan, Florida (designated #501, provided by Dr. Robert

Schmidt, University of Florida).

Data were collected from both rust screens 6 months after inoculation. For each

ramet with multiple shoots, the number of shoots with galls and the number of galls were

counted and recorded. In addition, two shoots with single galls were randomly chosen to

measure stem length, gall length, and gall width (in millimeters) for each ramet.

Data collected from both the ten-gall and one-gall screens were treated identically

for gall measurements. Gall measurement values were averaged by ramet if there was

more than one shoot with a single gall. Gall volume was calculated from gall length and

gall width data, assuming a fusiform shape:

Volume= ( 4)1 .igi (width)2









Ramets were scored as 0 (no gall) or 1 (at least one gall) for gall score. Ramets that

did not form galls were not included in the gall length, width, and volume data. Gall

length, width, and volume data sets were standardized using their respective phenotypic

standard deviations calculated from the linear model.

Estimation of Genetic Parameters

Variance components and genetic parameters were estimated by GAREML (Huber,

1993), which employs restricted maximum likelihood estimation (Patterson and

Thompson, 1971) and best linear unbiased prediction [(BLUP) Henderson, 1973]. The

same linear model was applied to the traits measured in all four disease screens, since the

experimental designs were identical. The linear model was:

Yjklm = u + R, + t(r), + ga k + gca + sca k + c(family )km + r fki + e iki
where:

yijklm is the mth observation of the kith cross in thejth tray of ith rep.

[t is the population mean.

Ri is the fixed resolvable replication, i=1-5.

t(r)ij is the random variable tray incomplete block -NID(O,o2t), j=1-21.

gcak is the random variable female general combining ability (GCA) -NID(0, 02gca) k=l-

32.

gcal is the random variable male general combining ability ~NID(0, 2cgca) 1=1-32.

scakl is the random variable specific combining ability (SCA) ~NID(0, 2 sca).

c(family)klm is the random variable clone within a family ~NID(0, 02, ,,.. ).

r*fikl is the random variable replication by family interaction ~NID(0, 2r*f)).

eijklm is the random variable error within the experiment -NID(0, 2 e).






13

The narrow- (h2) and broad-sense (H2) heritabilities were calculated according to

Falconer and Mackay (1996):

ca V(A)
6p V(P)
gca Sca dt)

where:

(2P is the phenotypic variance,

'(P) is the total phenotypic variance,

V(A) is the additive variance,

V(D) is the dominance variance,

V(I) is the epistasis variance.

To partition the broad sense heritability I calculated the ratio of dominance variance

to total phenotypic variance ( D ) and the ratio of epistatic variance to total phenotypic

variance ( u ), using the following formulas:

2 4sca V(D)
hD -



o V(P)
2 _B-(-0.5(tg +0.75c _=(I)


The broad sense heritability of clonal means (HC2) and family means (HF2) were

calculated using the formulae below:
(2* g + + 6()
C (2* ) +s + 2) + +( /r) + (6/r)










^ -2 (2*Qrca)+*sca
-(2*T2^2ca)+ ^ +' 2 +((T2/ / tr)+((.2 *C)


where r is the harmonic mean of ramets per clone and c is the harmonic mean of clones

per family.

Family deviations were predicted by summing the following BLUP estimates

produced by GAREML:

Family deviation = Predicted female value (gcak) + Predicted male value (gcal) +

Predicted specific combining ability (scakl)

Genetic Correlations

The genetic correlation between gall score and gall length at the parental, family,

and clonal levels, and the correlation among screens within and across diseases were

2
calculated on combined data sets by adding experiment x GCA ( ge), experiment by

2 2
family ( se) and experiment by clone(family) ( c(f)e) interaction factors to the linear

model and using the Type B genetic correlation formula (rB; Yamada, 1962):

^2
gca
(rB)PARENTAL c2 2
^-2 ^2
gca +ge

2( gca + sea
(B )FAMILY 2 2 ^2 2 2 2
26gc +O07Ca +252 +(2
gca sca ge se
^2 ^2 ^.2
22 gca + sea + C(F)
(rB )C) = ^2 ^2 ^2 2 + 2 + 2
252 + 22 +22 + 2 ^ 2 2
gca +sa + (f) ge +7se +c(f)e

Efficiency of using multiple ramets per genotype was calculated according to Huber et al.
(1992):
Efficiency = (1- H2) / r

where r is the number of ramets per clone.









Results

The mating design shown in Figure 1, coupled with clonal propagation, allowed

predictions of clonal, family, and parental genotypic values as well as population-wise

estimates of heritabilities and genetic correlations of disease traits for both pathosystems.

A total of 27,373 phenotypic data points were collected for lesion length (pitch canker),

gall score, gall length, and gall width (fusiform rust). I first present data on pitch canker

phenotypes, followed by fusiform rust, and finally a comparison of pitch canker and

fusiform rust resistance.

Pitch Canker Disease Resistance is Heritable

The pitch canker disease screens performed at UF and RSC resulted in 89% of the

ramets (i.e., rooted cuttings) showing measurable disease symptoms in each screen.

BLUP clonal values were predicted for each screen, and the resulting distributions are

shown in Figure 2a. The consistency of the disease rates and the shapes of the

distributions (i.e., skewed to the right) suggest that statistical comparisons between the

RSC and UF screens are appropriate. The genetic correlation between the RSC and UF

screens was 0.88 at the parental level, 0.76 at the family level, and 0.69 at the clonal

level. A scatter plot based on family ranks is presented in Figure 2b and reflects the

positive correlation between the two screens. Therefore, I conclude that parents, families,

and clones performed consistently across screens. After combining the data from the two

screens, the five most resistant and the five most susceptible full-sib families were

identified based on predicted family values and standard deviations for lesion length;

these are indicated in Figure 2- 2a by their ID number from Figure 2-1. The resistant tail

contains families 50 and 48, which have parent 8 in common. The resistant tail also

contains half-sib families 61 and 4, which share parent 32. Resistant family 44 is not










related to any of the other resistant families in the tail (Figure 2-1). The susceptible tail is

composed of three half-sib families. Susceptible family 12 has parent 30 in common with

family 53. Family 12 also shares parent 2 with susceptible family 69. Susceptible families


350
A .36 1, ,
300 48 36
250 69
250 i- I
S4 II
200 I 12
-o -- I *I
S150
E 100
50.
50 riJ .....
2 7 12 17 22 27 32 37 42 47
Lesion length (mm)
B

60
C,* *









0 20 40 60
RSC
Family ranks for lesion length


Figure 2-2 Frequency distributions and genetic correlation for pitch canker lesion length.
(A) Distribution of best linear unbiased prediction (BLUP)-predicted clonal
values for the USDA Forest Service Resistance Screening Center [(RSC)
black] and University of Florida [(UF) white] pitch canker screens. Above the
distribution are the predicted means and standard deviations of the five most
susceptible and resistant families identified by their family ID number. (B)
Ranks based on BLUP-predicted family values for RSC and UF were plotted
against each other (a rank of 1 is the most resistant and 63 the most
susceptible). A least squares regression line is shown after being forced
through the origin due to a non-significant intercept.
through the origin due to a non-significant intercept.









1 and 36 have parent 17 in common. Families in the resistant tail and families in the

susceptible tail did not have any parents in common, indicating no genetic relationships

across the classes.

Two Distinct Inoculation Procedures Reveal Similar Heritabilities for Lesion
Length

Using the RSC, UF, and pooled data, the heritabilities based on individual tree,

family, and clonal means were calculated to determine how much of the variation in

lesion length could be attributed to genetic variation and to determine the precision of the

predicted clonal and family means. The broad-sense heritabilities for the clonal (HC2) and

family (HF2) means were determined for both the individual and pooled pitch canker

screens to evaluate the precision of the clonal and family means predicted above. HC2

and HF2 were greater for RSC (0.75) than for UF (0.61; Figure 2-3), because the number

of ramets per clone and the number of clones per family were approximately three times

greater for the RSC screen compared to the UF screen (Table 1). Narrow-sense

heritabilities (h2) for both the RSC and UF datasets were 0.27. Broad-sense heritabilities

(H2) were similar for both the RSC (0.43) and the UF screens (0.37) (Figure 2-3). When

the RSC and UF data sets were pooled, heritabilities were not different from that

calculated for each screen individually (Figure 2-3). This is another indicator that results

from the two screens were comparable.

Disease Traits Associated with Fusiform Rust are Independently Inherited

The two fusiform rust screens are characterized by the type of inocula used, either

ten-gall or one-gall. There were 36% and 31% galled ramets for the ten-gall and one-gall

screens, respectively. A disease incidence (referred to as "score") dataset was generated










1.0

0.8

0.6

0.4

0.2

0.0
R C F R C F R C F
RSC UF Pooled


Figure 2-3 Heritability estimates for pitch canker lesion lengths. The bar graph shows the
heritabilities for individual ramets (R=H2) and the broad-sense heritabilities
for clonal (C=Hc2) and family (F=HF2) means for the RSC, UF, and pooled
data. Narrow-sense heritability [(h2) solid black], epistatic heritability [(hi2)
solid gray], and dominance heritability [(hD2) white] are stacked so that the y-
axis corresponding to the top of the bar is the broad-sense heritability.

by designating disease-free ramets as 0 and galled ramets as 1. Fusiform rust screens for

score are shown in Figure 2-4a. The distributions for score in both screens follow a

similar pattern, that is, there is a minor peak at a mean -0.1, and the distribution is

skewed to the right. In addition to disease incidence, gall length, and gall width were

measured for ramets with galls. In contrast to score, the predicted clonal means for gall

length revealed a normal distribution for both fusiform rust screens (Figure 2-4b).

Because the distributions and overall disease incidences were similar, scaling prior to

comparing the data from the two screens was not necessary for either trait.

Genetic correlations between the two screens were calculated for score and gall length in

order to determine if inoculum type might impact trait expression. The genetic correlation

for score was 0.80 at the parental level, 0.83 at the family level, and 0.86 at the clonal















IOU

" 120
0
S80
E
z 40

0


0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 1.0


Score (decimal)


o
U
40



C 20


LL.


* .
*


60
c*


S40


S20
0U

0
LL
0


20 40 60
Ten-gall
Family ranks for gall score


52 3
40 69
2 I 51 I

I I54

I 3 ,


20 40 60

Ten-gall
Family ranks for gall length


Figure 2-4 Frequency distributions and genetic correlations for fusiform rust disease
traits. Distribution of BLUP-predicted clonal values for gall score (A) and gall
length (B) in the ten-gall inoculum (black) and one-gall inoculum (white)
screens are shown. Above the distribution are the predicted values and
standard deviations of the five most susceptible and resistant families
identified by their family ID number. Ranks based on predicted family values
for gall score (C) and gall length (D) (1 = resistant, 63 = susceptible) are
plotted against each other. A least squares regression line is shown after being
forced through the origin due to a non-significant intercept

level, suggesting a general consistency in performance between the ten-gall and one-gall


mixes. For gall length, the genetic correlation between the two screens was 1.00 at the


parental level, 1.00 at the family level, and 0.76 at the clonal level, again indicating


general consistency in performance between the two fusiform rust screens. Despite the


high genetic correlations, I did observe "outlier" families that performed differently in


- nminiminiminimir riniiimimia m.m.* ** -. .
1 3 6 8 11 13 15 18 20 23
Gall length (mm)









the two screens, suggesting some potentially significant genotype by inoculum

interactions (Figure 2-4c, d).

Relationships among families with extreme phenotypes can reveal information

regarding inheritance. For score, the predicted family values for the five most resistant

and five most susceptible families are plotted on the graph in Figure 4a, along with their

within-family standard deviations and family ID numbers. The resistant tail contains two

half-sib family groups, that is, families 26, 40, and 49 that have parent 9 in common, and

families 52 and 53 that have parent 28 in common. The susceptible tail is composed of

two families that are half-sibs, that is, families 3, 20, and 51 have parent 22 in common,

and families 54 and 69 have parent 2 in common. Similarly for gall length, the five

families with the shortest galls and the five families with the longest galls are shown

above the distribution in Figure 4b. The short gall-forming tail includes families 31 and

62 that have parent 25 in common. The remaining three families in this tail are unrelated.

The five families with the longest galls comprise three half-sib families (13 and 29; 29

and 30; and 57, 13, and 57) that are related to one another through parents 21, 20, and 19,

respectively. Family 15 is unrelated to the others. For both score and gall length, familial

relationships within a given tail were common, whereas no such genetic relationships

among families in opposing tails were observed. This is consistent with both score and

gall length being heritable traits.

To evaluate how much of the trait variation associated with fusiform rust can be

attributed to genetic effects, heritabilities were calculated. Since the genetic correlations

for score and gall length were high across inocula (Figure 2-4c, d), data were pooled and

used for heritability calculations. Gall score was consistently more heritable than gall











A
1.0

e 0.8
0









R C F R C F R C F
Ten-gall One-gall Pooled
B
1.0

0.8
0)

j 0.6

S0.4

0.2

0.0
R C F R C F R C F
Ten-gall One-gall Pooled
C


60 4 *


o *
40 *4 *

Famiy ra

20 -' ,




0 20 40 60
Family ranks for gall length


Figure 2-5 Heritability estimates and family rank scatter plots for fusiform rust disease
traits. Gall score (A) and gall length (B) heritabilities for R, and C and F
means are shown. Heritability estimates for the ten-gall and one-gall pooled
screens are given for both traits. h2 (solid black), hi2 (solid gray), and hD2 are
stacked such that the y-axis corresponding to the top of the bar is the H2. (C)
Scatter plot of family ranks illustrates a lack of correlation between gall score
and gall length traits (1 = resistant, 63 = susceptible)









used for heritability calculations. Gall score was consistently more heritable than gall

length for the one-gall, ten-gall, and pooled datasets (Figure 2-5a, b).

Host Genes Underlying Resistance to Pitch Canker and Fusiform Rust are
Independent

Necrotrophic (i.e., F. circinatum) and biotrophic (i.e., C. quercuum) pathogens

have distinct life history properties. This implies that host genes underlying resistance

may be different for diseases incited by necrotrophic and biotrophic pathogens. To

determine whether host responses to F. circinatum and C. quercuum are independent, I

computed the genetic correlations between lesion length (pitch canker) and the various

gall characteristics (fusiform rust). There were no significant correlations between lesion

length and gall length (Figure 2-6), or between lesion length and gall score (data not

shown). The estimated genetic correlation between lesion length and gall length were

0.00 at the parental level, 0.00 at the family level, and 0.02 at the clonal level. No genetic

correlations were found between lesion length and gall volume or gall width (0.00 for all,

data not shown). Together, these results imply that resistance to pitch canker and

resistance to fusiform rust are controlled by different host genes.

Efficiency of Using Multiple Ramets per Genotype

Theoretically, if the number of ramets per genotype is high enough, heritability

estimates based on clonal means will be 1. To describe the relationship between the

number of ramets and HC2 for disease traits investigated in this study, the efficiencies

(Huber et al., 1992) for increasing number of ramets per genotype were plotted against

the number of ramets (Figure 2-7), where efficiency is calculated as the average reduction













60


| 40 *

20


. ..
S20- *

E **
LL 0
"- 0 ----


*


*4

V..


20 40 6
Fusiform rust
Family ranks for gall length


Figure 2-6 No genetic correlation between pitch canker and fusiform rust resistance.
Family rank-rank scatter plot based on predicted family means for pitch
canker (lesion length) and fusiform rust (gall length), fitted with a least
squares regression line (1= resistant, 63= susceptible)


0.8

0.7

0.6

S0.5

0.4

0.3

0.2

0.1

0
1 2 3 4 5 6 7 8 9 10
Number of ramets

Figure 2-7 Efficiency is inversely proportional to the number of ramets per genotype.
Efficiency of using multiple ramets in the estimation of HC2 plotted against
number of ramets for RSC-lesion length (filled circles), UF-lesion length
(open circles), ten-gall-gall score (filled squares), one-gall-gall score (open
squares), ten-gall-gall length (filled triangles) and one-gall-gall length (open
triangles). Efficiency was calculated as (1-H 2)/(number of ramets)

in error per ramet. For the different disease traits, the error associated with a clonal means

decreases at different rates, depending on the number of ramets used to represent

genotypes and H2. An increase in the experiment size above ca. five ramets per clone


;-









does not appreciably increase the precision of heritability estimates, suggesting that

future experiments of this type should be replicated to approximately the same extent as

this study.

Discussion

Loblolly pine exhibits considerable variation in resistance to both fusiform rust

(Kuhlman and Powers, 1988; Powers and Zobel, 1978) and pitch canker diseases

(Dwinell and Barrows-Broaddus, 1981; Kuhlman and Cade, 1985). The pathogens that

incite these diseases, the biotrophic fungus C. quercuum and the necrotrophic fungus F.

circinatum, have distinct life history strategies, reflected in the contrasting disease

symptoms visible on susceptible hosts. This study allowed a direct comparison of the host

resistance mechanisms to these distinct pathogens in a common set of host genotypes.

Consequently, it was possible to compare and contrast the genetic architecture of host

responses to both pathogens.

Complex trait analysis requires a reliable estimation of phenotypic values for

subsequent correlations with genotype. As a first step toward dissecting complex disease

traits in loblolly pine, I undertook this study to evaluate a variety of disease phenotypes in

a clonally propagated population generated via a circular mating design. Complex

pedigree structures such as these can be useful for mapping QTL (Jannink et al., 2001).

Genetic variation for pitch canker resistance

Pitch canker resistance was continuously distributed across clones, suggesting that

resistance may behave as a complex trait. Resistance to fungal necrotrophs is often

inherited as a complex trait in crop species including maize (Bubeck et al., 1993) and rice

(Wang et al., 1994). Another explanation for this continuous distribution is Mendelian

inheritance of resistance within families that appears continuous when examined across









families. If resistance were monogenic, some families would be expected to show a

bimodal distribution for lesion length. To assess this possibility, I tested individual

families for bimodal distributions of resistance. None of the within-family distributions

was bimodal; all showed continuous distributions. Since lesion length showed a

continuous distribution within families across the entire study, I infer that pitch canker

resistance is appropriate to analyze as a complex trait.

The repeatability of the pitch canker resistance screens was high, indicated by the

high genetic correlation between the two screens, one of which was based on hand-

inoculation in a warm environment (UF screen) and the other using established spray

inoculation methods in a cooler environment (RSC screen). The stability of H2 in the

pooled dataset relative to the individual screens also supports this conclusion. I do not

expect pathogenic variation to significantly change the resistance rankings of these

genotypes, even though these experiments were performed by inoculating hosts with a

single clonal isolate ofF. circinatum. This is because there is little evidence for specific

resistance in this pathosystem; families rank consistently when challenged with different

fungal isolates (G. Blakeslee, personal communication). The facultative nature of this

pathogen presumably creates little selection pressure for the evolution of gene-for-gene

specificity in this pathosystem. Consequently, these clonal rankings may be robust across

a broad range of pathogen isolates and predictive of rankings expected in the clonal field

trials established with these genotypes.

While narrow-sense heritability is an important metric for breeding applications,

our use of clonally replicated material allowed additional heritability calculations, H2,

HC2, and HF2 values, which take advantage of the mating and propagation designs used in









this study. Hc2 is an appropriate metric for association and quantitative trait loci studies,

because genotyping and phenotyping are both done at the clonal level. Accordingly, in

the RSC screen (which involved the most genotypes of the two pitch canker screens) ca.

75% of the variation in lesion length at the clonal mean level was due to genetic

variation. Therefore, I expect lesion length to be an appropriate phenotypic trait for future

QTL identification.

Gall score and gall length are the most heritable fusiform rust traits

Our analysis of gall score (i.e., disease incidence) revealed a non-normal, right-

skewed distribution with one major peak and several minor peaks. The major peak of

apparently "resistant" genotypes may reflect an overestimation of host resistance

because of the use of rooted cuttings. Studies comparing the responses of seedlings to

rooted cuttings have revealed that these two types of plant material behave differently in

response to pathogen challenge, with rooted cuttings showing enhanced resistance (Foster

and Anderson, 1989; Frampton et al., 2000). This enhanced resistance phenomenon has

been observed in other species and is often referred to as "age-dependent" resistance

because the developmental stage of the infected organ is the key driver of resistance, over

and above the action of specific resistance genes (Kus et al., 2002). As clonal host

materials become more widely used in research and plantation forestry, our

understanding of this phenomenon should improve.

Evidence for specific resistance in the loblolly pine-C. quercuum pathosystem has

been obtained using genomic mapping (Wilcox et al., 1996) and by inference based on

family rank changes in response to genetically distinct pathogen cultures (Kuhlman,

1992; Powers, 1980; Stelzer et al., 1997). Although the overall consistency among clonal

performances in our two screens was high, I observed a few family and clonal rank









changes for particular families and genotypes between the ten-gall and one-gall

inoculations (see outliers in Figure 2-4c), suggesting resistance genes in the host

population interacted with specific pathotypes in the inocula. The families showing rank

changes between the two inocula may provide a good starting point for identifying

additional resistance genes in loblolly pine.

Gall length was normally distributed and was heritable, although to a lesser extent

than gall score. Gall length could only be measured on a subset of the population (i.e., on

galled ramets), and this sampling effect may account in part for the reduced heritability

estimates. Our rationale for measuring gall size characteristics was based on work in

slash pine (Schmidt et al., 2000) suggesting that families exhibiting small (short) gall

phenotypes were expressing partial resistance to fusiform rust, based on their lack of

subsequent sporulation. Partial resistance may be a more durable form of resistance given

that it is often race nonspecific (Schmidt et al., 2000 and references therein). I observed

continuous variation in gall length in loblolly pine and found no changes in the relative

rankings of genotypes that formed galls in both screens as indicated by high genetic

correlations (Figure 2-4d). Thus, inoculum type did not appear to exert a major effect on

gall length. Studies involving a number of defined pathogen cultures will be required to

resolve the question of whether gall length is conditioned by (relatively late-acting)

specific resistance factors, or if gall length is a complex trait, potentially involving

multiple genes with small effects.

The relationship between gall score and gall length was of interest, because these

are distinct phenotypes whose genetic relationship is not well understood. The lack of

genetic correlation between gall score and gall length, and the lack of relatedness among









families in the tail distributions for gall score and gall length both suggest that distinct

gene systems condition these two traits. Previous studies have revealed that mean gall

length varies substantially in loblolly pine families phenotyped as "resistant" based on

score (Kuhlman, 1992), providing further support for the conclusion that gall score and

gall length are conditioned by distinct genetic mechanisms. Future identification of QTL

underlying gall length should help distinguish these loci from resistance genes known to

be associated with gall score in loblolly pine (Wilcox et al., 1996).

Resistance to pitch canker and fusiform rust are under the control of two different
mechanisms

The lack of genetic correlation between pitch canker resistance and fusiform rust

resistance (as measured by gall score, or gall length) is consistent with distinct genetic

architectures underlying host resistance to these two diseases. Biotrophic pathogens

suppress host defenses because they require living host cells for survival and nutrient

uptake. Hosts resistant to biotrophic pathogens often activate a localized cell death

response to prevent spread of the pathogen (Thomma et al., 2001). In contrast,

necrotrophic pathogens actively destroy host cells and utilize the released nutrients for

survival. Therefore, a host-cell death-response effective against biotrophic pathogens is

postulated to benefit necrotrophic pathogens by increasing nutrient availability through

accelerated host tissue destruction. I propose that resistance to the necrotrophic pathogen

F. circinatum is mechanistically distinct from resistance to the biotrophic pathogen C.

quercuum due to the differing strategies employed by the two pathogens to incite disease

in the host. This is supported by gene-expression array data, which revealed a lack of

regulation of rust-associated genes after challenge by Fusarium (Morse et al., 2004).

Although I identified families with excellent resistance to both diseases (Figure 2-6),









disease resistance to the two pathogens should be regarded as independent traits by

breeders.

Phenotyping for disease trait dissection in loblolly pine

The work described in this manuscript has assigned specific phenotypic values to

more than 1,000 loblolly pine genotypes, enabling the identification of genes and alleles

that condition resistance through association studies. Genotyping and association studies

are currently underway (ADEPT project Web site, Allele Discovery of Economically-

important Pine Traits, http://dendrome.ucdavis.edu/ADEPT/) for candidate loci (Morse et

al., 2004) thought to be involved in disease resistance in loblolly pine.

In this study, I increased the precision of phenotyping by using clonally propagated

genotypes and mixed linear modeling to adjust for environmental effects. Increasing the

number of ramets for a given clone will increase the clonal mean based heritability for

use in linkage or association studies. However, there is a point of diminishing returns

beyond which adding more ramets does not increase precision of phenotyping. This

population was an excellent starting point to evaluate the heritabilities and relationships

among disease traits. Furthermore, it should afford an opportunity to identify QTL by

linkage and linkage disequilibrium (i.e., association) mapping, which has been proposed

(Wu et al., 2002) and applied with success (Farnir et al., 2002; Meuwissen et al., 2002).














CHAPTER 3
FUSIFORM RUST RESISTANCE COSEGREGATES WITH AN FRI-LINKED
MARKER AND REVEALS VARIABLE PENETRANCE OF THE DISEASE
PHENOTYPE

Introduction

The economic value of pine in the southeastern United States exceeds $19 billion

annually with this region supplying more than half of the nation's pulpwood (McKeever

and Howard, 1996). Loblolly pine (Pinus taeda L.) is the primary pine species in the

region, covering 45% of the commercial forest land (Schultz, 1999) with annual

production of over 1 billion seedlings for planted in reforestation programs (McKeand et

al., 2003).

Successful plantation establishment in the southeastern United States is highly

dependent on the resistance of the planting stock to fusiform rust disease, which is incited

by the endemic pathogen, Cronartium quercuum Berk. Miyable ex Shirai f. sp. fusiforme

(Burdsall and Snow, 1977). The major symptom of fusiform rust disease is the formation

of stem galls which decrease survival, wood quality, and growth, causing an annual loss

ranging from $25-$135 million (Cubbage et al., 2000). Loblolly pine families exhibit

substantial genetic variation in resistance to fusiform rust disease (Kuhlman and Powers,

1988; McKeand et al., 1999) both in greenhouse and the field.

Genomic mapping has identified the region containing FrI (fusiform resistance-1)

conferring pathotype-specific resistance to fusiform rust (Wilcox et al., 1996). RAPD

marker J7_485A was linked to the FrI locus in progeny of a single loblolly pine parent.

Thus, the progeny that have this marker were resistant whereas the ones without the









marker were susceptible to fusiform rust incited by C. quercuum with the corresponding

avirulence allele (Avrl). This genetic marker was consistently predictive of fusiform rust

resistant trees in greenhouse (Kubisiak et al., 2005; Kuhlman et al., 1997), and field

screens (Wilcox et al., 1996).

Screening of fusiform rust disease on clonally propagated loblolly pine has

revealed the existence of "escapes"; ramets that are genetically susceptible yet do not

show any disease symptoms (Foster and Anderson, 1989; Frampton et al., 2000). The

basis for an "escape" can be a passive form of resistance; a random phenomenon where

some cuttings harden off faster than others because of local environment within an

experimental block. Alternatively, there may be a genetic basis for disease resistance,

which may occur, for example if certain genotypes develop succulent shoots in response

to hedging and fertilization at different rates or to different extents than other genotypes.

A genetic analysis can answer this question.

Biologically a clone is susceptible if it has at least one diseased ramet. In this study

I used the same approach which led us to use a genotype based analysis rather than a

ramet based. I used some of the DNA markers developed in previous mapping studies to

distinguish host genotypes that carry/lack the pathotype-specific Fr] allele. I tested the

hypothesis that the Fr] allele is predictive of resistance in greenhouse and field

experiments. Because these studies involved clonally propagated materials, I also

quantified the extent to which genetic and non-genetic factors influence disease

expression levels and escape rate in greenhouse and field trials.









Materials and Methods

Genetic Material

All the clones that were screened in the greenhouse and the field came from 63 full-

sib loblolly pine families obtained from a circular mating design among 32 unrelated

parents with some off-diagonal crossing. The genetic material, the propagation methods,

the inoculations and the data collection were described in (Kayihan et al., 2005). The

parents were from the Atlantic Coastal Plain and Florida provenances of loblolly pine.

Briefly, there were 7-21 clones per full-sib family depending on the family and

experiment and approximately 4 ramets per clone.

Genotyping Families 0 and 1 for Frl

Among 32 parents used in this study, parent number 17 was recognized as

heterozygous for pathotype-specific resistance gene Frl (Frl/Frl; (Wilcox et al., 1996)).

Full-sib families 0 and 1 were generated by crossing parent number 17 with parents 18

and 19 (Kayihan et al., 2005) which were known to be Frl/Frl (unpublished data) and a

total of 61 clones from these families were genotyped using the protocols described in

Wilcox et al. (1996). The J7_470 RAPD marker is linked to the FrI locus and therefore

could be used to predict seedling genotypes (Frl/Frl or Frl/Frl). The mating design

(Kayihan et al., 2005) coupled with clonal propagation allowed direct assessment of

marker-trait co-segregation. Because parents 18 and 19 are homozygous for the Fri allele

(recessive), families 0 and 1 are test-cross progeny and segregate 1:1 for resistance to Fri

avirulent (Avr-Frl) inoculum. Since the maternal parent is heterozygous,

megagametophyte samples were genotyped at the onset of the study to predict seedling

genotypes. At the conclusion of the greenhouse screen, foliage samples were collected









from galled ramets that had been initially genotyped as Frl/Frl, and the genotyping

reactions were repeated on the foliar DNA.

Greenhouse screen

The experimental design was a randomized complete block with single-tree plots

arranged in an alpha lattice with an incomplete block size of twenty. Propagation of

cuttings was described in Baltunis et al.(2005). A total of 63 families were used to

generate 1360 clones for the ten-gall screen and 699 clones for the one-gall screen (Table

3-1). The clones were replicated with a maximum number of five ramets per experiment

(Kayihan et al., 2005).

Table 3-1 Summary of the greenhouse and field screens reported in this study. The 63
families and most of the clones screened were the same across the ten gall,
one gall and field screens. Percentage of diseased ramets and clones are
reported as a measure of infection rate.
Screen # of families # of clones # of ramets % of clones galled % of ramets galled
Ten gall 63 1360 5473 62 36
One Gall 63 698 2743 49 31
Field 60 868 3362 51 26

The ten-gall test was inoculated with aeciospores pooled from a ten-gall collection

from Madison, FL (designated L-10-1-99) (Figure 3-1). The ten-gall inoculum was tested

for virulence against Fr]; I inoculated 100 open-pollinated seedlings derived from parent

17, using 50,000 basidiospores/ml and RSC standard methods (Knighten, 1988). Ninety-

four out of ninety-six seedlings that were chosen for DNA analysis were scorable for the

RAPD marker J7_470 and the marker data obtained from the megagametophytes of these

seedlings were used to detect virulence against Fr]. In order to choose an inoculum with

the least amount of genetic diversity for the one-gall screen, aeciospores collected from

single galls on slash pines in a field site (a generous gift from Dr. Robert Schmidt) were

assessed by Simple Sequence Repeat (SSR) markers using the methods described by









(Kubisiak et al., 2004). The analysis showed that all four of the single gall samples

contained at least four or more SSR haplotypes, indicating a minimum of four fungal

pathotypes in each gall (data not shown). The single gall spore collection (designated

#501) from Nassau, FL (Figure 3-1) was chosen from for inoculation because of its low

genetic diversity, however its virulence against FrI was not known prior to this study.

The artificial inoculation procedures are described in Kayihan et al. (2005) and ramets

were assessed for the presence (1) and absence (0) of a gall 6 months after the

inoculation.


Figure 3-1 The inoculum sources used in the ten-gall (Madison County, FL) and one-gall
(Nassau County, FL) greenhouse trials and field screens (Randolph County,
GA) mapped in Florida and Georgia along with the other areas that were
assessed for virulence. Virulence against FrI was not detected in the inoculum
obtained from these counties (S=Santa Rosa, A=Alachua, B=Bradford,
M=Madison, R=Randolph, N=Nassau; personal communication Dr. Henry
Amerson).









Field

The experimental design was a randomized complete block, with single-tree plots

arranged in an alpha lattice with row/column family restrictions and an incomplete block

size of five. A total of 868 clones from 60 families shared with the greenhouse screens

were evaluated for fusiform rust disease resistance in the field (Table 3-1). The field

location was Randolph, GA at latitude 31.78N, longitude 84.8W (Figure 3-1). Trees

were planted in 4 replications each with 40 columns and 110 rows, inoculation was

allowed to occur naturally and the cuttings were phenotyped for presence (1) or absence

(0) of fusiform rust galls during the second growing season. This is an area classified as

"high hazard" for fusiform rust disease based on south wide classification of >30% of

stems in 5 to 15 years old stands that are likely to have galls on main stems or on live

limbs that are fairly close to the main stem (Anderson et al., 1988).

Data Analysis

Conceptually, all the ramets from a resistant clone should be disease free and all

ramets from a susceptible clone should be diseased. In this study, a clone was labeled as

"resistant" when all of the ramets from that particular clone were disease free. However,

there were a considerable number of cases where only some of the ramets from a

susceptible clone were galled. For the purposes of this study, a single ramet bearing one

or more galls was sufficient to identify a susceptible clone. I converted "ramet based

score" data to a "clone based" dataset by classifying all the genotypes (i.e. clones) with

one or more galls as susceptible (1) and the ones with no galls at all resistant (0).

I used families 0 and 1 as a measure of resistance, since both of these families

included clones that were Frl/Frl or Frl/Frl. A few cases of no marker-trait co-

segregation could be explained by genetic recombination between the marker and Frl.







To evaluate the probability of a crossover between the molecular marker J7_485A and
the FrI locus in a given family I used the formula;

f (x) = Kp(1- p)),,-

where
n is the number of clones in the given family,
x is the number of clones that are putative recombinants in the given family,
p is the recombination fraction.


4


Figure 3-2 Diagrams illustrating genotype (clone) based phenotyping for disease
resistance, susceptibility and escape rate. (A)A clone was declared resistant
with five ramets were disease free. Resistant genotypes were not included in
calculations of 'escape rate'. (B) An illustration of a "susceptible" clone with
five galled ramets (i.e., an escape rate of 0%). (C)An illustration of a
susceptible clone with three galled and two disease free ramets (i.e., escape
rate of 40%).


A& ^


tid


;rd









I defined "escape rate" as the ratio of ramets that did not exhibit a gall, to the total

number of ramets, given a clone that had at least one galled ramet (Figure 3-2).


r-r


where,

Er is escape rate

rt is total number of ramets for a given clone, and

rg is the number of ramets with galls for a given clone

When all the ramets from a clone are disease free, without genetic marker

information one can not distinguish a true genotype-level escape from a clone harboring

additional resistance determinants. Therefore, genotypes that lacked diseased ramet were

deleted from this dataset. Since the escape rate dataset was formed using the percentage

of disease-free ramets in otherwise susceptible clones, this dataset was "clone based".

The "clone based" datasets from all 63 families from the two greenhouse screens

(ten-gall and one-gall) and field screen were analyzed to understand genetic control of

both score (susceptible or resistant) and escape rate. Variance components and genetic

parameters were estimated by GAREML (Huber, 1993) which employs restricted

maximum likelihood estimation (REML) (Patterson and Thompson, 1971) and best linear

unbiased prediction (BLUP) (Henderson, 1973). This approach also aided a more valid

comparison of the score and the escape rate datasets. The linear model used to analyze

the 'escape rate' and the score dataset was:

Yklm = + gLL + gcal + scak + ekm


where,









Yklm is the mth clone of the klth full-sib family,

i is the population mean,

gcak is the random variable female general combining ability (GCA) -NID(0,o2gca) k=l

to 32,

gcal is the random variable male general combining ability -NID(0,o2gca) 1=1 to 32,-

scakl is the random variable specific combining ability (SCA) -NID(,o2,sca),

eklm is the random variable error within the experiment -NID(0,o2e).

The narrow (h2) and broad (H2) sense heritabilities were calculated according to

(Falconer and Mackay, 1996) based on 0, 1 data for disease incidence (gall score) and

decimal equivalent for escape:

2ca V(4P


H2 +ca (V(4 + V(D))
H2 _
lp V(P)

where:

2p is the phenotypic variance,

Sis the total phenotypic variance,
V(P) is the total phenotypic variance,

V(A) is the additive variance,

V(D) is the dominance variance.

Genetic Correlations

The genetic correlation at the family level between the ten-gall and the field screens

was calculated on combined data sets by adding Experiment by GCA ( ) and
was calculated on combined data sets by adding Experiment by GCA ( -ge) and









2
Experiment by Family ( e )interaction factors to the linear model and using the Type B

genetic correlation formula (rB; Yamada 1962):

2 2gca + 2sca
)FAMLY 2 2 2 2
2"gca +O'sca 2ge + se

Asymptotic Z-test

In the field test, there was a possibility of uneven inoculation a result of natural

dispersion of inocula. With the purpose of investigating this possibility I run the

following model in ASREML(Gilmour et al., 2004);

Ykm = + r, + c, + wk + (r c), + ( L),k + a + g + sCalm + clone(family) (n + (r sca),m + eklmn


i is the population mean,

ri is the random replication -NID(0,o2r), i=1 to 4,

cj is the random variable column incomplete block -NID(0,o2c), j=l to 40,

Wk is the random variable row incomplete block -NID(0,oC2), j=l to 110,

gcal is the random variable female general combining ability (GCA) -NID(0,o2gca) k=l

to 32,

gcam is the random variable male general combining ability -NID(0,o2gca) 1=1 to 32,

scalm is the random variable specific combining ability (SCA) -NID(,20,sca),

clone(family)mn is the random variable clone within a family -NID(0,2.,,,,,, ,),

(r*sca) iim is the random variable replication by family interaction -NID(0,o2r*sca)),

eijklm is the random variable error within the experiment -NID(0,o2e).

Asymptotic Z-test for ri, cj, (r*c)ij, (r*w)ik and (r*sca)ilm were calculated by dividing

the variance of these components by corresponding standard deviation.









Results

Inheritance of Fusiform Rust Resistance in the Greenhouse and Field

In a previous report (Kayihan et al., 2005) I calculated fusiform rust disease

resistance for genotypes and disease incidence (score) based on the proportion of ramets

that were diseased for each clone. In this study a clone with at least one diseased ramet

was classified as susceptible (1) and a clone with no diseased ramet was classified as

resistant (0). Clone based resistance to fusiform rust disease (score) was an equally

heritable trait both in the greenhouse and the field screens. Narrow sense heritabilities

(h2) for gall score were the highest in one-gall screen and the lowest in the field screen

(Table 3-2). The broad sense heritabilities (H2) for both the greenhouse and the field

screen were high for this trait (Table 3-2). H2 was highest in one-gall screen which was

followed by ten-gall screen. The field screen yielded the lowest H2 among these screens.

Comparison of h2 with H2 within each screen showed that general combining ability

(GCA) was higher then specific combining ability (SCA) in all trials.

Table 3-2 Summary of score (disease incidence) and escape rate datasets along with
narrow sense heritabilities (h2) and broad sense heritability (H2) for escape
rate and score in ten-gall, one-gall and field fusiform rust screens. All the
families analyzed in one-gall and field were a subset of the families screened
in ten-gall.

Score dataset Escape dataset
# of families # of clones h2 H2 # of families # of clones h2 H2
Ten gall 63 1360 0.39 0.46 62 443 0.23 0.29
One gall 63 699 0.43 0.52 59 337 0.29 0.30
Field 60 868 0.31 0.36 61 439 0 0

The ten-gall and one-gall greenhouse screens were highly correlated (Kayihan et al,

2005). According to the genetic correlations I calculated between the ten-gall and the







41


field screen for score (disease incidence), the families performed consistently, yielding

high genetic correlation (rf= 0.83) (Figure 3-3).


41
29


16
19
56


/ 69
27 35


928


67
23
6S


7
50
46


3860


Figure 3-3 Scatter plot of ranks based on BLUP-predicted family genetic values for ten-
gall and field were plotted against each other (a rank of '1' is the most
resistant and '63' the most susceptible). A least squares regression line is
shown after being forced through the origin due to a not-significant intercept.
The numbers shown are the family identification codes for the full-sib
families.

Validation of Frl marker

Prior to the greenhouse screen using the rooted cuttings, the ten-gall inoculum was

tested for virulence against FrI with use of Frl/Frl parent 17 seedling progeny in North

Carolina State University (personal communication Dr Henry Amerson). From the 94

seedlings that were inoculated with ten-gall inoculum, 47 were genotyped FrI (resistant)









and 47 were Fr] (susceptible). Of the 47 resistant seedlings, none were galled 9 months

after inoculation and in the susceptible group 34 of 47 trees were galled. Hence infection

in the Fr] group was 72%, while infection in the Fr] group was 0%. There was no

evidence of virulence against Fr]. Also infection in Resistance Screening Center (RSC),

Asheville, NC standard susceptible check for loblolly pine (10-8-3) was 76%, so the

susceptible check lot and the Fr] group had the same amount of infection.

Fr] marker data can be used to make inferences about resistance to fusiform rust

disease in families from parent 17. I had three screens (ten-gall, one-gall and field) to

contrast and compare the relative resistance levels of two genotyped families 0 and 1.

Similar percentages of disease incidence in ten-gall, one-gall and field screens at both the

ramet (36%, 31% and 26% respectively) and the clone (62%, 49% and 51% respectively)

levels, gave us confidence to compare them without further adjustment (Table 3-1).

All megagametophytes from families 0 and 1 were genotyped and classified as

either Fr] or Fr]. The marker J7_485A cosegregated with presence/absence of galls in

families 0 and 1 in the greenhouse and field screens (Table 3-3). In families 0 and 1 most

of clones behaved as expected. In the ten-gall screen, twenty four out of twenty nine

clones that were Frl/Frl were disease free whereas twenty six clones were disease free

out of thirty clones that were Frl/Frl. In both the one-gall and the field screens, all the

clones from family 0 and 1 that were Frl/Frl, did not have any disease symptoms as

expected. Twenty two out of twenty five clones that were Frl/Frl were diseased in the

one gall and twelve out of sixteen that were genetically susceptible were galled in the

field test. As I stated before ten-gall inoculum collected from Madison County, FL was

avirulent to Frl.









Table 3-3 Segregation of marker J7_485A linked to Fri gene in families 0 and 1 across
ten-gall, one-gall and field screens (658 ramets combined) with disease
phenotype. Parent number 17 is heterozygous for pathotype-specific
resistance gene Fri. Family 0 is full sib test crosses between parent 17
(Frl/Frl) and 18 (Frl/Frl), whereas family 1 is test cross between 17
(Frl/Frl) and 19 (Frl/Frl). Gray cells highlight the clones that were not in
the expected class.
Inoculum Ten gall One gall Field
Genotype Frl (+) orfrl (-) + + +
Gall + + + + + + -
Family 0 2 13 11 3 0 14 10 0 0 9 7 1
Family 1 2 13 13 2 0 15 12 3 0 9 5 3
Overall observed 4 26 24 5 0 29 22 3 0 18 12 4
Overall expected 0 30 29 0 0 29 25 0 0 18 16 0

However neither the one gall inoculum (Nassau, FL), nor the naturally existing

inoculum in the field trial (Randolph, GA) was tested for avirulence. Yet, 0% of the Frl/-

clones from genotyped families 0 and 1 were diseased in either the one-gall trial or the

field experiment, and 83% of the Frl/Frl clones from these families were diseased in the

one-gall and field screens. This suggests that the inocula used in those screens were

avirulent to FrI, too. Furthermore, in an unpublished study varying the numbers of galls

collected from counties; Santa Rosa, Madison, Alachua and Bradford, Florida virulence

against Fri was found to be very low in Florida (personal communication Dr Henry

Amerson) (Figure 3-1).

There were a few cases where clones did not perform as anticipated. Four out of

thirty clones in the ten-gall screen developed galls even though they were genotyped as

Frl/Frl (Table 3-3). To investigate potential mislabeling problems needle samples from

all the ramets belonging to these four clones were re-genotyped with markers AJ4 420

and J7 470 to ensure their identity. Based on the marker information, all plants were

marker (-) for AJ4 420 and (+) for J7 470 confirming the genotypes previously assessed.









One possible explanation for the lack of marker-trait co-segregation in these cases was

genetic recombination between the marker and Fr]. To evaluate the likelihood of this

occurrence, I calculated the probability of recombination for both full sib families (0 and

1) and low values (9.3x10-6, 2. x10-4 respectively) suggested that a recombination event

between the marker and Fr] gene was extremely unlikely. On the other hand five out of

twenty nine clones in the ten-gall screen, three out of twenty five clones in the one-gall

screen and four out of sixteen clones in the field did not show disease although they were

genotyped as Frl/Frl (Table 3-3). These were potentially examples of physiological

"escapes" that are described in more detail in the next paragraph.

The Genetic Basis for "Escape Rate"

Clonal replication provided multiple observations of individual host genotypes and

helped identify "escapes" (disease free ramets from a susceptible clone, Figure 3-2).

Because most clones that were susceptible had at least one ramet was diseased (83 % of

Frl/Frl clones fit this category; Table 3-3), I used all of the families from greenhouse

and field screens in the analysis. In an idealized experiment in which the "escape rate"

(ER) was zero, all ramets from a given clone would be either gall-free or galled. If a

histogram is plotted illustrating the percentage of galled ramets for a given clone in an

experiment without any escapes I would have two bars; one at 0 percent (resistant); and a

second one at 100 percent (susceptible). To evaluate the distribution of percentage of

galled ramets per clone in each screen, I plotted histograms for the ten-gall, one-gall and

field tests (Figure 3-4). The existence of the bars at 20%-80% drew attention to the

significance of escapes in these screens. This suggested a role for one or more

environmental factors that prevented normal expression of the disease phenotype.





























A 0% 20% 40 0%60 80% 100%
Percentage of galled ramets per clone in Ten-gall screen

400

350

300

250

200

150

100

50

0
B 0% 20% 40% 60% 80% 100%
Percentage of galled ramets per clone in One-gall screen

500
450
400
350
300
250
200
150
100
50
0
C 0% 20% 40% 60% 80% 100%
Percentage of galled ramets per clone in the Field



Figure 3-4 Distribution of percentage of galled ramets by clone in the ten-gall (A), one-
gall (B) and field (C) screens. There were a total of 1471 genotypes (i.e.
clones) in all the experiments and each clone was replicated 1-5 times in each
experiment.


_IIL


_U






46












80O
10 10 0 & 0 40 LO2






Q O 6Wcpo 4








00 000
S00,


a oezo 0 0
0 0






0 10 20 30 40



Column


Figure 3-5 Random distribution of fusiform rust disease resistance performance of ramets
from clones that had at least one diseased ramet in Randolph, GA field trial.
Ramets that formed galls were illustrated as full circles whereas healthy
individuals were presented as empty circles. Column and row refers to the
exact location of each ramet as they were planted in the field. A lack of circle
indicates the position of a ramet in a disease-free genotype.









ER might be an environmentally driven event or it might be under control of

genetic factors. Replication of the clones as ramets and placing them randomly in

different blocks enabled us to calculate escape rate for each clone and compute

heritability for this trait. Clones with no galled ramets were not included in the analysis

since genetic and physiological resistance cannot be distinguished in these cases.

According to our calculations, ER was nearly as heritable as score (disease incidence) in

both greenhouse screens (Table 3-2). This was consistent with the explanation that escape

rate in the greenhouse was controlled in part by genes. In contrast, when I ran the field

data for the same trait, the heritability was zero. Thus, ER was only heritable when the

cuttings were in the greenhouse and it was non-genetically controlled in the field. I ruled

out the possibility that infection occurred in a specific pattern (i.e. only on the north side)

and show that infection was spread uniform in the field; I graphed the distribution of

ramets from susceptible clones in the field (Figure 3-5). I also tested distribution of

resistant and susceptible clones in field to find out if there was any non-random pattern to

their placement in field area. Asymptotic Z-test results were not significant implying that

the resistant and susceptible clones and ramets were distributed in field in a random

fashion.

Discussion

In this study I utilized a large, structured population of loblolly pine that had been

phenotyped in the greenhouse and in the field for resistance to fusiform rust disease,

which is an endemic pathosystem in which specific resistance has evolved.

Agreement Among Greenhouse and Field Screens

Greenhouse disease screens are performed to predict resistance classes the

genotypes will fall into in the field. In the greenhouse I can control nearly all conditions









whereas our influence on conditions in the field is limited to experimental design. One of

the most important parameter in disease screen is the inoculum; C quercuum which has

varying pathogenicity in geographically different field sites (Kuhlman, 1990; Powers and

Langdon, 1977; Powers Jr, 1985; Snow et al., 1975; Snow and Kais, 1970; Walkinshaw

and Bey, 1981). An implication of genetic variation at the pathogen side is unpredictable

performance of elite crosses across field sites (McKeand et al., 2003). Furthermore, it

was discovered recently that in the field, a single gall is usually induced by a single C.

quercuum basidiospore (Kubisiak et al., 2004), whereas concentrated basidiospore spray

system (CBS) inoculation allows multiple haplotypes to infect and form a single gall

(Kubisiak et al., 2005). Thus, the host can be screened for resistance against different

genotypes of virulent pathogen in a single experiment and resistant genotypes would

perform more consistently across the sites. In this study I not only conservatively score a

clone as susceptible if it had at least one ramet diseased but I also screened both in the

greenhouse and the field. This cautious approach aided us in identification of resistant

individuals with complete penetrance of disease resistance trait. Thus, concentrated

basidiospore spray system I used to inoculate the clonally propagated material in RSC

and transformation of continuous data clonall means) to binary scale (0< ramets diseased:

susceptible, 0=ramets diseased: resistant) conservatively predicted resistance/

susceptibility of the genotypes I was testing.

I found that greenhouse and field data showed high genetic correlation, presumably

due to genetic similarity of the inocula utilized in the greenhouse screen and the inoculum

in the natural ecosystem at the field site. These results support the earlier reports where

gall score in the field yielded similar heritabilities as greenhouse screens (de Souza et al.,









1990; Miller, 1983). Moreover the genetic correlation between greenhouse screens and

the field trial was very high suggesting greenhouse screens can be used to select elite

parents, to breed for fusiform rust resistance in the field. Fusiform rust rankings from the

Randolp, GA field site was obtained from 2nd year data which would be considered as

"preliminary" for this disease. Multiple field site data would be available in near future

and then I can compare the field fusiform rust rankings with the ones I calculated from

the greenhouse screens.

Marker-trait Cosegregation for Fusiform Rust Disease Resistance

Markers segregating with resistance genes have been used for selection purposes

over the last decade (Francia et al., 2005). This kind of information recently became

available for the fusiform rust-loblolly pine pathosystem (Wilcox et al., 1996). Host

genotyping with markers linked to the known pathotype-specific resistance gene Fr]

revealed marker-trait cosegregation in both greenhouse and field screens. I used this

RAPD marker data to predict performances of clones from the two families which were

segregating for Fr] resistance gene. The ten-gall inoculum used in the greenhouse

experiment was tested for virulence towards Fr] before this screen and found to be

avirulent (Avr) since after the challenge with this inoculum the clones with Frl/-

(resistant) genotypes were gall-free and Frl/Fr](susceptible) genotypes were mostly

diseased (personal communication Dr Henry Amerson). In the ten-gall screen clones that

were genotyped for Fr] performed as expected from their genotypes, confirming the

previous results. Cosegregation data in the one-gall screen showed that Frl/- genotypes

were disease free and Frl/Frl genotypes were mostly diseased, these results suggested

that the one-gall inoculum was also avirulent to Fr] resistance gene. I did not have

control over inoculum in the field as I did in the greenhouse screens. However, the data









collected in Randolph, GA on families 0 and 1 indicated that the marker was holding up

quite well. Thus, all the inocula I used for these screens were avirulent to Frl.

I identified some clones that were genotyped as Frl/- but expressed a susceptible

phenotype. I initially considered that these may have been cases in which Frl/Frl

genotypes were mislabeled as Frl/-. I repeated the genotyping reactions on diploid

tissues of diseased cuttings and verified that all of ramets for each of the 'unexpected'

classes of genotypes (Table 3-3, grey boxes) gave rise to the same genotypic classes that

were assigned to them based on megagametophyte genotyping. Hence I favor the

explanation that these exceptions to marker-trait cosegregation are due to a low level of

virulence in the inoculum, i.e., a low frequency of basidiospores with virulence to Frl.

Consistent with this view is the observation that all four cases of diseased FrI genotypes

in the ten-gall screen were due to single diseased ramets (data not shown). Moreover ten-

gall inoculum was a mixture often galls from a high hazard site; also the inoculation load

in greenhouse was much higher than in the field. Diseased Frl/- clones could be the

result of a recombination event between the marker J7 485A and the actual resistance

gene Fr]; however, our calculations show that the likelihood of having a recombination

between the marker and the resistance gene in the families 0 and 1 was very low, thus

unlikely. Thus, a much more likely explanation for these cases of disease in the presence

of the DNA marker was that these four genotypes were infected with virulent C.

quercuum genotypes that were present at relatively low frequency in the ten-gall

inoculum.

Penetrance of the Fusiform Rust Disease Phenotype

Clonal propagation enabled us to quantify the penetrance of the fusiform rust

disease phenotype in genotyped and non-genotyped families within the structured









population. Although the disease phenotype was expressed at a similar frequency in both

greenhouse and field screens as revealed by similar proportions of galled ramets and

clones the penetrance of the disease phenotype was dramatically reduced in both

greenhouse screens relative to the field. The basis for this conclusion is that escape rate

(the lack of disease symptom development in a susceptible host genotype) was heritable

and similarly so in magnitude across both greenhouse screens, but not heritable in the

field. The biological explanation for the reduced penetrance of the disease phenotype in

the greenhouse could be driven by pathogenic variation; there may be a low level of

avirulence in the ten-gall and one-gall inocula that correspond to unmonitored host

resistance genes that are segregating in the structured population. If this explanation is

correct, then the ten-gall and one-gall inocula must harbor similarly low levels of

avirulent pathotypes such that equivalent heritability estimates are obtained, and the

heritability of escape rate (ER) is being driven by segregation of resistance genes in the

structured population. Alternatively, the biological explanation for the reduced

penetrance of the disease phenotype in the greenhouse could be driven by host

physiological genetics; there may be inconsistent growth (shoot flush) rhythms among

genotypes that lead to a lack of infection in some genotypes at the time of inoculation. If

this model is correct, then the heritability of ER is being driven by segregation of genes

that directly or indirectly regulate production of shoot tissues that are potential infection

courts. Distinguishing between these competing models should be feasible when

sufficient marker coverage allows association testing between candidate genes and

disease phenotypes (Brown et al., 2004).









Pathogen Infection in an Ecologically Relevant Setting

Certainly there were many differences in both host and the pathogen dynamics in

the greenhouse compared to the field. Because the artificial inoculation with rust spores

in a greenhouse screening trial normally occurs within a narrow window of time

(Knighten, 1988), all ramets may not have succulent shoots that are susceptible to the

pathogen at that time. This phenomenon might stem from the fact rooted cuttings harden

off faster than seedlings, preventing an otherwise successful infection. C. quercuum

prefers young actively growing plant tissue to infect (Griggs and Schmidt, 1977), so a

genetically susceptible ramet might not be convenient for infection if it already had

hardened off. In contrast, field-grown trees may be exposed to inoculum periodically over

a much longer time span of several weeks after spring (Schmidt, 1998), so it is much

more likely for the pathogen to find the host in a succulent state. Random distribution of

more favorable micro sites with more water, fertilizer and sunlight likely affected the

escapes in the field. This approach reinforced the conclusions of previous reports (Foster

and Anderson, 1989; Frampton et al., 2000; Stelzer et al., 1998) where rooted cuttings

became physiologically equivalent to the seedlings after several years in field and had

higher resistance to fusiform rust then the seedlings. It is very likely that these changes in

both host and the pathogen lead to loss of heritability of escape rate in the field.

I also investigated the possibility of a non-random dispersion of basidiospores on

field. However, I did not find any evidence to support uneven dispersion of inoculum

over the field which would result in spatial pattern of susceptible ramets suggesting a

driver of zero heritability. The diseased plants were scattered all over the field site in a

random fashion. Thus, the lack ofER heritability in the field was not a consequence of









spatially nonrandom infection. I surmise that extended periods of inoculation may occur

under natural conditions, obscuring genetic influences of host shoot phenology on ER.

In this study I examined the architecture of fusiform rust disease resistance in a

large structured population in which I considered resistance as a binomial trait (i.e.,

resistant or susceptible) for each host genotype. This is in contrast to other studies in

which disease resistance was scored on a continuous scale for each genotype, based upon

the proportion of ramets that exhibited disease symptoms (Frampton et al., 2002; Isik et

al., 2004; Kayihan et al., 2005). Both approaches have value, based on objectives;

evaluating resistance at clonal level might aid answering biologically important questions

about disease inheritance whereas the clonal mean for disease incidence approach might

reveal quantitative genetically important questions such as the amount of epistasis. Both

approaches have reduced amount of error with the use of clonal propagation. Moreover,

the existence of these clonally replicated field trials presents an opportunity to monitor

potential shifts in pathogen virulence that may occur in the C. quercuum population. Such

shifts may occur in part due to increased planting of resistant genotypes in the field, and

create potentially greater risks if planting stock is clonal.















CHAPTER 4
TRANSCRIPT PROFILING REVEALS POTENTIAL MECHANISMS OF FUSIFORM
RUST DISEASE DEPENDENT SHIFTS IN PINE STEM DEVELOPMENT

Introduction

Loblolly pine is one of the most economically important tree species in the

southeastern United States since loblolly pine plantations cover nearly 13.4 hectares

(Schultz, 1999) in this region and over 1 billion seedlings are planted annually (McKeand

et al., 2003). These plantations have been threatened by the endemic fungus Cronartium

quercuum Berk. Miyable ex Shirai f. sp. fusiforme (Cqf) (Burdsall and Snow, 1977)

which incites fusiform rust disease. Fusiform rust is one of the most destructive fungal

diseases in the South causing damage in millions of dollars every year (Cubbage et al.

2000).

Cqf is a biotrophic fungus that induces gall formation on susceptible trees. The

pathogen causes a number of abnormal changes in the stem the galls themselves have

an organized cellular structure distinct from a normal stem when viewed using light

microscopy. The Cqf hyphae are intimately associated with cortical cells phloem and

xylem ray cells, and with cambial cells, with the hyperplasia (swelling) of the stem

apparently due to an increase in the number of xylem ray cells and vertical resin ducts in

the diseased stem relative to the healthy tissue (Jackson and Parker, 1958). There is

evidence that galls disrupt water transport in diseased trees based on nuclear magnetic

resonance imaging of galled and healthy stems (MacFall, 1994) and expression of

desiccation-associated genes in galled vs. healthy stems (Myburg et al., in press). Galls









weaken the structural integrity of stems such that diseased trees are more susceptible to

breakage by wind, resulting in reduced stocking in stands (Cubbage et al., 2000). Thus

the galls alter both structural and functional features of pine stems.

The interactions between the host and the pathogen during gall formation are far

from being completely understood in the loblolly pine-fusiform rust system. However

recently, transcript enrichment techniques (Warren and Covert, 2004) and microarray

analysis have begun to identify genes from both pine and Cqf that are differentially

expressed at infection, gall initiation and gall expansion stages (Myburg et al., in press).

The Myburg et al. (in press) study is of particular relevance to this chapter, in that I have

re-analyzed the data presented in Myburg et al. to extract information on the expression

profiles of differentially regulated genes. The Myburg et al. paper presented the overall

study design and the analysis focused on contrasts between selected time points to

identify genes with potential roles in specific stages of gall development (i.e., infection,

initiation and expansion). In this study, I focused on the actual expression profiles of

individual genes, which I define as the observed change in transcript abundance across

time intervals. This analysis allowed me to identify genes whose profiles differed by

treatment (pathogen vs. control), genotype (resistant vs. susceptible) and disease state

(diseased vs. healthy).

Materials and Methods

Plant Material, Genotyping and Harvesting

The genetic materials, fingerprinting and sample collection methods are described

in a previous study (Myburg et al. in press). Briefly, seedlings from the cross (10-5 Y x

4666-4 o) segregate for Fr] because it is a testcross between genotype 10-5, which is

heterozygous for the dominant resistance gene Fr] (Wilcox et al., 1996), and 4666-4









(Frl/Frl). The megagametophytes were harvested from each germinating seedlings and

the haploid tissue was screened for RAPD markers J7_470 (Wilcox et al., 1996) and

AJ4_420 that define a ~1 cM interval containing Fr]. The RAPD markers identified 350

resistant (Frl/Frl; +J7_470, -AJ4_420) and 350 susceptible (Frl/Frl; -J7_470,

+AJ4_420) individuals.

Among the 350 individuals in each resistance class 210 individuals (15 seedlings x

2 biological reps x 7 time points) were challenged with Cqf and 140 individuals (10

seedlings x 2 biological reps x 7 time points) were inoculated with distilled water (as

control). Twenty additional seedlings were water inoculated and included in the study as

index plants. These index plants were marked with two black ink spots, one immediately

below the apical bud and the other approximately 1.5 cm below the first spot where the

potential gall formation with take place. Using the references from the index plants,

tissue from this region was harvested before the onset of visible disease symptom. The

first harvest time point was 90 min after inoculation followed by additional harvests at

6hrs, 24hrs, 7 days, 28 days, 56 days and 112 days post inoculation.

Fungal Material and Inoculation

A single aecisospore isolate of C. quercuum that was avirulent to Fr] (SC 20-21,

obtained from E.G. Kuhlman, USDA-FS, retired) was used in the inoculations that were

performed at the Resistance Screening Center, Asheville, NC following their standard

inoculation protocol except inoculum was increased by 400% to minimize escapes.

Microarray

Experimental design, microarray preparation, target synthesis, microarray

hybridization and scanning were described in Myburg et al. (in press).









Statistical Analysis

The experiment was implemented in a balanced incomplete block design (Kerr and

Churchill, 2001) and analyzed using a mixed model approach (Wolfinger, 2001). In order

to compare the gene expression on several arrays treated with RNA from different

treatment X genotype combinations that were dyed with 2 different dyes I applied a

normalization data to the entire dataset. The normalization model was chosen with

respect to the significance of the effects that extracted from the full model where every

effect and the combinations of effects were tested. I assumed dye, genotype, time and

treatment did not change the overall level of gene expression in a single biological

sample. The resulting normalization model for the log2 transformed data (yijklm) was:

Yklm = + a, + D +Gk + T + M m + D *M + M T,, + e klm

where:

t is the population mean.

ai is the random variable array -NID(0,o2a), i=1-56

Dj is the fixed resolvable dye, j=1-2

Gk is the fixed resolvable genotype, k=1-2

T1 is the fixed resolvable treatment, 1=1-2

Mm is the fixed resolvable time, m=1-6

D*Mjm is the fixed resolvable dye by time

T*Mlm is the fixed resolvable treatment by time

eijklm is the random variable error within the experiment -NID(0, 2e).

Before residual values derived from this model were incorporated into the gene-

specific model I deleted the two early time points (1.5 hrs and 6hrs) since any of the three









time points within a 24 hr time period should represent the stage of stem development at

the time of the inoculation. The multiple early sample times were initially chosen to try

and capture gene expression shifts associated with a rapid hypersensitive response;

however this was unsuccessful (Myburg et al., in press). I chose to retain the 24 hr time

point since the inoculation procedure involves incubation of trees in a humid chamber for

6-12 hr, whereas the 24 hr (hereafter referred to as "Iday"), 7 day, 56 day and 112 day

samplings were all carried out in the greenhouse. The 1 day, 7 day, 56 day and 112 day

data were then analyzed using the gene model:

y,k = + a +(a) + Gk + +M +G* + +M* T,, +G*Tkl +G*T* f,.. +ekem

where:

[t is the population mean.

ai is the random variable array -NID(0, C2a), i=1-56.

s(a)ij is the random variable spot number(array) -NID(0, 2s(a), j=1-4.

Gk is the fixed resolvable genotype, k=1-2.

T1 is the fixed resolvable treatment, 1=1-2.

Mm is the fixed resolvable time, m=1-4.

G*Mkm is the fixed resolvable genotype by time.

T*Mim is the fixed resolvable treatment by time.

T*Mklm is the fixed resolvable genotype by treatment by time

eijklm is the random variable error within the experiment ~NID(0, a2e).

I used PROC MIXED in SAS (SAS Institute Inc. SAS/STAT Software version 9,

SAS Institute, Cary, NC) to run both the array and gene level models (Wolfinger et al.,

2001) (Figure 4-1). I identified genes that were significant for genotype, treatment, time,







59


genotype by time, genotype by treatment, time by treatment and genotype by time by

treatment using the ANOVA F-test combined with Bonferroni multiple testing correction

(p=0.01). Using this conservative criterion for significance, a total of 861 genes were

significant for one or more of these effects.

3,705 genes on 56 arrays


4 treatments in a 2x2 factorial arrangement
(C=water control, I=pathogen inoculated) x (R=Frl/frl, S=frl/frl)
4 time points (Iday, 7days, 56day, and 112days)

I
Array-level adjustment for the entire experiment
(Wolfinge et al., 2001)

ANOVA F-test + Bonferroni multiple testing correction (p=0.01)

861 significant genes
with 1 or more significant interactions
I
Gene*treatment interactions each assigned a "profile" across the three time
intervals based on increased (>+3SD), unchanged (+3SD>x>-3SD) or
decreased (<-3SD) transcript abundance across adjacent time points

All of 27 possible gene*treatment profiles
are occupied (Figure 4-2)


Identification of genes whose profiles vary by treatment


218 genes left


I I I I
Same profile C vs. I R vs. S H vs. D
all treatments
No. of genes: 72 20 29 97


Figure 4-1 Flow chart illustrating the procedure to identify significant and biologically
interesting gene expression profiles. ANOVA was performed for each of the
3705 genes. After experimentwise correction for multiple testing, significant
gene*treatment interactions were assigned "profiles" comprised of three time
intervals (ld to 7d; 7d to 56d; 56d to 112d) based on shifts in the LS mean for
transcript abundance at each adjacent time point. Biologically interesting
profile contrasts are explored; "SI vs. rest" is equivalent to "diseased vs.
healthy."









I established profiles of transcript abundance that these 861 genes followed on the

four time points (i.e. three intervals) (Figure 4-1). To establish these profiles I began with

the least square mean for each significant gene in each treatment and time point. Using

the "pdiff" option I identified mean differences and standard deviations associated with

the mean differences, for each consecutive time point within each treatment. At any given

time increment (e.g between 1 day and 7 days), expression of a gene can increase,

decrease or not change. A decision rule was applied in which expression for a gene was

declared "not changed" if the absolute mean difference between the adjacent time points

was within 3 standard deviations. However if the absolute mean difference of gene

expression was greater than 3 standard deviations within a time increment it was declared

"increased" or "decreased." I simply joined the three adjacent time intervals for a gene in

order to assign it a "gene profile." A gene profile is thus comprised of three integers

representing the change in gene expression across the three adjacent time intervals in the

experiment. For example, a gene that increases in expression at each interval has a profile

1 1 1, whereas a gene that decreases during the first interval and does not change

thereafter has a profile -1 0 0.

Results

Among 3705 genes that were evaluated for gene expression, 861 genes were

significant for time, genotype, treatment, time*treatment, treatment*genotype,

time*genotype, or time*genotype*treatment with Bonferroni corrections for multiple

testing (p=0.01) (Figure 4-1). Since there were only 3 time intervals and 3 possible trends

within each interval, there can be only 27 distinct profiles. The diversity of profile types

among significant genes was sufficiently high to occupy all 27 possible profile groups. I

















2- 2" 2- 2-

0 0" 0" 0"

-2 -2" -2" -2"
-3 -3 -3" -3
Time point Time point Time point Time point
A(N)=52 B(N)=176 C(N)=40 D(N)=121

2' 2 2 2
1 1

11 1 1
-2" -2 -2 -2
-3 -3 3 -3
Time point Time point Time point Time point
E(N)=606 F(N)=68 G(N)=66 H(N)=181




0 0 0 0
2 2" 2" 2 -

-2 -2 -2 -2
-3" -3 -3 -3
Time point Time point Time point Time point
I(N)=12 J(N)=50 K(N)=270 L(N)=57

2- 2" 2" 2"


1" 1" 1 1
0 0" 0 0
-2- -2" -2" -2"
-3 -3 -3" -3
Time point Time point Time point Time point
M(N)=38 N(N)=536 O(N)=152 P(N)=4





1 1 1 1
2" 2 2 2"

0 0 0 0

-2 -2 --2 --2
-3 -3 -3 -3
Time point Time point Time point Time point
Q(N)=156 R(N)=13 S(N)=29 T(N)=182

2 2" 2" 2 -



-2 -2 --2 .. --2
-3 -3 -3 -3
Time point Time point Time point Time point
U(N)=I 1 V(N)=80 W(N)=5 X(N)=399


2"
1



-3"
Time point
Y(N)=7


2



-2"
-3
Time point
Z(N)=132


2"
0


-3-

Time point
ZAA(N)=5


Figure 4-2. Analyses of mean gene expression data support 27 distinct profile groups,

"A" through "AA." The genes that were significant for time, genotype,

treatment, time*treatment, treatment*genotype, time*genotype, or

time*genotype*treatment (p=0.01) were categorized into 1 of 3 possible

profiles (i.e. up, down or unchanged) based on variation 3 standard deviations

from the mean. Mean gene expression (y-axis) was plotted across the time

intervals (x-axis) for each profile group. Time points are indicated as 1=1

day, 2=7 days, 56 days=3 and 112 days=4. N= the number of gene or gene

interactions that fall into a particular profile group.










described these profile groups and counted the gene-treatment combinations falling into

these groups (Figure 4-2). The frequency at which non-regulated gene and gene

interaction profiles were detected ("N"; N=536), however the number of genes that were

unregulated in all four treatments was much lower (N=5 genes; data not shown).

The criteria I used to distinguish profiles allowed me to investigate the trends of a

given gene expressed in resistant-control (RC), resistant-inoculated (RI), susceptible-

control (rC), susceptible-inoculated (rI) treatment conditions. There were 72 genes that

did not respond differently to the genotype-treatment combinations; in other words, the

profile of gene expression was not affected by genotype or by fungal inoculation (Figure

4-3).


50
unknown or no hits cellular development
S40 -

30

20



signal transduction
protein synthesis and stability

1 2 3 B
A
Time Interval

Figure 4-3. Genes with the same expression profile in all treatment combinations were
predominantly induced during the first time interval. (A) For each time
interval, profiles were summed across the significant genes and plotted. Time
interval 1=interval from time point 1 (Id) to time point 2 (7d), interval
2=interval from time point 2 (7d) to time point 3 (56d) and interval 3=interval
from time point 3 (56d) to time point 4 (112d). (B) Chart of the genes
significantly up-regulated in the first time interval but unchanged with respect
to treatment or genotype after categorization into functional groups.









As stated before each gene was assigned a series of numbers (1, 0 or -1) to describe

the trend within a treatment. The sum of these numbers (across genes) for a given time

interval illustrates the profile of overall expression for those genes (Figure 4-3A).

According to this analysis the genes that had the same profiles in all treatment

combinations were predominantly up-regulated within the first time interval (1 day-7

days). For the rest of the time intervals the expressions of the 72 genes were nearly

equally up and down regulated. There were 41 genes with a profile in which expression

increased in interval 1 (Figure 4-3 B); these genes were in category "E" for all gene-

treatment combinations (Figure 4-2).

I extended this analysis to identify potentially interesting genes whose profiles

changed according to treatment (Figure 4-4). Genes significantly regulated between

"control" vs. "inoculated" treatment classes differed mainly in the 3rd interval (56-112

days) (Figure 4-4A). In contrast, genes regulated between "resistant" vs. "susceptible"

genotypic classes differed mainly in the 1st time interval (1-7 days) (Figure 4-4B). In the

comparison of "healthy" (resistant-inoculated, resistant-control, susceptible-control) vs.

"diseased" (susceptible-inoculated) classes (Figure 4-4C) the 2nd time interval (7-56

days) was the interval during which most genes were regulated. Among these

biologically interesting comparisons only the healthy vs. diseased contrast revealed genes

of fungal origin. There were 13 known fungal genes regulated in this category most of

these occurring in the second time interval (7days-56 days) with possible additional

fungal genes classified as "unknown" origin (Figure 4-4C).













12.5 -
450-

10.0 -
I 1 I. 0 5
5.5 -
.30 -

2.5 I
10


El-


1 2 3 12 13 23 123 1 2 3 12 13 23 123
1 2 3 12 13 23 123
A Tim interval B Ti interval C Ti interval

Figure 4-4. Profile groups can be categorized into biologically interesting clusters with
distinct changes in gene expression patterns. Genes significantly regulated
across "control" and "inoculated" treatment classes (A), across "resistant" and
"susceptible" genotypic classes (B) and across "healthy" (resistant-inoculated,
resistant-control, susceptible-control) and "diseased" (susceptible-inoculated)
classes (C) are shown. Time intervals 1, 2, and 3 are as defined in Figure 2.
12 = intervals 1 and 2 combined, 23= 2 and 3 combined, 13= 1 and 3
combined, and 123= 1, 2 and 3 combined. The genes that originated from
host are represented by a black bar, genes with no known origin are
represented by a gray bar and fungal genes are represented by a white bar.

I identified genes that were differentially regulated in diseased vs. healthy tissues,

and found there were dramatic shifts in all three intervals. The three intervals had been

previously characterized as reflecting distinct phases of disease development, specifically

the 1st time interval = infection, 2nd time interval = gall initiation and 3rd time interval =

gall expansion (Myburg et al., in press). I contrasted the profiles in diseased vs. healthy

tissues in order to identify the direction of regulation imposed by the pathogen (Figure 4-

5). In the "infection" interval, most of the genes that are differentially regulated were

down-regulated in diseased seedlings. During gall "initiation" the majority of the genes

were up-regulated and 11 of the 50 genes that were in this class were fungal genes

whereas 27 of them were pine genes and the rest (12 genes) were unknown. The "gall

expansion" phase was dominated by pine genes, with half of the genes in this class up-

regulated and the other half down-regulated under the influence of Cqf
















NXCI 082 GO1 calcium ion binding
NXCI153 A02 F CLAVATA1 receptor kinase
NXLV 022 H08 F ras related protein RAB8 5
NXNV 096CO9 putative sparagine synthetase
NXNV 125 E12 F ASP1 (ASPARTATE AMINOTRANSFERASE 1)
NXNV147 G03 F cystein proteinase (by similarity)
NXPV 062 E04 F AUX1 AUXINN RESISTANT 1)
NXLV 100 F021 auxin:hydrogen sylporter
'infection' NXSI 059 G09 Adoosine kinase

24 genes NXSI 102 F11 putative UMP/CMP kinase a

(21 host, 3 unknown) NXSI 067 10 F putativenitrilase 2

SNXNV015H07 Calcium binding EF hand
NXNV 073 G04 Putative cellulose synthase catalytic subunit
NXNV 132 G1 F Cellulosesynthase



NXLV100 F02 F Auxin:hydrogen symporter
NXLV 012 A05 F Annexin
NXLV 049 GII F LMWheatshockprotein
NXNV135 EO01 F Dirigent like protein pDIR4
NXSI 099 1H06 F (+) alphapinenesynthse
NXSI 104 B11 FED A; electron transporter/iron ion binding






NXCI 018A08 pectatelyase
NXCI 027 G06 Inositol 3 phosphatesynthase

gallinitiation' NXCI 067 H1106 Sadenosyl methioninesterol C methyltransferase
NXCI 075 D09 Epoxidehydrolase
50 genes NXLV079GO7 F Leucine-rich repeat ansmembrane protein kinase 1

(27 host, 11 fungal, NXNV 122 C07 F Calciumdependent protein kinase
12 unknown)
NXSI 063 D01 Flavanone 3 hydroxylase
NXSI 076 E08 Putativemitochondrial dicarboxylate carrier protein
NXSI 101 BO Putative glycine-rich protein
NXSI 103 D F Putative sinapyl alcohol dehydrogenase
pi1341 Acyltransferase/carboxlic ester hydrolase/lip as e









NXNV 129 F06 PIN1 like auxin transportprotein
NXNV 163 F07 F Acyl CoAbinding
NXPV 068 E06 F Xyloglucan endotransglycosylase
NXRV064 CO7 F Putativexyloglucan endotransglycosyl*e
NXRV079 DO1 F Xyloglucan endotransglycosylase
NXSI 103 E12 F Xyloglucan endotransglycosylaseXET1
'gall elongation'

20 genes

(15 host, 2 fungal,
3 unknown)
NXLV103 E01 F Flavanone 3 hydroxylase 2
NXSI 008 GlT Late embryogenesis abundant protein
pi295 Typell proteinaseinhibitor family protein





Figure 4-5 Clustergram of gene profile differences (or contrasts) between diseased and

healthy treatments. Genes whose expression in diseased tissues are higher

than in healthy tissues for a given interval are shown in red; if lower, green; if

identical, black. Examples of genes in selected clusters are shown on the right.



The genes regulated during the "infection" phase are similar to genes that encode



proteins involved in auxin transport (NXLV_100 F02 = auxin:hydrogen symporter;









NXPV_062_E04 F = AUX1 auxin transporter) and auxin biosynthesis (NXSI_067F10OF

= nitrilase 2, an IAA biosynthetic gene). The "infection" phase also appears to involve

calcium fluxes as indicated by regulation of genes involved in calcium signaling

(NXCI_082_G01 = calcium ion binding protein). During gall "initiation" cell wall

modification enzymes appear to be induced under the influence of Cqf

(NXNV_132_G11F = cellulose synthase; NXCI_018A08 = pectate lyase;

NXSI_ 101 B01 = putative glycine-rich protein; NXSI_103_D 11F = putative sinapyl

alcohol dehydrogenase). During the gall "expansion" phase, four distinct members of the

xyloglucan endotransglycosylase family were down-regulated, potentially in association

with additional cell wall architecture modifications during gall growth (NXSI_103_E12F;

NXPV_068_E06F; NXRV064_C07_F; NXRV079_D01F = xyloglucan endotrans-

glycosylase).

Discussion

Cronartium quercuum is a biotrophic, macrocyclic, heteroecious fungus (Burdsall

and Snow, 1977) that incites abnormal changes in the fusiform rust susceptible pine stem

such as swollen phloem cells, and increase in the number of resin ducts and ray cells

(Jackson and Parker, 1958; Gray and Amerson, 1983, Jewell et al., 1962; Miller et al.,

1976). To identify genes and processes that may underlie the development of disease

symptoms, I used a microarray dataset derived from a time course analysis of fusiform

rust disease development. The genes on the microarray included ESTs and cDNAs

obtained from subtraction libraries and from genes that are expressed during the

interaction between the host and the pathogen, an approach that has been successful in

identifying genes that may condition disease phenotypes (Birch and Kamoun, 2000; Wan

et al., 2002). These powerful tools became available for the fusiform rust-loblolly pine









pathosystem recently (Warren and Cover, 2004; Myburg et al., in press). In this study I

investigated fungal and pine gene expression on the pine stems obtained from resistant

and susceptible seedlings that were inoculated with Cqf or water using an experimental

design that involved 4 time points. I captured a diverse array of expression profiles across

the genes that were significant for the main effects or interactions of the main effects I

was testing.

Transcription Profiling Reveals Differential Gene Expression

Diverse patterns of gene expression were observed in this study, such that all

possible combinations of 27 profiles were occupied by at least one gene or gene

interaction. The chip-level model that I used, effectively corrected for statistically

significant treatment and interaction effects at the chip level prior to the gene-level

analysis. This approach assumes that large-scale unidirectional shifts in gene expression

in any particular treatment or treatment combination are based on technical artifacts, not

biologically meaningful effects, in the microarray experiment. A chip-level adjustment

model with fewer terms may be more biologically appropriate for this study, in that it

may identify more genes whose expression is altered; on the other hand, such an

approach may lead to more false positives. The decision to use 3 SD as a criterion for

significance was based on striking a balance between identifying potentially interesting

biological mechanisms, while still being sufficiently conservative to exclude most false

positives. An indication that this criterion was reasonably conservative was the

observation that although all of the gene and gene interactions were highly significant

experimentwise (after Bonferroni correction), 648 profiles were declared biologically

non-significant based on the 3 SD criterion.









There were 72 genes that showed identical profiles regardless of treatment

(genotype or pathogen). The majority of the genes in this category belonged to group "E"

which is characterized by an increase in interval 1 and stable expression thereafter. The 1

day to 1 week period of pine seedling growth during this study is predicted to be a time of

active primary stem growth and development. The annotation of the genes in group "E"

for all treatments revealed that they belonged to functional groups such as metabolism,

protein stability, signal transduction and cellular development, which is consistent with

the kinds of functions expected in an actively growing seedling. Notably, these genes are

not affected by genotype or by pathogen challenge, so they are more relevant to stem

development in loblolly pine than to disease development per se. Such genes presumably

reflect the juvenile developmental state of the stem when the seedlings were harvested at

the earliest time point in the experiment.

When the transcriptomes of control and pathogen inoculated seedlings are

compared, one might expect that the difference in gene expression would be at the first

time interval since that would be the period during which pathogen spores are

germinating and contributing to the transcriptome (this might occur on both resistant and

susceptible seedlings). Interestingly, this analysis suggests otherwise, in that the effect of

Cqf inoculation is on the regulation of genes much later than the first time interval. In

fact, the effects of Cqf inoculation on gene expression profiles gradually increased

through the 3rd time interval. Thus, both resistant and susceptible plants gave the same

response to a fungal infection attempt that resulted in differences between the control and

the inoculated plants. It is intriguing to speculate that some kind of pathogen-induced

systemic response might be responsible for this observation. However, known examples









of long-term pathogen-induced responses (such as systemic acquired resistance and

induced systemic resistance) are typically incited by an incompatible interaction (resistant

host) distinct from the compatible interaction (susceptible host), which is not consistent

with my findings. In addition, the number of genes in this group was relatively small

(N=20), so further gene expression studies should be performed to confirm and resolve

this phenomenon.

The comparison of profiles in resistant vs. susceptible genotypes revealed

differential gene expression, regardless of whether the plants were inoculated with Cqf or

the water control. This is interesting in and of itself, since it suggests that allelic

differences at the Fr] locus (and loci linked within 1 cM of Fr]) can be detected at the

level of gene expression. The level of genetic resolution in this study was reasonably

high, in that flanking markers were used for genotyping (thus only double recombinants

within a 1 cM interval would be misclassified), and a relatively large number of seedlings

was bulked for each time point (50 Frl/Frl and 50 Frl/Frl seedlings for each time

point). Therefore the gene expression comparisons at each time point are likely to reflect

allelic differences at or near Fr] but mixtures of both alleles at unlinked loci. One would

expect that regulation could occur in cis- or in trans-. If cis-regulation is being observed

here, then markers within the regulated genes could be used to create a fine map of loci in

the Fr] interval. If trans-regulation predominates (e.g., Kirst et al. 2005), then allelic

configuration at or near Fr] may induce downstream signaling mechanisms that are

manifest on the microarrays.

Influences of Fusiform Rust Disease Development on Gene Profiles

The healthy vs. diseased comparison revealed over a hundred genes that were

differentially regulated across all time intervals. In this section, I elaborate on the types of









putative gene functions and physiological mechanisms that may be involved in the

development of the fusiform rust disease state.

In the first interval, a dramatic (down-) regulation of host regulatory genes and

auxin biosynthesis and transport genes, in the developing disease state. Nitrilase 2 is the

enzyme that catalyzes the conversion of indole-3-acetonitrile (IAN) to indole-3-acetic

acid (IAA) which is active auxin (Woodward and Bartel, 2005). The auxin:hydrogen

symporter is an efflux auxin carrier whereas AUX1 is an influx auxin carrier. Therefore,

in diseased seedlings it is feasible that auxin transport as well as biosynthesis is impaired

(or its transport is modified under the influence of the pathogen) within infected cells.

Ca+2 is a well-known second messenger acting downstream of many stimuli including

hormone signaling, and specifically in auxin signaling since calmodulin binding proteins

are encoded by members of the auxin response gene family (SAURs; Yang and Poovaiah,

2000). Interestingly, during gall expansion there is a similar, coordinated regulation of a

putative auxin transport protein-encoding gene (Pinl-like) and a gene encoding a protein

containing a putative calcium-binding EF-hand. This provides another potential

connection between auxin and calcium signaling in fusiform rust disease development. A

potentially fruitful area for future research would be to quantify hormones in developing

fusiform rust galls, since there is also evidence for a potential role of gibberellins in later

stages of gall development (Myburg et al., in press).

In the second interval, transcripts encoded by the fungal pathogen were detected.

The detection of fungal transcripts presumably reflects the establishment of a compatible

interaction with the host (Heath, 1997) whereby the pathogen has begun to disperse

within the host tissue (Walkinshaw, 1978) to manipulate the host sink to its needs. Since









Cronartium spp. are ecologically and economically important pathogens on pines (for

example, the white pine blister rust is incited by Cronartium ribicola J. C. Fisch; Jurgens

et al., 2003; Hudgins et al., 2005) there would be value in using transcript profiling

approaches to better understand gene expression shifts in the pathogen component of the

disease interaction in the future.

In the second interval I also observed disease-altered profiles of host genes

potentially involved in modifying cell wall architecture. Cellulose synthase and pectate

lyase were shown to be up-regulated in xylem (compared to leaves) (Paux et al., 2004),

and in Populus reaction wood (Andersson-Gunneras et al., 2006). Sinapyl alcohol

dehydrogenase is involved in the monolignol precursor pathway leading to lignin

biosynthesis (Anterola et al., 2002) and glycine-rich proteins are thought to be involved

in structural integrity and inducible reinforcement of plant cell walls (Ringli et al., 2001).

Thus, host genes related to cell wall synthesis may be induced as a suite of genes required

for rapid cell wall biosynthesis associated with the initiation of a gall.

In the third interval I observed regulation of a multigene family whose products are

involved in growth and cellular architecture (loosening and/or tightening of the cell wall).

Cellulose microfibrils are typically stabilized by xyloglucan moieties, which can be

cleaved via xyloglucan endotransglycosylase (XET) so that cell can alter its shape (Fry et

al., 1992). One of the family members is induced in the gall expansion phase, whereas

four members are repressed, suggesting the family members have non-redundant roles in

cell wall modification. It is interesting to note the rapid yet organized manner in which

specific cell types in fusiform rust galls develop ray cells increase in size and number,

whereas resin ducts increase in number only (Jackson and Parker, 1958) whereas the






72


attachments among parenchymal cells appear to loosen (Walkinshaw, 1978). Given the

complex dynamics of cell architecture changes, it is tempting to speculate that some of

the XETs may play roles in specific cell types and thus be performing distinct functions

accordingly.

These results validate the stages of disease development proposed by Myburg et al.

(in press) that seedlings can be analyzed according to the discrete disease phases of

infection (1-7days), gall initiation (7-56 days) and gall expansion (56-112 days).














CHAPTER 5
CONCLUSION

Throughout this study I investigated two fungal diseases that are threatening

loblolly pine plantations in southeastern US. The first disease is fusiform rust which is

incited by biotrophic fungus Cronartium quercuum (Burdsall and Snow, 1977), the

second one is pitch canker which is incited by the nectrorophic fungus Fusarium

circinatum (Nirenberg and ODonnell, 1998). Biotrophic and necrotrophic fungi have

different life cycles and infection mechanisms. Thus, a resistant host would respond

differently to these two fungal diseases. A fusiform rust resistant host would recognize C.

quercuum induce HR whereas F. circinatum, the pathogen can survive HR based

defense by detoxification which is a common protection for necrotrophic fungi (Mayer,

Staples, and Gil-ad 2001). Another major difference between C. quercuum and F.

circinatum is the fact that F. circinatum requires wound to enter the host (Kuhlman 1987)

while C. quercuum does not. Another host defense after the infection can be delimiting

fungal growth by forming cell appositions to stop the disease progression. It was

demonstrated that to C. quercuum triggers necrosis in resistant loblolly pine and forms

wall appositions that are partially composed of callose to prevent infection (Gray and

Amerson 1983). It was also reported that a distinct lesion formation is a type of resistance

which limits F. circinatum growth in loblolly pine (Barrows-Broaddus and Dwinell

1983). I propose C. quercuum and F. circinatum have very different life cycles and

infection styles as a result the way host responds to them should be distinct, too. Later in









the study I showed that in loblolly pine resistance to fusiform rust is not correlated to

resistance /susceptibility to pitch canker and vice versa.

I also investigated Fri marker co-segregation with the disease phenotype in two

families that were a result of a cross between Fri/frl andfrl/frl parents. The offspring

from the crosses were genotyped for Frl locus. I showed that Frl marker information is

predictive of resistance in the both greenhouse and the field as long as the inoculum is

Avr to Frl locus. Using clonally propagated material I increased the precision and also

identified the ramets that escape. I showed that escape is a heritable trait in the

greenhouse whereas it becomes inheritable in the field due to inoculum source that is

presented for longer periods of time compared to a few minutes in the artificial

inoculations.

In the last part of this study I switched gears to investigate gene expression in

resistant (Frl/frl) and susceptible (frlfrl) individuals that were inoculated with water or

C. quercuum in a time frame of 112 days. I showed that there are genes that have unique

expression profiles across control vs. inoculated, resistant vs. susceptible and healthy vs.

diseased. These results would be helpful to parse the disease development in susceptible

loblolly pine seedlings.

Neither fusiform rust nor pitch canker disease resistance is far from being

completely understood. Although this study brings us closer to the answers scientists are

seeking, more research on biology of these fungal diseases is needed.














APPENDIX A
SAS SCRIPTS FOR MICROARRAY ANALYSIS

/*Frl2 is the dataset that contains treatment genotype time_point genename log2

expression here I am dropping all the control genes since they are causing extra noise*/

data Frl2; set tmp 1.Fr] 1;

if gene='Blank water control' or gene='Blank' or gene='BLANK' or gene='Water

Control' or gene=' Control BAR' or gene=' Control GFP' or gene=' Control Globulin'

or gene='SP3' or gene='SP2' or gene='SP1' or gene='Sp3' or gene='Spl' or gene='Sp2' or

gene='Spike Control Spl' or gene='Spike Control Sp2' or gene='Spike Control Sp3' or

gene='Spike Control Sp4' then delete; run;

/*after control_genes are deleted, run the array-level analysis to get rid of major

effects like dye*/

proc sort data=Fr]2; by dye gene time; run;

proc mixed data=Fr]2; class array dye genotype treat time;

model log2i= dye genotype treat time dye*time treat*time/OutPred=no_contr;

random array; run;

/* after this delete time points=all, 6 hours, 90 mins*/

proc sort data=tmpl.Frno_contr; by genotype time; run;

data Fro_contr ri 6_90; set tmpl.Frno_contr;

if genotype='r' and treat='I' then delete;

if time='all' or time='6' or time='90' then delete; run;

/*now I can run the second level analysis where gene by gene*/









proc sort data=Frno_contr r i 6_90; by gene;

proc mixed; by gene; class array treat time dye spotnumber genotype;

model resid=time2 treat genotype time2*treat treat*genotype time2*genotype

time2*genotype*treat;

random array spotnumber(array); ods output Tests3=pvalno_cr i 6 90; run;

proc sort data=pvalno_cr i 6_90; by probf; run; data Geneno_c r i 6 90;

set pvalno_cri_6_90; proc sort; by ProbF;

where ProbF<3.85683E-07 ; run;

proc sort data=Geneno_c r i 6_90; by gene; run;

/* get the significant genes, only one copy of gene should be in the list*/

data once_sig_geneno_690ri; set Gene no c r i 6 90;

by gene; if first.gene; run;

/*here I merge the significant gene list with the residuals from the array level

analysis to get Ismeans and stdev associated with them*/

proc sort data=Frno_contr r i 6_90; by gene;run;

proc sort data=once_sig_geneno_690ri; by gene;run; data resno_690ri;

merge Fmo_contr r i 6_90 once_sig_geneno_690ri; by gene; run;

proc sort data=resno_690ri; by ProbF;

data clean res no 690ri; set res no 690ri; if ProbF<0 then delete; run;

proc sort data=cleanres no_690ri; by gene; run;

/*it is to get Ismeans and stdev*/

proc mixed; by gene; class array treat time dye spotnumber genotype;

model resid=time2 treat genotype treat*time2*genotype;










random array spotnumber(array); Ismeans treat*time2*genotype;

ods output LSMeans=woall_G T_Tmeans diffs=wo_all_pdiff; run;

proc sort data=Wo_all_pdiff; by time time;

data clean Wo all pdiff; set Wo all pdiff;


time2=l then delete; if time2=2 and time2=2 then delete;


if time:

if time:

if time:

if time:

if time:

if time:

if time:

if time:

if time:

if time:

if time:

if time:

if time:

if time:

if time:


=1 and

=3 and

=5 and

=1 and

=1 and

=2 and

=2 and

=3 and

=2 and

=4 and

=6 and

=4 and

=6 and

=5 and

=5 and


4 and

6 and

=1 and

1 and

2 and

3 and

4 and

3 and

5 and

3 and

5 and

4 and


time:

time:

time:

time:

time:

time:

time:

time:

time:

time:

time:

time:

time:

time:


time

time

time

time=

time=

time

time

time=

time=

time=

time=

time=

time=


=4 then delete;

=6 then delete;

=4 then delete;

6 then delete;

5 then delete;

=5 then delete;

=6 then delete;

1 then delete;

1 then delete;

2 then delete;

2 then delete;

3 then delete;

3 then delete;


time2=4 then delete;


if time2=6 and


time2=5 then delete; if treat='C' and treat='I' then delete;


if genotype='R' and _genotype='r' then delete; if treat='I' and _treat='C' then delete;

if genotype='r' and _genotype='R' then delete; run;


then delete; if time2=6 and

then delete; if time2=6 and


then delete; if time=

then delete; if time=

then delete; if time

then delete; if time=

then delete; if time=

then delete; if time=

then delete; if time=

then delete; if time=

then delete; if time=

then delete; if time=

then delete; if time=

then delete; if time=















APPENDIX B
ASREML SCRIPT FOR ASYMTOTIC Z-TEST

Test E fusiform rust only susceptible clones

clone 440 !A

family 61 !A

female 32 !P

male 32 !P

rep 4 !A

inc 110 !A

row 55 !A

col 20 !A

ncol 40 !A

nrow 110 !A

score

/gck/parped.txt !ALPHA

/gck/VINF.prn

score mu !r rep female and(male) family clone
















APPENDIX C
HEALTHY VS. DISEASED GENE LIST


CloneID Genotype Treatment First time Second time Third time
increment increment increment
J4 R C 0 0 0
J4 R I 0 0 0
J4 S C 0 0 0
J4 S I 0 -1 0
07E10 R C -1 0 0
07 E10 R I -1 0 0
07E10 S C -1 0 0
07 E10 S I -1 0 1
37 G12 R C -1 0 1
37 G12 R I -1 0 1
37 G12 S C -1 0 1
37 G12 S I 0 0 1
G1 R C 0 0 0
G1 R I 0 0 0
G1 S C 0 0 0
G1 S I 0 -1 0
G12 R C 0 0 0
G12 R I 0 0 0
G12 S C 0 0 0
G12 S I 0 -1 0
G16 R C 0 0 0
G16 R I 0 0 0
G16 S C 0 0 0
G16 S I 1 -1 -1
G17 R C 0 0 0
G17 R I 0 0 0
G17 S C 0 0 0
G17 S I 0 -1 0
G27 R C 0 0 0
G27 R I 0 0 0
G27 S C 0 0 0
G27 S I 0 -1 0
G30 R C 0 0 0
G30 R I 0 0 0









G30 S C 0 0 0
G30 S I 0 -1 0
G34 R C 0 0 0
G34 R I 0 0 0
G34 S C 0 0 0
G34 S I 0 -1 0
G35 R C 0 0 0
G35 R I 0 0 0
G35 S C 0 0 0
G35 S I 0 -1 0
G39 R C 0 0 0
G39 R I 0 0 0
G39 S C 0 0 0
G39 S I 0 -1 0
G41 R C 0 0 0
G41 R I 0 0 0
G41 S C 0 0 0
G41 S I 1 -1 0
G45 R C 0 0 0
G45 R I 0 0 0
G45 S C 0 0 0
G45 S I 1 -1 -1
G51 R C 0 0 0
G51 R I 0 0 0
G51 S C 0 0 0
G51 S I 0 -1 0
G56 R C 0 0 0
G56 R I 0 0 0
G56 S C 0 0 0
G56 S I 0 -1 0
G8 R C 0 0 0
G8 R I 0 0 0
G8 S C 0 0 0
G8 S I 0 -1 -1
H8 R C 0 -1 0
H8 R I 0 -1 0
H8 S C 0 -1 0
H8 S I 0 0 0
NXCI 002 E02 R C 0 -1 0
NXCI 002 E02 R I 0 -1 0
NXCI 002 E02 S C 0 -1 0
NXCI 002 E02 S I 0 0 0
NXCI 004 G05 R C -1 0 0
NXCI 004 G05 R I -1 0 0









NXCI 004 G05 S C -1 0 0
NXCI 004 G05 S I -1 -1 0
NXCI 018 A08 R C -1 0 0
NXCI 018 A08 R I -1 0 0
NXCI 018 A08 S C -1 0 0
NXCI 018 A08 S I -1 -1 0
NXCI 027 G06 R C -1 1 0
NXCI 027 G06 R I -1 1 0
NXCI 027 G06 S C -1 1 0
NXCI 027 G06 S I -1 0 0
NXCI 034 F04 R C 0 0 0
NXCI 034 F04 R I 0 0 0
NXCI 034 F04 S C 0 0 0
NXCI 034 F04 S I 0 -1 -1
NXCI 042 D04 F R C -1 0 0
NXCI 042 D04 F R I -1 0 0
NXCI 042 D04 F S C -1 0 0
NXCI 042 D04 F S I -1 -1 0
NXCI 057 B05 R C -1 0 0
NXCI 057 B05 R I -1 0 0
NXCI 057 B05 S C -1 0 0
NXCI 057 B05 S I 0 -1 0
NXCI 067 H06 R C -1 0 0
NXCI 067 H06 R I -1 0 0
NXCI 067 H06 S C -1 0 0
NXCI 067 H06 S I -1 -1 0
NXCI 070 D01 R C -1 0 0
NXCI 070 D01 R I -1 0 0
NXCI 070 D01 S C -1 0 0
NXCI 070 D01 S I -1 -1 0
NXCI 075 D09 R C 0 0 0
NXCI 075 D09 R I 0 0 0
NXCI 075 D09 S C 0 0 0
NXCI 075 D09 S I 0 -1 0
NXCI 075 Ell R C -1 -1 1
NXCI 075 Ell R I -1 -1 1
NXCI 075 Ell S C -1 -1 1
NXCI 075 Ell S I -1 -1 0
NXCI 082 E07 F R C 0 -1 1
NXCI 082 E07 F R I 0 -1 1
NXCI 082 E07 F S C 0 -1 1
NXCI 082 E07 F S I 0 -1 0
NXCI 082 G01 R C -1 0 0
NXCI 082 G01 R I -1 0 0









NXCI 082 G01 S C -1 0 0
NXCI 082 G01 S I 0 0 0
NXCI 099 A12 R C -1 0 0
NXCI 099 A12 R I -1 0 0
NXCI 099 A12 S C -1 0 0
NXCI 099 A12 S I 0 0 0
NXCI 111 C10 R C -1 0 0
NXCI 111 C10 R I -1 0 0
NXCI 111 C10 S C -1 0 0
NXCI 111 C10 S I 0 -1 0
NXCI 150 F06 F R C -1 1 -1
NXCI 150 F06 F R I -1 1 -1
NXCI 150 F06 F S C -1 1 -1
NXCI 150 F06 F S I -1 0 0
NXCI 153 A02 F R C -1 -1 0
NXCI 153 A02 F R I -1 -1 0
NXCI 153 A02 F S C -1 -1 0
NXCI 153 A02 F S I 0 -1 0
NXCI 164 A06 F R C -1 0 0
NXCI 164 A06 F R I -1 0 0
NXCI 164 A06 F S C -1 0 0
NXCI 164 A06 F S I 0 -1 0
NXLV100 F02 F R C -1 0 0
NXLV100 F02 F R I -1 0 0
NXLV100 F02 F S C -1 0 0
NXLV100 F02 F S I 0 0 0
NXLV103 E01 F R C -1 1 0
NXLV103 E01 F R I -1 1 0
NXLV103 E01 F S C -1 1 0
NXLV103 E01 F S I -1 0 -1
NXLV105 B02 F R C -1 0 0
NXLV105 B02 F R I -1 0 0
NXLV105 B02 F S C -1 0 0
NXLV105 B02 F S I -1 1 0
NXLV129 C12 F R C -1 0 0
NXLV129 C12 F R I -1 0 0
NXLV129 C12 F S C -1 0 0
NXLV129 C12 F S I -1 -1 0
NXLV 012 A05 F R C 0 -1 0
NXLV 012 A05 F R I 0 -1 0
NXLV 012 A05 F S C 0 -1 0
NXLV 012 A05 F S I 0 0 0
NXLV 022 H08 F R C -1 0 0
NXLV 022 H08 F R I -1 0 0









NXLV 022 H08 F S C -1 0 0
NXLV 022 H08 F S I 0 0 -1
NXLV 023 D12 F R C -1 -1 0
NXLV 023 D12 F R I -1 -1 0
NXLV 023 D12 F S C -1 -1 0
NXLV 023 D12 F S I 0 0 0
NXLV 049 G11 F R C -1 0 -1
NXLV 049 G11 F R I -1 0 -1
NXLV 049 G11 F S C -1 0 -1
NXLV 049 G11 F S I -1 0 0
NXLV 079 G07 F R C -1 0 0
NXLV 079 G07 F R I -1 0 0
NXLV 079 G07 F S C -1 0 0
NXLV 079 G07 F S I -1 1 0
NXNV 129 F06 R C 1 0 0
NXNV 129 F06 R I 1 0 0
NXNV 129 F06 S C 1 0 0
NXNV 129 F06 S I 1 0 1
NXNV015H07 R C -1 1 -1
NXNV015H07 R I -1 1 -1
NXNV015H07 S C -1 1 -1
NXNV015H07 S I -1 1 0
NXNV027F 10 R C 0 1 0
NXNV027F 10 R I 0 1 0
NXNV027F10 S C 0 1 0
NXNV027F 10 S I 0 0 0
NXNV 073 G04 R C -1 1 0
NXNV 073 G04 R I -1 1 0
NXNV 073 G04 S C -1 1 0
NXNV 073 G04 S I 0 0 0
NXNV 096 C09 R C 0 0 0
NXNV 096 C09 R I 0 0 0
NXNV 096 C09 S C 0 0 0
NXNV 096 C09 S I 0 0 -1
NXNV 118 E06 R C -1 0 0
NXNV 118 E06 R I -1 0 0
NXNV 118 E06 S C -1 0 0
NXNV 118 E06 S I 0 -1 0
NXNV 122 C07 F R C -1 0 0
NXNV 122 C07 F R I -1 0 0
NXNV 122 C07 F S C -1 0 0
NXNV 122 C07 F S I 0 0 0
NXNV 125 E12 F R C -1 0 0
NXNV 125 E12 F R I -1 0 0









NXNV 125 E12 F S C -1 0 0
NXNV 125 E12 F S I -1 -1 0
NXNV 132 G1l F R C -1 1 0
NXNV 132 G11 F R I -1 1 0
NXNV 132 G11 F S C -1 1 0
NXNV 132 G1l F S I -1 0 0
NXNV 135 E01 F R C -1 -1 0
NXNV 135 E01 F R I -1 -1 0
NXNV 135 E01 F S C -1 -1 0
NXNV 135 E01 F S I 0 -1 0
NXNV 147 G03 F R C -1 1 0
NXNV 147 G03 F R I -1 1 0
NXNV 147 G03 F S C -1 1 0
NXNV 147 G03 F S I 0 0 0
NXNV 159 G03 R C 0 -1 0
NXNV 159 G03 R I 0 -1 0
NXNV 159 G03 S C 0 -1 0
NXNV 159 G03 S I 0 0 0
NXNV 163 F07 F R C -1 0 0
NXNV 163 F07 F R I -1 0 0
NXNV 163 F07 F S C -1 0 0
NXNV 163 F07 F S I 0 0 0
NXNV 173 B11 F R C -1 1 0
NXNV 173 B11 F R I -1 1 0
NXNV 173 Bll F S C -1 1 0
NXNV 173 B11 F S I 0 0 0
NXPV 062 E04 F R C -1 0 0
NXPV 062 E04 F R I -1 0 0
NXPV 062 E04 F S C -1 0 0
NXPV 062 E04 F S I 0 0 0
NXPV 068 E06 F R C -1 1 -1
NXPV 068 E06 F R I -1 1 -1
NXPV 068 E06 F S C -1 1 -1
NXPV 068 E06 F S I -1 1 0
NXPV 076 C12 F R C -1 1 -1
NXPV 076 C12 F R I -1 1 -1
NXPV 076 C12 F S C -1 1 -1
NXPV 076 C12 F S I -1 1 0
NXRV064 C07 F R C -1 0 -1
NXRV064 C07 F R I -1 0 -1
NXRV064 C07 F S C -1 0 -1
NXRV064 C07 F S I -1 0 0
NXRV079 D01 F R C -1 1 -1
NXRV079 D01 F R I -1 1 -1









NXRV079 DO1 F S C -1 1 -1
NXRV079 D01 F S I -1 1 0
NXRV118 B08 F R C -1 0 0
NXRV118 B08 F R I -1 0 0
NXRV118 B08 F S C -1 0 0
NXRV118 BO8 F S I -1 -1 0
NXSI 008 G11 R C 0 -1 0
NXSI 008 G11 R I 0 -1 0
NXSI 008 G11 S C 0 -1 0
NXSI 008 G11 S I 1 -1 -1
NXSI 059 G09 R C 0 -1 0
NXSI 059 G09 R I 0 -1 0
NXSI 059 G09 S C 0 -1 0
NXSI 059 G09 S I 1 -1 0
NXSI 102 F11 R C -1 -1 0
NXSI 102 F11 R I -1 -1 0
NXSI 102 F11 S C -1 -1 0
NXSI 102 F11 S I 0 -1 0
NXSI 013 C04 R C 0 -1 0
NXSI 013 C04 R I 0 -1 0
NXSI 013 C04 S C 0 -1 0
NXSI 013 C04 S I 1 -1 0
NXSI 027 G10 R C -1 0 0
NXSI 027 G10 R I -1 0 0
NXSI 027 G10 S C -1 0 0
NXSI 027 G10 S I -1 -1 0
NXSI 040 C01 R C -1 0 0
NXSI 040 C01 R I -1 0 0
NXSI 040 C01 S C -1 0 0
NXSI 040 C01 S I -1 -1 0
NXSI 041 B01 R C 0 -1 0
NXSI 041 B01 R I 0 -1 0
NXSI 041 B01 S C 0 -1 0
NXSI 041 B01 S I 0 0 0
NXSI 055 H08 R C -1 0 0
NXSI 055 H08 R I -1 0 0
NXSI 055 H08 S C -1 0 0
NXSI 055 H08 S I -1 1 0
NXSI 060 E02 R C -1 0 0
NXSI 060 E02 R I -1 0 0
NXSI 060 E02 S C -1 0 0
NXSI 060 E02 S I 0 -1 0
NXSI 063 D01 R C -1 1 0
NXSI 063 D01 R I -1 1 0









NXSI 063 D01 S C -1 1 0
NXSI 063 D01 S I -1 0 0
NXSI 064 A03 R C 1 -1 0
NXSI 064 A03 R I 1 -1 0
NXSI 064 A03 S C 1 -1 0
NXSI 064 A03 S I 1 -1 -1
NXSI 067 F10 F R C -1 0 0
NXSI 067 F10 F R I -1 0 0
NXSI 067 F10 F S C -1 0 0
NXSI 067 F10 F S I 0 0 0
NXSI 069 F12 F R C -1 1 -1
NXSI 069 F12 F R I -1 1 -1
NXSI 069 F12 F S C -1 1 -1
NXSI 069 F12 F S I -1 0 0
NXSI 076 E08 R C -1 1 0
NXSI 076 E08 R I -1 1 0
NXSI 076 E08 S C -1 1 0
NXSI 076 E08 S I -1 0 0
NXSI 092 E10 R C -1 -1 1
NXSI 092 E10 R I -1 -1 1
NXSI 092 E10 S C -1 -1 1
NXSI 092 E10 S I -1 -1 0
NXSI 098 C01 R C 1 -1 0
NXSI 098 C01 R I 1 -1 0
NXSI 098 C01 S C 1 -1 0
NXSI 098 C01 S I 1 -1 -1
NXSI 099 H06 F R C -1 0 0
NXSI 099 H06 F R I -1 0 0
NXSI 099 H06 F S C -1 0 0
NXSI 099 H06 F S I -1 1 0
NXSI 101 B01 R C -1 0 0
NXSI 101 B01 R I -1 0 0
NXSI 101 B01 S C -1 0 0
NXSI 101 B01 S I -1 -1 0
NXSI 103 D11 F R C -1 1 0
NXSI 103 D11 F R I -1 1 0
NXSI 103 D11 F S C -1 1 0
NXSI 103 D11 F S I -1 0 0
NXSI 103 E12 F R C -1 1 -1
NXSI 103 E12 F R I -1 1 -1
NXSI 103 E12 F S C -1 1 -1
NXSI 103 E12 F S I -1 1 0
NXSI 104 B11 R C 0 0 1
NXSI 104 B11 R I 0 0 1









NXSI 104 Bll S C 0 0 1
NXSI 104 Bll S I 0 1 1
NXSI 114 A04 R C 0 -1 1
NXSI 114 A04 R I 0 -1 1
NXSI 114 A04 S C 0 -1 1
NXSI 114 A04 S I 0 -1 0
NXSI 115 A12 F R C -1 0 0
NXSI 115 A12 F R I -1 0 0
NXSI 115 A12 F S C -1 0 0
NXSI 115 A12 F S I 0 0 0
NXSI 133 G11 R C 0 -1 0
NXSI 133 G11 R I 0 -1 0
NXSI 133 G11 S C 0 -1 0
NXSI 133 G11 S I 0 0 0
pi134-1 R C 0 1 0
pi134-1 R I 0 1 0
pi134-1 S C 0 1 0
pi134-1 S I 0 0 0
pi150-2 R C 0 -1 1
pi150-2 R I 0 -1 1
pi150-2 S C 0 -1 1
pi150-2 S I 0 -1 0
pi226 R C 0 0 0
pi226 R I 0 0 0
pi226 S C 0 0 0
pi226 S I 0 -1 0
pi243 R C 0 0 0
pi243 R I 0 0 0
pi243 S C 0 0 0
pi243 S I 1 -1 -1
pi266 R C 0 -1 1
pi266 R I 0 -1 1
pi266 S C 0 -1 1
pi266 S I 0 0 1
pi295 R C 0 0 0
pi295 R I 0 0 0
pi295 S C 0 0 0
pi295 S I 1 0 -1
pi310 R C 1 0 0
pi310 R I 1 0 0
pi310 S C 1 0 0
pi310 S I 1 -1 0
pi64-9 R C 0 0 0
pi64-9 R I 0 0 0









pi64-9 S C 0 0 0
pi64-9 S I 0 -1 -1