Discovery of Candidate Effectors Involved in Cronartium Quercuum F. Sp. Fusiforme Infection of Pine and Oak

MISSING IMAGE

Material Information

Title:
Discovery of Candidate Effectors Involved in Cronartium Quercuum F. Sp. Fusiforme Infection of Pine and Oak
Physical Description:
1 online resource (59 p.)
Language:
english
Creator:
Smith, Katherine E
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Master's ( M.S.)
Degree Grantor:
University of Florida
Degree Disciplines:
Plant Molecular and Cellular Biology
Committee Chair:
Davis, John M
Committee Members:
Rollins, Jeffrey A
Smith, Jason A.

Subjects

Subjects / Keywords:
cronartium -- effector -- fungi -- gene-expression -- gene-for-gene -- genome -- pathogen
Plant Molecular and Cellular Biology -- Dissertations, Academic -- UF
Genre:
Plant Molecular and Cellular Biology thesis, M.S.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
Cronartium quercuum (Berk.) Miyabe ex Shirai f. sp. fusiforme (Burds. & Snow) – Cqf – is a biotrophic rust fungus that infects both pine and oak trees. It is the causative agent of fusiform rust which is the major disease of southern pine forests. For decades, the large economic impact of fusiform rust has motivated a large volume of research on managing the disease. Since the southern pine host is economically important to the sawtimber and pulpwood industries, the primary research focus has been on the pine host. To date, the major research advances have been the identification of pine families with genetic resistance and the discovery of specific interactions between the pine host and the pathogen. Pathogen effectors are a diverse set of secreted proteins that enhance virulence, often by mechanistically thwarting host defenses. Fungal effectors that have been characterized in other pathosystems are typically small secreted proteins, localized to either the host cytoplasm or apoplast. Identification of the genes encoding secreted proteins provides insight into the pathogenicity of this fungus, as well as a source of candidate fungal effectors. Avirulence genes encode a subset of effectors that are recognized by the host, which in turn allows the host to block disease development. Candidate effectors identified through the annotation of genome sequence can be a valuable complement to marker based avirulence gene identification strategies. A Cqf sequencing project has been undertaken at the United States Department of Energy (DOE), Joint Genome Institute (JGI), which provided a project midpoint draft, as well as a second and final,genomic sequence assembly enabling the annotation of a complete set of genes. This project contributed to the annotation and used it as a tool to identify secreted proteins and measure fungal transcriptome level gene expression in both hosts. Major contributions of this work include an in silico predicted secretome, a list of host-specific candidate effectors and a physical map of the AVIRULENT TO FUSIFORM RUST 1 (AVR1) locus that contains avirulence gene candidates.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Katherine E Smith.
Thesis:
Thesis (M.S.)--University of Florida, 2012.
Local:
Adviser: Davis, John M.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2013-02-28

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2012
System ID:
UFE0044468:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 DISCOVERY OF CANDIDATE EFFECTORS INVOLVED IN Cronartium quercuum f. sp fusiforme INFECTION OF PINE AND OAK By KATHERINE E. SMITH A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLM ENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2012

PAGE 2

2 2012 Katherine E. Smith

PAGE 3

3 For the late bloomers

PAGE 4

4 ACKNOWLEDGMENTS I would like to thank the Plant Molecular and Cellular Biology Program for having the flexibility to allow me to complete this degree on a part time basis. I also would like to thank my committee members John Davis, Jeff Rollins and Jason Smith for their time and expert advice. I especially want to thank my colleagues Chris Dervinis and Ali son Morse for their help, advice and encouragement during my tenure in the Forest Genomics Laboratory.

PAGE 5

5 TABLE OF CONTENTS p age ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABL ES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 ABSTRACT ................................ ................................ ................................ ..................... 9 CHAPTER 1 INTRODUCTION AND LIT ERATURE REVIEW ................................ ..................... 11 Biotrophy and Fungal Effectors in Plants ................................ ................................ 11 Cronartium quercuum f. sp. fusiforme ................................ ................................ ..... 13 Heteroecious Lifecycle ................................ ................................ ..................... 13 Fusiform Rust Disease ................................ ................................ ..................... 15 2 RNA SEQUENCING SUPPORTS PREDICTION OF CQF GENE MODELS AND A SECRETOME ................................ ................................ ................................ ...... 21 Background ................................ ................................ ................................ ............. 21 Materials and Methods ................................ ................................ ............................ 23 Fungal and Plant Material ................................ ................................ ................. 23 RNA Extraction ................................ ................................ ................................ 23 RNA Sequencing ................................ ................................ .............................. 24 Gene Model Prediction ................................ ................................ ..................... 24 Functional Annotation ................................ ................................ ....................... 25 Bioinformatic Secretome ................................ ................................ .................. 25 Results ................................ ................................ ................................ .................... 26 Cqf Gene Models ................................ ................................ ............................. 26 Cqf Secretome ................................ ................................ ................................ 27 Discussion ................................ ................................ ................................ .............. 28 3 MEASURING CQF GENE EXPRESSION IN PINE AND OAK INFECTION ........... 36 Background ................................ ................................ ................................ ............. 36 Materials and Methods ................................ ................................ ............................ 37 Fungal and Plant Material ................................ ................................ ................. 37 RNA Extraction ................................ ................................ ................................ 38 Microarray Experimental Design ................................ ................................ ...... 38 Statistical Analysis ................................ ................................ ............................ 39 Results ................................ ................................ ................................ .................... 40 Discussion ................................ ................................ ................................ .............. 41

PAGE 6

6 4 MAPPING AVR1 CANDIDATE GENES ................................ ................................ .. 46 5 CONCLUSIONS ................................ ................................ ................................ ..... 52 APPENDIX: FUNCTIONALLY ENRICHED GO ANNOTATIONS IN THE CQF SECRETOME ................................ ................................ ................................ ......... 53 LIST OF REFERENCES ................................ ................................ ............................... 54 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 59

PAGE 7

7 LIST OF TABLES Table page 2 1 List of GO terms enriched in the Cqf secretome (FDR .001) .............................. 35 4 1 AVR1 candidate genes by location on scaffold 20 ................................ .............. 49 4 2 AVR1 marker sequence identity to final assembly scaffolds .............................. 51 A 1 Complete list of GO terms enriched in the Cqf secretome (FDR .05) ................. 53

PAGE 8

8 LIST OF FIGURES Figure page 1 1 Schematic o f the haustorium infection structure taken from Panstruga and Dodds, 2009. ................................ ................................ ................................ ...... 18 1 2 The lifecycle of Cronartium quercuum modified from Phelps and Czabator 1978 ................................ ................................ ................................ .................. 19 1 3 et al. 1997. ................................ ................................ ................................ .......... 20 1 4 Starkey et al. 1997. ................................ ................................ ............................. 20 2 1 BLASTP results of ab intio only gene models compared to the NCBI protein database (expect value cutoff E 6 ). ................................ ................................ ..... 31 2 2 The n umber of top BLASTP hits by species. All species shown have a sequenced genome. The others category includes all species with less than 10 top hits (expect value cutoff E 6 ) ................................ ................................ .... 32 2 3 BLASTP comp arison of Cqf gene models to the NCBI protein database (expect value cutoff E 6 ). ................................ ................................ ..................... 33 2 4 Comparison of the number of Cqf secreted proteins predicted by two different in silico methods from Jol y et al., 2010 and Min, 2010. ................................ ...... 33 2 5 The Cqf secretome is enriched for proteins with no similarity to known proteins (BLASTP with the NCBI protein database, expect value cutoff E 6 ). ..... 34 3 1 The majority of Cqf genes are expressed in both hosts. ................................ ..... 44 3 2 Distributions of log2 transformed pine (red) and oak (blue) backgrou nd subtracted signal intensities. Each line is a single replicate sample. .................. 44 3 3 Genes encoding secreted proteins are enriched among genes with greater than 4 fold increased expression in one host over the other. ............................. 45 3 4 Genes encoding lineage specific SSPs are enriched among genes with greater than 4 fold increased expression in one host over the other (BLASTP, expect value cutoff E 6 ). ................................ ................................ ..... 45 4 1 Genetic (A) and physical (B) maps of the Avr1 locus indicate a physical/genetic distance ratio of 10.4kb/cM (460kb/44.2cM). ............................ 48

PAGE 9

9 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science DISCOVERY OF CANDIDATE EFFECTORS INVOLVED IN Cronartium quercuum f. sp fusiforme INFECTION OF PI NE AND OAK By Katherine E Smith August 201 2 Chair: John Davis Major: Plant Molecular and Cellular Biology Cronartium quercuum (Berk.) Miyabe ex Shirai f. sp f usiforme ( Burds. & Snow ) Cqf is a biotrophic rust fungus that infects both pine and oak trees It is the causative agent of fusiform rust which is the major disease of southern pine forests. For decades, the large economic impact of fusiform rust has motivated a large volume of research on managing the disease Since the southern pine host i s economically important to the sawtimber and pulpwood industries the primary research focus has been on the pine host. To date, the major research advances have been the identification of pine families with genetic resistance and the discovery of specifi c interaction s between the pine host and the pathogen Pathogen e ffectors are a diverse set of secreted proteins that enhance virulence, often by mechanistically thwarting host defenses. F ungal effect or s that have been characterized in other pathosystem s a re typically small secreted proteins l ocalized to either the host cytoplasm or apoplast I dentification of the genes encoding secreted proteins provides insight into the pathogenicity of this fungus, as well as a source of candidate fungal effectors. Avir ulence genes encode a subset of effectors that are

PAGE 10

10 recognized by the host, which in turn allow s the host to block disease development Candidate effectors identified through the annotation of genome sequence can be a valuable compl e ment to marker based avi rulence gene identification strategies. A Cqf sequenc ing project has been undertaken at the United States Department of Energy (DOE) Joint Genome Institute (JGI) which provided a project midpoint draft as well as a second and final, genomic sequence asse mbly enabling the annotation of a complete set of genes. This project contributed to the annotation and used it as a tool to identify secreted proteins and measure fungal transcriptome level gene expression in both hosts Major contributions of this work i nclude an in silico predicted secretome a list of host specific candidate effectors and a physical map of the AVIRULENT TO FUSIFORM RUST 1 ( AVR 1 ) locus that contains avirulence gene candidates

PAGE 11

11 CHAPTER 1 INTRODUCTION AND LITERATURE RE VIEW Biotrophy and Fungal Effectors in Plants Biotrophic fungi can persist inside their hosts for long periods of time causing large scale changes to host cell morphology as is the case for Cronartium quercuum (Berk.) Miyabe ex Shirai f. sp f usiforme ( Burds. & Snow ), refer red to as Cqf Unlike necrotrophs that kill their host s and feed on dying or dead tissue, biotrophs have evolved mechanisms for evad ing host defenses and feeding off living cells. In order to obtain the nutrients necessary for proliferation within the host rust fungi such as Cqf form host cell penetration structures called appressoria (Gray et al., 1982) that apply pressure to the host cell wall allowing sp ecialized infection structures called haustoria to grow into the host cell without actually making contact with host cytoplasm (Figure 1 1) Discovery of expressed gene s specific to haustoria have shed light on the acquisitio n of metabolites during the biotrophic interaction. Some of the planta induced genes (PIGS) of the bean rust fungus Uromyces fabae include amino acid and hexose transporters, as well as ATPases (Hahn and Mendgen, 1997) which are suspected to provide the energy necessary for haustoria specific amino acid permeases to transport metabolites into haustoria cells (Hahn and Mendgen, 2001) Other PIGS are involved in general defense, such as a metallothionein that may provide protection from oxidative stress and a cytochrome P 450 monooxygenase that may detoxify damaging plant compounds Fungi avoid host defenses and establish disease thr ough the action of an array of secreted proteins referred to as effectors These proteins interact with the host to alter host cell morphology and function (Gan et al., 2010) The Cqf pine host interaction is an

PAGE 12

12 obvious example since not only must the fun gus suppress host defenses to grow within pine stems, but also host cells must change morphology to form galls. Some fungal effectors function outside of host cells in the apoplastic space between plant cells or in the extrahaustorial matrix (Figure 1 1) For example, some e ffector s mechanistically block host defenses by inhibit ing host chitinases and proteases For example, Cladosporium fulvum a biotrophic fungus that causes tomato leaf mold, secretes a chitinase i nhibitor (Avr4) and a plant cysteine protease inhibitor (Avr2) (van Esse, 2008) during infection Avr4 binds chitin in the fungal cell wall and apparently protects the wall from degradation by tomato chitinases (van den Burg et al., 2006) AvrP123, from the flax rust fungus Melampsora lini is related to K azal serine protease inhibitors (Catanzariti, 2006) In a ddition, there are many more fungal effectors that function inside host cells sometimes entering the host nucle us and likely affecting gene expression. For example, the Uromyces fabae haustoria specific protein Rust Transferred Protein 1 was shown by immu nofluorescence to be localized to the extrahaustorial matrix (Figure 1 1) and the interior of plant cells, including the nucleus (Kemen et al., 2005) A subset of effectors conform to predictions made by the gene for gene hypothesis meaning they are avi rulence genes with alleles that specifically interact with host resistance gene alleles resulting in a resistance phenotype. This phenomenon was first reported by Flor in Melampsora lini (Flor, 1955) and subsequent genetic studies have revealed at least 30 AVR genes with corresponding resistance genes in flax (Ellis et al., 2007) Coevol ution of rust fungi with their hosts has led to the proliferation of patho type specific avirulence genes and interacting host resistance gene s that has been described as an arms race C lone d avirulence genes of Melampsora lini are

PAGE 13

13 contained in four loci, each with AVR genes similar to each other but with no known homologs in other species. Most known avirulence effectors including those from Melampsora lini enter host cells and interact directly with resistance gene products Effector proteins encoded b y genes on the AvrL567 locus have variants that differ in amino acids on the protein surface and have been shown in yeast two hybrid studies to directly interact with resistance proteins (Stergiopoulos and de Wit, 2009) Exactly how secreted effectors get into host cells remains unclear In rust fungi, haustoria penetrate into the host cell such that the fungal cell wall does not make contact with host cytoplasm, but is separated from it by an extrahaustorial matrix surrounded by an extrahaustorial membrane (Figure 1 1) Secreted proteins containing an N te rminal secretion signal are delivered through the haustorial cell wall into the extrahaustorial matrix via the eukaryotic secretory system. They must then enter the plant cell by either a plant derived mechanism or a fungal derived mechanism. There is evid ence in fungal like oomyc ota for a plant derived system since t he conserved effector domain, RXLR is both necessary and sufficient for oomycte effectors to enter host cells and this domain is present in plant proteins i nvolved in membrane trafficking (Panstruga and Dodds, 2009) Howeve r, biotrophic fungal effectors have no such conserved motif and most have little similarity to known proteins (Gan et al., 2010) Cronartium quercuum f. sp. f usiforme Heteroecious Lifecycle Cronartium quercuum defined as the rust fungus that cause s galls on pine and colonizes both pine and oak as hosts, has been proposed to be divided into special forms ( formae speciales, f.sp .) so that f.sp. fusiforme (loblolly longleaf and slash pine) f.sp. banksianae (jack pine), f.sp. echinatae (shortleaf pine) and f.sp. virginia na e

PAGE 14

14 (Virginia pine) are each distinguished by the pine host (Burdsall and Snow, 1977) Phylogenetic analys i s of pine stem rusts from the genera Cronartium and Peridermium based on nuc leotide similarity of the internal transcribed region of nuclear ribosomal genes placed Cqf nearest to the other Cronartium quercuum special forms and Endocronartium harknessii (Vogler and Bruns, 1998) E harknessii is a species that cause s galls on pine, however, it is autoecious, meaning all stages of its lifecycle occur on pine. Figure 1 2 illustrates the complex Cqf lifecycle that requires two hosts, oak and pine, includes five different spore stages and takes up to two years to complete (Phelps, 1978) Timing of the various stages is influenced by weather conditions, such as temperature, wind and rainfall. In the late spring, young pine trees are infected primarily by penetration of germinated basidiospores (Schmidt, 1998) into stems resulting in stem and branch galls within 6 9 months In the fall, galls produce bright orange pycnial droplets that contain haploid spores. Presumably i nsects and perhaps other organism s mov e pycnia within a gall or from one gall to another causing spermatization (Kubisiak et al., 2005) leading to the dikaryo tic ( N + N ) spore type, aeciospores, that are produced by galls in the spri ng. Aeciospores are bright yellow and coat the outside of the gall. These spores are dispersed by the wind and infect succulent leaves of species in the group red oak, primarily through stomata. Oaks are minimally affected by the fungus (Schmidt, 1998) Infected o ak leaves have lo ng cylindrical teli o spores that germinate and produc e bas i diospores (Phelps, 1978) Basidiospores carried by the wind then infect young pines and the cycle begins again. M eiosis occur s in the telial columns (2N) while on the oak host giving rise to haploid basidio spores (N) that are

PAGE 15

15 each the result of different recombination even ts of the original N + N aeciospore genotype that infected the oak. Oak leaf infection by aeciospores may also give rise to pustules that contain ured i ni ospores, the dikaryo t ic repeating stage on oak. There is no repeating stage on pine, since only basidio spores can infect pine. High genetic diversity of the fungal population is thought to be maintained at least in part by yearly genetic recombination on oaks. Fusiform Rust Disease Fusiform rust disease caused by the fungus Cronartium quercuum f. sp. f usif orme occurs solely on the North American continent and primarily in the southern United States, from northern Florida, west to the eastern edge of Texas and Oklahoma and north as far as Maryland Cqf affects primarily the southern pine species, Pinus elli o t tii (slash pine) and Pinus taeda (loblolly pine), although Pinus palustris ( longleaf pine ) and Pinus serotina ( pond pine ) show moderate susceptibility (Dwinell, 19 76) It has been estimated that over 35 million dollars are lost to fusiform rust disease annually (Anderson et al., 1986) Figure 1 3 and F igure 1 4 show the widespread incidence of fusiform rust measured by the USDA Forest Service, across the entire natural range of s lash and loblolly pine, respectively (Starkey, 1997) The warm and moist climate of the southeastern United States is ideal for Cqf spore production and germination. The major symptom is gall format ion on the stems and branches of young, actively growing pines. These galls damage and deform trees often leading to death, but galls can also persist for many years. Galls that persist put trees in risk of breakage and wind damage (Phelps, 1978) The management of fusiform rust disease has included two general strategies amendi ng silviculture practices and planting rust resistant pine families. Silviculture

PAGE 16

16 s trategies have included removing infected seedlings from nurseries removing infected stems from young pines, as well as efforts to reduce the number of oaks at a field site Perhaps the more effective and sustainable control will be p lanting resistant pines. At first, resistant trees came from field trials or were survivors from heavily rust infected areas. In 1973, the USDA Forest Service established the Rust Resistance Scr eening Center (Asheville, North Carolina) in response to the rust epidemic. There an artificial and accelerated inoculation technique is used on six week old seedlings that allows quick testing of a large number of genotypes. Based on a performance index, greenhouse inoculation correlates well with field trials. The s creening c enter has also aided research on the genetics of pine resistance (Schmidt, 2003) Studies have shown that the Cqf pine host interaction operates in a gene for gene manner (Flor, 1955) and the first resistance gene, Fr1, was mapped in loblolly pine (Wilcox et al., 1996) Recently, the corresponding avirulence gene, A VR 1 was mapped in the fungus (Kubisiak et al., 2011) This work produced 421 mapped ma rkers throughout the Cqf genome 14 of which are linked to AVR 1 and define a genetic interval for the gene. In addition to Fr1, 8 more resistance genes have been mapped in pine (Amerson et al. in preparation ) Take n together the research surrounding major gene resistance in pine has led to the proposal of a new approach to fusiform rust disease management. Nelson et al. (2010) proposed pine resistance screening combined with monitor ing of AVR gene allele frequencies in the field by geographic region to gui de selection of resistant pine families. The authors present ed two methods by which this could be achieved. One is through the identification and use of single genotype isolates of the fungus that are informative when tested on pine s that are

PAGE 17

17 heterozygous and homozygous recessive for the corresponding resistance genes These material s could be used to monitor genes present in Cqf samples or to screen pine families for breeding programs. The second method would be to expand the number of marker s on the genet ic maps of Cqf and loblolly pine such that resistance genes and avirulence genes, respectively, could be definitively identified in test subjects (Nelson et al., 2010) This is much easier to achieve in the fungu s compared to the pine host The huge genome size of loblolly pine, 21,658Mb, compared to ~90Mb in Cqf makes it more difficult to obtain the marker coverage necessary to definitively identify resistance genes. In addition, the Cqf reference genome sequen ce will make identification of avirulence genes and subsequent development of fully diagnostic markers more straight forward.

PAGE 18

18 Figure 1 1. Schematic of the ha us torium infection structure taken from Panstruga and Dodds, 2009.

PAGE 19

19 Figure 1 2 The lifecyc le of Cronartium quercuum modified from Phelps and Czabator 1978

PAGE 20

20 Figure 1 3 from Starkey et al 1997. Figure 1 4 rate from Starkey et al. 1997.

PAGE 21

21 CHAPTER 2 RNA SEQUENCING SUPPORTS PREDICTION OF CQF GENE MODELS AND A SECRETOME Background Researchers from the USDA Forest Service, Southern Institute of Forest Genetics and the University of Florida School of Forest Resources and Conservation, were awarded a sequencing grant by the Joint Genome Institute (JGI) Community Sequencing Program, to sequence the Cronartium quercuum f. sp fusiforme genome ( http://w ww.jgi.doe.gov/sequencing/why/cronartium.html ) P ine trees are of interest to JGI because they are fast grow ing and produce large amount s of biomass ; therefore, hold promise as a source of biofuel Fusiform rust greatly diminishes yields of southern pine forests and a complete genome sequence of Cqf could provide major insights leading to improved control of the disease. Elucidating genes involved in the pathogenicity of Cqf is a first step in the development of new disease control strategies (Nelson et al., 2010) In July 2011, the project midpoint draft assembly of a genomic reference sequence was completed which consists of 2084 scaffolds totaling 89.1MB, very close to the ~90Mb genome size predicted using fl ow cytometry (Anderson et al., 2010) Gene models i ncluding translation start and stop sites, introns, exons and untranslated regions were annotated on tho se scaffolds with the support of RNA sequencing data. The JGI Cqf sequencing project was supplied with 6 RNA samples from various fungal stages for sequencing The goal was to provide gene expression data to support ab initio gene model prediction with the most complete coverage possible. M ore stages of the fungus sampled would enable better coverage. Expressed transcript sequences were compared to gene models to delineate introns, exons and translation start and

PAGE 22

22 stop sites It is likely that an essentially complete set of genes was annotated using the project midpoint assembly This is because the total scaffold length was similar to the expected genome size and stretches present in the assembly were considered to be repetitive non genic sequence. Given an ever increasing number of sequenced genomes, it is now a common strategy to look for fungal effectors as well as other proteins important to the disease process, in the secretome of a pathogen. Sec retomes are predicted in sil ico by searching for N terminal signal peptides in protein sequences derived from gene models. Seventy nine of t he 426 secreted proteins of the biotrophic smut fungus, Ustilago maydis were located in clusters throughout the genome (Kamper et al., 2006) I n this paper, i nformation gained from c luster expression pattern s during infection and virulence of cluster deletion mutants supported the utility of predicting secreted proteins The rice blast fungus, Magnaporthe grisea was predicte d to have 739 secreted proteins, 8 of which may encode cutinases (Dean et al., 2005) One of those 8 putative cutinases CUT2, was later shown to be important in host penetration and required fo r full virulence ( Skamnioti, 2 007) A greater number of sec reted proteins were predicted in the larger rust genomes of Melampsora larici populina and Puccinia graminis 1 549 and 1 852 respectively. H ierarchical clustering of these proteins identified 8 protein families as likely effectors and good candidates for fu rther functional testing (Saunders et al., 2012) Here, a set of Cqf secreted proteins is identified from the project midpoint assembly gene models and described using BLAST (Basic Local Alignment Search Tool) comparison and functional enrichment analys is

PAGE 23

23 Materials and Methods Fungal and Plant Material S amples for RNA sequencing were collected from both the pine and the oak host s of Cq f G alls were collected from 5 year old s lash pine trees in October as pycniospores were forming and again in April as aeciospores were forming. Galls were freeze dried for 1 week. To obtain tissue for RNA extraction, yellow colored aeciospore (spring) or orange colored pycniospore (fall) hymenial layers were chipped away from the outside of freeze dried galls using a scal pel. Aeciospores were collected by knocking them off the surface of spring collected pine galls They were then stored at 20 C. Oak associated tissues included infected oak leaves with attached telial columns, telial columns removed from oak leaves and ba sidiospores collected onto pH 2 .0 water wetted filters ( to prevent germination ) Oak leaves and telial columns were stored at 20 C and basidiospores were stored in pH 2 .0 water, at 4 C for up to 4 days. Pine associated tissues were collected in the field, at the University of Florida in Gainesville, Florida Oak associated tissues were collected from greenhouse inoculated open poll i nated wild northern red oak ( Quercus rubra ) seedlings at the USDA Forest Service, Resistance Screening Center in Asheville, N orth Carolina RNA E xtraction RNA was extracted from the following tissue s: aeciospores, basidiospores, telial columns, infected oak leaves and the spring and fall hymen i al layers of pine galls. A previously described c etyltrimethylammonium b romide ( CTAB ) buffer method (Chang et al., 1993) was used, with different gr i nd ing procedures for each tissue. Infected oak leaves were frozen and ground in liquid nitrogen. The following tissues were ground in CTAB buffer pre war med to 65 C using a Geno/Grinder 2000 homogenizer (BT&C

PAGE 24

24 Incorporated) : 1) Aeciospores were ground in 4ml round bottom vials containing ~20mg of spores and a 1.0cm stainless steel ball; 2) Basidiospores were ground in 1.5ml E ppendorf tubes containing ~20mg of spores, 150mg zircon beads and 12.5mg diatomaceous earth; 3) Telial columns were ground in 1.5ml E ppendorf tubes containing ~50mg of telial columns, 150mg zircon beads, 12.5mg diatomaceous earth and a single 2mm tungsten bead; 4) Chips of both pycniospo re and aeciospore hymenial layers were preprocessed in a coffee grinder and then ground in 4ml round bottom vials containing CTAB buffer and a 1.0cm steel ball. In all cases e xtracted RNA was treated for 30 minutes at 37 C with RQ1 RNase f ree DNase (Promeg a, M6101) and then purified using a Qiagen RNeasy Mini Spin Column. RNA Sequencing RNA for library production and sequencing was analyzed on an Agilent 2100 Bioanalyzer and only RNA with a minimum RNA integrity score (RIN) of 6.3 was provided to JGI. Libr aries were constructed using RNA from the following five sam ples: 1) Pycnial hymenial layer; 2) Aecial hymenial layer; 3) Aeciospores; 4) Basidiospores; and 5) Infected oak leaves with telial columns. All libraries were constructed to enable Illumina 75 b ase pair, paired end high throughput sequencing which generated a total of ~ 406 million reads. Gene Model Prediction Gene models were predicted as part of the Cqf genome sequencing project by the Yandell laboratory (www.yandell lab.org/) at the University of Utah. The genome annotation pipeline Maker 2 was used (Holt and Yandell, 2011) Maker2 was designed to predict gene models from genome sequence data without the benefit of preexisting known gene models to use as train ing data for gene finding programs. The pipeline

PAGE 25

25 u ses SNAP, Augustus and GeneMark to produce gene models, then incorporates RNA sequence data and BLAST data to refine gene models Functional Annotation The publicly available genomic research tool, BLAST 2G O was used to functionally annotate genes in the C qf genome (Conesa et al., 2005) BLAST 2G O performed a BLAST similarity search on each protein sequence with a significance level cutoff of E 6 BLAST 2G O compiles can didate GO (Gene Ontology, http://www.geneontology.org ) terms associated with gene identifiers (GI) of the hit s, along with the accompanying evidence code s (EC) Annotation assignments are made by applying an an notation rule that takes into account sequence similarity and node relatedness among the GO term candidates, as well as experimental evidence Experimental evidence is weighted more highly than electronic evidence in the assignment of GO terms. Enrich ment for specific GO terms in the Cqf secretome was tested using a module within BLAST 2GO that integrates the protein structure comparison tool GOSSIP (Global Structural Superposition of Proteins), to compute ate (FDR) of 0.05. Bioinformatic Secretome Protein s were designated as secreted using previously published method s (Joly et al., 2010) Signal peptides were identified using both TargetP 1.1 (www.cbs.dtu.dk/services/TargetP/) and SignalP 3.0 (www .cbs.dtu.dk/services/SignalP 3.0) online software P roteins with transmembrane domains were remov ed using the online software TMHMM 2.0 (www.cbs.dtu.dk/services/TMHMM/) To account for the fact that T MHMM 2.0 may have difficulty distinguishing transmembrane domains from signal peptides (Krogh et al., 2001) protein s with a signal peptide predicted by TargetP

PAGE 26

26 and SignalP and a single transmembrane domain occurring with in 40 amino acids of the N terminus were desig nated as secreted (Joly et al., 2010) The Joly et al. method was validated using a second metho d developed by Min (2010) This method combined SignalP 4.0 (www.cbs.dtu.dk/services/SignalP) Wo LF P SORT (http:/wolfpsort.org) and Phobius (http:/phobius.sbc.su.se) online software, to locate signal peptides. TMHMM 2.0 was used to remove transmembrane proteins and Scan Prosite (http:/prosite.expasy.org/scanprosite) was used t o remove proteins targeted to the endoplasmic reticulum (Min, 2010) Results Cqf Gene Models The Cqf sequencing project generated 8782 total in silico gene model s. Nearly all of these ( 81 61 ) were present in th e RNA sequenc ing libraries and therefore had evidence of expression as well as support for the determination of a coding region The 6 21 gene models that were not present in the RNA sequence libraries wer e designate d as predicted by ab in i tio evidence only One hundred and fifty nine of the g ene models in th is category encoded proteins with homology to proteins in the NCBI (National Center for Biotechnology Information) database (Figure 2 1 ) Approximately 75% had no hits in t he database and five with no hits could be assigned GO terms Proteins encoded in all predicted gene models were compared to the NCBI protein database by BLAST P T op hits broken down by species reflect that the two fungi that are most closely related to Cq f and have sequenced genomes are Melampsora larici populina and Puccinia graminis (Figure 2 2 ) Both of these fungi are rusts and Melampsora larici populina alternates infect ion between Populus an angiosperm tree species lik e Querc us and Larix a gymnosperm tree species like Pinus Roughly half of

PAGE 27

27 all Cqf proteins were similar to proteins of known function, w hile the other half either had no similarity at all (~25%) or similarity to proteins present in other organisms but wi th no known function (~25%) (Figure 2 3 ) B LAST p rotein sequence comparison and GO term assignment using the BLAST 2G O platform resulted in the functional annotation of 3841 genes or 47% of genes in the Cqf genome Cqf Secreto me The 804 member Cqf secretome was determined by a bioinformatic prediction method used by Joly et al. to identify secreted proteins in several closely related Melampsora leaf rust species (Joly et al., 2010) A second validation method was also applied. This m ethod was developed in an effort to predict secretion while minimizing false positives and false nega tives by using a manually curated set of known fungal proteins (Min, 2010) A Venn diagram comparison shows the methods to have similar results and a large degree of overlap with the validation method predict ing more se creted proteins, 1053 in total (Figure 2 4 ) The validation method is less restrictive and therefore likely to contain more false positives. The Cqf secreted proteins were characterized by sequence comparison and by function enrichment testing based on GO terms. BLAST P results for the secretome compared to the entire proteome show an increase in the proportion of proteins that share similarity to unknown proteins or share no similarity to any proteins Fifty percen t of small secreted peptides (SSPs, less than 300 amino acids) have no hits in the NCBI protein database (Figure 2 5 ) Secreted proteins with known functions were heavily enriched for GO terms associated with carbohydrate met abolism lipases and proteases (Table 2 1)

PAGE 28

28 Discussion The RNA sequencing data provided good support for defining gene model s Overall 93% of gene models were expressed in the six fungal stages sampled. Melampsora larici popul ina and Puccinia graminis two closely related and sequenced rust fungi account ed for about 71% of the top hits by species (Figure 2 3) The total number of Cqf gene models, 8782, is curiously less than the number reported fo r Melampsora larici populina and Puccinia graminis 16,399 and 17,773 respectively (Duplessis et al., 2011b) However, the proportion of Cqf proteins predicted to be secreted, 9.2%, is similar to Melampsora larici p opulina and Puccinia graminis which have 9.2% and 10.1% predicted secreted proteins respectively (Saunders et al., 2012) I t remains to be seen what accounts for the large difference in number of genes in these similarly sized rust genomes. The current gene count for Cqf is well support ed, since Maker2 uses three different ab initio gene find ing programs and the predicted genes have excellent evidence for expression. G ene models without RNA sequence support are possibly genuine Cqf genes but expressed in stages of the fungus that w ere not sampled f or example time points earlier in infection Once validated, t hese genes could be involved in early host recognition or act as host range determinates Among the 75 proteins in this category that are predicte d to be secreted and therefore have greater potential as effectors, 69 have no homology to known proteins or are homologous to proteins of unknown function (Figure 2 2) Only six proteins have known function and all six could be involved in infection by one or more of the following modes; cell wall degradation, nutrient acquisition or signaling. These proteins include are 3 glycoside hydrolases from

PAGE 29

29 3 different hydrolase families 1 subtilisin protease one triacylglycerol lip ase and one serine/threonine kinase. The predicted secretome of Cqf provide s a list of potential virulence determina nt s and potential effectors that can be used to gain insight into the pathogenicity of the fungus. One Cqf secreted protein shares 29% ident ity with the TRI14 protein of the hemibiotroph Colletotrichum higginsianum TRI14 is a virulence determinant in Fusarium graminearum involved in the synthesis of the mycotoxin trichothecene (Dyer et al., 2005) Four were similar to rust transferred protein from Uromyces viciae fabae (Kemen et al., 2005) with amino acid identities between 41 % and 43 % Two proteins were similar to two of the haustorially expressed secreted proteins of flax rus t ( Melampsora lini ) (Catanzariti, 2006) hesp379 with 60% identity and hesp 735 with 58% identity. In addition, the Cqf secretome is enrich ed for candidate avirulence effectors These proteins coevolve with hosts and known fungal avirulence effectors commonly have no sequence similarity to proteins in other fungi and are short secreted, cysteine rich peptides (Stergiopoulos and de Wit, 2009) The Cqf secretome includes 514 SSPs (small secreted peptides less than 300 amino acids) 220 of which have no similarity to other proteins (Figure 2 6). Taken as a group the SSPs are more cysteine rich, with an average percent of cysteine res idues for the proteome as whole of 1.43 compared to 1.74 for those proteins in the secretome an d 2.24 for the SSPs. The 118 SSPs with greater than 3% cysteine residues, 106 of which have no assignable function are good candidates for avirulence effectors. Five out of the six s ecreted proteins that contain a cysteine rich CFEM domain (CFEM common i n fungal extracellular and membrane ) are under 325 amino acids. CFEM domains are suspected

PAGE 30

30 to be involved in pathogenesis as cell surface receptors, signal transducers or adhesion molecules (Kulkarni et al., 2003) Using an e value cut of 1E 50 16 % of C qf genes encoding secreted proteins have par alogs an indication of diversifying selection The most paralogous genes are 1 set of five paralogs and 2 sets of four, all of which encode unknown proteins. This secretome data combined with avirulence gene mapping could be useful in the identification o f the 9 (at least) expected avirulence genes in Cqf pine interaction. Proteins potentially important during infection exist among the 210 secreted proteins that could be functionally annotated by the BLAST 2GO platform Using a false discovery rate (FDR) of .001 t hese proteins were heavily enriched in process and activity GO terms involved in cell wall degradation and carbohydrate metabolism. These enriched GO terms support the idea that Cqf can restructure its own cell wall as it grows within a host, indica ted by the enrichment of terms for both chitin metabolic and catabolic processes In addition, protein degradation functions such as peptidase and serine peptidase activity are enriched. These functions are common in fungal pathogens because they obtain nu trient amino acids from their hosts.

PAGE 31

31 Figure 2 1 BLAST P results of ab intio only gene models compared to the NCBI protein database ( expect value cutoff E 6 )

PAGE 32

32 Figure 2 2 The number of top BLAST P hits by species. All spec i es shown have a sequenced genome. The others category includes all species with less than 10 top hits ( expect value cutoff E 6 )

PAGE 33

33 Figure 2 3 BLAST P comparison of Cqf gene models to the NCBI protein database ( expect value cutoff E 6 ) Figure 2 4 C omparison of the number of Cqf secreted proteins predicted by two different i n silico methods from Joly et al., 2010 and Min, 2010.

PAGE 34

34 Figure 2 5 The Cqf secretome is enriched for proteins with no similarity to known proteins ( B LAST P with the NCBI protein database, expect value cutoff E 6 )

PAGE 35

35 Table 2 1 L ist of GO terms enriched in the Cqf secretome (FDR 0 01 ) Category Term FDR # in test group # in reference group Process Carbohydrate Metabolic Process 8.61E 33 76 281 Metabolic Process 9.38E 13 171 2152 Polysaccharide Catabolic Process 1.77E 10 14 22 Primary Metabolic Process 3.80E 09 139 1669 Polysaccharide Metabolic Process 4.77E 05 16 46 Carbohydrate Catabolic Process 1.72E 04 15 60 Disaccharide Metabolic Process 2. 78E 04 12 40 Glycoside Metabolic Process 3.17E 04 12 41 O ligosaccharide Metabolic Process 3.66E 04 12 42 Aminoglycan Catabolic Process 3.72E 04 6 9 Chitin Catabolic Process 3.72E 04 6 9 Function Hydrolase Activity, Acting o n Glycosyl Bonds 3.24E 21 32 64 Catalytic Activity 4.95E 18 168 1892 Hydrolase Activity 1.48E 17 97 728 Hydrolase Activity, Hydrolyzing O Glycosyl Compounds 6.09E 16 24 47 Cellulase Activity 2.82E 04 7 12 Peptidase Activity 3.66E 04 23 140 Serine Type Peptidase Activ ity 3.72E 04 11 36 Serine Hydrolase Activity 3.72E 04 11 36 Chitinase Activity 3.72E 04 6 9 Lipase Activity 4.57E 04 9 24 Phospholipase Activity 5.82E 04 8 19

PAGE 36

36 CHAPTER 3 MEASURING CQF GENE EXPRESSION IN PINE AND OAK INFECTION Background Studies o f Cqf gene expression to date have focused on only a handful of genes (Warren and Covert, 2004; Baker et al., 2006) Using t he gene models obtained from the Cqf sequencing project w hole genome m icroarrays were used to measure gene expression during the vegetative stage of infection in both the oak and pine hosts The fungus infects the pine stem, leading to gall formation By contrast, the fungus infects oak through leaf stomata and does not stimulate gall formation Th e vegetative growth stage of the non sporulating fungus was chosen as the most comparable between these two different infections. T he fungus is actively obtaining nutrients from the host to support its own growth and in order to do so, actively circumv enting host defenses. Since e ffector loci co evolve with host resistance loci it would not be surprising if separate sets of effectors evolved for each host i n the genome of a single fungus. Recent microarray studies in the closely related Melampsora larici populina have focused on the telial host only, comparing infection time points, different zone s of infected leaves or infected versus uninfected tissues These data show that small encod ing genes are expressed more highly in infected tissues compared to spores (Duplessis et al., 2011b) In addition, candidate effectors are expressed to a higher level in haustoria containing host mesophyll cells compared to areas where sporulation is taking place (Hacquard et al., 2010) Melampsora larici populina also expresses distinct sets of these genes along the time course of infection (Duplessis et al., 2011a; Hacquard et al., 2012) Th e mircoarray experimen t described

PAGE 37

37 in this thesis offers a unique direct comparison of the same fungus actively growing in two different hosts. In order to obtain Cqf gene expression information while the fungus is interacting with each host, the sampled tissues contained both C qf and host transcripts, either pine stem or oak leaf. The use of m icroarrays offer s advantage s over high throughput RNA seque ncing in this experimental design Since essentially all genes in the genome have probes on the microarrays all genes were sampled By contrast RNA sequencing would require enough sequenc ing depth to accommodate the highly expressed genes both host and pathogen, and detect less common transcripts Lack of sequencing depth could result in skewing the observed global expression patte rn. The m icroarrays filter h ost transcripts because host transcripts do not hybridize to Cqf probes. We used the microarray data to test whether or not the Cqf transcriptome i s substantially different, or substantially similar, as the fungus infects it s two hosts. Materials and Methods Fungal and Plant Material Inoculations of both oak and pine seedlings were done at the USDA Forest Service, Resistance Screening Center in Asheville, North Carolina. A single uredinial spore isolate called SC20 21 was use d to inoculate 15 open poll i nated northern red oak ( Quercus rubra ) seedlings. Three infected leaves w ere pooled and harvested from each of 8 plants five days after inoculation before lesions were visible and before telia formed. The 3 leaves were combined into a single sample and immediately frozen on dry ice. Inoculated o ak plants that were not harvest ed were used : 1) T o monitor the subsequent level of infection ; and 2) To collect basidiospores for pine inoculations Heavy infection was observed and ample basidiospores were collected. Fifty open

PAGE 38

38 pollinated sus ceptible slash pine ( Pinus elliottii ) seedlings were inoculated. G all ed stem sections, of uniform size, were collected from 8 pine seedlings 18 weeks after inoculation and well before sporulation. Ind ividual s tem s ample s were immediately frozen on dry ice. RNA E xtraction Pine stem samples were freeze dried for 4 days before extraction. They were broken into small pieces using a coffee grinder The small pieces were further gr ound to a fine powder using three 5/32 inch stainless steel balls in microcentrifuge tubes processed in a Geno /G rinder 2000 homogenizer Oak leaf samples were ground in liquid nitrogen to a fine powder. RNA was extracted from approximately 200mg of ground oak or pine sample using a previously described c etyltrimethylammonium b romide ( CTAB ) buffer method (Chang et al., 1993) Microarray Experimental Design The microarray experiment was a two dye control design (Churchill, 2002) Two Agilent 4 X 44K microarray slides populated with custom probes were used. P robes web bas ed eArray software. Of the 8782 Cqf gene models eArray designed from one to five 60 mer oligonucleotide probes for 8692 genes. All probes were used and e ach microarray had 43803 gene features with 26525 probes present once and 8639 probes present twice. L abeled target cRNA ( compl e mentary RNA) was generat ed using Agilent Low Input Quick Amp Labeling Kit such that oak and pine samples were labeled with either cy3 or cy5 an equal number of times across the experiment Each microarray was hybridized with la beled cRNA target derived from a single oak sample and labeled cRNA target derived from a single pine sample. There were a total of eight oak sample replications and eight pine

PAGE 39

39 sample replications. Target hybridization and scanning was performed by the Uni versity Research using standard procedures and an Agilent M icroarray S canner. Statistical Analysis and to calculate ba ckground subtracted feature signal intensities based on local background. Features were flagged and removed from differential expression analysis if any of the following software determined conditions were met: 1) The feature could not be found; 2) The fe ature was saturated; 3) The feature was non uniform; 4) The feature was an outlier; or 5) The feature was not positive and significant over hybridized gene features were Background subtracted signal intensities were log 2 transformed and normalized by setting microarray mean s to zero with a standard deviation of 1. Least squared means of gene expression levels in oak and pine were calculated with a mix ed model analysis of variance (ANOVA) using PROC MIXED in SAS (SAS Institute, Cary, NC, USA) where the effect of host was fixed and the effects of probe, dye and array were random. Q values on the estimates between oak and pine were calculated using the st atistical software R (The R Foundation for Statistical Computing) with a false discovery rate of 0.01 (Storey and Tibshirani, 2003)

PAGE 40

40 Results Figure 3 1 depicts the tissue expression patterns of all 8692 genes on the microarrays based on results from the feature extraction software. The majority of genes are expressed in both tissues. The mixed tissue samples, oak leaf/ Cqf and pine stem/ Cqf had the effect of limiting the amount of Cqf target hybridizing to the microarrays. Signal intensities across all mi croarrays were observed to be low based on the fact that 40% of gene features were removed as not positive and significant above background. Despite the large number of features below background only 111 genes were completely removed by flags and designate d undetected. This was due to the large number of sample replicates (8) for each host and the fact that most genes had multiple probes with 6865 genes having 3 or more probes after flagged features were removed. In addition, there was evidence for expressi on of 572 out of 621 ab in i tio genes obtained with microarrays, 95 of which were expressed 2 fold or more in one host over the other. Figure 3 2 shows the distribution of log2 transformed, background subtracted signal intens ities of Cqf expression in pine and oak hosts. The mean transcript abundance was higher in pine compared to oak, and the variability of transcript abundance was hi gher in pine compared to oak (i.e., there was a wider bell shaped curve in pine compared to o ak). A total of 5077 genes showed evidence of significantly higher expression in one host compared to the other with 3 068 transcripts higher in pine and 2009 transcripts higher in oak. Applying a s tringent criteri on of statistical significance (FDR<0.01) in addition to 4 fold higher transcript abundance in on e host compared to the other, 73 genes were expressed to a higher level in pine and 100 genes were expressed higher in oak. Genes encoding secreted proteins, as well as

PAGE 41

41 lineage specific SSPs, were enri ched among these highly differentially expressed genes (Figure 3 3 Figure 3 4) Discussion The microarray experiment was designed to test whether transcript abundance reflected involvement of d istinct, nonoverlapping sets of Cqf effectors during development of leaf rust (on oak) and stem galls (on pine). However, almost all Cqf genes showed evidence of expression in both Quercus and Pinus hosts, suggesting that the same set of genes is involved in both infections (Figure 3 1) It therefore seems feasible that the heteroecious life cycle of Cqf is enabled by similar genes acting to establish disease states in both hosts. Perhaps Cqf effectors target host processes th at are conserved between angiosperms and gymnosperms. Alternatively, the host alternation of Cqf may be conditioned by post transcriptional processes that are not reflected by differential transcript abundance. Interestingly, Cqf colonizing oak leaves appe ars to show a lower mean and less variation in transcript abundance compared to Cqf colonizing pine (Figure 3 2) The reason for the observed difference is not obvious, and could be due to Cqf abundance and RNA extractability in the samples. Alternatively there could be an underlying biological explanation, perhaps reflecting distinct host specific constraints on Cqf gene regulation The most highly differentially expressed Cqf genes between oak and pine colonized tissues are likely to be important to host specific infection This assertion is supported by the fact that these categories are enriched in genes encoding secreted proteins (Figure 3 3) Of the 100 proteins encoded by genes expressed 4 fold or more in oak over pine, 37 are predicted to be secreted and 22 are lineage specific SSPs

PAGE 42

42 (small secreted proteins less than 300 amino acids) Of the 73 proteins encoded by genes expressed 4 fold or more in pine over oak, 28 are predicted to be secr eted and 10 are lineage specific SSPs (Figure 3 3 Figure 3 4) These are candidate effectors and include known putative effectors The 4 rust transferred protein paralogs show an interesting ex pression pattern ; 2 are overexpressed in oak with 3.7 and 6.6 fold increases in expression and 2 are overexpressed in pine with 2.6 and 4.5 fold increases in expression. In bean rust this effector is known to enter the nucleus of host cells, presumably to alter host gene expression in some way (Kemen et al., 2005) In Cqf different paralogs of the same effector might be preferentially altering host gene expression in two different hosts. Gene differentially expressed between oak and pine, yet not predicted to be secreted may still be important to the infection process. Among the 63 genes encoding proteins with 4 fold or more increased expression in oak and not predicted to be secreted there are 44 unknown genes and 19 genes with putative functions These includ e a multi copper oxidase laccase like protein involved in lignin degradation 2 proteins involved in transport and 3 proteins encod ing signaling proteins. The genes with unknown functions include 1 with an RNA binding domain and 1 with the transcription factor activity GO term. Am ong the 45 genes encoding proteins with 4 fold or more increased expression in pine and not predicted to be secreted there are 24 unknown genes and 21 genes with putative functions These include s ix transporters and 1 identified as a n m RNA binding post tr anscriptional regulator In addition, the gene encoding the known mycotoxin synthesis protein TRI14 is expressed 4 fold more in pine

PAGE 43

43 than oak. Relative gene expression determine d with microarrays has identified genes that putatively function preferentially during infection of one host or the other.

PAGE 44

44 Figure 3 1 The majority of Cqf genes are expressed in both hosts. Figure 3 2 Distributions of log2 transformed pine ( red ) and oak ( blue ) background subtracted signal intensi ties. Each line is a single replicate sample.

PAGE 45

45 Figure 3 3 Genes encoding secreted proteins are enriched among genes with greater than 4 fold increased expression in one host over the other. Figure 3 4 Genes encoding lineage specific SSPs are e nriched among genes with greater than 4 fold increased expression in one host over the other ( BLAST P expect value cutoff E 6 )

PAGE 46

46 CHAPTER 4 MAPPING AVR1 CANDIDATE GENES Given that the most promising strategy for fusiform rust disease control is the monitorin g of AVR alleles in natural populations, one of the expected outcomes of the Cqf sequencing project was faster identification of AVR genes. The release of the JGI final assembly led to the construction of a physical map containing the Avr1 locus (Figure 4 1) and the identification of Avr1 candidate genes (Table 4 1 ) reduced the number of scaffolds from 2084 to 1198 by joining previously separate scaffolds. M arker sequences of polymorphic bands for 6 amplified fragment length polymorphism (AFLP) markers and two randomly amplified polymorphic DNA (RAPD) markers shown to be linked to Avr1 (Kubisia k et al., 2011) were previously obtained along with a dditional sequence surrounding the RAPD marker sequences generated using the GenomeWalker Universal kit (Clontech #63894). These sequences were compared to the scaffold assembly and the results are su mmarized in Table 4 2 Two AFLP derived and one large RAPD derived marker sequence, as well as a simple sequence repeat (SSR) were localized to scaffold 20 in an order predicted by the Cqf genetic map (Kubisiak et al., 2011 ; Figure 4 1) The RAPD marker BB07 matched to scaffold 20 with 2229 of 2235 bp showing 100% identity. The AFLP marker E13M6 matched to scaffold 20 with a n exact match in 74 of 74bp while E6M7 matched to scaffold 20 with 99 of 461 bp and 96% identity. The primer sequences of SSR DN_058 occurred in scaffold 20 at a distance and orientation that would predict a polymerase chain reaction (PCR) band of 3 31 bp that includes 20 repeats of the dinucleotide pai r TG, as expected based on marker development experiments in Cqf (Burdine et al.,

PAGE 47

47 2007) Th ese data provide good evidence that the Avr1 gene is likely to reside between markers BB07 and E13M6 on the physical map. This interval contains 25 genes, 3 of which are predicted to encode secreted proteins (Table 4 1) The most logical candidates are genes encoding secreted proteins since avirulence proteins must leave fungal cells in order to interact with the host pro teins and lead to the no disease phenotype, however, prediction methods are not foolproof and all genes should be considered as candidates.

PAGE 48

48 Figure 4 1 Genetic (A) and physical (B) ma ps of the Avr1 locus indicate a physical/genetic distance ratio of 1 0 .4 kb/cM (460kb/44.2cM)

PAGE 49

49 Table 4 1 AVR1 candidate genes by location on s caffold 20 Scaffold 20 Coordinates BLASTP Description BLASTP E value Protein Length Secretion Prediction Microarray Expression (FDR 0.01) 4676 4749 E13M6 Marker 11036 5318 hypothetical protein [Melampsora larici populina ] 1.6e 173 1397 expressed in pine and oak 13459 13033 no hits 26 not tested 18285 22499 nuclear elongation and deformation protein 1 0 1115 expressed in pine and oak 41077 43578 hypothetical protein [Melampsora larici populina] 0 312 expressed in pine and oak 74051 70543 developmental regulator 6.1e 31 371 expressed in pine and oak 79309 77744 ribosomal RNA processing protein 8 2.0e 144 377 expressed in pine and oak 85282 86634 hypothetical p rotein [Puccinia graminis f. sp. tritici] 5.7e 103 356 yes expressed in pine and oak 106082 109381 hypothetical protein [Melampsora larici populina] 1.4e 101 763 expressed in pine and oak 118826 116211 aspartyl aminopeptidase 0 521 expressed in pine and oak 143532 133336 cation transporting ATPase 0 2418 yes expressed in pine and oak 145410 144897 no hits 103 4 fold up in oak 146710 147654 unknown protein 6.4e 20 173 not tested 164104 164978 peptidyl prolyl cis trans isomerase NIMA interac ting 4 2.9e 53 140 expressed in pine and oak 165764 168672 no hits 508 yes 2 fold up in pine

PAGE 50

50 Table 4 1 C ontinued Scaffold 20 Coordinates BLASTP Description BLASTP E value Protein Length Secretion Prediction Microarray Expression (FDR 0.01) 17 5775 177281 unknown protein 6.6e 22 336 2 fold up in oak 178998 179097 E6M7 Marker 194077 195314 no hits 103 4 fold up in oak 203178 200574 hypothetical protein [Melampsora larici populina] 1.75e 75 582 2 fold up in oak 205083 204227 ring box protein 1 9.9e 37 93 expressed in pine and oak 214869 213003 g protein alpha subunit 1.6e 141 376 expressed in pine and oak 225168 218906 myosin 5 0 1689 expressed in pine and oak 227391 228236 guanine nucleotide binding protein alpha 2 subuni t 1.1e 21 121 expressed in pine and oak 241466 242464 uracil phosphoribosyltransferase 6.0e 115 199 expressed in pine and oak 243491 244153 hypothetical protein [Melampsora larici populina] 7.8e 9 220 expressed in pine and oak 252449 251812 hypoth etical protein [Melampsora larici populina] 2.7e 59 186 expressed in pine and oak 253915 255134 scf complex subunit skp1 1.0e 91 158 expressed in pine and oak 251681 254246 BB07 Marker

PAGE 51

51 Table 4 2 AVR 1 marker sequence identity to final assembl y scaffolds Marker Type Sequence Length Top Hit Scaffold Top Hit E value Next Best Hit Scaffold/E value E6M7* AFLP 461 906 0 204/8E 52 E7M6 AFLP 470 6 0 229/E 163 E9M7 AFLP 437 39 0 202/0.10 E8M22 AFLP 473 47 0 53/0 E12M8 AFLP 201 398 E 110 34/0.7 2 E13M6 AFLP 74 20 2E 35 34/0.23 BB07 RAPD 2235 20 0 22/E 117 AZ17 RAPD 903 32 0 309/0 DN_058 forward primer** SSR 20 20 1.4 26/5.7 DN_058 reverse primer** SSR 20 20 0.092 28/1.4 *E6M7 has homology to scaffold 20 with and E value of 2E 40 **Scaffold 20 contains a 331bp interval between the forward and reverse SSR primers that includes the dinucleotide repeat TG, repeated 20 times.

PAGE 52

52 CHAPTER 5 CONCLUSIONS This project has provided gene expression evidence that Cqf uses essentially one set of genes to ca use two very different infections; pine gall formation and oak leaf colonization. The identification of encoded secreted proteins in the Cqf genome will serve as a reference for future experiments evaluating effectors and their roles in pathogenicity. Secr etome information and gene expression results can be combined to provide valuable insight into candidate effectors. For example, two different families of paralogous genes that encode secreted proteins (one set of 4, and one set of 5) show at least 2 fold increased expression during pine infection compared to oak infection. The fact that these genes are unique to Cqf and show expression bias make them good candidates for effectors important to fusiform rust disease. In addition, this work has provided a seq uence interval that greatly enhances the possibility of identifying the fusiform rust avirulence gene, Avr1 (Kubisiak et al., 2011) that interacts genetically with the resistance gene Fr1 in loblolly pine (Wilcox et al., 1996) Two strategies are being taken to pinpoint this important effector. First, experiments are underway to narrow the candidate gene interval by iden tifying additional genetic markers linked to Avr1 using high throughput sequencing data. Also in progress is a re sequencing effort to identify alleles of candidate genes associated with the disease resistance and susceptibility by screening unrelated pycn iospores from pines genotyped for the resistance locus. In a broader context, there are plans to integrate the previously published genetic map with the scaffold sequence in order to obtain a clearer picture of the Cqf genome and aid in the generation of p hysical maps for other AVR genes (Kubisiak et al., 2011)

PAGE 53

53 APPENDIX FUNCTIONALLY ENRICHED GO ANNOTATIONS IN THE CQF SECRETOME Table A 1 Complete l ist of GO terms enriched in the Cqf secretome (FDR .05 ) Category Term FDR Proce ss Carbohydrate Metabolic Process 8.61E 33 Metabolic Process 9.38E 13 Polysaccharide Catabolic Process 1.77E 10 Primary Metabolic Process 3.80E 09 Polysaccharide Metabolic Process 4.77E 05 Carbohydrate Catabolic Process 1.72E 04 Disaccharide Me tabolic Process 2.78E 04 Glycoside Metabolic Process 3.17E 04 O ligosaccharide Metabolic Process 3.66E 04 Aminoglycan Catabolic Process 3.72E 04 Chitin Catabolic Process 3.72E 04 Cellular Carbohydrate Metabolic Process 0.00121 Phospholipid Catab olic Process 0.001306 Cell Wall Macromolecule Catabolic Process 0.001435 Cellular Polysaccharide Metabolic Process 0.002117 Proteolysis 0.003736 Chitin Metabolic Process 0.003945 Cell Wall Macromolecule Metabolic Process 0.004693 Lipid Cataboli c Process 0.005398 Cellular Lipid Catabolic Process 0.006946 Starch Metabolic Process 0.006946 Sucrose Metabolic Process 0.006946 Cellular Glucan Metabolic Process 0.015064 Glucan Metabolic Process 0.015064 Macromolecule Catabolic Process 0.016 761 Function Hydrolase Activity, Acting o n Glycosyl Bonds 3.24E 21 Catalytic Activity 4.95E 18 Hydrolase Activity 1.48E 17 Hydrolase Activity, Hydrolyzing O Glycosyl Compounds 6.09E 16 Cellulase Activity 2.82E 04 Peptidase Activity 3.66E 04 Se rine Type Peptidase Activity 3.72E 04 Serine Hydrolase Activity 3.72E 04 Chitinase Activity 3.72E 04 Lipase Activity 4.57E 04 Phospholipase Activity 5.82E 04 Ion Binding 0.002612 Cation Binding 0.002612 Peptidase Activity, Acting o n L Amino A cid Peptides 0.003028 Endopeptidase Activity 0.046954 Compartment Extracellular Region 0.001435

PAGE 54

54 LIST OF REFERENCES Anderson, C.L., Kubisiak, T.L., Nelson, C.D., Smith, J.A., and Davis, J.M. 2010. Genome size variation in the pine f usiform rust pathogen Cronartium quercuum f.sp. fusiforme as determined by flow cytometry. Mycologia 102:1295 1302. Anderson, R.L., McClure, J.P., Cost, N., and Uhler, R.J. 1986. Estimating f usiform r ust l osses in f ive s outheast s tates. Southern Journal of Applied Forestry 10:237 240. Baker, L.G., Spaine, P., and Covert, S.F. 2006. Effect of surface wettability on germination and gene expression in Cronartium quercuum f. sp. fusiforme basidiospores. Physiol Mol Plant P 68:168 175. Burdine, C.S., Kubisiak, T .L., Johnson, G.N., and Nelson, C.D. 2007. Fifty two polymorphic microsatellite loci in the rust fungus, Cronartium quercuum f.sp. fusiforme Molecular Ecology Notes 7:1005 1008. Burdsall, H.H., Jr., and Snow, G.A. 1977. Taxonomy of Cronartium quercuum and C. fusiforme Mycologia 69:503 508. Catanzariti, A. M., Dodds, Peter N., Lawrence, Gregory J., Ayliffe, Michael A. and Ellis, Jeffrey G. 2006. Haustorially e xpressed s ecreted p roteins from f lax r ust a re h ighly e nriched for a virulence e licitors The Plant Cell 18:243 256. Chang, S., Puryear, J., and Cairney, J. 1993. A simple and efficient method for isolating RNA from pine trees. Plant Molecular Biology Reporter 11:113 116. Churchill, G.A. 2002. Fundamentals of experimental design for cDNA microarrays. Nat Genet. 32:490 495. Conesa, A., G tz, S., Garca Gmez J.M., Terol, J., Taln M., and Robles, M. 2005. B LAST 2GO: a universal tool for annotation, visualization and analysis in functional genomics research. Bioinformatics 21:3674 3676. Dean, R.A., Talbot, N.J., Ebbole, D.J., Farman, M.L., Mitchell, T.K., Orbach, M.J., Thon, M., Kulkarni, R., Xu, J. R., Pan, H., Read, N.D., Lee, Y. H., Carbone, I., Brown, D., Oh, Y.Y., Donofrio, N., Jeong, J.S., Soanes, D.M., Djonovic, S., Kolomiets, E., Rehmeyer, C., Li, W. Harding, M., Kim, S., Lebrun, M. H., Bohnert, H., Coughlan, S., Butler, J., Calvo, S., Ma, L. J., Nicol, R., Purcell, S., Nusbaum, C., Galagan, J.E., and Birren, B.W. 2005. The genome sequence of the rice blast fungus Magnaporthe grisea Nature 434:980 9 86. Duplessis, S., Hacquard, S., Delaruelle, C., Tisserant, E., Frey, P., Martin, F., and Kohler, A. 2011a. Melampsora larici populina t ranscript p rofiling d uring g ermination and t imecourse i nfection of p oplar l eaves r eveals d ynamic e xpression p atterns a ss ociated with v irulence and b iotrophy. Mol Plant Microbe In 24:808 818.

PAGE 55

55 Duplessis, S., Cuomo, C.A., Lin, Y. C., Aerts, A., Tisserant, E., Veneault Fourrey, C., Joly, D.L., Hacquard, S., Amselem, J.l., Cantarel, B.L., Chiu, R., Coutinho, P.M., Feau, N., Fiel d, M., Frey, P., Gelhaye, E., Goldberg, J., Grabherr, M.G., Kodira, C.D., Kohler, A., Kes U., Lindquist, E.A., Lucas, S.M., Mago, R., Mauceli, E., Morin, E., Murat, C., Pangilinan, J.L., Park, R., Pearson, M., Quesneville, H., Rouhier, N., Sakthikumar, S ., Salamov, A.A., Schmutz, J., Selles, B., Shapiro, H., Tanguay, P., Tuskan, G.A., Henrissat, B., Van de Peer, Y., Rouze, P., Ellis, J.G., Dodds, P.N., Schein, J.E., Zhong, S., Hamelin, R.C., Grigoriev, I.V., Szabo, L.J., and Martin, F. 2011b. Obligate bio trophy features unraveled by the genomic analysis of rust fungi. Proceedings of the National Academy of Sciences. Dwinell, L.D. 1976. Biology of f usiform r ust. i n m anagement of f usiform r ust in s outhern p ines, Symposium Proceedings of the Southern Forest D isease and Insect Research Council, R.J.S. Dinus, Robert A., ed (University of Florda, Gainesville, Flordia), pp. 18 24. Dyer, R.B., Plattner, R.D., Kendra, D.F., and Brown, D.W. 2005. Fusarium graminearum TRI14 Is Required for High Virulence and DON Produ ction on Wheat but Not for DON Synthesis in Vitro. J Agr Food Chem 53:9281 9287. Ellis, J.G., Dodds, P.N., and Lawrence, G.J. 2007. Flax r ust r esistance g ene s pecificity is b ased on d irect r esistance a virulence p rotein i nteractions. Annual Review of Phytop athology 45:289 306. Flor, H.H. 1955. Host p arasite i nteraction in f lax r ust its g enetics and o ther i mplications Phytopathology 45:680 685. Gan, P.H.P., Rafiqi, M., Hardham, A.R., and Dodds, P.N. 2010. Effectors of biotrophic fungal plant pathogens. Fun ctional Plant Biology 37:913 918. Gray, D.J., Amerson, H.V., and Dyke, C.G.V. 1982. An ultrastructural comparison of monokaryotic and dikaryotic haustoria formed by the fusiform rust fungus Cronartium quercuum f.sp. fusiforme. Canadian Journal of Botany 60 :2914 2922. Hacquard, S., Delaru elle, C., Legu V., Tisserant, E., Kohler, A., Frey, P., Martin, F., and Duplessis, S. 2010. Laser c apture m icrodissection of u redinia f ormed by Melampsora larici populina r evealed a t r anscriptional s witch b etween b iotrophy and s porulation. Mol Plant Microbe In 23:1275 1286. Hacquard, S.p., Joly, D.L., Lin, Y. C., Tisserant, E., Feau, N., Delaruelle, C., Legue, V., Kohler, A., Tanguay, P., Petre, B., Frey, P., Van de Pe er, Y., Rouz P., Martin, F.M., Hamelin, R.C., and Dupl essis, S.b. 2012. A comprehensive analysis of genes encoding small secreted proteins identifies candidate effectors in Melampsora larici populina (poplar leaf rust). Mol Plant Microbe In. Hahn, M., and Mendgen, K. 1997. C haracterization of in planta induce d rust genes isolated from a haustorium specific cDNA l ibrary. Mol Plant Microbe In 10:427 437.

PAGE 56

56 Hahn, M., and Mendgen, K. 2001. Signal and nutrient exchange at biotrophic plant fungus interfaces. Curr Opin Plant Biol 4:322 327. Holt, C., and Yandell, M. 20 11. MAKER2: an annotation pipeline and genome database management tool for second generation genome projects. Bmc Bioinformatics 12:491. Joly, D., Feau, N., Tanguay, P., and Hamelin, R. 2010. Comparative analysis of secreted protein evolution using express ed sequence tags from four poplar leaf rusts ( Melampsora spp.). BMC Genomics 11:422. Kamper, J., Kahmann, R., Blker M., Ma, L. J., Brefort, T., Saville, B.J., Banuett, F., Kronstad, J.W., Gold, S.E., Muller, O., Perlin, M.H., Wosten, H.A.B., de Vries, R. Ruiz Herrera, J., Reynaga Pena, C.G., Snetselaar, K., McCann, M., Perez Martin, J., Feldbrugge, M., Basse, C.W., Steinberg, G., Ibeas, J.I., Holloman, W., Guzman, P., Farman, M., Stajich, J.E., Sentandreu, R., Gonzalez Prieto, J.M., Kennell, J.C., Molina L., Schirawski, J., Mendoza Mendoza, A., Greilinger, D., Munch, K., Rossel, N., Scherer, M., Vranes, M., Ladendorf, O., Vincon, V., Fuchs, U., Sandrock, B., Meng, S., Ho, E.C.H., Cahill, M.J., Boyce, K.J., Klose, J., Klosterman, S.J., Deelstra, H.J., Ort iz Castellanos, L., Li, W., Sanchez Alonso, P., Schreier, P.H., Hauser Hahn, I., Vaupel, M., Koopmann, E., Friedrich, G., Voss, H., Schluter, T., Margolis, J., Platt, D., Swimmer, C., Gnirke, A., Chen, F., Vysotskaia, V., Mannhaupt, G., Guldener, U., Mnst erktter M., Haase, D., Oesterheld, M., Mewes, H. W., Mauceli, E.W., DeCaprio, D., Wade, C.M., Butler, J., Young, S., Jaffe, D.B., Calvo, S., Nusbaum, C., Galagan, J., and Birren, B.W. 2006. Insights from the genome of the biotrophic fungal plant pathogen Ustilago maydis Nature 444:97 101. Kemen, E., Kemen, A.C., Rafiqi, M., Hempel, U., Mendgen, K., Hahn, M., and Voegele, R.T. 2005. Identification of a Protein from Rust Fungi Transferred from Haustoria into Infected Plant Cells. Mol Plant Microbe In 18:11 30 1139. Krogh, A., Larsson, B.r., von Heijne, G., and Sonnhammer, E.L.L. 2001. Predicting transmembrane protein topology with a hidden markov model: application to complete genomes. J Mol Biol 305:567 580. Kubisiak, T.L., Amerson, H.V., and Nelson, C.D. 2 005. Genetic interaction of the fusiform rust fungus with resistance gene F r1 in loblolly pine Phytopathology 95:376 380. Kubisiak, T.L., Anderson, C.L., Amerson, H.V., Smith, J.A., Davis, J.M., and Nelson, C.D. 2011. A genomic map enriched for markers li nked to Avr1 in Cronartium quercuum f.sp. fusiforme Fungal Genet Biol 48:266 274. Kulkarni, R.D., Kelkar, H.S., and Dean, R.A. 2003. An eight cysteine containing CFEM domain unique to a group of fungal membrane proteins. Trends Biochem Sci 28:118 121.

PAGE 57

57 Min X.J. 2010. Evaluation of computational methods for secreted protein prediction in different eukaryotes Journal of Proteomics and Bioinformatics 3:143 147. Nelson, C.D., Kubisiak, T.L., and Amerson, H.V. 2010. Unravelling and managing fusiform rust disea se: a model approach for coevolved forest tree pathosystems. Forest Pathol 40:64 72. Panstruga, R., and Dodds, P.N. 2009. Terrific protein traffic: the mystery of effector protein delivery by filamentous plant pathogens Science 324:748 750. Phelps, W.R., Czabator, F.L. 1978. Fusiform rust of southern pines. in forest insect and disease Leaflet, F orest Service, United States Department of Agriculture, ed. Saunders, D.G.O., Win, J., Cano, L.M., Szabo, L.J., Kamoun, S., and Raffaele, S. 2012. Using hierarchic al clustering of secreted protein families to classify and rank candidate effectors of rust fungi PLoS ONE 7:e29847. Schmidt, R.A. 1998. Fusiform rust disease of southern pines: biology, ecology and management In University of Florida, Institute of Food and Agriculture, Technical Bulletin, pp. 1 14. Schmidt, R.A. 2003. Fusiform rust of southern pines: a major success for forest disease management Phytopathology 93:1048 1051. Skamnioti, P., Gurr, Sarah J. 2007. Magnaporthe grisea Cutinase2 mediates appres sorium differentiation and host penetration and is required for full virulence The Plant Cell 19:2674 2689. Starkey, D.A., Anderson, Robert L. Young, Carol H., Cost, Noel D., Vissage, John S., May, Dennis M., Yockey, Edwin K. 1997. Monitoring incidence of fusiform rust in the south and change over time In Protection Report (Atlanta, Georgia: United States Department of Agriculture, Forest Service). Stergiopoulos, I., and de Wit, P.J.G.M. 2009. Fungal e ffector p roteins. Annual Review of Phytopathology 47:2 33 263. Storey, J.D., and Tibshirani, R. 2003. Statistical significance for genomewide studies. Proceedings of the National Academy of Sciences 100:9440 9445. van den Burg, H.A., Harrison, S.J., Joosten, M.H.A.J., Vervoort, J., and de Wit, P.J.G.M. 2006. C ladosporium fulvum Avr4 protects fungal cell walls against hydrolysis by plant chitinases accumulating during infection Mol Plant Microbe In 19:1420 1430. van Esse, H.P., van't Klooster, John W., Bolton, Melvin D, Yadeta, Koste A., van Baarlen, Peter, Bo eren, Sjef, Vervoort, Jacques, de Wit, Pierre J.G.M. and Thomma, Bart P.H.J. 2008. The Cladosporium fulvum virulence pr otein Avr2 inhibits host proteases required for basal def ense The Plant Cell 20:1948 1963.

PAGE 58

58 Vogler, D.R., and Bruns, T.D. 1998. Phylogenetic r elationships among the pine stem rust fungi ( Cronartium and Peridermium spp.). Mycologia 90:244 257. Warren, J.M., and Covert, S.F. 2004. Differential expression of pine and Cronartium quercuum f. sp. fusif orme genes in fusiform rust gall s. Appl Environ Microb 70:441 451. Wilcox, P.L., Amerson, H.V., Kuhlman, E.G., Liu, B.H., OMalley, D.M., and Sederoff, R.R. 1996. Detection of a major gene for resistance to fusiform rust disease in loblolly pine by genomic mapping. P Natl Acad Sci USA 93:3859 3864.

PAGE 59

59 BIOGRAPHICAL SKETCH Katherine Smith has held a diverse array of laboratory technician positions since her 1988 graduation from Virginia Tech, with a Bachelor of Science in biology. She has been working in the f ield of plant pathology since 1996 and began working on fusiform rust disease when she was hired by the USDA Forest Service in 2000. She contributed to the work on genetic markers that surround the first Cqf Avr gene, Avr1 That research in combination wit h genomic sequencing could eventually lead to the cloning of Avr1 While working for the Forest Service, she began taking graduate courses in the fall of 2006 and entered the PMCB program in spring 2010, where she earned a Master of Science degree studying Cqf fungal effectors.