Quantifying Efficiency and Accuracy in the Washington State

Home , Aurora Golden Gala, Beacon (apple), Braeburn, Cosmic Crisp, Cripps Pink, Haralson (apple)

QUANTIFYING EFFICIENCY AND ACCURACY IN THE WASHINGTON STATE

UNIVERSITY APPLE BREEDING PROGRAM

JULIA MAE HARSHMAN

A dissertation submitted in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

WASHINGTON STATE UNIVERSITY Department of Horticulture

DECEMBER 2015

To the Faculty of Washington State University:

The members of the Committee appointed to examine the dissertation of

JULIA MAE HARSHMAN find it satisfactory and recommend that it be accepted.

______Katherine M. Evans, Ph.D., Chair

______Amit Dhingra, Ph.D

______Craig M. Hardner, Ph.D

______James P. Mattheis, Ph.D

______Michael O. Pumphrey, Ph.D

Acknowledgements

I would first like to thank my advisor, Kate Evans. The serendipitous meeting in 2011 and the leap of faith on your part has led to this culmination of my education and graduate career. It has been an honor to be your student and the experiences you have shared will forever be invaluable to me. No one could have prepared me to be a plant breeder or a professional as well as you have, and I look forward to trying.

I would like to thank Craig Hardner for his guidance and extreme patience in teaching me equation-based quantitative genetics, as well as the life-changing experience of living abroad in

Australia. I would like to thank Amit Dhingra, Jim Mattheis and Mike Pumphrey for their input, guidance and frank conversations over the course of this doctoral program. A special thank you to Amit Dhingra who tried to persuade me to attend WSU as a doctoral student when I was an undergraduate in 2008. You were the first person to suggest I pursue my doctorate. I would like to thank two of my Masters committee members, Chris Walsh and Wayne Jurick II, for their continued support, guidance and “checking in” during my doctoral program. I would like to thank Jay Norelli for his technical guidance and hosting me at his research station.

I would like to thank Lisa Brutcher, Nancy Buchanan, and Bonnie Schonberg for answering all of my infinite questions, and all of the fun times in the lab, field and outside of work. Lastly, I would like to thank my family and friends who have cheered me on all this time, as well as the rest of the Tree Fruit Research and Extension staff, students and professors.

iii

QUANTIFYING EFFICIENCY AND ACCURACY IN THE WASHINGTON STATE

UNIVERSITY APPLE BREEDING PROGRAM

Abstract

by Julia Mae Harshman, Ph.D. Washington State University December 2015

Chair: Katherine M. Evans

The Washington State University Apple Breeding Program just passed its 20th anniversary, released its third new variety and moved under new direction with breeder Kate

Evans. The major program goal is to release a portfolio of new and improved cultivars suited to production in central Washington. Improved cultivars should have exceptional storability, to compliment the state’s unrivaled long term cold storage facilities, while engendering an exceptional consumer experience. Apple breeding is a long and expensive process and therefore exploring how well the program meets its goal—releasing the best new cultivars—can validate the expenditures of the program while also potentially finding methods that would improve discovery rate and cost efficiency. The three main subprojects and conclusions are as follows: 1)

Quantify moldy core susceptibility in currently available germplasm to inform crossing decisions and discover predisposing characteristics to inform culling decisions. Factors that predispose

iv selections for moldy core were confirmed and appear heritable. 2) Quantify fire blight resistance in wild relative Malus sieversii to further inform crossing decisions and provide data for future association mapping. Multiple M. sieversii accessions were found to be resistant and will be used in 2016 as parents. 3) Analyze the cost structure of the current breeding program and the accuracy of trait evaluations in the replicated data collection phase of the program to explore the most efficient design structure for identifying elite selections. An alternative design that would allow more selections to be evaluated in that phase was proposed and is under consideration for the 2016 planting.

TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS ...... iii

ABSTRACT ...... iv

LIST OF TABLES ...... ix

LIST OF FIGURES ...... xii

LIST OF ABBREVIATIONS ...... xiv

DEDICATION ...... xvi

CHAPTER ONE: Introduction ...... 1

Apple Origin and Cultivation ...... 1

Washington State University Apple Breeding Program ...... 2

Storability ...... 6

Resistance to Diseases ...... 7

Phase 2 Trial Design ...... 11

Dissertation Overview and Hypotheses ...... 14

Literature Cited ...... 15

CHAPTER 2: Survey of Moldy Core Incidence in Germplasm from Three U.S. Apple

Breeding Programs ...... 26

Abstract ...... 26

Introduction ...... 27

Materials and Methods ...... 29

Results ...... 31

Author Contributions...... 34

Acknowledgements ...... 34

Literature Cited ...... 36

CHAPTER THREE: Resistance to Erwinia amylovora in Wild Accessions of Malus sieversii

...... 42

Abstract ...... 42

Introduction ...... 43

Materials and Methods ...... 47

Results ...... 51

Discussion ...... 56

Conclusion ...... 62

Author Contributions...... 62

Acknowledgements ...... 62

Literature Cited ...... 64

CHAPTER FOUR: Consideration of cost and accuracy for advanced breeding trial designs in apple ...... 96

Abstract ...... 96

vii

Introduction ...... 97

Methods ...... 100

Results ...... 106

Cost ...... 107

Effect on Accuracy ...... 108

Effect on response to selection ...... 109

Correlated Response ...... 110

Changing Intensity ...... 111

Discussion ...... 111

Conclusion ...... 117

Author Contributions...... 118

Acknowledgements ...... 118

Literature Cited ...... 119

Appendices ...... 126

CHAPTER FIVE: Summary and Larger Impact ...... 130

viii

LIST OF TABLES

CHAPTER ONE

Ch 1-1. Traits evaluated by the WSU Apple Breeding Program in Phases 1 and 2.

Evaluation at 2 months is after 8 weeks in 2°C storage + 1 week at room

temperature. Evaluation at 4 months is after 16 weeks in 2°C storage +

1 week at room temperature………………………………………………………….25

CHAPTER TWO

Ch 2-1. Core opening, calyx opening and incidence of moldy core at both harvest and

after 10 weeks in 2°C RA storage of RosBREED germplasm evaluated in 2010

and 2011. Incidence of moldy core is presented both as the number of individuals

in each core opening/calyx opening class and as a percentage………………………...39

Ch 2-2. Moldy core susceptibility of cultivars tested at Cornell University 1, University

of Minnesota 2, and Washington State University 3. Trademarked names are in

parenthesis……………………………………………………………………………...40

Ch 2-3. Number of seedling individuals evaluated and incidence of moldy core in

families as part of the RosBREED project in 2010 and 2011…………………….……41

CHAPTER THREE

Ch 3-1. Description of Central Asian collection sites for M. sieversii (Data from

Forsline et al. 2003). * indicate the sites represented by accessions in this

study……………………………………………………………………………………74

Ch 3-2. Congruence of resistance ratings based on natural occurrence of field infection (GRIN)

and ratings based on controlled inoculation in the field of selected M. sieversii

accessions and controls in USDA-ARS Kearneysville, WV for both 2013 and 2014.

GRIN resistance ratings: 1) very resistant - no occurrence; 2) moderately resistant –

only light rating; 3) intermediate – light to medium rating; 4) moderately susceptible –

medium to heavy rating; 5) very susceptible – very heavy rating. For the controlled

inoculations, qualitative resistance ratings of highly susceptible, susceptible,

moderately susceptible, resistant and highly resistant were assigned to accessions based

on %SLB. See Supplemental Table 1 for individual accession assignment. …...……..75

Ch 3-3. Congruence of resistance ratings of selected M. sieversii accessions between

controlled inoculation in the greenhouse at USDA-ARS-PGRU Geneva, NY

and in the field at USDA-ARS Kearneysville, WV. Accessions were considered

resistant if less than 20% of the total shoot length was blighted. The number of

accessions rated as resistant and susceptible are presented for both 2013 and 2014…..76

Ch 3-S1. Genetic cluster, collection site, fire blight resistance (FB Res.) for both

natural occurrence (_Field) and controlled inoculation (_GH Inoc.) at

USDA-ARS-PGRU Geneva, NY as well as number of shoots inoculated,

average percent of current season’s shoot growth blighted (%SLB), and SE,

Logit transformed %SLB and SE for both WA and WV in 2013 and 2014……….…..77

CHAPTER FOUR

Ch 4-1. Costs of phase 2 trial components for a single candidate standardized to the

current design, total standardized Phase 2 trial cost and the total number of

candidates that could be evaluated for the current total program cost for the

current WABP Phase 2 trials and alternative Phase 2 trial designs…………………..122

Ch 4-2. Estimated phenotypic variance (v.P) and percentage of estimated variance

components for individual random effects (G: candidate, H: harvest, L: Location,

E: residual error) and interactions from the analysis of individual fruit quality traits

assessed as part of the WABP Phase 2 trials. Full details of traits given in text and

in Table 1. Zero variance component indicates source of variation was not

significant Also shown is the repeatability for the fruit quality traits. The single

largest source of variance for each trait is emboldened………………………………123

Ch 4-3. The critical percentage difference (CPD) required between sample means to

reject the hypothesis that two candidates have the same true mean with 95%

confidence for traits assessed during Phase 2 trials by the WABP. Trait averages

are presented for the current design in the unit of the trait as is the CPD as a

percentage. CPD presented for alternative designs were subtracted from the

current design to give the degree of change rather than absolute value………….…..124

Ch 4-4. Response to selection (RS) for traits assessed during Phase 2 trials by the

WABP and correlated response to selection (CRS) for four selected pairs

of traits under the current trial design and alternative designs. Selection

intensity (SI) is 10%, unless noted. RS is in the unit of the trait and CRS is

in the unit of the sensory trait……………………………………………………...…125

LIST OF FIGURES

CHAPTER ONE

Ch 1-1. Diagram of the timeline and activities of the WABP…………………………….…....24

CHAPTER TWO

Ch 3-1. Map showing 12 collection regions of Malus sieversii germplasm

(Data from Forsline et al. 2003). As a reference, site 1 was at Latitude 39°N/

Longitude 68°E………………………………………………………………..……….68

Ch 3-2. Evaluation of fire blight resistance of M. sieversii GMAL4028.h. A:

In 34 of 35 fire blight shoots inoculated in 2013 and 2014 no evidence of

infection could be observed 6 weeks after inoculation. B: In 1 of 35 fire

blight inoculated shoots, fire blight progressed into 2 year-old wood of

central leader………………………………………………………………...…………69

Ch 3-3. Average proportion of the shoot length blighted (SLB), including standard

errors, after controlled inoculation with E. amylovora for 64 accessions evaluated

in both 2013 and 2014 and in both Wenatchee, WA and Kearneysville, WV...... 71

Ch 3-4. Average fire blight infection for accessions separated by their collection site.

Fire blight infection measured as the proportion of current season’s shoot length

blighted after inoculation with E. amylovora. The number of accessions collected

from each site is listed below the site number...... 72

Ch 3-5. Average fire blight infection for accessions by genetic cluster. Fire blight

infection measured as the proportion of current season’s shoot length blighted

after inoculation with E. amylovora. The number of accessions in each genetic

xii cluster is listed below the site number……………………………………………...….73

xiii

LIST OF ABBREVIATIONS

ARS: Agricultural Research Services

BCE: Before Common Era

CA: Controlled Atmosphere

CPD: Critical Percentage Difference

CRS: Correlated Response to Selection

CU: Cornell University

DARE: Durable Apple Resistance in Europe

FRTD: Fruit Diameter

FRTW: Fruit Weight

GBS: Genotype by Sequencing

GRIN: Germplasm Resources Information Network

HiDRAS: High-quality Disease Resistant Apples for a Sustainable Agriculture

LRT: Likelihood Ratio Test

LSD: Least Significant Difference

MAB: Marker Assisted Breeding

MASS: Marker Assisted Seedling Selection

NPGS: National Plant Germplasm System

PGRU: Plant Genetic Resources Unit

QTL: Quantitative Trait Loci

RA: Regular Atmosphere

RS: Response to Selection

SLB: Shoot Length Blighted

xiv

SPI: Starch Pattern Index

SSC: Soluble Solids Content

TA: Titratable Acidity

UMN: University of Minnesota

USDA: U.S. Department of Agriculture

WABP: Washington State University Apple Breeding Program

WSU: Washington State University

WTFRC: Washington Tree Fruit Research Commission

1-MCP: 1- methylcyclopropene

Dedication

To my parents, for raising me to be the person I am today and always encouraging me to

pursue a career that I am passionate about, even when it moved me 3,000 miles away.

Your unwavering love and support has made the distance seem so much less.

To my closest and dearest friend Demetra Skaltsas, for your continual support— the motivational

talks, the tough-love, the untold number of hours on the phone, and bottles of wine, that

we have shared in joint pursuit of our doctoral degrees.

To my high school sweetheart and partner Joseph Detz, for your eager and loving participation in

this journey over the past ten years. Your absolute faith in my abilities to achieve my

goals and dreams inspires me daily. You moved across the country without a second

thought and have been here to help with absolutely anything I have needed. We have

joked that you should get a conciliatory degree, but this dedication will have to suffice.

xvi

CHAPTER ONE: Introduction

Apple Origin and Cultivation

Domestic apple (Malus × domestica Borkh.) is an interspecific hybrid derived primarily from Malus sieversii (Ledeb.). Apple is a member of the Rosaceae family, along with other horticulturally important crops such as pear, cherry, peach, almond, raspberry, strawberry and rose (Brown 1975). Apples are pome fruits, along with pears and quince, characterized by fruit with carpellate tissues enclosed in fleshy accessory tissue (Brown 1975). Apples, cultivated since

6500 BCE (Morgan and Richards 2003), are currently grown worldwide in temperate regions and certain subtropical regions (Luby 2003). The majority of production is for fresh consumption; however, there is also a processing and growing cider market (O’Rourke 2003).

Washington State is the largest producer of apples in the U.S. (USDA NASS 2015) with the majority of production focused in central WA. Washington apples are shipped to all 50 states and more than 60 countries (Lyons 2014) with the 2014 crop worth approximately $1.8 billion

(USDA NASS 2015).

Apples are grown as composite plants with the desired fruit variety as the scion vegetatively propagated by grafting or budding onto a rootstock. Grafting allows continued production of a desired variety, and has been practiced since at least 1000 BCE (Mudge et al.

2009). Seedling rootstocks were used historically, but the current industry standard is dwarfing, precocious rootstocks (Wertheim and Webster 2003). Apples have a long juvenility phase; seedlings on their own roots can take as long as ten years to set their first crop (Brown 1975).

Dwarfing, precocious rootstocks have reduced that time to 2 or 3 years from time of planting.

Such rootstocks have made an industry-wide switch to higher density “fruiting walls” possible

(Robinson et al. 2014; Tustin 2014), which has major implications for the future of mechanization, labor and production efficiency in apple (Robinson 2008). Rootstocks have also been selected for cold-hardiness and disease resistance (Robinson et al. 2008; Wertheim and

Webster 2003).

Washington State University Apple Breeding Program

History and Breeding Targets. Apple scion breeding through directed hybridization was first reported by Thomas Andrew Knight in the early 1800’s (Brown 1975). Today, there are more than 10,000 named varieties of apple (Janick 1996), with scion breeding programs adding to the list annually (Gasic and Preece 2014). Apple breeding is a 20+ year process, from cross to release of commercial quality scion varieties (Brown 1975; Evans 2013; Janick 1996).

The WSU Apple Breeding Program (WABP) joins two active public U.S. university apple breeding programs, Cornell University and University of Minnesota, and one retired program, the co-operative PRI program (Purdue University, Rutgers University and the

University of Illinois). Each of these programs has made varietal contributions to the U.S. apple industry. ‘Cortland’ (1915), ‘Empire’ (1966), ‘Jonagold’ (1968) ‘NY-2’ (RubyFrost™, 2013), and ‘NY-1’ (SnapDragon™, 2013) have been released by the Cornell program (Brown and

Terry, 1997; Garris 2013). The very successful ‘Honeycrisp’ (1988) and ‘Minneiska’

(SweetTango™, 2008) have been released by the University of Minnesota program (Luby 1991).

‘Jonafree’ (1979), ‘Enterprise’ (1993), ‘GoldRush’, (1993), and ‘Co-Op 39’ (CrimsonCrisp™,

2006) have been released by PRI (Janick 2000).

By the early 1990’s, many varieties from other breeding programs were being released as

‘club’ varieties (ie. SweetTango™), and were unavailable to the majority of WA growers. Clubs

2 restrict who can grow the varieties and control the quality of the packed apples with the goal of maintaining a high grade standard and thus price in the market (Brown and Maloney, 2009).

Limited access to new varieities and the fact that no varieties were being bred and selected in the environmental conditions of the major production region in central Washington justified establishment of the WABP. With support from the Washington Tree Fruit Research

Commission (WTFRC), Washington State University (WSU) began an apple breeding program in 1994 (Evans 2013). ‘WA 2’ was the first release from this program in 2009 (Evans et al.

2010), followed by ‘WA 5’ in 2010 (Evans et al. 2011a). The program has just released its third variety, ‘WA 38’ (Cosmic Crisp™, 2013; Evans et al. 2012).

The aim of the WABP is to create a portfolio of new, improved unique cultivars that are adapted to the environment of central WA and accessible to WA growers (Evans 2013). The specific breeding targets are attractive appearance coupled with excellent eating quality, storability, regular cropping, high yield and resistance to sunburn, fire blight (Erwinia amylovora

(Burill.) and powdery mildew (Podosphaera leucotricha (Ell. and Ev.) Salm.). Storability

(section 1.3) and fire blight (section 1.4) are the focus of presented studies and thus additional information has been provided.

The WABP was the first apple breeding program to employ marker-assisted breeding

(MAB) for fruit quality, using both marker-assisted parent selection to guide crossing decisions and marker-assisted seedling selection (MASS) to cull seedlings with poor genetic potential. The first two genetic tests deployed were ethylene evolution genes (Zhu and Barritt 2008). The program currently has several tests focused primarily on fruit quality to use for these purposes

(Longhi et al. 2013; Verma 2014), and is part of a national effort to develop and deploy additional genetic markers (Guan 2013; Iezzoni et al. 2010; Ru et al. 2015).

Structure. The WABP is divided into four stages: seedling production, Phase 1, Phase 2, and Phase 3 (Evans 2013; Figure 1). Production of seedling trees takes 5 years, starting with pollen collection, flower emasculation and making planned crosses in the spring. Apples are harvested in the fall and seeds from those planned crosses are extracted. Seeds are stratified for no less than two months and germination begins in January of year two. Seedlings are grown in the greenhouse where they are culled for susceptibility to powdery mildew, and some are screened for resistance to fire blight or with DNA markers for fruit quality traits. Surviving seedlings are sent to Willow Drive Nursery (Ephrata, WA) and propagated onto dwarfing, precocious M.9 rootstocks. Budding onto dwarfing rootstocks shortens the juvenility period, allows for closer planting of the trees, reduces the need to prune and adds a measure of standardization to genetically diverse seedlings. Grafted seedlings are grown in the nursery until they are planted into Phase 1 (Evans 2013).

Phase 1 is the seedling selection phase where fruit is assessed for the first time. Grafted seedlings are planted at 0.3m in-row spacing, minimally pruned and treated with a standard spray program each year. Fruit collection typically begins in the second year in the field. Two members of the breeding team evaluate seedlings in the orchard for visual appeal and fruit eating quality. Ten fruit are harvested at starch level 3 on the 8-point Cornell starch chart (Blanpied and

Silsby 1992), stored for two months at 2°C and set out at room temperature (approximately

21°C) for one week prior to further assessment. Appearance, organoleptic eating quality, and several instrumental measures are recorded at that point (Table 1). At the end of year eight,

4 selected seedlings are propagated on G.41 rootstocks (Russo et al. 2007) for planting in Phase 2.

There are currently approximately 24,000 seedlings in Phase 1 trials (Evans 2013).

Propagating trees is a two year process; budwood is typically collected in late summer after bud swell, and either grafted or budded onto the desired rootstock. The grafted trees are grown by the nursery through the winter and the following year. Trees are dug in the second winter and planted in early spring as bare-rooted trees (Wertheim and Webster 2003). In the

WABP, this two year delay occurs when propagating seedlings for Phase 1, Phase 1 seedlings for

Phase 2, and Phase 2 selections for Phase 3 (Evans 2013). Along with juvenility, this is one of the reasons apple breeding is a 20+ year process.

Phase 2 is the data collection phase where clones of each seedling advanced from Phase 1 are planted in replicated trials. Five clones of each seedling are planted at each of three sites in the growing region and fruit is evaluated for no less than three years (Evans 2013). Further fruit quality and storage data (Table 1) are collected as well as yield efficiency and field performance.

Consumer taste panels are conducted as more fruit becomes available from the selections (Ross

2008). Decisions are made about selections to advance by the end of year thirteen from crossing.

There are currently 35 advanced selections planted in Phase 2 trials, with an additional 19 due to be planted in 2016. Phase 2 will be further discussed in section 1.5.

Phase 3 focuses on commercial viability of potential releases. At the point of submitting a selection to Phase 3, budwood is also sent to the Clean Plant Center (Prosser, WA) in order to be certified virus tested. Approximately 75 trees of each elite selection are planted at each of several grower orchard sites (Evans 2013). At this stage, growers manage the trees using their preferred system, with some guidance on crop load from the WABP. Harvest guidelines are

5 developed for each selection based on starch, soluble solid content, and color. Fruit is stored for three, four, six or ten months in RA (0.5°C), controlled atmosphere (CA; 1°C, 1% CO2, 2.5% O2) and CA + 1-methylcyclopropene (1-MCP) (Hanrahan et al. 2012). Data collected on 100-fruit samples after storage includes firmness, flavor, storage disorder incidence and greasiness.

Larger fruit samples are assessed for their ability to withstand a commercial packing line without damage. Trained sensory panelists also evaluate the selection compared to commercial cultivars and consumer evaluation trials are initiated when appropriate (Evans 2013; Ross 2008). Release decisisons are usually made by the end of year eighteen. There are currently five elite selections being evaluated in Phase 3 trials with one additional selection to be planted in 2017.

The WABP has been operational for 21 years, has released three varieties and has had one succession in lead breeder. The breeding program also serves as a wider resource to other researchers at WSU and the USDA-Agricultural Research Service (ARS) Wenatchee station (De

Kleine 2014; Dhingra et al. 2009; Dhingra et al. 2014; Guan 2013; Jung et al. 2014; Peace et al.

2013; Peace et al. 2014; Ru et al. 2015; Ross 2008; Verma 2014).

Storability

Consistent quality through long-term storage is one of the most important breeding targets of the WABP since the majority of the fruit produced in WA is stored for 3 to 12 months prior to commercial sale. Year-round supply of apples is made possible by cold storage, modified atmosphere storage and more recently ethylene blocking 1-MCP (Mattheis 2008) and has led to an extensive infrastructure of cold storage facilities in WA. Excellent storability is the ability of fruit to maintain or improve in flavor and texture while maintaining freedom from storage disorders and diseases. Variety, conditions during the growing season and maturity at harvest all

6 contribute to an apple’s success in long term cold storage (Ferguson et al. 1999; Hampson and

Kemp 2003; Watkins 2003). Important disorders include superficial scald, bitter pit, senescent breakdown, low temperature disorders (soft scald, internal browning, brown core), and modified atmosphere storage disorders (low O2 injury, CO2 injury) (Watkins, 2003). The two most commonly seen storage diseases, blue mold (Penicillium spp.) and grey mold (Botrytis cinerea;

Kim and Xiao 2008), are typically controlled with fungicide drenches prior to storage

(Rosenberger 1997; Watkins 2003). Economic loss due to these pathogens is important; however, few consumers ever see infected fruit as those would be culled after storage but prior to packing.

Moldy core is more difficult to manage as infection occurs during the floral stage, therefore post-harvest drenching is ineffective (Hickey 1990). Moldy cores can discourage consumers that slice their fruit from buying that variety again or worse, from buying apples in general. As moldy core is economically less significant and less prevalent than other diseases, there is less information about the characteristics that predispose a fruit for infection and relatively few varieties have been screened for susceptibility (Combrink et al. 1985; Silveira et al. 2013; Spotts 1999). Consequently, there is no information for breeders to utilize about the heritability of moldy core susceptibility or the fruit characteristics associated with susceptibility.

As such, breeders may cull selections unnecessarily due to assumptions about open calyxes increasing the likelihood of infection. Moldy core was deemed important enough to include in the RosBREED apple phenotyping protocol (Iezzoni et al. 2010; Evans et al. 2011b).

Resistance to Diseases

Genetic resistance to diseases is one of the most sustainable disease control options

(Brown 1975) and incorporating disease resistance is a major breeding goal for breeding programs around the world (Baumgartner et al. 2015; Broggini et al. 2013; Brown and Maloney

2003; Brown and Terry 1997; Bus et al. 2000; Kellerhals et al. 2011a). The three most important diseases—apple scab, powdery mildew and fire blight—have all been the focus of genetic resistance research.

Apple scab is a major disease, caused by Venturia inaequalis, for almost every production region in the world, but is rarely seen in WA due to the semi-arid climate. The most typical symptoms are brown lesions (‘scabs’) on leaves and deformed fruit with corky lesions

(Gessler et al. 2006). Apple scab resistance has been identified in diverse germplasm (Cova et al.

2015; Gessler et al. 2006; Crosby et al. 1990), but most breeding efforts have focused on the

Rvi6 (Vf) gene from Malus floribunda 821. One race of the pathogen has overcome the resistance conferred by Rvi6, highlighting the importance of both diversifying sources of resistance and pyramiding genetic resistance especially in a long-term perennial like apple (Baumgartner et al.

2015; Gessler et al. 2006; Grove et al. 2003).

Powdery mildew, caused by Podosphaera leucotricha, is a disease of the tree and fruit that can reduce both yield and fruit quality (Marine et al. 2010). Symptoms include defoliation, stunted growth and die-back of the vegetative tissues; infected fruit develop a russetted and discolored appearance (Marine et al. 2010). Incorporating powdery mildew resistance is a major apple breeding target (Baumgartner et al. 2015; Evans and James 2003; Janick et al. 1996;

Sestras et al. 2011), and with scab resistance, was the motivation for two multi-institute

European projects: ‘Durable Apple Resistance in Europe’ (DARE) (Lespinasse et al. 2000) and

‘High-quality Disease Resistant Apples for a Sustainable Agriculture’ (HiDRAS)

(Gianfranceschi and Soglio 2003). P. leucotricha populations can and have developed resistance to fungicides, but is usually adequately controlled with a diverse mode-of-action spray program

(Marine et al. 2010). As less susceptible or resistant cultivars are the most sustainable way of controlling mildew infection (Sestras et al. 2011), the WABP actively culls seedlings for powdery mildew susceptibility as seedlings following natural infection both in the greenhouse and in the nursery prior to planting in Phase 1 (Evans 2013).

Fire blight, caused by Erwinia amylovora, is also a destructive disease in most production regions; it has historically been less destructive in WA due to the reliance on older varieties (i.e.

‘Delicious’) which have moderate resistance, and environmental conditions less favorable for epidemic development. E. amylovora can infect all parts of the tree, but blossom, shoot and rootstock infections are the most common. During bloom, E. amylovora is spread by rain and pollinating insects (van der Zwet et al. 2012). Blossom blight kills the flower cluster, but leads to little long term damage if controlled, as blooms are typically thinned for crop load management anyway. Shoot blight and rootstock blight are much more problematic, as they can severely damage the structure of a tree or cause tree death (Norelli et al. 2003; van der Zwet et al. 2012).

Shoot blight can develop from blossom blight or can be caused by a wound to the shoot or leaves, such as with hail damage, which grants the epiphytic bacteria access. Rootstock blight is the most insidious, as the tree can appear asymptomatic until the fall (van der Zwet et al. 2012), and when symptoms appear they can be ambiguous, such as early leaf drop. Water-soaked lesions below the graft union are sometimes observed (van der Zwet et al. 2012). Damage to the tree is irreparable at this stage and the tree is dead in the spring.

The increase in plantings of more susceptible varieties grafted on susceptible rootstocks has led to an increase in fire blight epidemics (Norelli et al. 2003), including in WA. Control options for fire blight are less effective than those for apple scab or powdery mildew, and tree death is not uncommon. Antibiotics have been the most effective means of control since the

1950’s; however, the level of control achieved with antibiotics never exceeds approximately

80% and antibiotic resistant E. amylovora is an increasing threat (van der Zwet et al. 2012). The use of antibiotics in agriculture is contentious (McManus et al. 2002), and the use of

Streptomycin in organic orchards was revoked this year by the National Organic Program Board

(USDA 2015).

Genetic resistance is an especially preferred solution to address fire blight due to the low efficacy of current control methods, and considerable progress in identifying sources of resistance (Kellerhals et al. 2011b). Quantitative trait loci (QTL) have been identified for fire blight resistance from multiple sources: a linkage group (LG) 7 QTL from ‘Fiesta’ explains 35-

40% of phenotypic variation (Khan et al. 2006; 2007), a LG3 QTL from Malus robusta 5 explains 67-83% of phenotypic variation (Peil et al. 2007; 2008), a LG12 QTL from ‘Evereste’ explains 50-70% of phenotypic variation (Durel et al. 2009), a LG10 QTL from ‘Florina’ explains 15% of phenotypic variation (Le Roux et al. 2010), and a LG10 QTL from Malus fusca explains 60% of phenotypic variation (Emeriewen et al. 2014). The QTL conferring the most resistance are both from crabapples (‘Evereste’ and Malus × robusta 5) with small, highly astringent fruit. Using traditional breeding methods to incorporate resistance from these sources would likely require additional generations in the introgression process to achieve high fruit quality levels. Resistant rootstocks help prevent tree death and are an important component of

10 fire blight management (Norelli et al. 2003; Robinson et al. 2008). Introgressing fire blight resistance into rootstocks has been more successful (Norelli et al. 2003) due to the irrelevance of fruit quality.

Malus sieversii is the main progenitor species of domestic apple and is notable for its large fruit size and palatable flavor (Forsline et al. 2003). After M. sieversii was recognized as an underrepresented species in the Malus collection at USDA-ARS-Plant Genetic Resources Unit

(PGRU), Geneva, NY, the USDA-ARS National Plant Germplasm System (NPGS) sponsored several trips to multiple sites in Kazakhstan, Kyrgyzstan, Tajikistan, and Uzbekistan to collect

M. sieversii germplasm. More than 130,000 seeds from 892 M. sieversii trees were imported

(Forsline 2003). Of those, 1,200 seeds from 106 families were planted at the USDA-ARS-

PGRU in Geneva, NY (Forsline et al. 2008), 1,410 seeds from more than 32 families were planted in Minnesota (Forsline et al. 2008), and 963 seeds from 51 families were planted in New

Zealand (Luby et al. 2001). These plantings have been evaluated for several traits including their resistance to the natural occurrence of fire blight infection. Individuals from each family have shown resistance to fire blight (Forsline et al. 2008; Luby et al. 2001). Numerous seedlings were inoculated in a greenhouse at USDA-ARS-PGRU in Geneva, NY; 29% were resistant to fire blight (Momol et al. 1999). M. sieversii is an exceptional prospective source of fire blight resistance, as potentially fewer generations would be needed to reach commercially acceptable fruit size and quality.

Phase 2 Trial Design

The current WABP Phase 2 trial design has remained constant since the first Phase 2 trial was planted in 2004. Each trial is planted as a randomized complete block design. Each selection

11 is harvested in three consecutive weeks each year, over three years at three sites in the growing region. Standard cultivars are planted as controls and for standardizing comparisons between trial sites. The locations of the Phase 2 trials have changed, with plantings at a total of five locations in WA, and the three varieties used as standards have similarly changed, with a total of seven different varieties used over the course of trials. The data collected has changed slightly since 2004, and now consists of multiple appearance, flavor, texture and disorder traits (Table 1) collected at harvest, after short term storage (2 months at 2°C RA + 1 week at room temperature

(approximately 21°C) to mimic shelf-life) and after medium term storage (4 months at 2°C RA +

1 week at room temperature to mimic shelf-life). In 2010, the breeding team began investigating ways in which the data being collected in Phase 2 could be more fully utilized to inform selection decisions.

It is widely recognized that the best and most objective estimate of the genetic potential of an individual is the mean genetic potential predicted using all available information (Falconer and Mackay 1996). As part of several WTFRC-funded projects, software was developed that allows the WABP to routinely predict genetic potential of selections in Phase 2 (Peace et al.

2013; 2014). This software allows the WABP to calculate the predicted clonal value of fruit quality traits for candidates assessed in Phase 2, as well as outputs tests of normality, and ranks candidates for each trait. This information is now used regularly when considering advancement decisions.

In order to create the software, the program structure and the genetic architecture of the traits evaluated in Phase 2 was analyzed (Hardner at al. submitted). Data for 23 postharvest traits evaluated in Phase 2 trials collected from 2005 to 2012 was used to develop a model for the

12 analysis of individual trait observations across locations, seasons, harvests and storage duration and to estimate the variance components (Hardner et al. submitted). There were few significant interactions between genetic effects and the design elements (locations, seasons, harvests and storage durations); the study concluded that a less intensive assessment design could be used to predict candidate performance (Hardner et al. submitted).

Accuracy and Efficiency of Trial Design. Identifying selections superior to current cultivars is a function of the size of the selection population and the accuracy with which the available data predicts the genetic potential of a candidate selection (Falconer and Mackay

1996), leading the WABP to explore the feasibility of evaluating additional selections in Phase 2 trials. It is also anticipated that more seedlings in Phase 1 will be of sufficient quality to advance to Phase 2 as more traits are pre-selected using MASS (K. Evans, personal communication).

Increasing the selection intensity, by increasing the size of the selection population but advancing the same number to Phase 3, would increase the directional progress made in trait values from Phase 2 to Phase 3.

Candidate trait evaluations are approximations of the “true” mean candidate trait value.

Increasing replication can improve the accuracy of approximations of the true candidate trait value. Accurate prediction of genetic potential is achieved through replicated trials of potential selections in environments highly correlated with future commercial planting environments

(Cooper et al. 1993). However, maintaining replicated trials of apple candidates is expensive

(Weber 2008), and there are trade-offs between maximizing accuracy and minimizing cost to the program with the limited resources breeding programs face.

Methods have been developed in broad-acre crops and forestry species to evaluate the effect of changes in trial design on the accuracy of trait evaluations (Brennan et al. 1998; Cullis et al. 1996a, b; Patterson et al. 1977) and this type of analysis has led to design changes in expansive international breeding programs (Arief et al. 2015). These methods have never been evaluated in horticultural crops, such as tree fruits.

Dissertation Overview and Hypotheses

Each of the following chapters addresses a different aspect of the WSU apple breeding program, but all of them function under the umbrella hypothesis of improving efficiency or accuracy through quantification. The first two chapters hypothesize that new potential parents can be identified with additional information about susceptibility. The last chapter hypothesizes that the efficiency of advanced breeding trials could be improved without increasing the total program cost.

Chapter 2: Increasing knowledge about the susceptibility to moldy core of currently available germplasm will lead to more informed crossing decisions and ultimately, releases from the program that have less susceptibility.

Chapter 3: Screening Malus sieversii accessions for resistance to fire blight will lead to more informed crossing decisions and ultimately, releases from the program that have greater resistance.

Chapter 4: The accuracy with which traits are evaluated in advanced trials can be nominally reduced in a way that allows the program to evaluate more selections in those trials for the same total program cost, which increases the likelihood of releasing better apple varieties.

Literature Cited

Arief, V. N., DeLacy, I. H., Crossa, J., Payne, T., Singh, R., Braun, H-J., Tian, T., Basford, K. E., and Dieters, M. J. 2015. Evaluating testing strategies for plant breeding field trials: Redesigning a CIMMYT international wheat nursery. Crop Sci. 55:164-177.

Baumgartner, I. O., Patocchi, A., Frey, J. E., Peil, A., and Kellerhals, M. 2015. Breeding elite lines of apple carrying pyramided homozygous resistance genes against Apple Scab and resistance against Powdery Mildew and Fire Blight. Plant Mol. Biol. Rep. 33:1-11.

Blanpied, G. D., and Silsby, K. J. 1992. Predicting Harvest Date Windows for Apples. Cornell Coop. Ext. Bul. 221.

Brennan, J. P., Kingwell, R. S., Thomson, F. M., and Cullis, B. R. 1998. Optimal strategies for regional cultivar testing. Aust. J. of Agric. and Resource Economics 42:209-226.

Broggini, G. A. L., Durel, C. E. Vergne, E., Chevreau, E., Fahrentrapp, J., Vanblaere, T., Peil, A., Flachowsky, H., Hanke, M. V., Krens, F. A., Schouten, H. J., and Gessler, C. 2013. Cisgenic approach for improved disease resistance in apple. Acta Hortic. 974:117-121.

Brown, A. G. 1975. Apples. Pages 3-37 in: Advances in Fruit Breeding. J. Janick and J.N. Moore, eds. Purdue University Press, Indiana.

Brown, S. K., and Terry, D. E. 1997. The Cornell University apple breeding program: past, present, and future. Fruit Var. J. 51:199-204.

Brown, S. K., and Maloney, K. E. 2003. Genetic improvement of apple: breeding, markers, mapping and biotechnology. Pages 31-59 in: Apples: Botany Production and Uses. D. C. Ferree and I. J. Warrington, eds. CABI Publishing, Massachusetts.

Brown, S. K. and K. E. Maloney. 2009. Making sense of new apple varieties, trademarks and clubs: current status. New York Fruit Q. 17: 9-12.

Bus, V., Bradley, S., Hofstee, M., Alspach, P., Brewer, L., Luby, J., Geibel, M., Fischer, M., Combrink, J. C., Kotze, J. M., and Visagie, T. R. 1985. Colonization of apples by fungi causing core rot. Hort. Sci. 2:9-13.

Bus, V., Bradley, S., Hofstee, M., Alspach, P., Brewer, L., and Luby, J. 2000. Increasing genetic

diversity in apple breeding to improve the durability of pest and disease resistance. Acta Hortic. 538:185-190.

Combrink, J. C., Kotze, J. M., Wehner, F. C., and Grobbelaar, C. J. 1985. Fungi associated with core rot of Starking apples in South Africa. Phytophylactica 17:81-83.

Cooper, M., Delacy, I. H., Byth, D. E., and Woodruff, D. R. 1993. Predicting grain-yield in Australian environments using data from CIMMYT international wheat performance trials. 2. The application of classification to identify environmental relationships which exploit correlated response to selection. Field Crops Res. 32:323-342.

Cova, V., Bandara, N. L., Liang, W., Tartarini, S., Patocchi, A., Troggio, M.,Velasco, R., and Komjanc, M. 2015. Fine mapping of the Rvi5 (Vm) apple scab resistance locus in the ‘Murray’ apple genotype. Mol. Breed. 35:200.

Crosby, J. A., Janick, J., Pecknold, P. C., Korban, S. S., O'Connor, P. A., Ries, S. M., Goffreda, J., and Voordeckers, A. 1990. Breeding apples for scab resistance: 1945–1990. Acta Hortic. 317:43-70.

Cullis, B. R., Thomson, F. M., Fisher, J. A., Gilmour, A. R., and Thompson, R. 1996a. The analysis of the NSW wheat variety database. 1. Modelling trial error variance. Theor. Appl. Genet. 92:21-27.

Cullis, B. R., Thomson, F. M., Fisher, J. A., Gilmour, A. R., and Thompson, R. 1996b. The analysis of the NSW wheat variety database. 2. Variance component estimation. Theor. Appl. Genet. 92:28-39.

De Kline, M. 2014. Semi-automated end-effector concepts for localized removal and catching of fresh-market apples in fruiting wall orchards. PhD diss., Washington State University.

Dhingra, A., and Barritt, B. 2009. Establishing trait – gene relationships and gene discovery in apples. Washington Tree Fruit Res. Commission Res. Rep. Retrieved 15 October 2015 from http://jenny.tfrec.wsu.edu/wtfrc/.

Dhingra, A., and Evans, K. 2014. Reduction of generation cycle in apple breeding. Washington Tree Fruit Res. Commission Res. Rep. Retrieved 15 October 2015 from http://jenny.tfrec.wsu.edu/wtfrc/.

Durel, C-E., Denancé, C., and Brisset, M-N. 2009. Two distinct major QTL for resistance to fire

blight co-localize on linkage group 12 in apple genotypes 'Evereste' and Malus floribunda clone 821. Genome 52:139-147.

Emeriewen, O., Malnoy, M., Richter, K., Kilian, A., Hanke, M-V., and Peil, A. Evidence of a major QTL for fire blight resistance in the apple wild species Malus fusca. Acta Hortic. 1056:289-293.

Evans, K., and James, C. 2003. Identification of SCAR markers linked to Pl-w mildew resistance in apple. Theor. Appl. Genet. 106:1178-1183.

Evans, K. M., Barritt, B. H., Konishi, B. S., Dilley, M. A., Brutcher, L. J., and Peace, C. P. 2010. ‘WA 2’ Apple. HortScience 45:668-669.

Evans, K. M., Barritt, B. H., Konishi, B. S., Dilley, M. A., Brutcher, L. J., and Peace, C. P. 2011a. ‘WA 5’ Apple. HortScience 46:958-960.

Evans, K., Guan, Y., Luby, J., Clark, M., Schmitz, C., Brown, S., Orcheski, B., Peace, C., van de Weg, E., and Iezzoni, A. 2011b. Large-scale standardized phenotyping of apple in RosBREED. Acta Hortic. 945:233-238.

Evans, K. M., Barritt, B. H., Konishi, B. S., Brutcher, L. J. and Ross, C. F. 2012. ‘WA 38’ apple. HortScience 47:1177-1179.

Evans, K. 2013. Apple breeding in the Pacific Northwest. Acta Hort 976:75-78.

Falconer, D. S, and Mackay, T. F. C. 1996. Introduction to Quantitative Genetics, 4th ed. Pearson Education Limited. Harlow, England.

Ferguson, I., Volz, R., and Woolf, A. 1999. Preharvest factors affecting physiological disorders of fruit. Postharvest Biol. and Technol. 15:255-262.

Forsline, P. L., Aldwinckle, H. S., Hale, C., and Mitchell, R. 2002. Natural occurrence of fire blight in USDA apple germplasm collection after 10 years of observation. Acta Hortic. 590:351-358.

Forsline P. L., Aldwinckle, H.S., Dickson, E.E., and Hokanson, S.C. 2003. Collection, maintenance, characterization, and utilization of wild apples from central Asia. Hort Rev 29:1–61.

Forsline, P. L., Luby, J. J., and Aldwinckle, H. S. 2008. Fire blight incidence on Malus sieversii grown in New York and Minnesota. Acta Hortic. 793:345-350.

Garris, A. 2013. SnapDragon and RubyFrost are new apple varieties. Cornell Chronicle. Retrieved on 15 October 2015. http://news.cornell.edu/stories/2013/08/snapdragon-and- rubyfrost-are-new-apple-varieties.

Gasic, K., and Preece, J. E. Register of new fruit and nut cultivars list 47. HortScience 49:396- 421.

Gessler, C., Patocchi, A., Sansavini, S., Tartarini, S., and Gianfranceschi, L. 2006. Venturia inaequalis resistance in apple. Crit. Rev. Plant Sci. 25:473-503.

Gianfranceschi, L., and Soglio, V. 2003. The European project HiDRAS: innovative multidisciplinary approaches to breeding high quality disease resistant apples. Acta Hortic. 663:327-330.

Grove, G. G., Eastwell, K. C., Jones, A. L., and Sutton, T. B. 2003. Diseases of Apple. Pages 459-488 in: Apples: Botany Production and Uses. D.C. Ferree and I.J. Warrington, eds. CABI Publishing, Massachusetts.

Guan, Y. 2013. QTL identification and DNA marker development for individual sugars in apple. PhD diss., Washington State University.

Hampson, C. R., and Kemp, H. 2003. Characteristics of important commercial apple cultivars. Pages 61-89 in: Apples: botany, production and uses. D.C. Ferree and I.J. Warrington, eds. CABI Publishing, Massachusetts.

Hanrahan, I., Auvil, T., and Evans, K. 2012. WA 2 Update. Good Fruit Grower, June: 20.

Hardner, C., Evans, K., Peace, C., and Brien, C. Submitted. Genetic architecture of fruit quality traits in apple following short-term storage and implications for genetic breeding and selection. Tree Genet. Genomes.

Harker, F. R., Roger, F., Gunson, A., and Jaeger, S. R. 2003. The case for fruit quality: an interpretive review of consumer attitudes, and preferences for apples. Postharvest Biol. and Technol. 28:333-347.

Hickey, K. D. 1990. Moldy core and core rot. Pages 29-30 in: Compendium of apple and pear

diseases. H. S. Aldwinckle, ed. Amer. Phytopathol. Soc., Minnesota.

Iezzoni, A., Weebadde, C., Luby, J., Chengyan, Y., van de Weg, E., Fazio, G., Main, D., Peace, C. P., Bassil, N. V., and McFerson, J. 2010. RosBREED: Enabling marker-assisted breeding in Rosaceae. Acta Hortic. 859:389-394.

Janick, J. 2000. History of the PRI apple breeding program. Acta Hortic. 595:55-60.

Janick, J. 1996. Apples. Pages 1-78 in: Fruit breeding, tree and tropical fruits. Vol. 1. J. Janick and J. N. Moore, eds. John Wiley & Sons, New Jersey.

Jung, S., Ficklin, S. P., Lee, T., Cheng, C.H., Blenda, A., Zheng, P., Yu, J., Bombarely, A., Cho, I., Ru, S., Evans, K., Peace, C., Abbott, A. G., Mueller, L. A., Olmstead, M. A., and Main, D. 2014. The genome database for rosaceae (GDR): year 10 update. Nucleic Acids Res. 42:1237-1244.

Kellerhals, M., Baumgartner, I.O., Leumann, L., Frey, J. E., and Patocchi, A. 2011a. Progress in pyramiding disease resistances in apple breeding. Acta Hort 976:487-491.

Kellerhals, M., L. Franck, I.O. Baumgartner, A. Patocchi, and J.E. Frey. 2011b. Breeding for Fire Blight resistance in apple. Acta Hortic. 896:385-389.

Khan, M.A., Duffy, B., Gessler, C., and Patocchi, A. 2006. QTL mapping of fire blight resistance in apple. Mol. Breed. 17: 299-306.

Khan, M. A., Durel, C., Duffy, B., Drouet, D., Kellerhals, M., Gessler, C., and Patocchi, A. 2007. Development of molecular markers linked to the 'Fiesta' linkage group 7 major QTL for fire blight resistance and their application for marker-assisted selection. Genome 50:568-577.

Kim, Y. K., and Xiao, C. L. 2008. Distribution and incidence of Sphaeropsis rot in apple in Washington State. Plant Dis. 92:940-946.

Lespinasse, Y., Durel, C. E., Laurens, F., Parisi, L., Chevalier, M., Pinet, C., Geibel, M., Fischer, M., and Fischer, C. 2000. A European project: DARE-durable apple resistance in Europe (FAIR5 CT97-3898) durable resistance of apple to scab and powdery-mildew: one step more towards an environmental friendly orchard. Acta Hortic. 538:197-200.

Le Roux, P-M. F., Khan, M. A., Broggini, G. A. L., Duffy, B., Gessler, C., and Patocchi, A.

2010. Mapping of quantitative trait loci for fire blight resistance in the apple cultivars 'Florina' and 'Nova Easygro'. Genome 53:710-722.

Longhi, S., Hamblin, M. T., Trainotti, L., Peace, C. P.,Velasco, R., and Costa, F. 2013. A candidate gene based approach validates Md-PG1 as the main responsible for a QTL impacting fruit texture in apple (Malus x domestica Borkh). BMC Plant Biol. 13:37.

Luby, J. J. 1991. Breeding cold-hardy fruit crops in Minnesota. HortScience 26:507-512.

Luby, J., Forsline, P., Aldwinckle, H., Bus, V., and Geibel. M. 2001. Silk road apples— collection, evaluation, and utilization of Malus sieversii from Central Asia. HortScience 36:225-231.

Luby, J.J. 2003. Taxonomic Classification and Brief History. Pages 1-14 in: Apples: Botany Production and Uses. D.C. Ferree and I.J. Warrington, eds. CABI Publishing, Massachusetts.

Lyons, R. 2014. International. Washington Apples. Retrieved on 17 October 2015. http://bestapples.com /international/index.aspx

Marine, S. C., Yoder, K. S., and Baudoin, A. 2010. Powdery mildew of apple. The Plant Health Instructor. Online publication. DOI:10.1094/PHI-I-2010-1021-01

Mattheis, J. P. 2008. How 1-methylcyclopropene has altered the Washington State apple industry. HortScience 43:99-101.

McManus, P. S., Stockwell, V. O., Sundin, G. W., and Jones, A. L. 2002. Antibiotic use in plant agriculture. Annu. Rev. Phytopathol. 40:443-465.

Momol, M. T., Aldwinckle, H. S., Forsline, P. L., and Lamboy, W. F. 1999. Fire blight resistance and horticultural evaluation of wild Malus populations from Central Asia. Acta Hortic. 489:229-234.

Morgan J., and Richards, A. 2003. The new book of apples. Ebury Press, London.

Mudge, K., Janick, J., Scofield, S., and Goldschmidt, E. E. 2009. A History of Grafting. Hortic. Rev. 35:437-493.

Norelli, J. L., Jones, A. L., and Aldwinckle, H. S. 2003. Fire blight management in the twenty-

first century. Plant Dis. 87:756-765.

O’Rourke, D. 2003. World Production, Trade, consumption and Economic Outlook for Apples. Pages 15-30 in: Apples: Botany Production and Uses. D.C. Ferree and I.J. Warrington, eds. CABI Publishing, Massachusetts.

Patterson, H. D., Silvey, V. Talbot, M. and Weatherup, S. T. C. 1977. Variability of yields of cereal varieties in UK trials. J. Agric. Sci. 89:239-245.

Peace, C. P., Hardner, C., Evans, K., and Bink, M. 2013. Increasing decision confidence in cultivar development and adoption. Washington Tree Fruit Res. Commission Res. Rep. Retrieved 15 October 2015 from http://jenny.tfrec.wsu.edu/wtfrc/.

Peace, C. P., Hardner, C., and Evans, K. 2014. Support systems to deliver elite new cultivars for extended storage. Washington Tree Fruit Res. Commission Res. Rep. Retrieved 15 October 2015 from http://jenny.tfrec.wsu.edu/wtfrc/.

Peil, A., Garcia-Libreros, T., Richter, K., Trognitz, F.C., Trognitz, B., Hanke, M.V. and Flachowsky, H. 2007. Strong evidence for a fire blight resistance gene of Malus robusta located on linkage group 3. Plant Breed. 126:470-475.

Peil, A., Hanke, M. V., Flachowsky, H., Garcia-Libreros, T., Horner, M., Bus, V., Richter, K., Celton, J. M., and Gardiner, S. 2008. Confirmation of the fire blight QTL of Malus x robusta 5 on linkage group 3. Acta Hortic. 793:297-303.

Robinson, T. 2008. The evolution towards more competitive apple orchard systems in the USA. Acta Hortic. 772:491-500.

Robinson, T. L., Hoying, S. A., and Fazio, G. 2008. Performance of Geneva® rootstocks in on- farm trials in New York state. Acta Hortic. 903:249-255.

Robinson, T. L., Hoying, S.A., Sazo, M. M., Dominguez, L.I., and J.C. Fachinello, J. C. 2014. Yield, Fruit quality and mechanization of the tall spindle apple production system. Acta Hortic. 1058:95-104.

Rosenberger, D.A. 1997. Recent research and changing options for controlling postharvest decays of apples. In Proc. Harvesting, Handling, and Storage Workshop. Northeast Reg. Agric. Eng. Serv. Publ. NRAES-112, 14.

Ross, C. 2008. Sensory profiles and consumer acceptance of apple breeding selections. Washington Tree Fruit Res. Commission Res. Rep. Retrieved 15 October 2015 from http://jenny.tfrec.wsu.edu/wtfrc/.

Ru, S., Main, D., Evans, K., and Peace, C. 2015. Current applications, challenges, and perspectives of marker-assisted seedling selection in Rosaceae tree fruit breeding. Tree Genet. Genomes 11:1-12.

Russo, N. L., Robinson, T. L., Fazio, G., and Aldwinckle, H. S. 2007. Field evaluation of 64 apple rootstocks for orchard performance and fire blight resistance. HortScience 42:1517- 1525.

Sestras, A. F., Pamfil, D., Dan, C., Bolboaca, S. D., Jäntschi, L., and Sestras, R. E. 2011. Possibilities to improve apple scab (Venturia inaequalis (Cke.) Wint.) and powdery mildew [Podosphaera leucotricha (Ell. et Everh.) Salm.] resistance on apple by increasing genetic diversity using potentials of wild species. Aust. J. Crop Sci. 5:748- 755.

Silveira, F. N., Kretzschmar, A. A., Rufato, L., Bogo, A., and Fioravanco, J. C. 2013. Relationship between fruit morphological chararcterisitcs and incidence of moldy core in ‘Gala’ and ‘Fuji’ apple clones on different rootstoocks. Rev. Bras. Frutic. 35:75-85.

Spotts, R. A., Cervantes, L. A., and Mielke, E. A. 1999. Variability in postharvest decay among apple cultivars. Plant Dis. 83:1051-1054.

Tustin, D. S. 2014. Future Orchard Planting Systems- Do we need another revolution?. Acta Hortic. 1058:27-36.

USDA, National Organic Program (NOP). 2015. Sunset 2015 Amendments to the National List. Agricultural Marketing Service. 80: 146. Online publication. http://www.regulations.gov/#!documentDetail;D=AMS-NOP-15-0015-0001.

USDA, National Agricultural Statistics Service (NASS). 2015. National Statistics for Apple. Retrieved 16 October 2015. Washington, DC. van der Zwet, T., Orolaza-Halbrendt N., and Zeller, W. 2012. Fire blight: history, biology, and management. Amer. Phytopath. Soc. Press, Minnesota.

Verma, S. 2014. A predictive genetic knowledge for apple. PhD diss., Washington State

University.

Watkins, C. B. 2003. Principles and Practices of Postharvest Handling. Pages 585-614 in: Apples: Botany Production and Uses. D.C. Ferree and I.J. Warrington, eds. CABI Publishing, Massachusetts.

Weber, M. 2008. Sustainable apple breeding needs sustainable marketing and management. Proceedings from the 13th International Conference on Cultivation Technique and Phytopathological Problems in Organic Fruit-Growing, 232-237.

Webster, A. D., and Wertheim, S. J. 2003. Apple Rootstocks. Pages 91-124 in: Apples: Botany Production and Uses. D.C. Ferree and I.J. Warrington, eds. CABI Publishing, Massachusetts.

Wertheim, S. J. and A. D. Webster. 2003. Propagation and Nursery Tree Quality. Pages 125-152 in: Apples: Botany Production and Uses. D.C. Ferree and I.J. Warrington, eds. CABI Publishing, Massachusetts.

Zhu, Y., and Barritt. 2008, B. H. Md-ACS1 and Md-ACO1 genotyping of apple (Malus x domestica Borkh.) breeding parents and suitability for marker-assisted selection. Tree Genet. Genomes 4:555-562.

Figure 1: Diagram of the timeline and activities of the WABP

Table 1. Traits evaluated by the WSU Apple Breeding Program in Phases 1 and 2. Evaluation at

2 months is after 8 weeks in 2°C storage + 1 week at room temperature. Evaluation at 4 months is after 16 weeks in 2°C storage + 1 week at room temperature.

Trait Scale Type Phase 1 Phase 2 Harvest 2 months Harvest 2 months 4 months Starch Ordinal, 1-8 Visual   Size Ordinal, 1- 5 Visual   Shape Ordinal, 1- 5 Visual   Color Hue Ordinal, 1- 10 Visual   Color 6 Descriptors Visual   Adjective Red color % Ordinal, 1- 5 Visual   Red color type Ordinal, 1- 3 Visual   Ground color Ordinal, 1- 3 Visual   Russett Ordinal, 1- 5 Visual   Lenticels Ordinal, 1- 5 Visual   Hardness Ordinal, 1- 5 Sensory   Crispness Ordinal, 1- 5 Sensory   Juciness Ordinal, 1- 5 Sensory   Sweet Ordinal, 1- 5 Sensory   Acid Ordinal, 1- 5 Sensory   Aromatic Ordinal, 1- 5 Sensory   Appearance Ordinal, 1- 5 Visual    Summary Eating Quality Ordinal, 1- 10 Sensory    Summary Overall Ordinal, 1- 3 Sensory    Browning Ordinal, 1- 3 Visual    Bitterpit Ordinal, 1- 5 Visual    Soluble Solids Continuous, Instrumental   °Brix Firmness Continuous, Instrumental     lbs Crispness Continuous Instrumental     Diameter Continuous, Instrumental     cm Weight Continuous, g Instrumental     Titratable Continuous Instrumental     Acidity pH Continuous Instrumental    

CHAPTER 2: Survey of Moldy Core Incidence in Germplasm from Three U.S. Apple

Breeding Programs

Julia M. Harshman1, Kate Evans1

1Department of Horticulture, Washington State University Tree Fruit Research and Extension

Center, 1100 North Western Avenue, Wenatchee, WA 98801

This chapter has been published: Harshman, Julia M., and Kate Evans. 2015. Survey of Moldy Core Incidence in Germplasm from the Three US Apple Breeding Programs. Journal of the American Pomological Society 69(1): 51-57.

Abstract

Moldy core, mainly caused by Alternaria spp., in apples has been studied in a limited number of cultivars, and susceptibility is attributed to open sinuses and calyxes in the fruit. In three U.S. apple breeding programs, a diverse germplasm collection was characterized for core opening, calyx opening, and moldy core incidence at several time points during storage. Ten cultivars showed signs of moldy core, all had open cores while only three had open calyxes. Fruit with either an open core or an open calyx increased the likelihood of moldy core incidence. Two susceptible cultivars, ‘Gingergold’ and ‘Pinova’, also had progeny with high incidence of moldy core. A separate project screened 707 seedlings with diverse parentage in the Washington State

University apple breeding program for core opening, calyx opening, sinus opening and moldy core incidence. Only four of the seedlings had open sinuses, and all failed to develop moldy core.

The cultivar survey information presented here may be useful to other breeders, horticulturists, and pathologists interested in determining the heritability for moldy core susceptibility.

Introduction

Moldy core (mouldy core, dry core rot) is caused primarily by Alternaria spp., but has also been reported as a complex consisting of Stemphylium spp., Cladosporium spp., Ulocladim spp., Epicoccum spp., Coniothyrium spp. and Pleospora herbarum (Pers.) Rabenh., with

Alternaria alternata (Fr.) Keissl. as the dominant species (Combrink et al. 1985a; Hickey 1990;

Niem et al. 2007). Only a few commercial apple (Malus × domestica Borkh.) cultivars including

‘Fuji’, ‘Red Delicious’ and sports of ‘Red Delicious’ such as ‘Starking’ are reported to be affected; their open sinus is thought to be responsible as a port on entry for inoculum (Combrink et al. 1985a; Miller 1959). Combrink et al. (1985b) found 38-66% of ‘Red Delicious’ fruit cut at harvest had moldy core. It has recently been reported that A. alternata and A. tenuissima (Kunze)

Wiltshire can also cause external lesions on stored ‘Nittany’ apple fruit that manifests during storage (Jurick et al. 2014; Kou et al. 2013). Research on Alternaria spp., Stemphylium spp. and

Cladosporium spp. has shown that their spores are ubiquitous in temperate zones, with concentrations affected by diurnal fluctuations, the time of year and nearby sporulating plants

(Bergamini et al. 2004; Rodriguez-Rajo et al. 2005; Sreeramulu 1959).

The disease cycle starts soon after blossoms open. Spores germinate and mycelia grow down the open style/ sinus to the developing core where colonization of the locular region occurs. Sensitivity to colonization drops significantly after petal fall (Reuveni et al. 2002).

During fruit development, as cells divide and enlarge and fruit tissue swells, the style/sinus opening closes. An open sinus could be a characteristic of the cultivar, as in the case of ‘Red

Delicious’ (Combrink et al. 1985a), or could be caused by rapid, irregular growth or when early season dry weather is followed by heavy rain (Hickey 1990). Warmer weather during early fruit development has also been indicated in increased susceptibility, as well as reducing the length of

27 the fruit (Sugar 2002). Mycelial growth is generally restricted to the locules and carpellate region, but can develop into dry rot if it colonizes the surrounding mesoderm (Hickey 1990).

Wet core rot is distinct from moldy core/dry core rot in that it is a postharvest disease caused mainly by Penicillium expansum Link, other Penicillium spp. and Mucor piriformis Scop. (Li et al. 2014; Peter et al. 2012; van der Walt et al. 2010; Warner 2006).

While environmental conditions and inoculum density likely contribute to moldy core incidence, host physiology is the main driver of susceptibility during bloom (Combrink et al.

1984; Shtienberg 2012). Larger calyx openings, smaller length to diameter ratio and larger fruit have all been positively correlated to moldy core susceptibility in ‘Gala’ and ‘Fuji’ (Silveira et al. 2013). In ‘Red Delicious’ and ‘Golden Delicious’, sinus opening is inversely related to the length-diameter ratio of fruit (Spotts 1990). Open calyx and open core cultivars are considered more susceptible to moldy core than cultivars with a closed calyx, though research has focused mainly on ‘Red Delicious’ (Combrink et al. 1985a, b; Hickey 1990). The heritability of this trait has been difficult to discern as studies have focused on only a few commercial cultivars. While symptoms typically do not affect the edible portions, it is off-putting to consumers that slice their apples to find mold in the center of the fruit. There are currently no postharvest fungicides labelled for use on pome fruit to control moldy core, and this coupled with the fact that

Alternaria spp. also produce mycotoxins such as alternariol and alternariol monomthyl ether, make this an important disease to combat (Biggs 1994; Scott 2012; Warner 2006). At least one cultivar has been released and largely rejected by industry due to its susceptibility to moldy core

(Janick et al. 2000). Most apple breeding programs should or currently select against moldy core susceptible cultivars.

In this survey, data were collected over three years in pursuit of an overarching trend for moldy core incidence in two large sets of genetically diverse individuals from the USDA-

Specialty Crops Research Initiative project, RosBREED (Iezzoni et al. 2010) and the

Washington State University apple breeding program (WABP; Evans, 2013).

Materials and Methods

Plant Materials. The RosBREED apple Crop Reference Set included 154 cultivars and elite selections as well as 313 seedlings. Subsets of these individuals were grown at the Cornell

University (CU) New York State Agricultural Experiment Station in Geneva, NY; at the

University of Minnesota (UMN) Horticultural Research Center near Chaska, MN; and at the

Washington State University (WSU) Tree Fruit Research & Extension Center Sunrise Orchard in

Rock Island, WA. The RosBREED material grown at CU were on their own roots and planted in single rows at approximately 1 m within-row and 5.8 m between-row spacing (3’ x 19’).

Seedlings grown at UMN were grafted on B9 rootstocks, were in double rows at 0.6 m x 0.6 m within-row (2’ x 2’) and 3.7 m (12’) between-row spacing. The cultivars and elite selections at

UMN were a mix of older accessions and test plantings grafted on M26, with most planted at approximately 3 m within-row and 4.8 m between-row spacing (10’ x 16’). The RosBREED material grown at WSU were grafted on M9-337 and planted at 1.8 m within-row and 3.7 m between-row spacing (6’ x 12’). There was little overlap of cultivars in the population structure between sites (Peace et al. 2014); however, in 2010 ‘Silken’ was evaluated at WSU and UMN, and seven cultivars in 2011 were available for evaluation at more than one location (CU/UMN:

‘Autumn Crisp’; CU/WSU: ‘Empire’, ‘Fuji’; UMN/WSU: ‘Arlet’, ‘Ginger Gold’, ‘Zestar’;

CU/UMN/WSU: ‘Honeycrisp’).The WABP seedling set comprised of 29 families derived from crosses using the following parents: ‘Ambrosia’, ‘Aurora Golden Gala’, ‘Braeburn’, ‘Chinook’,

‘CrimsonCrisp’, ‘Cripps Pink’, ‘Delblush’, ‘GoldRush’, ‘Honeycrisp’, ‘Huaguan’, ‘Pinova’,

‘Sciros’, ‘Splendour’ and several WSU breeding selections. Seedlings were vegetatively propagated on to M 9 rootstocks and planted approximately 0.3 m (12”) apart on a three-wire trellis in groups by family. The 707 seedlings were randomly selected from progeny of the 29 families. At the time of fruit evaluation in 2013, seedlings were in their third or fourth leaf, depending on the family.

Fruit Evaluation. Fruit maturity of the RosBREED individuals was monitored weekly and a 25-fruit sample was harvested when the starch index was between three and four (Blanpied and Silsby, 1992). Five fruits were evaluated at three time points: harvest, 10 weeks 2°C regular atmosphere (RA) storage with one week shelf-life at room temperature (approx. 20°C), and 20 weeks in RA storage with one week shelf-life at room temperature. Calyx openings were evaluated in five fruit samples at harvest only and scored as “closed” when the calyxes of all five fruits were closed, “mixed” when the calyxes were a mix of open and closed, and “open” when the calyxes of all five fruits were open. Core opening was also evaluated at harvest only as open or closed. A core was considered closed when there was no break in the endocarp, thus allowing no transfer from one locule to another. Moldy core was evaluated at all three time points. Moldy core was considered present if mycelium could be seen in the core region of any of the five fruits without the aid of a dissecting microscope. Moldy core was scored as absent if there was no visible infection. Fruit evaluation was based on the RosBREED protocol for apple (Evans et al.,

2012). Photographs of the scoring guides for calyx opening, core opening and moldy core can be found at http://www.rosbreed.org/sites/default/files/files/RosBREED_2010Phenotyping_ protocol_Malus.pdf

Twenty fruit were harvested from the WABP seedlings in 2013 at a starch index between three and five. Fruit evaluations occurred at four time points: harvest, 8 weeks 2°C RA storage with one week at room temperature shelf-life, 16 weeks 2°C RA storage with one week at room temperature shelf-life, and 24 weeks in 2°C RA storage. At harvest, all 20 fruit were evaluated visually for calyx opening. At 8 weeks, subsets of five fruits were evaluated using the

RosBREED protocol for moldy core. At 16 weeks, subsets of five fruits were examined for moldy core, core opening and sinus opening. To evaluate sinus opening, each fruit was cut transversely approximately 2 cm from the calyx end and scored either as open or closed. At 24 weeks, subsets of five fruits were scored only for moldy core. As some fruit succumbed to storage rots over the course of the experiment, not all seedlings could be scored for the traits of sinus opening, core opening and moldy core incidence.

Statistical Analysis. Due to the low incidence of moldy core in both sets of data, traditional statistical methods (chi-square, zero-inflated Poisson regression, data transformation, etc.) were not robust. Hence, the results presented here represent observations drawn from the data without the aid of statistical analysis. RosBREED data is presented as a combination of data taken in 2010 and 2011. The WABP seedling data is from 2013.

Results

The number of individuals within the RosBREED germplasm with open cores was similar to the number with closed cores (Table 1). Closed calyxes, however, were more numerous than either open calyxes or mixed calyxes. Incidence of moldy core was highest in those apples with open cores or open calyxes both at harvest and after 10 weeks of storage. At harvest, open core/open calyx individuals had the highest incidence of moldy core but after ten

31 weeks in storage, the open core/open calyx individuals showed a similar level of incidence to the open core/closed calyx and closed core/open calyx individuals (Table 1). When the data is separated by sites, UMN and CU had similar trends, with open core/open calyx individuals having the highest incidence while open core/closed calyx individuals had the highest incidence of moldy core in WSU (data not shown). Differences between sites are likely due to the different germplasm evaluated at each site, availability of fungal inoculum, and the different weather conditions, particularly differences in humidity and precipitation.

Of the approximately 70 cultivars evaluated, ten showed symptoms of moldy core:

‘Fortune’, ‘Ginger Gold’, ‘Honeygold’, ‘Hudson’s Golden Gem’, NJ90, ‘Pinova’, ‘Sawa’,

‘Sonya’, ‘SunCrisp’ and ‘Zestar’ (Table 2). All but ‘Ginger Gold’ had open cores; ‘Ginger

Gold’ had an open core at UMN and a closed core at WSU. The calyxes of ‘Fortune’, ‘Ginger

Gold’, ‘Honeygold’, NJ90, ‘Pinova’, ‘Sawa’, and ‘Zestar’ were closed while ‘Hudson’s Golden

Gem’, and ‘SunCrisp’ had open calyxes. ‘Fuji’ and ‘Red Delicious’, two cultivars known for their susceptibility, did not develop moldy core in the two years of evaluation. ‘Red Delicious’ was evaluated only at WSU and ‘Fuji’ was evaluated at WSU and CU (Table 2). This could be a result of lack of viable pathogen and/or unfavourable weather conditions.

In the 707 WABP seedlings, there were more open core individuals (88%) than closed core individuals (12%). There were also more open calyx individuals (59%) than either mixed

(22%) or closed (19%) calyx individuals. There was extremely low incidence of moldy core in

2013, with only 49 of the total 707 individuals (6.9%) showing moldy core. Only four selections had open sinuses out of those evaluated, all of which had open cores and open calyxes. None of those selections had moldy core at any of the evaluation points which could again be attributed to inoculum load or environmental conditions.

Few of the cultivars shown to be susceptible to moldy core had been used as parents in the families screened, however there were three families derived from ‘Pinova’. In the two New

York families with ‘Pinova’ as the female parent (NY-A, NY-F), several of the progeny showed susceptibility to moldy core with the highest incidence being 17 of the 24 NY-F seedlings in

2010 (Table 3). Three of the 22 WABP ‘Cripps Pink’ × ‘Pinova’ seedlings had moldy core.

Although interesting to breeders, no conclusions can be drawn regarding heritability due to the small sample size in each case.

A second susceptible parent, ‘Ginger Gold’, was used at both CU and UMN with different results. In NY-D with ‘Ginger Gold’ as the pollen parent, ten progeny in 2010 and eight progeny in 2011 showed susceptibility to moldy core. In two UMN families with ‘Ginger Gold’ as the female parent (MN-G, MN-H), no susceptibility was seen in 2010 and only one progeny from each family showed susceptibility in 2011. This seeming discrepancy could be the result of the effect of the other parent - ‘Braeburn’ at CU and two seedlings at UMN - or the effect of the environment, or inoculum load. A study using controlled inoculation would be required to better predict the cause of the discrepancies or the heritability of moldy core susceptibility in these families.

The RosBREED germplasm was structured to include a large number of individuals but only a small number of fruit sampled for each individual, primarily for discovery of quantitative trait loci associated fruit quality traits (Peace et al. 2014). With the low incidence of moldy core in this fruit, statistically sound conclusions were not possible. However, as general observations, the data have value to breeders and pathologists to confirm and dispel some previous conclusions regarding fruit morphology and disease incidence. Open core/open calyx individuals from diverse backgrounds are more susceptible to moldy core than other core/calyx combinations;

33 however, an open core is more likely to lead to moldy core incidence than an open calyx. Larger diameter fruit, which are often the individuals targeted in breeding programs, are more likely to have open cores than smaller diameter fruit. Open sinuses in ‘Red Delicious’ and its sports have been attributed for the relatively high incidence of moldy core in this cultivar, however open sinuses are relatively rare, certainly in the germplasm within this study. A more definitive way of determining whether the calyx is open or closed, such as that used by Spotts et al. (1999) where air pressure is applied to the flesh side of a transversely cut apple submerged in water while the calyx end is monitored for air bubbles, could be used in future studies. In order to further explore this concept, it would be interesting to examine the sinuses of the cultivars found to be susceptible at different developmental stages of the fruit to determine whether the sinus is particularly slow to close in these cultivars compared to those that show no symptoms.

Author Contributions

JH and KE designed the study. JH collected and interpreted the data. JH drafted the majority of the manuscript. Both authors read and approved of the final document.

Acknowledgements

This project was supported by funds from the Specialty Crop Research Initiative

Competitive Grant 2009-51181-05808 of the USDA’s National Institute of Food and

Agriculture, and the Washington Tree Fruit Research Commission Grant AP-13-106.

The authors wish to thank Nancy Buchanan for help with harvest, fruit processing and training in postharvest evaluation of fruit. We wish to also thank Lisa Brutcher for help with fruit processing, Paul Sandefur for help with harvest and fruit processing, and the RosBREED

34 apple team, Yinghzu Guan (WSU), Jim Luby, Cari Schmitz, Matt Clark (UMN), Susan Brown and Ben Orcheski (CU) for fruit harvest and evaluation.

Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by Washington State University.

Literature Cited

Bergamini, B. M., Grillenzoni, S., Andreoni, A. D., Natali, P., Ranzi, A., and Bertolani, M. F. 2004. Alternaria spores at different heights from the ground. Allergy 59:746-752.

Biggs, A.R. 1994. Mycelial growth, sporulation, and virulence to apple fruit of Alternaria alternata isolates resistant to iprodione. Plant Dis. 78:732-735.

Blanpied, G., and Silsby, K. 1992. Predicting harvest date windows for apples. Cornell Coop. Ext. Bul. 221.

Combrink, J., Visagie, T., and Grobbelaar, C. 1984. Variation in the incidence and occurence in different areas of core rot of Starking apples. Deciduous Fruit Grower:88-89.

Combrink, J. C., Kotze, J. M., and Visagie, T. R. 1985a. Colonization of apples by fungi causing core rot. Hort. Sci. 2:9-13.

Combrink, J. C., Kotze, J. M., Wehner, F. C., and Grobbelaar, C. J. 1985b. Fungi associated with core rot of Starking apples in South Africa. Phytophylactica 17:81-83.

Evans, K. 2013. Apple breeding in the pacific northwest. Acta Hortic. 976:75-78.

Evans, K., Guan, Y., Luby, J., Clark, M., Schmitz, C., Brown, S., Orcheski, B., Peace, C., van de Weg, E., and Iezzoni. A. 2012. Large-scale standardized phenotyping of apple in RosBREED. Acta Hortic. 945: 233-238.

Hickey, K. D. 1990. Moldy core and core rot. Pages 29-30. in: Compendium of apple and pear diseases, H.S. Aldwinckle (ed.). Amer. Phytopathol. Soc., Minnesota.

Iezzoni, A., Weebadde, C., Luby, J., Yue, C. Y., Peace, C. P., Bassil, N., and McFerson, J. 2010. RosBREED: Enabling marker-assisted breeding in Rosaceae. Acta Hortic. 859: 389-394.

Janick, J., Goffreda, J. C., and Korban, S. S. 2000. ‘Coop 25’ (Scarlet O’Hara™) apple. HortScience 35:150-151.

Jurick, W. M., Kou, L. P., Gaskins, V. L., and Luo, Y. G. 2014. First report of Alternaria alternata causing postharvest decay on apple fruit during cold storage in Pennsylvania. Plant Dis. 98:690-690.

Kou, L. P., Gaskins, V. L., Luo, Y. G., and Jurick, W. M. 2014. First Report of Alternaria tenuissima causing postharvest decay on apple fruit from cold storage in the United States. Plant Dis. 98:690-690.

Li, J., Gaskins, V. L., Yan, H. J., Luo, Y. G., and Jurick, W. M. 2014. First report of mucor rot on stored ‘Gala’apple fruit caused by Mucor piriformis in Pennsylvania. Plant Dis. 98: 1157.

Miller, P. 1959. Open calyx tubes as a factor contributing to carpel discoloration and decay of apples. Phytopathology 49:520-523.

Niem, J., Miyara, I., Ettedgui, Y., Reuveni, M., Flaishman, M., and Prusky, D. 2007. Core rot development in Red Delicious apples is affected by susceptibility of the seed locule to Alternaria alternata colonization. Phytopathology 97:1415-1421.

Peace, C. P., Luby, J. J., van de Weg, W. E., Bink, M. C. A. M., and Iezzoni, A. F. 2014. A strategy for developing representative germplasm sets for systematic QTL validation, demonstrated for apple, peach, and sweet cherry. Tree Genet. Genomes 10:1679-1694.

Peter, K. A., Vico, I., Gaskins, V. L., Janisiewicz, W. J., and Jurick, W. M. 2012. Identification and characterization of two new Pencillium species causing blue mold of stored apple fruit in the United States. Phytopathology 102:92-92.

Reuveni, M., Ben-Arie, R., Prusky, D., Sheglov, D., and Sheglov, N. 2002. Sensitivity of Red Delicious apple fruit at various phenologic stages to infection by Alternaria alternata and moldy-core control. Eur. J. of Plant Pathol. 108:421-427.

Rodriguez-Rajo, F.J., Jato, V., and Iglesias, I. 2005. Variation assessment of airborne Alternaria and Cladosporium spores at different bioclimatical conditions. Mycol. Res. 109:497-507.

Scott, P. M. 2012. Recent research on fumonisins: a review. Food Addit. Contam. 29:242-248.

Shtienberg, D. 2012. Effects of host physiology on the development of core rot, caused by Alternaria alternata, in Red Delicious apples. Phytopathology 102:769-778.

Silveira, F. N., Kretzschmar, A. A., Rufato, L., Bogo, A., and Fioravanco, J.C. 2013. Relationship between fruit morphological chararcterisitcs and incidence of moldy core in ‘Gala’ and ‘Fuji’ apple clones on different rootstoocks. Rev. Bras. Frutic. 35:75-85.

Spotts, R. A., Cervantes, L. A., and Mielke, E. A. 1999. Variability in postharvest decay among apple cultivars. Plant Dis. 83:1051-1054.

Sreeramulu, T. 1959. The diurnal and seasonal periodicity of spores of certain plant pathogens in the air. Transact. Brit. Mycol. Soc. 42:177-184. van der Walt, L., Spotts, R. A., Visagie, C. M., Jacobs, K., Smit, F. J., and McLeod A. 2010. Penicillium species associated with preharvest wet core rot in South Africa and their pathogenicity on apple. Plant Dis. 94:666-675.

Warner, G. 2006. Shedding light on core rots. Good Fruit Grower 57:20-21.

Table 1. Core opening, calyx opening and incidence of moldy core at both harvest and after 10 weeks in 2°C RA storage of RosBREED germplasm evaluated in 2010 and 2011. Incidence of moldy core is presented both as the number of individuals in each core opening/calyx opening class and as a percentage.

Incidence at Incidence after 10 Core at Calyx at No. harvest weeks storage harvest harvest apples Number % Number % Open Closed 296 32 10.8 34 11.5 Mixed 161 16 9.9 12 7.5 Open 126 21 16.7 14 11.1 Closed Closed 279 8 2.9 5 1.8 Mixed 136 5 3.7 7 5.1 Open 119 10 8.4 14 11.8

Table 2. Moldy core susceptibility of cultivars tested at Cornell University 1, University of

Minnesota 2, and Washington State University 3. Trademarked names are in parenthesis.

Parents with Moldy Core

Fortune 2 Nevson (Sonya™) 3 Ginger Gold 2,3 NJ55 (SunCrisp™) 1 Honeygold 2 NJ90 3 Hudson's Golden Gem 1 Pinova 3 Minewashta (Zestar™) 2,3 Sawa 2

Parents without Moldy Core

8S6923 (Aurora Golden Gala™) 3 Granny Smith 3 Akane 2 Haralson 2 Ambrosia 3 Honeycrisp 1,2,3 23 2 Arlet Keepsake Autumn Crisp 1,2 Kerr 2 Beacon 2 Macoun 2 Braeburn 3 Malinda 2 Cameo 3 Mantet 2 Co-Op 15 3 McIntosh 2 Co-Op 29 (Sundance™) 3 MN 447 (Frostbite™) 2 Co-Op 33 (Pixie Crunch™) 2 NY1 (SnapDragon™) 1 Co-Op 39 (CrimsonCrisp™) 3 NY2 (RubyFrost™) 1 Cortland 2 Oriole 2 Cripps Pink (Pink Lady™) 3 Pitmaston Pineapple 2 Cripps Red (Sundowner™) 3 Red Delicious 3 Delblush (Tentation™) 3 Regent 2 Delorgue 3 Scifresh (Jazz™) 3 Discovery 2 Scired (Pacific Queen™) 3 Dolgo 2 Sciros (Pacific Rose™) 3 Duchess of Oldenburg 2 SPA 440 (Nicola™)3 Early Cortland 2 Splendour 3 Empire 1,3 StateFair 2 Fiesta 3 Sunrise 2 Fireside 2 Sweet Sixteen 2 Fuji 1,3 Tsugaru 2 Gala 3 WA 38 (Cosmic Crisp™) 3 Golden Delicious 3 Wealthy 2 GoldRush 3 Wildung (SnowSweet™) 2

Table 3. Number of seedling individuals evaluated and incidence of moldy core (MC) in families as part of the RosBREED project in 2010 and 2011.

Family Female parent Pollen parent No. MC No. MC Code seedlings incidence seedlings incidence 2010 2010 2011 2011 MN-A Honeycrisp Jonafree 3 1 10 1 MN-B Honeycrisp Monark 10 1 23 4 MN-C Honeycrisp Pitmaston 2 0 17 1 Pineapple MN-D Honeycrisp Regent 3 0 6 0 MN-E Zestar BC-8S-27-43 4 0 9 2 MN-F Sunrise GMAL4329 3 0 19 0 MN-G Ginger Gold GMAL4328 3 0 15 1 MN-H Ginger Gold GMAL4332 1 0 22 1 MN-I SnowSweet Pixie Crunch 8 1 11 2 MN-J Dayton Zestar 4 0 25 3 MN-K Honeycrisp GMAL4327 9 0 19 0 MN-L Sweet16 BC-8S-27-43 3 0 4 1 MN-M Honeycrisp Akane 1 0 22 4 MN-N Honeycrisp Silken 6 1 15 2

WA-A Delicious Honeycrisp 7 0 9 0 WA-B Honeycrisp Splendour 11 2 11 2 WA-C Cripps Pink Honeycrisp 8 0 8 1 WA-D W7 W5 19 1 23 5 WA-E Aurora Golden Gala Arlet 13 0 13 0 WA-F Aurora Golden Gala Enterprise 13 2 13 2 WA-G Aurora Golden Gala Honeycrisp 15 2 25 2 WA-H Aurora Golden Gala Granny Smith 11 0 0 0 WA-I Enterprise Arlet 9 0 9 2 WA-J Honeycrisp Arlet 16 2 16 1 WA-K Cripps Pink W7 23 5 23 4 WA-L Cripps Pink Aurora Golden 12 0 12 0 Gala

NY-A Pinova Cameo 13 3 14 4 NY-B Sansa Granny Smith 18 6 15 2 NY-C Hudson NY-S1 29 3 NY-D Braeburn Ginger Gold 25 10 35 8 NY-E Autumn Crisp Fuji 9 5 10 0 NY-F Pinova NY-S2 24 17 18 4 NY-G NY-S3 Sonya 3 1 4 0 NY-H Sonya NY-S4 4 1 3 0 NY-I Fuji NY-S5 9 5 10 1 NY-J Honeycrisp NY-S6 1 0 1 0

CHAPTER THREE: Resistance to Erwinia amylovora in Wild Accessions of Malus sieversii

Julia M. Harshman1, Kate M. Evans1, Haley Allen1, Michael E. Wisniewski2, John L. Norelli2

1Department of Horticulture, Washington State University, Tree Fruit Research and Extension

Center, 1100 North Western Avenue, Wenatchee, WA 98801

2United States Department of Agriculture, Agricultural Research Service, Appalachian Fruit

Research Laboratory, 2217 Wiltshire Road, Kearneysville, WV 25430

Abstract

Fire blight, caused by Erwinia amylovora, is a destructive bacterial disease of domestic apple (Malus × domestica). Epidemics are becoming more common as modern varieties and rootstocks are more susceptible, while control is becoming more difficult with the development of antibiotic-resistant strains. Sources of resistance currently being utilized by breeding programs tend to have very poor fruit size and flavor characteristics. Malus sieversii is the progenitor species of domestic apple and is notable for its relatively large, palatable fruit. Previous studies have indicated that some accessions of M. sieversii have resistance to fire blight. In this study, nearly 200 accessions and controls were inoculated with E. amylovora in both Washington and

West Virginia to identify resistant accessions. Resistance to fire blight was significantly affected by the test environments. Several accessions showed a unique resistance response where the accessions had low incidence of infection, but high severity once infection was initiated. This response has not been reported in domestic apple. Twelve resistant to highly resistant M. sieversii accessions were identified, some of which will be used as parents in the 2016 crossing

42 season of the WSU apple breeding program. These resources will enable faster introgression of novel and diverse resistance to a serious disease in apple.

Introduction

Fire blight, caused by the bacterium Erwinia amylovora (Burrill) Winslow et al., is a destructive disease in apple (Malus × domestica Borkh.) that can kill young trees outright or result in permanent structural damage to mature trees. Fire blight remains a challenging disease to control, even with use of cultural practices to reduce primary inoculum and sprays of copper, biological controls, and/or antibiotics to limit pathogen multiplication (van der Zwet 2012). The efficacy of streptomycin to control fire blight is threatened by the development of streptomycin- resistant E. amylovora strains (McManus et al. 2002; McManus and Jones 1994; Russo et al.

2008) and antibiotic applications were recently revoked for use by U.S. organic growers (USDA,

NOP 2015). Fire blight epidemics often develop explosively and the limited number of effective management practices available to growers makes it difficult to slow or stop their progress. Most commercially successful apple cultivars introduced in recent decades, such as ‘Braeburn’, ‘Fuji’,

‘Gala’, and Pink Lady™ (‘Cripps Pink’), and newly released cultivars such as RubyFrost™

(‘NY-2’), are more susceptible to fire blight than older cultivars such as ‘Delicious’ (Breth 2014;

Brown 2012). Recent plantings of susceptible scion cultivars on susceptible rootstocks have increased the danger of fire blight epidemics in apple orchards to unprecedented levels (Norelli et al. 2003).

Host plant resistance is one of the most effective and sustainable options for managing fire blight (Peil et al. 2009), however most resistant and highly resistant domesticated apple cultivars do not meet industry horticultural and fruit quality standards (Peil et al. 2009). QTL

43 linked to resistance have been identified in several Malus species that explain phenotypic variance ranging from 15 to 83% (Durel et al. 2009; Emeriewen et al. 2014; Gardiner et al.

2012; Khan et al. 2007; Le Roux et al. 2010; Peil et al. 2008). The QTLs conferring the strongest resistance are from crab apples with undesirable fruit qualities such as small size, and bitter or astringent flavors. These sources are actively being used in rootstock breeding and several scion breeding programs in Europe that are utilizing accelerated breeding strategies (Fazio 2008;

Flachowsky et al. 2011; Khan et al. 2012; Khanizadeh et al. 2000; Le Roux 2012). The U.S. apple industry is dependent on growing high value scion cultivars prized for their fruit quality and the negative effects on fruit quality of introgressing wild germplasm are time-consuming to overcome through multiple generations of conventional breeding.

M. sieversii (Lebed.) M. Roem. is the primary progenitor species of Malus × domestica

(Luby 2003). Native to central Asia, this species has long been of interest for increasing genetic and phenotypic diversity in domestic apple (Janick 1996; Volk et al. 2015). The U.S. Department of Agriculture (USDA) Agricultural Research Service’s (ARS) National Plant Germplasm

System (NPGS) sponsored four expeditions to collect M. sieversii between 1989 and 1996 to 12 climatically diverse sites in Kazakhstan, Kyrgyzstan, Tajikistan, and Uzbekistan (Forsline et al.

2003; Table 1; Fig. 1). These expeditions resulted in an extensive collection of over 130,000 seeds from 892 trees, which were distributed to several states and countries (Luby et al. 2000).

Approximately 1,200 M. sieversii seedlings from 106 half-sibling families (ie. seeds collected from the same mother tree) were planted in an orchard at the USDA-ARS-Plant Genetic

Resources Unit (PGRU) in Geneva, NY (Forsline et al. 2008). Some of those planted at USDA-

ARS-PGRU have resulted in permanent accessions designated in Germplasm Resources

Information Network (GRIN; USDA-ARS-NPGS 2008).

Twenty-five phenotypic traits were recorded from fruit harvested from the mother trees at the time of collection (Forsline et al. 2003); phenotypic evaluation has continued in the U.S. and

New Zealand (Luby et al. 2002; Fazio et al. 2008; Forsline and Aldwinckle 2004), including resistance to biotic and tolerance to abiotic stresses.

Fire blight resistance has been evaluated using both controlled inoculations and natural occurrence in established plantings. Natural incidence of infection was recorded on 1,151 seedlings planted in 1997 and 1998 at USDA-ARS-PGRU Geneva, NY (Forsline and

Aldwinckle 2002) and 1,410 accessions planted in 1998 in MN (Forsline et al. 2008). A total of

124 families were evaluated, with 32 families represented at both sites. There was low incidence of fire blight susceptibility in 12 families, and a total of 535 seedlings were rated as highly resistant or resistant (Fazio et al. 2009). In New Zealand, 936 seedlings from 52 families were evaluated for natural infection and just 13% showed signs of fire blight infection (Forsline et al.

2003; Luby et al. 2002). There were no significant differences between collection site and fire blight susceptibility in the New Zealand study (Luby et al. 2000) while evaluations at NY and

MN found that seedlings from site 6 were more resistant than those from site 9 (Forsline et al.

2008). Accessions that appear resistant to natural occurrence of fire blight could have escaped infection, therefore confirmation of resistance with inoculation tests is prudent.

Greenhouse-grown seedlings from 79 families were inoculated in two consecutive years at USDA-ARS-PGRU Geneva, NY (Momol et al. 1999) and a total of 29% were rated as resistant (< 20% of total shoot length blighted). Seedlings from collection sites 1, 4, 5 and 6 had lower disease severity with site 1 having the greatest proportion of resistant seedlings (Momol et al. 1999). Additional controlled inoculations have been undertaken in the greenhouse to confirm

45 resistance of those that had little natural occurrence (<30% shoot blight). Of the 289 accessions inoculated, 60% were confirmed to be resistant (Forsline et al. 2008), but the performance of specific accessions has not been published.

The combination of high levels of fire blight resistance and large, palatable fruit make M. sieversii accessions excellent potential parental germplasm for breeding disease resistant scions

(Forsline et al. 2002; 2008) as compared to other wild Malus species which tend to have poorer fruit quality (Luby et al. 2001). Several M. sieversii seedlings reported as highly resistant also have fruit with desirable quality attributes: GMAL3208.h, GMAL3256.c, GMAL3256.f,

GMAL3331.h, GMAL3333.a (Forsline et al. 2003; Momol et al. 1999).

Richards et al. (2009b) characterized 949 M. sieversii accessions from eight collection sites for genetic diversity and population structure using seven microsatellite loci. The genetic diversity was significantly different between collection sites and levels of within-site variation also differed between sites with family having the largest effect on genetic structure. The greatest intra-site genetic variation was found at collection sites 5, 11 and 12, whereas as sites 4 and 9 showed the least diversity. Four genetic clusters were identified when accessions were clustered by their genotype, regardless of family or collection site. Cluster 1 and 2 were common and found at all sites. Individuals from collection sites 10, 11 and 12 clustered commonly to group 3 and individuals from sites 11 and 12 commonly clustered to group 4. A stratified core set of 110 accessions were selected that captured the majority of character and genetic variation (Richards et al. 2009a).

Plantings of M. sieversii that represented most of the genetic diversity within the species and included many accessions reported to be fire blight resistant were established in East

Wenatchee, WA and Kearneysville, WV. The goal of this study was to obtain quantitative

46 measures of fire blight resistance following controlled inoculation of a set of M. sieversii accessions to aid in parent identification and facilitate association mapping of the genetic determinants of resistance.

Materials and Methods

Plant material and orchard establishment. Based on previous phenotypic and genetic evaluation of M. sieversii accessions grown at the USDA-ARS-PGRU Geneva, NY, 197 M. sieversii accessions were selected for use in this study. The primary selection factor was the presence or absence of naturally occurring fire blight shoot infections on M. sieversii accessions evaluated at the USDA-ARS-PGRU Geneva, NY. Accessions were selected with the aim of obtaining a ratio of 1 susceptible: 1 resistant accession for subsequent association mapping studies. Accessions with fire blight shoot infections were presumed to be susceptible to fire blight, however absence of fire blight could be due to either resistance or escape from infection.

Based on greenhouse inoculation studies (Forsline et al. 2008), the number of presumed resistant accessions was increased by 40% unless supporting data was available from the greenhouse evaluation. The accessions in the study included 110 individuals (92%) from the stratified core sets previously selected by Richards et al. (2009a) to represent the majority of phenotypic and genetic diversity of the M. sieversii collected in central Asia. Additional accessions were selected based on resistance to scab (caused by Venturia inaequalis), resistance to cedar apple rust

(caused by Gymnosporangium juniper-virginianae), high fruit soluble solids content, fruit acidity

(avoiding low and very high acidity), later fruit harvest dates, fruit weight and fruit flavor

(avoiding bitter, sour and astringent flavors).

Six controls were included at different resistance levels: highly resistant (M. × robusta 5), resistant (‘Delicious’), moderately resistant (‘Empire’, ‘Golden Delicious’) and highly susceptible (‘Jonathan’, ‘Gala’) (Aldwinckle and Preczewski 1976; Gardner et al. 1980; van der

Zwet and Beers 1992). In addition, three modern cultivars reported to be resistant, ‘Goldrush’

(Crosby et al. 1994), ‘Fiesta’ (Khan et al. 2007) and ‘Splendour’ (van der Zwet and Beers 1992) were included for comparison to M. sieversii accessions. ‘Fiesta’, ‘Splendour’ and ‘Goldrush’ are cultivars widely reported as resistant to fire blight (Crosby et al. 1994; van der Zwet and Beer

1999); however, field performance and controlled inoculation studies are scarce. Fire blight resistance loci in ‘Fiesta’ were genetically mapped (Khan et al. 2007).

Budwood of selected accessions and control cultivars was collected at the USDA-ARS-

PGRU in Geneva, NY and budded on M.7 rootstocks at Van Well Nursery, East Wenatchee, WA in fall 2010. The material was planted at two locations in spring 2012. A total of 206 genotypes were planted at the WSU Columbia View orchard site in East Wenatchee, WA (WA). The orchard was planted in two sections, with 103 accessions in the first, 124 accessions in the second, and 21 accessions duplicated in both sections. Each section comprised three complete randomized blocks of single tree plots per accession at a density of 4.9m between rows and 1m in-row spacing. A total of 202 genotypes were planted at the orchard at USDA-ARS,

Kearneysville, WV (WV) in 4 randomized complete blocks of single tree plots per accession at

6m between rows and 2.1m in-row spacing. 197 M. sieversii accessions, 7 M. domestica cultivars and M. × robusta 5 were common across sites (Supplemental Table 1). There were unique accessions at WV (GMAL4054.z, GMAL4002.p and ‘Fiesta’) and WA (GMAL4209.g).

All controls were scored both years in WV. In WA, only ‘Golden Delicious’ and ‘Jonathan’ had sufficient growth to inoculate and were scored both years.

Inoculum and Inoculation. Freeze-dried E. amylovora strain Ea153n was used for inoculation (Stockwell 1998) and prepared following methods outlined by Stockwell (1998). The same freeze-dried Ea153 was used for inoculum preparation in both locations in each year.

Desired inoculum concentrations (5.2x1011 CFU/g in 2013 and 7.5x1011 CFU/g in 2014) were obtained by diluting with 0.01M dibasic phosphate buffer, pH 7.

Up to ten shoots (2013) or 20 shoots (2014) per tree were selected for inoculation.

Selected shoots were actively growing, had at least 10cm of new growth and were on independent branches from other selected shoots (Norelli et al. 2003). Trees were inoculated by bisecting the shoot tip with scissors dipped in E. amylovora inoculum as described by Norelli et al. (1988). To determine the current season’s shoot length at the time of inoculation, shoots were measured from the 1st bud scale scar to either the apical meristem (if measured prior to inoculation, WA) or to the first cut leaf (WV). Trees were inoculated on June 4th at WV and June

12th at WA in 2013 and on June 3rd at WV and May 30th at WA in 2014.

Disease Measurement. The total length of necrotic fire blight lesions was recorded once disease progression had stopped. Lesions did not necessarily girdle the stem and length of non- continuous lesions was the total sum of necrosis length. The age of wood that the lesion extended into was recorded as 0= no infection, 1= current season’s shoot growth, 2= previous season’s growth, etc. In addition, naturally occurring blossom and shoot infections that were not the result of controlled fire blight inoculation were also recorded and monitored.

Statistical analysis. The measure of fire blight resistance in this study was the proportion of the current season’s shoot growth that was blighted (%SLB) following inoculation with E. amylovora Ea153n. The %SLB data was logit transformed to satisfy the assumption that residual variation is normally distributed. Prior to transformation, zero percentage data was randomly

49 assigned a value between 0 and 1. The logit transformed measure of %SLB (Logit%SLB) was used for statistical analysis.

All analyses of variance were performed using the GLIMMIX procedure of SAS 9.4

(SAS; SAS Institute Inc., Cary, NC, USA). Analysis of variance was performed on the complete dataset to test the fixed effects of evaluation location (WA, WV), evaluation year (2013, 2014), and accessions on Logit%SLB. The following analysis of variance was performed on accessions scored in WV: 1) to test the fixed effects of evaluation year (2013, 2014), and accessions on

Logit%SLB; 2) to test the fixed effects of evaluation year, and genetic cluster (Richards 2009b) on Logit%SLB; 3) to test the fixed effects of evaluation year, and collection site (Forsline et al.

2008) on Logit%SLB. Every collection site had sufficient accessions to include in the analysis except site 3 (3 accessions) and site 7 (2 accessions). One analysis of variance was performed on

64 accessions scored in both WV and WA in both years to test the fixed effects of evaluation location (WV, WA), evaluation year (2013, 2014), and accessions on Logit%SLB. Interaction terms that were not significant were removed from the model. Block/plot within location (WV and WA), rep within location (WA) and row position within the location were included as random effects. Mean separations were performed using the Tukey option at the p < 0.05 level.

Genetic cluster of the accession was obtained from published results (Richards 2009b) and collection site was obtained from USDA Germplasm Resources Information Network

(GRIN) database (USDA-ARS-NGRP, 2008). Natural occurrence of fire blight was obtained from published results (Forsline et al. 2003; USDA-ARS-NGRP, 2008). For the accessions included in this study, 188 were previously evaluated for natural occurrence of infection in established field plantings (USDA-ARS-NPGS, 2008) using a 1 to 5 scale (1: very resistant - no occurrence; 2: moderately resistant – only light rating; 3: intermediate – light to medium rating;

4: moderately susceptible – medium to heavy rating; 5: very susceptible – very heavy rating

(Forsline and Aldwinckle 2002; 2008). Accessions inoculated at WV were assigned resistance categories (HR, R, M, S, HS) based on performance of controls and mean separations to facilitate comparison with GRIN data (Supplemental Table 1). Data from greenhouse inoculation at USDA-ARS-PGRU Geneva, NY was obtained from personal communication (G. Volk); accessions were considered resistant if average shoot blight ratings were less than 20%. To facilitate comparison, inoculation data in this study was similarly reduced to resistant (<20%) and susceptible (>20%).

Results

Fire blight resistance of full panel. The interaction between accession, year and location was significant (p<0.0001), so each location was analyzed individually. The interaction between accession and year was significant at WV (p<0.0001) and at WA (p<0.0001), so each location- year combination was analyzed individually (WA13, WA14, WV13, WV14).

Fire blight resistance of standard varieties in West Virginia. Control cultivars generally performed as expected in WV in 2013, except the susceptible control ‘Gala’ which was comparable with resistant controls M. × robusta 5 and ‘Golden Delicious.’ One replicate of

‘Jonathan’ developed large lesions, resulting in tree death. In 2014, M. × robusta 5, ‘Golden

Delicious’, ‘Empire’ and ‘Gala’ performed as expected. ‘Delicious’ was less resistant than expected and ‘Jonathan’ was less susceptible than expected (Supplementary table 1). ‘Fiesta’ was evaluated in 2014 only at WV and was not significantly different from resistant control M. × robusta 5. Both ‘Goldrush’ and ‘Splendour’ were inoculated in both 2013 and 2014 at WV; neither were significantly different from resistant controls M. × robusta 5 or ‘Delicious’.

Fire blight resistance of M. sieversii accessions in West Virginia. The interaction between accession and year was significant at WV (p<0.001), so each year was analyzed individually. The effect of accession on logit transformed percent shoot length blight

(Logit%SLB) was significant for both the 2013 (p<0.0001) and 2014 (p<0.0001) seasons.

Diverse responses among the wild apple accessions were observed after inoculating vigorously- growing shoot tips with E. amylovora, ranging from highly susceptible to highly resistant.

Typical fire blight infections were observed on many of the accessions. M. sieversii accessions exhibited greater variability in their susceptibility to fire blight than M. × domestica controls

(Supplemental Table 1). Three accessions were significantly (p<0.05) more susceptible to fire blight than ‘Jonathan’ in 2013 and 25 accessions were significantly more susceptible than ‘Gala’ in 2014.

M. sieversii accessions were not significantly more resistant than M. × domestica controls due to the relative lack of fire blight shoot infection on resistant controls, however 41 and 98 accessions had lower %SLB values than ‘Delicious’ in 2013 and 2014, respectively

(Supplemental Table 1). The fire blight resistance of GMAL3608.h, GMAL4002.k, PI657115, and PI657116 appeared equivalent to the highly resistant control M. × robusta 5 in more than one test (Supplemental Table 1). GMAL3616.o, GMAL4002.m, GMAL4028.h, GMAL4211.d,

PI657054, and PI657085 developed minimal fire blight infections in one of the tests, but were comparable with M. × robusta 5 in resistance. GMAL3544.b, GMAL3688.c, GMAL3975.c,

GMAL3989.c, GMAL4059.a, GMAL4211.a, GMAL4211.e, GAML4211.g, and PI657044 developed fire blight comparable with ‘Delicious’, a resistant control cultivar in one of the WV tests (Supplemental Table 1).

Multiple accessions appeared to be resistant in one year and susceptible in the other. For example, in 2013 only two of the 20 shoots inoculated on PI633921 developed a lesion, and both were just 4cm in length or 10% of the total shoot length. In 2014, 29 of the 32 shoots developed infections; 11 of those progressed into 2nd year wood and one into 3rd year wood. GMAL4032.i had a similar response; only one out of 20 shoots inoculated in 2013 developed a lesion, which was 14 %SLB. In 2014, 36 shoots were inoculated and 15 of them developed lesions; eight of those progressed in to 2nd year wood and three progressed into 3rd year wood.

Unique resistance responses to E. amylovora. Atypical from observations in most domesticated apple cultivars (Aldwinckle and Preczewski 1976), several M. sieversii accessions appeared resistant to the initiation or incidence of fire blight infection, but were highly susceptible to the severity of infection when an infection did occur, and infections progressed into older wood in a single growing season. In the case of GMAL4028.h, infection was initiated in only one of 35 shoots inoculated in 2013 and 2014, but that infection progressed through 2 year-old wood into the central leader, resulting in the eventual death of the tree (Fig. 2). The one infected shoot had 100% SLB, but the average %SLB and average age of wood infected for this accession was near zero. The response of GMAL4055.w was similar to that of GMAL4028.h.

Only two of 25 inoculated GMAL4055.w shoots developed infection, but in both shoots the infections progressed into 2 and 3 year-old wood. Neither of these accessions was inoculated in

WA.

The fire blight resistance of some accessions differed in WV in comparison to WA. The

WA site sustained heavy deer damage in the spring and early summer of 2013 which had a lasting detrimental effect on growth so not all accessions were evaluated in every test. A

53 sufficient number of shoots with adequate growth for reliable fire blight evaluation was seen on

79 accessions in 2013 and 83 in 2014.

For the 64 accessions scored in both locations and in both years, the interaction between accession, year and location was significant (p<0.0001). Separated by year, the interaction between accession and location was significant for both 2013(p<0.0001) and 2014 (p<0.0001).

Each location and year combination was analyzed individually, and accession was significant

(p<0.0001) in each analysis. The standard error was larger in both years in WA than in WV (Fig.

3). In general, accessions performed similarly across the years and locations with a few notable exceptions. PI633921 was fairly resistant in every trial except WV 2014 where the average was

87 %SLB. Several accessions, such as GMAL4047.o, GMAL4047.k, GMAL4039.y and

‘Jonathan’, developed less disease in WA than in WV in both years. ‘Jonathan’ had less disease development in WV in 2014 due to the death of one replicate in 2013. In 2013 PI657068,

GMAL6383.e and PI657097 were more susceptible in WA than in WV, however no accessions were found that were consistently more susceptible in WA in both years.

Controlled inoculations in WV were comparable to natural infection in NY. There was general agreement between inoculations in this study and natural occurrence of infection in field plantings (USDA-ARS-NPGS, 2008), with 94 accessions in 2013 and 99 in 2014 that scored in similar categories (ie. HR or R in 1 or 2; Table 2). Those accessions in category 1 or 2 rated as susceptible or highly susceptible in the evaluation likely escaped infection in the field. However, there were nine accessions in 2013 and 11 in 2014 that were rated as 4 or 5 based on natural occurrence of infection but appeared highly resistant or resistant after inoculation.

Controlled inoculations in WV were comparable to controlled inoculation in NY.

Evaluation following controlled inoculation of 121 accessions in the greenhouse at USDA-ARS-

PGRU Geneva, NY (G. Volk, personal communication) indicated 86 were resistant; meaning less than 20% of the shoot length was blighted. Sixty-seven accessions identified as resistant based on greenhouse inoculation were found to also have < 20 % SLB in WV in both 2013 and

2014 following inoculation (Table 3; Supplemental Table 1). Seven accessions in 2013 and eight accessions in 2014 were rated as susceptible in both tests. Of those rated resistant in the greenhouse test, 19 in 2013 and 18 in 2014 were susceptible in the WV test. Similarly, 28 in

2013 and 26 in 2014 of those rated as susceptible in the greenhouse test were rated as resistant in the WV field inoculation.

Collection site and genetic cluster was associated with the fire blight resistance of accessions. There was a significant interaction (p<0.0001) between collection site and evaluation year at WV, therefore each year was analyzed separately. Fire blight resistance was significantly affected by collection site for both 2013 (p<0.0001) and 2014 (p<0.0001). The highest %SLB in both years was exhibited by accessions collected from site 12 (Fig. 4). Site 12 was significantly different from other sites in 2013 (Fig. 4). In 2014, accessions from site 12 remained the most susceptible, but were not significantly different from those collected from sites 11 and 5. The lowest %SLB was observed on accessions collected from sites 4 and 10 in 2013 and on accessions from sites 4 and 6 in 2014.

Genetic cluster data was available for 185 of the selected accessions (Richards et al.

2009b). The interaction between genetic cluster and evaluation year was significant (p<0.0001) at WV, therefore each year analyzed separately. Logit%SLB was significantly affected by the genetic cluster of accessions in both 2013 (p<0.0001) and 2014 (p<0.0001). Accessions in genetic clusters 1 and 2 had significantly less fire blight than those in genetic cluster 3 in both years (Fig. 5). Accessions in genetic cluster 4 were not significantly different from those in

55 cluster 3 in 2013, yet appeared more resistant in 2014 and not significantly different from accessions in genetic clusters 1 and 2. None of the other three clusters grouped with cluster 3.

Discussion

Successful management of fire blight hinges on the severity of the disease. High incidence but low severity of individual infections can be effectively managed with chemical sprays and cultural practices (van der Zwet et al. 2012) while high severity infections are harder to manage and the source of major economic loss. Shoot blight infection was chosen as the measure of susceptibility, rather than blossom blight, as shoot blight typically leads to structural damage of the tree and longer term economic losses. Blighted shoots are typically pruned out, resulting in economic losses from yield loss. Shoot blight can also lead to severe structural damage or tree death.

A high concentration of the fire blight pathogen was used in evaluating resistance in these apple accessions to increase the chance that an infection would be initiated, and then resistance was evaluated based upon disease progression in the infected shoot. The proportion of current season’s growth blighted was chosen as the measure of fire blight susceptibility so that data from this phenotypically diverse material could be compared. Tree size, branch angles, internode length, growth rate, growth habit, leaf shape, thickness and color, blossom color, precocity, and fruit characteristics varied widely. The location WV was chosen as the focus of the analysis and presented results as all nine controls were evaluated and overall tree health and vigor were substantially better than WA.

Resistance to fire blight. In general, there was a high level of agreement between the results in this study using controlled inoculations and published reports of natural occurrence of

56 fire blight on M. sieversii accessions (Forsline et al. 2008). However, there were a total of fourteen accessions rated as resistant after inoculation in this study but susceptible based on natural occurrence of infection (Forsline et al. 2008). Four of these appeared highly resistant under field inoculations (GMAL4059.a, GMAL4211.a, GMAL4211.e, and GAML4211.g) but were rated as either 4 or 5 based on natural field infection at USDA-ARS-PGRU Geneva, NY.

There are several possible explanations. Blossom blight and shoot blight are generally in agreement, but there are known discrepancies. The natural occurrence ratings could have been increased (rated as more susceptible) based on blossom blight. Furthermore, susceptibility to fire blight is highly dependent on host vigor, and the genotypes may have appeared to be resistant if the host was not receptive to infection when inoculated. There have been other reported discrepancies between natural occurrence of infection at different locations (Forsline et al. 2008;

Luby et al. 2002) and between natural occurrence and inoculation (Luby et al. 2002; Momol et al. 1999); however, those discrepancies were between half-siblings collected from the same mother tree. The discrepancies listed above are for the same clonal genotypes.

There are 28 accessions in total that were rated as susceptible in greenhouse inoculations

(Momal et a. 1999) and resistant under field inoculations. This could be due to the different inoculation methods, different strains used, or the somewhat arbitrary cut-off level of 20 %SLB for susceptible and resistant. Fifteen of those 28 developed more than 5 %SLB and should likely be considered moderately susceptible.

Incidence (the number of infections per plant) and severity (the relative size and effect of the infection) are generally well correlated. In several accessions of M. sieversii, a high severity of fire blight was observed with low incidence which is atypical in M. × domestica. This is potentially an interesting infection mechanism. The host is able to thwart infection by the

57 pathogen in most cases, but is unable to prevent the spread of the pathogen once infection has been initiated. This type of resistance is less useful for breeding by itself as it would result in economic losses under severe infection. However, it could be useful if pyramided with other mechanisms of resistance to fire blight as it would lead to reduced frequency of infections and thus inoculum load.

Effect of collection site and genetic cluster on resistance. Evaluation of plantings in NY and MN have shown significant differences between collection sites, focusing on accessions from site 6 and site 9 (Forsline et al. 2008), while evaluations in New Zealand did not show significant differences between collection sites (Luby et al. 2000). Controlled inoculation of accessions from sites 1 to 8, indicated that accessions from collection sites 3 and 6 were more susceptible than other sites. Material evaluated at MN, NY and New Zealand was grown from seeds. Those seeds from the same family would be half-siblings, but still genetically unique and diverse.

Results from this study indicate that the accessions from collection site 4 were more resistant than those at other sites, and those from collection site 12 were the most susceptible.

Collection site 4 is humid temperate mixed forests. It is logical that genotypes from that type of climate would have some resistance to bacterial infection. Comparisons between this study and the material evaluated in NY are most appropriate, as the seeds grown out in NY turned into the formalized accessions from which this study chose (Forsline et al. 2008; Forsline and

Aldwinckle 2002). However, accessions chosen for this study were not selected randomly or with consideration for collection site; they were chosen based on known resistance to achieve a 1 resistance to 1 susceptible ratio. The analysis could be biased for some of the collection sites, however, this is unlikely, given the similar number of accessions from each collection site. Also,

58 accessions from collection site 9 were among the most susceptible and accessions from collection site 6 showed a similar level of resistance to the most resistant site, which is in agreement with published data from USDA-ARS-PGRU Geneva NY (Forsline et al. 2008;

Momol et al. 1999).

There have been no published comparisons between genetic cluster and fire blight resistance to date. Accessions in genetic clusters 1 and 2 had the highest level of fire blight resistance. Genetic clusters 1 and 2 were the most common and found at all sites. The majority of individuals at sites 4, 5 and 9 were in genetic cluster 1 while the majority of individuals collected at site 6 were in genetic cluster 2. Genetic clusters 3 and 4 were less represented than cluster 1 and 2; therefore the analysis could be biased. Of course, the clusters may also not reflect the genetic diversity of the genes controlling fire blight resistance.

Host plant and environmental effects on resistance. Disease severity observed after inoculation with a pathogen is an interaction between the pathogen, the host plant and the environment. Fire blight severity is strongly influenced by the environment and predictions of host genetic resistance that are not confounded with environmental variation are most useful. The

M. sieversii accessions were evaluated in two very different environments, Wenatchee, WA and

Kearneysville, WV, over two different years. In many cases, an accession may have appeared highly resistant in some of the tests, but susceptible in others. This is the result of resistance that is strongly influenced by environment and of less use in the breeding program. Although approximately 10-30% of the accessions may have appeared highly resistant in any individual test, only 6%, or 12 accessions, were consistently, highly resistant in multiple tests.

The efficacy of inoculation is strongly influenced by host vigor. Accessions would ideally be actively growing with at least 10cm of growth at the time of inoculation. While the

59 growth stage of the M. × domestica cultivars was readily determined, the growth stage of the M. sieversii material was far more challenging to determine. A number of accessions had atypical growth forms and grew very little each year. Trees in WA that were compromised by deer damage showed generally low plant vigor and growth stage was particularly difficult to determine. This partially explains the discrepancy in disease severity between WA and WV.

With the diversity of accessions screened in this test, it is likely that some were inoculated either before or after the ‘active’ growth stage.

Inoculation efficacy is greatly influenced by the environmental conditions during inoculation. In WA, conditions were sub-optimal during inoculation. Infection of E. amylovora occurs between 18.5°C and 35°C, but most readily between 21-27°C. Rain or periods of high humidity have long been associated with fire blight infection, with 80% relative humidity (RH) as the optimum (van der Zwet et al. 2012). The lower limits of RH sufficient for infection have not been reported, but infection has been induced at as low as 50% RH (van der Zwet et al.

2012). The average temperature and humidity in WA during inoculation were 17°C / 43% RH in

2013 and 20°C / 44% RH in 2014. There was no precipitation preceding or after inoculation either year. Temperatures were within the range for fire blight infection, but the lower than optimum RH may have hindered fire blight infection in WA. Conversely, the average temperature and humidity in WV during inoculation were 18.7°C / 53% RH in 2013 and 23°C /

68 % RH. In the three days following inoculation, there was 35.5 cm of rain in 2013 and 25 cm of rain in 2014.

Use of M. sieversii accessions in breeding fire blight resistant apple cultivars. Multiple accessions were identified with similar or greater resistance than ‘Delicious’ that was stable over multiple tests: GMAL3544.b, GMAL3608.h, GMAL3616.o, GMAL3688.c, GMAL3975.c,

GMAL3989.c, GMAL4002.m, GMAL4028.h, GMAL4059.a, GMAL4211.a, GMAL4211.d,

GMAL4211.e, GAML4211.g, GMAL4211.h, PI657044, PI657054, and PI657085 (Supplemental

Table 1). Those accessions scored as susceptible based on natural occurrence (GMAL4059.a,

GMAL4211.a, GMAL4211.e, and GMAL4211.g) in the field are probably neither truly resistant nor advisable to use in breeding efforts. GMAL4028.h showed unique susceptibility, with low incidence and high severity, and is likely not a source of resistance to pursue initially for breeding. Slight natural infection was observed on PI657085. The other accessions were all rated as 1 or 2, and are likely resistant to fire blight.

The advantage of using M. sieversii accessions as sources of fire blight resistance is that they tend to have larger fruit size and a greater flesh to core ration than other sources of resistance (ie. M. × robusta 5). For all of the accessions listed above, fruit size is larger than crab apples, levels of soluble solids (11 to 13.9°Brix) are acceptable, and some were in desirable harvest windows. Phenotypic GRIN data is not available for all of the accessions and further fruit quality data is being collected by the WSU apple breeding program.

The challenge remains that fruit size and quality of M. sieversii accessions are still far below industry standard, but it is anticipated that fewer introgression generations will be required especially if introducing this resistance in to a background such as ‘Splendour’ or ‘Delicious’ that already show some level of resistance. The long term objective will be to pyramid multiple resistance genes to create high quality, commercially competitive apple varieties with durable, and sustainable resistance to fire blight. It is particularly important that the resistance is durable in a perennial tree crop as plantings last more than 20 years.

Conclusion

Identification of fire blight resistant sources with larger fruit and no unpleasant flavors will be critical to allow rapid progress in combining superior fruit quality and resistance. The data available for fire blight susceptibility of M. sieversii accessions has typically been qualitative and based on incidence. Precise quantitative data, as provided by this study, on severity as well as incidence is required for mapping studies; genotype by sequencing (GBS) data has been collected for these individuals. Beyond association mapping, resistance-trait- specific DNA tests will augment breeding efforts to incorporate and pyramid resistance. Fast- track breeding can further accelerate introgression efforts, and these methods have been successfully demonstrated with other sources of fire blight resistance (Le Roux et al. 2014).

This project has identified several excellent sources of fire blight resistance for use in apple scion breeding programs. Broadening the diversity of resistance sources available, especially those with different mechanisms of resistance, will enable the strategic development of durable multi-gene resistance to fire blight.

Author Contributions

JN and KE designed the study. JH, HA, ME and JN collected the data. JH and JN analyzed the data. JH drafted the majority of the manuscript. All authors read and approved of the final document.

Acknowledgements

The authors thank Jade Flamenco, Ryan Potts and Roger Lewis of the USDA-ARS,

Kearneysville, WV and Bonnie Schonberg, Lisa Brutcher, Nancy Buchanan of the WSU-TFREC for their expert technical assistance in the project. They thank Gayle Volk of the USDA-ARS,

Plant Germplasm Preservation Research Unit, Fort Collins, CO and Phillip Forsline (retired), C.

Thomas Chao, and staff of the USDA-ARS, Plant Genetic Resources Unit, Geneva, NY for providing data on the collection and initial characterization of Malus sieversii accessions, as well as supplying budwood of accessions for plant propagation. They also thank the farm crews at

USDA-ARS Kearneysville, WV and WSU Columbia View Orchard for orchard maintenance.

This project was partially funded by the Washington Tree Fruit Research Commission

Projects CP-10-110 and CP-12-104, and AFRI NIFA Pre-doctoral Fellowship 123154-001 awarded to J. Harshman.

Literature Cited

Aldwinckle, H. S., and Preczewki, J. L. 1976. Reaction of terminal shoots of apple cultivars to invasion by Erwinia amylovora. Phytopathology 66:1439-1444.

Breth, D. 2014. Managing Fire Blight in Young RubyFrost Plantings. Empire State Producers Expo Proceedings. Online Publication. http://www.hort.cornell.edu/expo/proceedings/ 2014/Tree%20Fruit/Fire%20 Blight%20in%20Ruby%20Frost%20Breth.pdf.

Brown, S. 2012. Apple. Pages 329-367 in: Fruit Breeding. M. L. Badenes and D.H. Byrne (eds.). Springer Science & Business Media, New York.

Calenge, F., D. Drouet, C. Denancé, W. E. Van de Weg, M-N. Brisset, J-P. Paulin, and C-E. Durel. 2005. Identification of a major QTL together with several minor additive or epistatic QTLs for resistance to fire blight in apple in two related progenies. Theor. Appl. Genet. 111:128-135.

Calenge, F., and Durel, C. 2006. Both stable and unstable QTLs for resistance to powdery mildew are detected in apple after four years of field assessments. Mol. Breed. 17:329- 339.

Crosby, J. A., Janick, J., Pecknold, P. C., Goffreda, J. C., and Korban, S. S. 1994. ‘Gold Rush' Apple. HortScience 29:827-828.

Durel, C-E., Denancé, C., and Brisset, M-N. 2009. Two distinct major QTL for resistance to fire blight co-localize on linkage group 12 in apple genotypes' Evereste' and Malus floribunda clone 821. Genome 52:139-147.

Emeriewen, O., Malnoy, M., Richter, K., Kilian, A., Hanke, M-V., and Peil, A. Evidence of a major QTL for fire blight resistance in the apple wild species Malus fusca. Acta Hortic. 1056:289-293.

Fazio, G., Aldwinckle, H. S., Volk, G. M., Richards, C. M., and Janisiewicz, W. J., and Forsline, P. L. 2007. Progress in evaluating Malus sieversii for disease resistance and horticultural traits. Acta Hortic. 814:59-66.

Fazio, G., Aldwinckle, H.S., Robinson, T.L., and Wan, Y. 2008. Implementation of molecular marker technologies in the apple rootstock breeding program in Geneva-challenges and successes. Acta Hortic. 903:61-68.

Fischer, M. 1996. The Pillnitz apple rootstock breeding methods and selection results. Acta Hortic. 451:89-98.

Flachowsky, H. K., P‐M. Le Roux, A. Peil, A. Patocchi, K. Richter, and M‐V. Hanke. 2011. Application of a high‐speed breeding technology to apple (Malus × domestica) based on transgenic early flowering plants and marker‐assisted selection. New Phytologist 192:364-377.

Forsline, P. L., H. S. Aldwinckle, C. Hale, and R. Mitchell. 2002. Natural occurrence of fire blight in USDA apple germplasm collection after 10 years of observation. Acta Hortic. 590:351-358.

Forsline P. L., H. S. Aldwinckle, E. E. Dickson, and S. C. Hokanson. 2003. Collection, maintenance, characterization, and utilization of wild apples from central Asia. Hortic. Rev. 29:1-61.

Forsline P. L., and Aldwinckle, H. S. 2004. Evaluation of Malus sieversii seedling populations for disease resistance and horticultural traits. Acta Hortic. 663:529–534.

Forsline, P. L., Luby, J. J., and Aldwinckle, H. S. 2008. Fire blight incidence on Malus sieversii grown in New York and Minnesota. Acta Hortic. 793:345-350.

Gardner, R. G., Cummins, J. N., and Aldwinckle, H. S. 1980. Fire blight resistance in the Geneva apple rootstock breeding program. J. Am. Soc. Hortic. Sci. 105:907-912.

Gardiner, S.E., J.L. Norelli, N. de Silva, G. Fazio, A. Peil, M. Malnoy, M. Horner, D. Bowatte, C. Carlisle, C. Wiedow, Y. Wan, C. L Bassett, A.M. Baldo, J. Celton, K. Richter, H.S. Aldwinckle, and V. Bus. 2012. Putative resistance gene markers associated with quantitative trait loci for fire blight resistance in Malus ‘Robusta 5’accessions. BMC genetics 13:25.

Gessler, C., and Pertot, I. 2012.Vf scab resistance of Malus. Trees 26: 95-108.

Janick, J. 1996. Apples. Pages 1-78 in: Fruit breeding, tree and tropical fruits. Vol. 1. J. Janick and J. N. Moore, eds. John Wiley & Sons, New Jersey.

Kellerhals, M., I. O. Baumgartner, L. Leumann, J. E. Frey, and A. Patocchi. 2011. Progress in pyramiding disease resistances in apple breeding. Acta Hortic. 976: 487-491.

Khan, M. A., C. Durel, B. Duffy, D. Drouet, M. Kellerhals, C. Gessler, and A. Patocchi. 2007. Development of molecular markers linked to the 'Fiesta' linkage group 7 major QTL for fire blight resistance and their application for marker-assisted selection. Genome 50:568- 577.

Khan, M. A., Y. F. Zhao, and S. S. Korban. 2012. Molecular mechanisms of pathogenesis and resistance to the bacterial pathogen Erwinia amylovora, causal agent of fire blight disease in Rosaceae. Plant Mol. Biol. Rep. 30:247-260.

Khanizadeh, S., Groleau, Y., Granger, R., Cousineau, J., and Rousselle, G.L. 2000. New hardy rootstocks from the Quebec apple breeding program. Acta Hortic. 538:719-721.

Le Roux, P., H. Flachowsky, M. Hanke, C. Gessler, and A. Patocchi. 2012. Use of a transgenic early flowering approach in apple (Malus × domestica Borkh.) to introgress fire blight resistance from cultivar Evereste. Mol. Breed. 30:857-874.

Le Roux, P. M., Flachowsky, H., Jänsch, M., Kellerhals, M., Patocchi, A., and Hanke, M-V. 2014. Development of Apple Pre-Breeding Genotypes Highly Resistant to Fire Blight by Early Flowering. Acta Hortic. 1048:55-64.

Luby, J., P. Forsline, H. Aldwinckle, V. Bus, V., and M. Geibel. 2000. Silk road apples— collection, evaluation, and utilization of Malus sieversii from Central Asia. HortScience 36:225-231.

Luby, J., Alspach, P. A., Bus, V. G. M., and Oraguzie, N. C. 2002. Field Resistance to fire blight in a diverse apple (Malus sp.) germplasm collection. J. Amer. Soc. Hort. Sci. 127:245- 253.

Luby, J. 2003. Taxonomic Classificatio and Brief History. Pages 1-14 in: Apples: botany, production and uses. D.C. Ferree and I. J. Warrington, eds. CABI publishing, Massachusetts.

McManus, P. S., and Jones, A. L. 1994. Epidemiology and genetic analysis of streptomycin resistant Erwinia amylovora from Michigan and evaluation of oxytetracycline for control. Phytopathology 84:627-633.

McManus, P.S., Stockwell, B. O., Sundin, G. W., and Jones, A. L. 2002. Antibiotic use in plant agriculture. Annu. Rev. Phytopathology 40: 443-465.

Momol, M. T., Aldwinckle, H. S., Forsline, P. L., and Lamboy, W. F. 1999. Fire blight resistance and horticultural evaluation of wild Malus populations from Central Asia. Acta Hotic. 489: 229-234.

Norelli, J. L., Aldwinckle, H. S., and Beer, S. V. 1988. Virulence of Erwinia amylovora strains to Malus sp. Novole plants grown in vitro and in the greenhouse. Phytopathology 78: 1292-1297.

Norelli, J. L., Holleran, H. T., Johnson, W. C., Robinson, T. L., and Aldwinckle, H. S. 2003. Resistance of Geneva and other apple rootstocks to Erwinia amylovora. Plant Dis. 87: 26-32.

Norelli, J. L., Jones, A. L., and Aldwinckle, H. S. 2003. Fire blight management in the twenty- first century. Plant Dis. 87:756-765.

Peil, A., Bus, V. G. M., Geider, K., Richter, K., Flachowsky, H. and Hanke, M-V. 2009. Improvement of fire blight resistance in apple and pear. Int. J. Plant Breed. 3:1-27.

Richards, C. M., Volk, G. M., Reeves, P. A., Reilley, A. A., Henk, A. D., Forsline, P. L., and Aldwinckle, H. S. 2009a. Selection of stratified core sets representing wild apple (Malus sieversii). J. American Soc. Hortic. Sci. 134:228-235.

Richards, C. M., Volk, G. M., Reilley, AHenk, A. D., Lockwook, D. R., Reeves, P. A, and Forsline, P. L. 2009b. Genetic diversity and population structure in Malus sieversii, a wild progenitor species of domesticated apple. Tree Genet. Genomes 5:339-347.

Russo, N. L., Burr, T. J., Breth, D. I., and. Aldwinckle, H. S. 2008. Isolation of Streptomycin- resistant isolates of Erwinia amylovora in New York. Plant Dis. 92:714-718.

Stockwell, V. O., Johnson, K. B., and Loper, J. E. 1998. Establishment of bacterial antagonists of Erwinia amylovora on pear and apple blossoms as influenced by inoculum preparation. Phytopathology 88:506-513.

USDA, National Organic Program (NOP). 2015. Sunset 2015 Amendments to the National List. Agricultural Marketing Service. 80:146. Online Publication. http://www.regulations.gov/ #!documentDetail;D=AMS-NOP-15-0015-0001

USDA, ARS, National Genetic Resources Program (NGRP). 2008. Germplasm Resources

Information Network (GRIN). [online database] National Germplasm Resources Laboratory, Beltsville, MD. 21 October 2015. . van Der Zwet, T., and Beer, S.V. 1992. Fire blight: its nature, prevention, and control: a practice guide to integrated disease management. Agriculture information bulletin (USA). van der Zwet, T., Orolaza-Halbrendt, N., and Zeller, W. 2012. Fire blight: history, biology, and management. American Phytopathological Society Press, Minnesota.

Volk, G. M., Chao, C. T., Norelli, J., Brown, S. K., Fazio, G., Peace, C., McFerson, J., Zhong, G-Y., and Bretting, P. 2015. The vulnerability of US apple (Malus) genetic resources. Genet. Resour. Crop Evol. 62:765–794.

Figure 1. Map showing 12 collection regions of Malus sieversii germplasm (Data from Forsline et al. 2003). As a reference, site 1 was at Latitude 39°N/ Longitude 68°E.

Figure 2: Evaluation of fire blight resistance of M. sieversii GMAL4028.h. A: In 34 of 35 fire blight shoots inoculated in 2013 and 2014 no evidence of infection could be observed 6 weeks after inoculation. B: In one of 35 fire blight inoculated shoots, fire blight progressed into 2 year- old wood of central leader.

A B

Arrows point to site of original fire Fire blight infection blight challenge in 2 yr-old wood of central leader

Figure 3. Average proportion of the shoot length blighted (SLB), including standard errors, after controlled inoculation with E.

amylovora for 64 accessions evaluated in both 2013 and 2014 and in both Wenatchee, WA and Kearneysville, WV.

2.00 1.50 1.00 0.50

0.00 SLB, WV 2013 WV SLB, -0.50 2.00 1.50 1.00 0.50

0.00 SLB, WV 2014 WV SLB, -0.50

71 2.00

1.50

1.00

0.50

SLB, WA 2013 WA SLB, 0.00 -0.50 2.00 1.50 1.00 0.50

0.00 SLB, WA 2014 WA SLB,

-0.50

Jonathan

PI633919 PI657095 PI633921 PI657041 PI657110 PI657058 PI657102 PI657081 PI657068 PI657107 PI657111 PI657097 PI657086 PI657114 PI657001 PI657088 PI657089 PI657076 PI633799 PI657047 PI657090 PI657118 PI657117 PI650945 PI657098 PI657119 PI637780

GMAL4002.i GMAL3999.i GMAL3635.i

GMAL3614.f

GMAL4039.s GMAL4051.s

GMAL3975.c GMAL4211.a GMAL3688.c GMAL3989.c GMAL3975.e GMAL3683.e GMAL3685.c GMAL3691.c GMAL4056.a

GMAL3616.o GMAL4002.k GMAL4047.h GMAL4211.d GMAL3683.k GMAL4028.k GMAL4053.b GMAL4032.p GMAL4053.o GMAL4051.q GMAL4047.q GMAL4047.o GMAL4047.k GMAL4309.g GMAL4039.v GMAL4056.p

GMAL4039.L GMAL4036.L

GMAL4054.m GMAL4051.m Golden Golden Delicious

Figure 4. Average fire blight infection for accessions separated by their collection site. Fire blight infection measured as the proportion of current season’s shoot length blighted after inoculation with E. amylovora. The number of accessions collected from each site is listed below the site number.

0.4 a

0.35 x

0.3 x x 0.25 b y b 0.2 y bc c 0.15 2013

ProportionSLB cd z 0.1 2014 d z 0.05

0 4 10 6 5 9 11 12 Collection Site

n=20 n=10 n= 37 n=30 n=29 n=32 n=29

Figure 5. Average fire blight infection for accessions by genetic cluster. Fire blight infection measured as the proportion of current season’s shoot length blighted after inoculation with E. amylovora. The number of accessions in each genetic cluster is listed below the site number.

0.40

0.35 x

0.30 a a 0.25 b y 0.20 y y b 0.15

2013 Proportion SLB 0.10 2014 0.05

0.00 1 2 3 4 Genetic Cluster

n=88 (2013) n=54 (2013) n= 26 n=16 n=87 (2014) n=53 (2014)

Table 1. Description of Central Asian collection sites for M. sieversii (Data from Forsline et al. 2003). * indicate the sites represented by accessions in this study.

Site Area Elevation Annual Climate Notes (m) Precipitation (mm) 1 Tajikistan NA NA NA

2 Uzbekistan NA NA NA

3*/8 Kazakhstan- 1170-1690 700 Humid temperate, mixed forest Zailisky

4* Kazakhstan- 1170-1760 800 Humid temperate, mixed forest Djungarsky

5* Kazakhstan- 1190-1360 850 Humid temperate, mixed forest Djungarsky

6* Kazakhstan- 600-910 250 Xeric, mixed scrub forest with Karatau diverse riparian habitat

7* Kyrgyzstan 1300-1500 1300 Very humid, temperate mixed forest 9* Kazakhstan- 870-1120 450 Dry continental forest (40°C Tarbagatai max/ -40 min)

10* Kazakhstan- 1600-1700 650 Semi-dry, temperate mixed Ketmen forest

11* Kazakhstan- 780-1230 250 Xeric mixed scrub forest with Karatau diverse riparian habitat

12* Kazakhstan- 1000-1025 320 Dry canyon, mixed forest, N. Talasky slope of 300 m canyon

Table 2. Congruence of resistance ratings based on natural occurrence of field infection

(GRIN) and ratings based on controlled inoculation in the field of selected M. sieversii accessions and controls in USDA-ARS Kearneysville, WV for both 2013 and 2014. GRIN resistance ratings: 1) very resistant - no occurrence; 2) moderately resistant – only light rating; 3) intermediate – light to medium rating; 4) moderately susceptible – medium to heavy rating; 5) very susceptible – very heavy rating. For the controlled inoculations, qualitative resistance ratings of highly susceptible, susceptible, moderately susceptible, resistant and highly resistant were assigned to accessions based on %SLB. See Supplemental

Table 1 for individual accession assignment.

GRIN Total Highly Resistant Resistant Moderate Susceptible Highly Rating Accessions % % % % Susceptible % 2013 2014 2013 2014 2013 2014 2013 2014 2013 2014 1 97 (2013) 15.5 18.9 13.4 21.1 19.6 22.1 32.0 22.1 19.6 15.8 95 (2014) 2 26 15.4 19.2 30.8 26.9 38.5 15.4 11.5 30.8 3.8 7.7 3 30 3.3 3.3 13.3 10.0 13.3 16.7 46.7 36.7 23.3 33.3 4 25 12.0 12.0 16.0 16.0 32.0 24.0 16.0 28.0 24.0 20.0 5 8 25.0 12.5 0.0 37.5 25.0 12.5 37.5 25.0 12.5 12.5

Table 3. Congruence of resistance ratings of selected M. sieversii accessions between controlled inoculation in the greenhouse at USDA-ARS-PGRU Geneva, NY and in the field at USDA-ARS Kearneysville, WV. Accessions were considered resistant if less than 20% of the total shoot length was blighted. The number of accessions rated as resistant and susceptible are presented for both 2013 and 2014.

NY Inoculations WV Inoculations 2013 2014 Resistant < 20 % SLB 67 67 (< 20 % SLB) > 20 % SLB 19 18 Susceptible < 20 % SLB 28 26 (> 20% SLB) > 20 % SLB 7 8

Supplementary Table 1. Genetic cluster, collection site, fire blight resistance (FB Res.) for both natural occurrence (_Field) and controlled inoculation (_GH Inoc.) at USDA-ARS-

PGRU Geneva, NY as well as number of shoots inoculated, average percent of current season’s shoot growth blighted (%SLB), and SE, Logit transformed %SLB and SE for both

WA and WV in 2013 and 2014.

West Virginia 2013 West Virginia 2014

Accession Genetic Cluster Collection Site FB Res._ Field FB Res._ Inoc. GH No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Rating Res. No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Rating Res. ------11 2.08 0.13 -5.01 0.90 WXYZ(2)ABCDEFG M Fiesta HLMNOPQRS GMAL3636.c 1 4 1 NA ------GMAL4290.g NA 12 5 NA ------PI633921 4 11 1 S 20 0.0 0.10 -5.54 0.64 (2)CDE HR 34 86.6 0.08 1.14 0.54 ABCDEFGHIJKLMN HS 4 11 2 R 20 -4.3 0.10 -5.50 0.64 (2)CDE HR 36 58.6 0.08 -2.03 0.53 NOPQRSTUVWXYZ( S GMAL4032.i 2)ABCDEFGHLM 1 9 3 NA 17 0.6 0.11 -5.60 0.69 (2)BCDE HR 33 16.7 0.08 -2.47 0.57 OPQRSTUVWXYZ(2 S GMAL4039.a )ABCDEFGHLMNOP Q 3 12 5 NA 20 1.2 0.10 -5.76 0.64 (2)CDE HR 39 7.1 0.08 -4.18 0.53 YZ(2)ABCDEFGHL M GMAL4296.e MNOPQRS 2 6 4 NA 18 0.5 0.10 -5.52 0.67 (2)BCDE HR 20 27.1 0.10 -4.18 0.67 XYZ(2)ABCDEFGHL M GMAL4054.ac MNOPQRS 4 11 1 S 20 5.2 0.10 -5.41 0.64 (2)BCDE HR 39 9.8 0.08 -4.26 0.52 Z(2)ABCDEFGHLM M

78 PI657047 NOPQRS

------20 3.3 0.10 -5.54 0.64 (2)CDE HR 22 6.7 0.10 -4.44 0.65 XYZ(2)ABCDEFGHL M Goldrush MNOPQRS 1 9 1 S 20 4.0 0.10 -5.66 0.64 (2)CDE HR 33 18.4 0.08 -4.44 0.57 (2)BCDEFGHLMNOP M PI657095 QRSU 1 5 1 R 20 2.3 0.10 -5.51 0.64 (2)CDE HR 32 9.4 0.08 -4.68 0.56 (2)CDEFGHLMNOP R PI657068 QRS 2 6 1 S 20 1.8 0.10 -5.58 0.64 (2)CDE HR 29 -0.4 0.08 -4.71 0.57 (2)CDEFGHLMNOP R PI657106 QRS 1 7 1 R 18 0.3 0.10 -6.20 0.67 (2)E HR 33 3.4 0.08 -4.75 0.55 (2)DEFGHLMNOPQ R PI633798 RS PI657113 1 5 1 S 20 -3.9 0.10 -5.87 0.64 (2)DE HR 39 5.5 0.08 -4.92 0.52 (2)GHLMNOPQRS R GMAL4179.b 1 5 4 NA 17 0.8 0.10 -5.49 0.68 Z(2)BCDEF HR 36 1.8 0.08 -5.09 0.53 (2)LMNOPQRS HR PI657044 1 5 1 S 20 -2.5 0.10 -5.46 0.64 (2)BCDE HR 36 -0.3 0.08 -5.18 0.54 (2)LMNOPQRS HR GMAL4211.a 3 12 4 NA 20 2.1 0.10 -5.66 0.64 (2)CDE HR 28 -3.2 0.09 -5.19 0.62 (2)GHLMNOPQRS R 4 11 1 NA 22 -1.4 0.09 -5.65 0.62 (2)CDE HR 20 -1.8 0.11 -5.21 0.72 (2)ACDEFGHLMNO M GMAL3544.b PQRSU GMAL4211.e 1 4 5 NA 20 -0.3 0.10 -5.92 0.64 (2)DE HR 35 1.1 0.08 -5.21 0.57 (2)LMNOPQRS R 1 6 1 R 19 2.4 0.10 -5.58 0.65 (2)CDE HR 8 0.0 0.15 -5.27 1.01 VWXYZ(2)ABCDEF M PI657014 GHLMNOPQRST

Cont’d West Virginia 2013 West Virginia 2014

Accession Genetic Cluster Collection Site FB Res._ Field FB Res._ Inoc. GH No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Rating Res. No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Rating Res. GMAL4002.k 2 6 2 R 20 1.4 0.10 -5.50 0.64 (2)BCDE HR 40 1.8 0.08 -5.45 0.53 (2)NOPQRS HR GMAL3999.i NA 6 2 S 19 -3.3 0.10 -5.61 0.66 (2)CDE HR 38 -7.1 0.08 -5.47 0.54 (2)NOPQRS HR PI657085 2 12 1 R 19 -2.8 0.10 -5.65 0.65 (2)CDE HR 37 0.2 0.08 -5.48 0.53 (2)OPQRS HR PI633919 1 10 1 S 20 -1.5 0.10 -5.50 0.64 (2)CDE HR 38 1.5 0.08 -5.50 0.52 (2)OPQRS HR GMAL4028.h 1 5 2 NA 15 -4.6 0.11 -6.19 0.73 (2)DE HR 15 0.7 0.11 -5.64 0.74 (2)GHLMNOPQRS R PI657116 2 11 1 R 15 -0.6 0.11 -5.72 0.73 (2)BCDE HR 15 1.0 0.12 -5.64 0.78 (2)EFGHLMNOPQRS R PI657115 1 5 1 R 20 0.4 0.10 -5.73 0.64 (2)CDE HR ------PI657100 2 6 1 R 16 3.1 0.10 -5.63 0.70 (2)BCDE HR ------2 6 4 NA 20 114.9 0.10 4.58 0.64 AB HS 28 147. 0.09 4.29 0.59 A HS GMAL4056.p 2 2 6 4 NA 20 164.9 0.10 5.17 0.64 A HS 16 171. 0.11 4.28 0.76 ABC HS GMAL4056n 4 1 9 1 R 20 100.1 0.10 3.14 0.64 ABC HS 37 139. 0.08 3.99 0.53 AB HS PI637780 1

79 NA 9 5 NA 20 99.6 0.10 1.71 0.64 ABCDEFGH HS 26 149. 0.09 3.41 0.60 ABCD HS

GMAL4238.c 0 1 11 1 R 20 111.9 0.10 0.62 0.64 BCDEFGHIJKLM HS 39 126. 0.08 2.99 0.55 ABCDE HS GMAL3682.i 9 GMAL4039.v 3 12 3 NA 20 78.1 0.10 0.30 0.64 CDEFGHIJKLMN HS 30 99.8 0.08 2.81 0.58 ABCDEF HS NA 12 NA R 20 140.5 0.10 5.04 0.64 A HS 32 118. 0.08 2.80 0.57 ABCDEF HS GMAL4311.c 9 GMAL4047.L 2 4 3 NA 15 76.8 0.10 3.03 0.73 ABCD HS 26 48.8 0.08 2.68 0.60 ABCDEFG HS 1 9 1 R 20 82.6 0.10 2.41 0.64 ABCDEFG HS 25 116. 0.09 2.41 0.60 ABCDEFGHIJ HS PI657098 7 1 9 1 R 20 39.7 0.10 -1.26 0.64 DEFGHIJKLMNO HS 39 92.1 0.08 2.24 0.52 ABCDEFGHI HS PI657090 PQRSTUVWXY(2) A GMAL4312.d NA 4 NA S 20 80.8 0.10 2.47 0.64 ABCDE HS 39 79.1 0.08 2.01 0.52 ABCDEFGHJ HS GMAL3638.c 1 4 1 R 20 85.8 0.10 1.14 0.64 ABCDEFGHIJKL HS 40 79.9 0.08 1.73 0.52 ABCDEFGHIJK HS 1 4 1 R 20 44.1 0.10 -0.69 0.64 CDEFGHIJKLMN HS 39 70.6 0.08 1.69 0.54 ABCDEFGHIJKL HS GMAL3631.m OPQRSTUVW GMAL4024.u 1 5 2 R 20 101.1 0.10 1.27 0.64 ABCDEFGHIJK HS 37 78.9 0.08 1.40 0.53 ABCDEFGHIJKLM HS 2 11 1 S 20 97.5 0.10 0.30 0.64 CDEFGHIJKLMN HS 25 126. 0.09 1.10 0.61 ABCDEFGHIJKLMN HS PI657119 O 6 OP

Cont’d West Virginia 2013 West Virginia 2014

Rating

Accession Genetic Cluster Collection Site FB Res._ Field FB Res._ Inoc. GH No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Rating Res. No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Res. 1 9 4 NA 20 49.3 0.10 -0.06 0.64 CDEFGHIJKLMN HS 36 74.4 0.08 0.95 0.54 ABCDEFGHIJKLMN HS GMAL4056.a OPQR O 2 6 1 R 20 73.4 0.10 1.40 0.64 ABCDEFGHIJ HS 37 67.2 0.08 0.85 0.54 ABCDEFGHIJKLMN HS PI650945 OP 1 5 3 NA 15 41.5 0.11 -1.60 0.73 DEFGHIJKLMNO HS 29 64.5 0.09 0.66 0.58 ABCDEFGHIJKLMN HS GMAL4047.q PQRSTUVWXYZ( OPQ 2)ABCD 1 4 3 NA 20 34.8 0.10 -2.08 0.64 HIJKLMNOPQRS HS 22 88.9 0.09 0.34 0.64 BCDEFGHIJKLMNO HS GMAL4053.o TUVWXYZ(2)AB PQRS CDE 1 6 1 R 18 38.8 0.10 -1.18 0.67 DEFGHIJKLMNO HS 37 51.9 0.08 0.19 0.54 CDEFGHIJKLMNOP HS PI657099 PQRSTUVWXYZ QR 1 5 3 NA 20 44.5 0.10 -1.98 0.64 HIJKLMNOPQRS HS 30 62.5 0.08 -0.03 0.57 DEFGHIJKLMNOPQ HS GMAL4047.o TUVWXYZ(2)AB RSTU CD 1 10 3 NA 20 23.7 0.10 1.60 0.64 ABCDEFGHI HS 34 10.1 0.08 -0.31 0.54 EFGHIJKLMNOPQR HS GMAL4047.k STUVW

80 4 12 3 NA 20 31.6 0.10 -1.41 0.64 EFGHIJKLMNOP HS 32 43.8 0.08 -0.60 0.56 FGHIJKLMNOPQRS HS

GMAL4039.y QRSTUVWXYZ(2) TUVWX AB 3 12 1 S 18 50.5 0.10 -0.44 0.67 CDEFGHIJKLMN HS 35 42.3 0.08 -0.99 0.53 HIJKLMNOPQRSTU HS GMAL3610.c OPQRSTU VWXYZ(2)B 1 4 1 R 20 40.9 0.10 -0.49 0.64 CDEFGHIJKLMN HS 31 41.4 0.08 -1.24 0.57 JKLMNOPQRSTUV S GMAL3635.i OPQRST WXYZ(2)ABCDI 3 12 NA S 20 74.8 0.10 0.16 0.64 CDEFGHIJKLMN HS 39 37.1 0.08 -1.56 0.52 LMNOPQRSTUVWX S GMAL4309.g OPQ YZ(2)ABCDE 1 9 1 R 15 33.4 0.11 -2.32 0.72 HIJKLMNOPQRS HS 21 20.3 0.10 -2.22 0.66 MNOPQRSTUVWXY S PI633799 TUVWXYZ(2)AB Z(2)ABCDEFGHLM CDE NOPQ 4 11 4 NA 15 118.3 0.11 2.76 0.74 ABCDEF HS 21 28.3 0.10 -2.41 0.68 MNOPQRSTUVWXY S GMAL4054.g Z(2)ABCDEFGHLM NOPQRS 1 11 1 R 20 26.5 0.10 -2.20 0.64 HIJKLMNOPQRS HS 32 28.7 0.08 -2.55 0.56 PQRSTUVWXYZ(2) S PI657076 TUVWXYZ(2)AB ABCDEFGHLMNOP CDE QR 3 12 1 S 20 35.0 0.10 -1.69 0.64 HIJKLMNOPQRS HS 32 23.7 0.08 -2.82 0.56 QRSTUVWXYZ(2)A S GMAL3614.e TUVWXYZ(2)AB BCDEFGHLMNOPQ C RS 1 4 1 R 20 106.5 0.10 2.50 0.64 ABCDE HS 26 36.4 0.09 -2.82 0.60 QRSTUVWXYZ(2)A S PI633920 BCDEFGHLMNOPQ RS

Cont’d West Virginia 2013 West Virginia 2014

Accession Genetic Cluster Collection Site FB Res._ Field FB Res._ Inoc. GH No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Res. Rating No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Res. Rating 2 6 4 NA 10 20.8 0.14 -2.60 0.90 GHIJKLMNOPQR HS 12 35.9 0.12 -2.99 0.81 NOPQRSTUVWXYZ( S GMAL4054.c STUVWXYZ(2)AB 2)ABCDEFGHLMNO CDE PQRS 1 9 4 NA 19 29.9 0.10 -1.71 0.65 FGHIJKLMNOPQ HS 28 40.1 0.09 -3.08 0.59 RSTUVWXYZ(2)AB S GMAL4054.t RSTUVWXYZ(2)A CDEFGHLMNOPQR BC S 2 11 1 R 20 48.8 0.10 -0.61 0.64 CDEFGHIJKLMN HS 34 2.2 0.08 -4.00 0.56 XYZ(2)ABCDEFGHL M PI657118 OPQRSTUV MNOPQRS 2 11 1 NA 20 47.4 0.10 -0.43 0.64 CDEFGHIJKLMN HS 38 -7.4 0.08 -4.38 0.53 (2)ACDEFGHLMNO M PI657117 OPQRS PQRSU 4 12 1 R 20 41.4 0.10 -0.99 0.64 DEFGHIJKLMNO HS 37 5.7 0.08 -4.38 0.53 (2)ACDEFGHLMNO M GMAL3688.j PQRSTUVWX PQRSU ------19 188.5 0.10 0.28 0.65 CDEFGHIJKLMN HS 18 8.4 0.10 -4.43 0.70 XYZ(2)ABCDEFGHL M Jonathan OP MNOPQRS 4 12 1 S 20 26.1 0.10 -2.01 0.64 HIJKLMNOPQRS HS 36 8.3 0.08 -4.49 0.53 (2)ACDEFGHLMNO M PI657082 TUVWXYZ(2)AB PQRSU CD 1 4 2 R 20 19.6 0.10 -4.18 0.64 STUVWXYZ(2)AB M 39 82.8 0.08 1.69 0.52 ABCDEFGHIJKL HS 81 GMAL4032.p CDE 3 12 4 NA 20 10.6 0.10 -4.25 0.64 STUVWXYZ(2)AB M 37 33.0 0.08 -1.19 0.54 JKLMNOPQRSTUV S GMAL4211.b CDE WXYZ(2)ABIK 3 -1.1 0.13 M 36.1 0.14 GHIJKLMNOPQRST HS GMAL4039.w TUVWXYZ(2)AB UVWXYZ(2)ABCDE 3 12 NA 10 -5.36 0.88 CDE 11 -2.18 0.91 FGHLMNOPQRS NA 5 3 NA 15 -5.9 0.11 -5.05 0.74 WXYZ(2)ABCDE M 30 7.4 0.08 -2.45 0.58 OPQRSTUVWXYZ(2 S GMAL4051.t )ABCDEFGHLMNOP Q 2 10 2 R 12 11.8 0.12 -4.74 0.80 STUVWXYZ(2)AB M 31 33.9 0.08 -3.10 0.56 RSTUVWXYZ(2)AB S GMAL4011.u CDE CDEFGHLMNOPQR S 3 12 2 R 20 1.5 0.10 -4.43 0.64 STUVWXYZ(2)AB M 38 10.5 0.08 -3.12 0.53 RSTUVWXYZ(2)AB S GMAL4032.m CDE CDEFGHLMNOPQR S 2 3 4 NA 19 8.2 0.10 -4.74 0.65 WXYZ(2)ABCDE M 33 16.2 0.08 -3.14 0.56 RSTUVWXYZ(2)AB S GMAL4054.q CDEFGHLMNOPQR S 2 6 1 S 20 8.4 0.10 -4.63 0.64 VWXYZ(2)ABCD M 40 14.7 0.08 -3.19 0.52 STUVWXYZ(2)ABC S PI657017 E DEFGHLMNOPQRS 3 11 5 S 20 6.7 0.10 -4.40 0.64 STUVWXYZ(2)AB M 40 8.9 0.08 -3.38 0.53 TUVWXYZ(2)ABCD S GMAL4290.f CDE EFGHLMNOPQRS

Cont’d West Virginia 2013 West Virginia 2014

Accession Genetic Cluster Collection Site FB Res._ Field FB Res._ Inoc. GH No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Rating Res. No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Rating Res. 1 5 2 R 19 7.7 0.10 -4.62 0.65 TUVWXYZ(2)AB M 33 12.5 0.08 -3.41 0.57 STUVWXYZ(2)ABC S GMAL4011.x CDE DEFGHLMNOPQRS 3 12 4 NA 20 6.4 0.10 -4.85 0.64 XYZ(2)ABCDE M 12 24.2 0.13 -3.43 0.85 OPQRSTUVWXYZ(2 S GMAL4054.m )ABCDEFGHLMNOP QRS 1 9 3 NA 20 0.9 0.10 -4.97 0.64 XYZ(2)ABCDE M 27 16.1 0.09 -3.65 0.59 UWXYZ(2)ABCDEF S GMAL4035.t GHLMNOPQRST 1 5 1 S 20 10.4 0.10 -4.46 0.64 STUVWXYZ(2)AB M 32 16.1 0.08 -3.68 0.57 WXYZ(2)ABCDEFG S PI657086 CDE HLMNOPQRS 1 5 1 R 20 8.0 0.10 -4.49 0.64 STUVWXYZ(2)AB M 38 6.7 0.08 -3.85 0.52 XYZ(2)ABCDEFGHL M PI657114 CDE MNOPQRS 1 10 3 NA 12 -0.1 0.12 -5.00 0.81 STUVWXYZ(2)AB M 16 20.9 0.11 -4.16 0.77 STUVWXYZ(2)ABC S GMAL4039.s CDE DEFGHLMNOPQRS 1 9 4 NA 9 14.2 0.14 -4.87 0.91 QRSTUVWXYZ(2) M 16 9.8 0.11 -4.30 0.75 VWXYZ(2)ABCDEF S GMAL4055.w ABCDE GHLMNOPQRST 1 4 2 S 19 8.9 0.10 -4.73 0.65 WXYZ(2)ABCDE M 39 -0.6 0.08 -4.30 0.54 Z(2)ABCDEFGHLM M GMAL4032.r NOPQRS

82 2 11 2 S 20 5.4 0.10 -4.59 0.64 UVWXYZ(2)ABC M 38 1.2 0.08 -4.44 0.53 (2)ACDEFGHLMNO M

GMAL4028.k DE PQRSU

4 7 NA NA 20 0.9 0.10 -4.99 0.64 XYZ(2)ABCDE M 37 8.2 0.08 -4.56 0.55 (2)ACDEFGHLMNO M GMAL4309.f PQRSU 2 6 2 S 15 7.2 0.11 -4.70 0.74 STUVWXYZ(2)AB M 26 3.5 0.09 -4.59 0.62 Z(2)ABCDEFGHLM M GMAL4002.p CDE NOPQRS 4 11 1 R 19 4.8 0.10 -4.87 0.65 XYZ(2)ABCDE M 35 7.5 0.08 -4.64 0.54 (2)CDEFGHLMNOP R GMAL3552.v QRS PI657101 1 6 1 R 20 2.1 0.10 -4.78 0.64 WXYZ(2)ABCDE M 32 0.8 0.08 -4.79 0.55 (2)EFGHLMNOPQRS R 1 9 1 R 13 2.9 0.12 -5.21 0.79 WXYZ(2)ABCDE M 19 -0.1 0.10 -4.84 0.70 Z(2)ABCDEFGHLM M PI657041 NOPQRS 3 12 1 R 7 2.6 0.16 -5.40 1.05 QRSTUVWXYZ(2) M 13 7.3 0.11 -4.85 0.76 YZ(2)ABCDEFGHL M GMAL3608.h ABCDE MNOPQRS PI657008 1 9 1 S 19 2.2 0.10 -4.87 0.65 XYZ(2)ABCDE M 38 2.0 0.08 -4.86 0.52 (2)FGHLMNOPQRS R 2 10 1 R 17 5.6 0.10 -4.57 0.69 STUVWXYZ(2)AB M 26 3.6 0.09 -4.92 0.64 (2)CDEFGHKLMNO R GMAL3619.a CDE PQRS PI657056 1 3 1 R 20 6.4 0.10 -4.90 0.64 XYZ(2)ABCDE M 36 3.1 0.08 -4.94 0.53 (2)GHLMNOPQRS R 2 11 1 R 18 7.8 0.10 -4.34 0.67 STUVWXYZ(2)AB M 25 -4.4 0.09 -4.95 0.62 (2)CDEFGHLMNOP R GMAL3683.k CDE QRS 1 5 5 NA 20 8.5 0.10 -4.80 0.64 XYZ(2)ABCDE M 36 2.0 0.08 -4.97 0.53 (2)GHLMNOPQRS R GMAL4211.g

Cont’d West Virginia 2013 West Virginia 2014

Accession Genetic Cluster Collection Site FB Res._ Field FB Res._ Inoc. GH No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Res. Rating No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Res. Rating 3 11 1 S 15 2.1 0.11 -5.00 0.73 VWXYZ(2)ABCD M 37 4.6 0.08 -5.06 0.53 (2)LMNOPQRS HR PI657049 E GMAL4177.d 1 10 4 NA 19 4.1 0.10 -4.97 0.66 XYZ(2)ABCDE M 27 -0.9 0.09 -5.08 0.62 (2)EFGHLMNOPQRS R GMAL3608.n 3 12 1 NA 20 4.3 0.10 -4.87 0.64 XYZ(2)ABCDE M 34 0.8 0.08 -5.12 0.56 (2)HLMNOPQRS R 1 1 R 20 8.9 0.10 -4.49 0.64 STUVWXYZ(2)AB M 32 1.7 0.08 -5.17 0.57 (2)LMNOPQRS HR PI657064 4 CDE 1 11 1 R 12 8.3 0.12 -4.68 0.78 STUVWXYZ(2)AB M 28 3.2 0.09 -5.18 0.58 (2)LMNOPQRS HR PI657073 CDE 2 6 2 R 15 -1.3 0.11 -4.55 0.73 STUVWXYZ(2)AB M 26 0.6 0.09 -5.20 0.64 (2)EFGHLMNOPQRS R GMAL4002.i CDE 1 9 4 NA 14 5.6 0.11 -5.02 0.75 VWXYZ(2)ABCD M 9 9.8 0.15 -5.22 1.03 TUVWXYZ(2)ABCD M GMAL4055.m E EFGHLMNOPQRS 2 6 1 R 13 -0.3 0.12 -4.91 0.77 STUVWXYZ(2)AB M 33 - 0.08 -5.25 0.58 (2)LMNOPQRS HR PI657009 CDE 10.2 1 11 1 R 18 4.6 0.10 -4.59 0.67 STUVWXYZ(2)AB M 38 -4.2 0.08 -5.26 0.54 (2)MNOPQRS HR GMAL3685.c CDE ------20 5.3 0.10 -4.55 0.64 TUVWXYZ(2)AB M 39 -2.2 0.08 -5.26 0.53 (2)MNOPQRS HR 83 Empire CDE GMAL3781.a 1 9 2 R 20 3.4 0.10 -4.71 0.64 WXYZ(2)ABCDE M 33 -0.3 0.08 -5.28 0.55 (2)LMNOPQRS R PI633801 4 11 1 R 20 8.1 0.10 -4.98 0.64 XYZ(2)ABCDE M 36 1.9 0.08 -5.35 0.54 (2)MNOPQRS HR 1 9 4 NA 20 3.9 0.10 -4.62 0.64 UVWXYZ(2)ABC M 34 -0.1 0.08 -5.37 0.56 (2)MNOPQRS HR GMAL4054.z DE 2 6 4 R 18 7.5 0.10 -4.62 0.67 STUVWXYZ(2)AB M 30 3.3 0.09 -5.39 0.58 (2)MNOPQRS HR GMAL4177.b CDE GMAL3688.c 2 11 1 R 14 -0.5 0.11 -5.10 0.75 WXYZ(2)ABCDE M 27 -0.2 0.09 -5.57 0.60 (2)MNOPQRS HR GMAL3975.e 2 6 2 S 20 3.1 0.10 -4.79 0.64 WXYZ(2)ABCDE M 37 -2.4 0.08 -5.93 0.53 (2)QRS HR 2 11 1 R 20 2.8 0.10 -5.06 0.64 XYZ(2)ABCDE R 30 20.9 0.08 -1.98 0.58 MNOPQRSTUVWXY S GMAL3683.e Z(2)ABCDEFGHLM N 1 9 3 NA 20 139.9 0.11 -5.23 0.64 YZ(2)ABCDE R 37 108. 0.09 -2.42 0.54 OPQRSTUVWXYZ(2 S GMAL4047.h 1 )ABCDEFGHLMNOP 4 11 3 NA 20 3.7 0.10 -5.24 0.64 YZ(2)ABCDE R 36 25.4 0.08 -2.68 0.53 QRSTUVWXYZ(2)A S GMAL4053.k BCDEFGHLMNOPQ R 1 9 2 S 18 1.2 0.10 -5.41 0.67 YZ(2)ABCDE R 38 8.3 0.08 -3.30 0.53 STUVWXYZ(2)ABC S GMAL3691.g DEFGHLMNOPQRS ------19 5.1 0.10 -5.10 0.67 XYZ(2)ABCDE R 34 28.5 0.08 -3.72 0.55 WXYZ(2)ABCDEFG S Gala HLMNOPQRS

Cont’d West Virginia 2013 West Virginia 2014

84 1 4 3 S 14 0.9 0.12 -5.32 0.76 XYZ(2)ABCDE R 32 8.3 0.08 -4.66 0.56 (2)CDEFGHLMNOP R GMAL4032.t QRS 1 9 2 R 20 -0.4 0.10 -5.07 0.65 XYZ(2)ABCDE R 39 0.9 0.08 -4.67 0.54 (2)CDEFGHLMNOP R GMAL3975.c QRS 2 6 2 R 19 0.2 0.10 -5.10 0.66 XYZ(2)ABCDE R 38 4.4 0.08 -4.75 0.56 (2)DEFGHLMNOPQ R GMAL3989.c RS Golden ------14 6.3 0.11 -5.37 0.75 XYZ(2)ABCDE R 36 2.3 0.08 -4.80 0.53 (2)EFGHLMNOPQRS R Delicious GMAL3691.i 1 5 2 R 20 -2.3 0.10 -5.37 0.64 (2)BCDE R 34 3.2 0.08 -4.83 0.55 (2)EFGHLMNOPQRS R GMAL3691.c 1 9 2 R 20 9.7 0.10 -5.13 0.64 YZ(2)ABCDE R 38 3.2 0.08 -4.83 0.53 (2)EFGHLMNOPQRS R PI657109 1 10 1 R 20 -7.4 0.10 -5.35 0.64 YZ(2)ABCDE R 37 16.5 0.08 -4.86 0.53 (2)EFGHLMNOPQRS R GMAL4198.d 1 5 4 NA 20 3.6 0.10 -5.28 0.64 YZ(2)ABCDE R 38 2.8 0.08 -4.91 0.53 (2)GHLMNOPQRS R GMAL4211.c 1 4 1 NA 20 1.7 0.10 -5.02 0.64 XYZ(2)ABCDE R 31 0.6 0.08 -5.03 0.57 (2)GHLMNOPQRS R PI657054 2 10 1 R 20 0.0 0.10 -5.26 0.64 YZ(2)ABCDE R 40 0.2 0.08 -5.07 0.52 (2)LMNOPQRS HR GMAL4059.a 2 6 4 NA 15 -2.8 0.11 -5.55 0.74 YZ(2)ABCDE R 31 2.4 0.08 -5.08 0.57 (2)HLMNOPQRS R GMAL4051.o 1 5 3 NA 20 2.7 0.10 -5.14 0.64 YZ(2)ABCDE R 35 -0.7 0.08 -5.09 0.53 (2)LMNOPQRS HR PI633923 1 5 1 R 18 0.9 0.10 -5.20 0.67 YZ(2)ABCDE R 35 0.3 0.08 -5.16 0.54 (2)LMNOPQRS HR GMAL4011.f 2 6 2 R 20 6.6 0.10 -5.13 0.64 YZ(2)ABCDE R 28 -2.3 0.09 -5.30 0.62 (2)LMNOPQRS HR PI657081 4 12 1 S 20 4.6 0.10 -5.14 0.64 YZ(2)ABCDE R 36 -1.4 0.08 -5.51 0.54 (2)OPQRS HR

Cont’d West Virginia 2013 West Virginia 2014

85 1 9 3 NA 20 25.5 0.10 -2.77 0.64 KLMNOPQRSTU S 38 35.0 0.08 -0.99 0.54 HIJKLMNOPQRSTU HS

GMAL4036.L VWXYZ(2)ABCD VWXYZ(2)B E 2 6 1 R 15 10.3 0.11 -3.14 0.74 KLMNOPQRSTU S 37 35.0 0.08 -1.04 0.52 IKLMNOPQRSTUV HS PI657105 VWXYZ(2)ABCD WXYZ(2)BI E 1 4 1 R 19 16.5 0.10 -3.39 0.66 MNOPQRSTUVW S 26 39.2 0.09 -1.17 0.64 GHIJKLMNOPQRST HS GMAL3637.d XYZ(2)ABCDE UVWXYZ(2)ABCDE F 3 10 2 R 18 8.7 0.10 -3.62 0.67 NOPQRSTUVWX S 37 34.4 0.08 -1.18 0.54 JKLMNOPQRSTUV S GMAL4011.w YZ(2)ABCDE WXYZ(2)AIJ 2 6 1 R 15 5.7 0.11 -3.87 0.74 NOPQRSTUVWX S 39 40.9 0.08 -1.30 0.53 KLMNOPQRSTUVW S PI657107 YZ(2)ABCDE XYZ(2)ABC 2 3 1 R 17 15.0 0.10 -3.05 0.68 LMNOPQRSTUV S 21 46.9 0.10 -1.42 0.67 HIJKLMNOPQRSTU HS GMAL3623.i WXYZ(2)ABCDE VWXYZ(2)ABCDEF GHL 3 11 3 NA 14 11.4 0.11 -3.18 0.75 KLMNOPQRSTU S 34 27.0 0.08 -1.59 0.55 KLMNOPQRSTUVW S GMAL4039.t VWXYZ(2)ABCD XYZ(2)ABCDEFG E 1 9 3 NA 17 8.9 0.10 -3.76 0.68 NOPQRSTUVWX S 20 25.0 0.10 -1.80 0.65 KLMNOPQRSTUVW S GMAL4039.d YZ(2)ABCDE XYZ(2)ABCDEFGHL MNO

Cont’d West Virginia 2013 West Virginia 2014

Accession Genetic Cluster Collection Site FB Res._ Field FB Res._ Inoc. GH No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Res. Rating No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Res. Rating 4 12 3 R 18 14.9 0.10 -3.37 0.67 MNOPQRSTUVW S 26 38.4 0.09 -2.01 0.61 MNOPQRSTUVWXY S GMAL4032.w XYZ(2)ABCDE Z(2)ABCDEFGHLM NO 3 12 1 R 20 13.6 0.10 -3.98 0.64 RSTUVWXYZ(2)A S 34 18.6 0.08 -2.05 0.54 NOPQRSTUVWXYZ( S PI633802 BCDE 2)ABCDEFGHLM 2 11 5 NA 20 18.5 0.10 -3.09 0.64 MNOPQRSTUVW S 37 19.4 0.08 -2.38 0.54 OPQRSTUVWXYZ(2 GMAL4290.e XYZ(2)ABCDE )ABCDEFGHLMNOP S 2 6 2 NA 20 19.5 0.10 -3.15 0.64 MNOPQRSTUVW S 30 24.2 0.08 -2.49 0.57 OPQRSTUVWXYZ(2 S GMAL4068.b XYZ(2)ABCDE )ABCDEFGHLMNOP Q 3 12 1 R 20 13.5 0.10 -3.28 0.64 MNOPQRSTUVW S 34 17.4 0.08 -2.75 0.56 QRSTUVWXYZ(2)A S GMAL3607.k XYZ(2)ABCDE BCDEFGHLMNOPQ RS 4 11 3 NA 14 9.9 0.11 -4.41 0.75 RSTUVWXYZ(2)A S 31 13.7 0.08 -2.78 0.56 QRSTUVWXYZ(2)A S GMAL4053.b BCDE BCDEFGHLMNOPQ RS 1 9 1 R 20 15.0 0.10 -3.08 0.64 MNOPQRSTUVW S 38 9.2 0.08 -2.89 0.53 RSTUVWXYZ(2)AB S PI657089 XYZ(2)ABCDE CDEFGHLMNOPQR 86 S 3 12 1 R 13 16.6 0.12 -3.75 0.78 MNOPQRSTUVW S 29 18.5 0.09 -2.99 0.59 QRSTUVWXYZ(2)A S PI633922 XYZ(2)ABCDE BCDEFGHLMNOPQ RS 1 5 1 S 20 15.5 0.10 -3.63 0.64 NOPQRSTUVWX S 34 14.1 0.08 -2.99 0.56 RSTUVWXYZ(2)AB S PI657067 YZ(2)ABCDE CDEFGHLMNOPQR S 1 4 1 R 18 13.8 0.10 -3.85 0.67 NOPQRSTUVWX S 34 25.1 0.08 -3.18 0.55 RSTUVWXYZ(2)AB S GMAL3629.m YZ(2)ABCDE CDEFGHLMNOPQR S 2 11 1 R 15 7.4 0.11 -3.29 0.74 LMNOPQRSTUV S 27 18.6 0.09 -3.24 0.61 RSTUVWXYZ(2)AB S GMAL3684.h WXYZ(2)ABCDE CDEFGHLMNOPQR S 2 6 3 NA 20 1.7 0.10 -2.64 0.64 JKLMNOPQRSTU S 38 18.7 0.08 -3.30 0.53 STUVWXYZ(2)ABC S GMAL4047.j VWXYZ(2)ABCD DEFGHLMNOPQRS E 2 11 1 NA 14 10.2 0.11 -3.94 0.75 NOPQRSTUVWX S 13 13.9 0.12 -3.32 0.83 NOPQRSTUVWXYZ( S PI657120 YZ(2)ABCDE 2)ABCDEFGHLMNO PQRS 2 11 1 R 20 5.4 0.10 -4.14 0.64 RSTUVWXYZ(2)A S 30 27.7 0.08 -3.53 0.57 TUVWXYZ(2)ABCD S GMAL3607.h BCDE EFGHLMNOPQRS 2 6 2 R 19 17.4 0.10 -3.51 0.65 NOPQRSTUVWX S 37 15.2 0.08 -3.55 0.53 WXYZ(2)ABCDEFG S GMAL4011.i YZ(2)ABCDE HLMNOPQRS

Cont’d West Virginia 2013 West Virginia 2014

Accession Genetic Cluster Collection Site FB Res._ Field FB Res._ Inoc. GH No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Rating Res. No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Rating Res. 3 11 1 R 20 22.5 0.10 -2.55 0.64 JKLMNOPQRSTU S 36 11.7 0.08 -3.58 0.55 VWXYZ(2)ABCDEF S PI657069 VWXYZ(2)ABCD GHLMNOPQRST E 2 6 1 R 20 27.6 0.10 -3.41 0.64 MNOPQRSTUVW S 39 10.1 0.08 -3.70 0.53 XYZ(2)ABCDEFGHL M PI657021 XYZ(2)ABCDE MNOPQRS 4 11 1 R 20 15.4 0.10 -3.73 0.64 NOPQRSTUVWX S 34 14.7 0.08 -3.73 0.56 WXYZ(2)ABCDEFG S GMAL3544.j YZ(2)ABCDE HLMNOPQRS 1 5 1 NA 19 8.5 0.10 -2.83 0.65 KLMNOPQRSTU S 32 9.8 0.08 -3.82 0.56 XYZ(2)ABCDEFGHL M PI657112 VWXYZ(2)ABCD MNOPQRS E 1 5 3 NA 20 7.0 0.10 -3.87 0.64 QRSTUVWXYZ(2) S 36 3.1 0.08 -3.92 0.54 XYZ(2)ABCDEFGHL M GMAL4051.p ABCDE MNOPQRS 1 5 3 NA 20 28.9 0.10 -2.65 0.64 JKLMNOPQRSTU S 35 12.3 0.08 -3.93 0.55 XYZ(2)ABCDEFGHL M GMAL4051.q VWXYZ(2)ABCD MNOPQRS E 1 5 3 NA 20 19.1 0.10 -3.34 0.64 MNOPQRSTUVW S 29 2.8 0.09 -3.99 0.58 XYZ(2)ABCDEFGHL M GMAL4051.s XYZ(2)ABCDE MNOPQRS

87 2 12 1 R 16 40.2 0.11 -2.89 0.71 JKLMNOPQRSTU S 22 0.6 0.09 -4.04 0.64 WXYZ(2)ABCDEFG S

GMAL3689.g VWXYZ(2)ABCD HLMNOPQRS E 1 5 1 R 20 11.1 0.10 -3.49 0.64 NOPQRSTUVWX S 36 9.6 0.08 -4.12 0.54 YZ(2)ABCDEFGHL M PI657088 YZ(2)ABCDE MNOPQRS 2 9 1 R 16 16.2 0.11 -3.37 0.71 MNOPQRSTUVW S 37 2.7 0.08 -4.13 0.52 YZ(2)ABCDEFGHL M PI657096 XYZ(2)ABCDE MNOPQRS 1 5 4 NA 20 2.4 0.10 -4.00 0.64 RSTUVWXYZ(2)A S 35 5.6 0.08 -4.20 0.54 YZ(2)ABCDEFGHL M GMAL4177.e BCDE MNOPQRS 1 5 3 NA 19 28.7 0.10 -3.68 0.65 NOPQRSTUVWX S 34 13.7 0.08 -4.23 0.54 YZ(2)ABCDEFGHL M GMAL4051.m YZ(2)ABCDE MNOPQRS NA 6 3 NA 19 17.3 0.10 -3.35 0.65 MNOPQRSTUVW S 40 12.9 0.08 -4.31 0.51 (2)ACDEFGHLMNO M GMAL4039.c XYZ(2)ABCDE PQRSU 1 9 1 R 17 11.6 0.10 -3.92 0.68 NOPQRSTUVWX S 31 6.3 0.08 -4.48 0.56 (2)ACDEFGHLMNO M PI657093 YZ(2)ABCDE PQRSU 3 12 1 R 20 26.3 0.10 -2.53 0.64 JKLMNOPQRSTU S 36 1.9 0.08 -4.51 0.55 (2)ACDEFGHLMNO M GMAL3689.a VWXYZ(2)ABCD PQRSU E 2 6 3 NA 20 21.3 0.10 -2.86 0.64 LMNOPQRSTUV S 32 -2.1 0.08 -4.66 0.56 (2)CDEFGHLMNOP R GMAL4039.n WXYZ(2)ABCDE QRS 3 11 3 NA 18 22.9 0.10 -3.23 0.67 MNOPQRSTUVW S 30 3.3 0.08 -4.67 0.57 (2)CDEFGHJLMNOP R XYZ(2)ABCDE QRS GMAL4039.u

Cont’d West Virginia 2013 West Virginia 2014

Accession Genetic Cluster Collection Site FB Res._ Field FB Res._ Inoc. GH No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Res. Rating No. Shoots Inoc. %SLB SE Logit %SLB SE Mean Separation Res. Rating 2 11 1 R 15 18.2 0.11 -2.81 0.73 IJKLMNOPQRST S 30 -0.8 0.09 -4.68 0.59 (2)ACDEFGHLMNO R PI657072 UVWXYZ(2)ABC PQRSU DE 2 6 1 S 20 20.9 0.10 -3.17 0.64 MNOPQRSTUVW S 38 3.5 0.08 -4.80 0.54 (2)EFGHLMNOPQRS R PI657001 XYZ(2)ABCDE 2 12 1 S 20 13.5 0.10 -3.75 0.64 NOPQRSTUVWX S 39 4.9 0.08 -4.81 0.55 (2)EFGHLMNOPQRS R GMAL3689.d YZ(2)ABCDE 1 5 4 NA 20 11.8 0.10 -3.67 0.64 NOPQRSTUVWX S 23 3.5 0.10 -4.88 0.66 (2)ACDEFGHLMNO M GMAL4054.f YZ(2)ABCDE PQRSU 3 12 5 NA 18 10.4 0.10 -3.84 0.67 NOPQRSTUVWX S 29 6.3 0.09 -4.97 0.59 (2)EFGHLMNOPQRS R GMAL4304.a YZ(2)ABCDE 3 12 1 S 16 12.7 0.11 -4.34 0.71 RSTUVWXYZ(2)A S 34 -1.5 0.08 -5.07 0.54 (2)LMNOPQRS HR GMAL3614.f BCDE 1 4 1 S 15 13.5 0.11 -4.12 0.73 OPQRSTUVWXY S 33 -1.2 0.08 -5.11 0.55 (2)LMNOPQRS HR GMAL3682.d Z(2)ABCDE 1 4 1 R 20 7.3 0.10 -3.95 0.64 QRSTUVWXYZ(2) S 33 0.7 0.09 -5.12 0.58 (2)GHLMNOPQRS R PI657111 ABCDE 1 4 1 R 20 19.9 0.10 -3.30 0.64 MNOPQRSTUVW S 15 3.4 0.11 -5.27 0.76 (2)ACDEFGHLMNO R

88 GMAL3629.g XYZ(2)ABCDE PQRSU

2 6 1 S 20 15.2 0.10 -3.67 0.64 NOPQRSTUVWX S 19 -8.4 0.10 -5.75 0.70 (2)LMNOPQRS HR GMAL4054.aa YZ(2)ABCDE 1 5 5 NA 19 38.0 0.10 -3.86 0.65 PQRSTUVWXYZ( S 31 1.0 0.08 -5.89 0.56 (2)QRS HR GMAL4304.g 2)ABCDE

Cont’d Washington 2013 Washington 2014 FB No. No. Genetic Collection FB Res._ Logit Mean Logit Mean Accession Res._ Shoots %SLB SE SE Shoots %SLB SE SE Cluster Site GH Inoc. %SLB Separation %SLB Separation Field Inoc. Inoc. Fiesta ------GMAL3636.c 1 4 1 NA 3 52.22 0.32 0.17 1.92 ABCDEF ------GMAL4290.g NA 12 5 NA 3 39.01 0.26 -0.76 1.81 ABCDEF ------PI633921 4 11 1 S ------10 2.3 0.17 -4.72 1.55 AB GMAL4032.i 4 11 2 R ------GMAL4039.a 1 9 3 NA ------6 7.3 0.23 -4.77 2.05 AB GMAL4296.e 3 12 5 NA ------4 3.4 0.24 -4.87 2.16 AB GMAL4054.ac 2 6 4 NA ------PI657047 4 11 1 S ------Goldrush ------PI657095 1 9 1 S 5 7.92 0.26 -1.70 1.58 ABCDEF 8 4.0 0.17 -5.47 1.53 B PI657068 1 5 1 R ------6 0.6 0.23 -5.58 2.05 AB PI657106 2 6 1 S ------

89 PI633798 1 7 1 R ------

PI657113 1 5 1 S ------GMAL4179.b 1 5 4 NA ------PI657044 1 5 1 S ------GMAL4211.a 3 12 4 NA ------3 5.9 0.23 -5.13 2.07 AB GMAL3544.b 4 11 1 NA ------GMAL4211.e 1 4 5 NA ------PI657014 1 6 1 R ------GMAL4002.k 2 6 2 R 4 0.00 0.25 -5.62 1.63 DEF 4 0.6 0.24 -6.14 2.16 AB GMAL3999.i NA 6 2 S ------3 0.7 0.25 -5.31 2.26 AB PI657085 2 12 1 R ------PI633919 1 10 1 S 5 9.33 0.20 -3.41 1.38 CDEF 8 4.0 0.17 -5.45 1.54 B GMAL4028.h 1 5 2 NA ------PI657116 2 11 1 R ------4 7.8 0.24 -4.28 2.16 AB PI657115 1 5 1 R 4 3.14 0.30 -4.91 1.74 CDEF 4 0.6 0.24 -5.08 2.16 AB

Cont’d Washington 2013 Washington 2014 FB No. No. Genetic Collection FB Res._ Logit Mean Logit Mean Accession Res._ Shoots %SLB SE SE Shoots %SLB SE SE Cluster Site GH Inoc. %SLB Separation %SLB Separation Field Inoc. Inoc. PI657100 2 6 1 R ------GMAL4056.p 2 6 4 NA 15 24.76 0.14 -1.67 0.88 BCDEF 16 66.4 0.14 -1.13 1.24 AB GMAL4056n 2 6 4 NA ------5 -0.9 0.20 -3.66 1.78 AB PI637780 1 9 1 R 11 126.19 0.17 4.92 1.03 A 5 9.6 0.23 -3.83 2.09 AB GMAL4238.c NA 9 5 NA ------5 104.9 0.20 4.60 1.80 A GMAL3682.i 1 11 1 R 4 113.58 0.25 5.80 1.63 AB ------GMAL4039.v 3 12 3 NA 6 23.78 0.19 -2.75 1.30 BCDEF 6 5.4 0.19 -6.00 1.70 B GMAL4311.c NA 12 NA R 3 118.52 0.26 4.02 1.81 ABCDE 3 0.4 0.21 -5.81 1.88 AB GMAL4047.L 2 4 3 NA ------PI657098 1 9 1 R 3 74.30 0.32 2.72 1.92 ABCDEF 3 1.2 0.25 -5.43 2.26 AB PI657090 1 9 1 R 6 48.08 0.29 0.64 1.53 ABCDEF 4 51.9 0.20 -0.83 1.79 AB GMAL4312.d NA 4 NA S ------GMAL3638.c 1 4 1 R ------

90 GMAL3631.m 1 4 1 R 8 84.96 0.22 3.99 1.24 ABC ------GMAL4024.u 1 5 2 R 7 85.97 0.22 1.98 1.32 ABCDEF 5 24.1 0.19 -1.82 1.66 AB PI657119 2 11 1 S ------4 25.6 0.21 -2.96 1.88 AB GMAL4056.a 1 9 4 NA 7 8.59 0.19 -2.52 1.24 BCDEF 8 20.6 0.16 -2.52 1.39 AB PI650945 2 6 1 R ------4 41.5 0.24 -0.34 2.16 AB GMAL4047.q 1 5 3 NA 5 35.25 0.22 -0.97 1.45 ABCDEF 5 7.4 0.23 -2.76 2.09 AB GMAL4053.o 1 4 3 NA 4 0.00 0.25 -5.77 1.63 DEF 3 51.0 0.25 0.18 2.26 AB PI657099 1 6 1 R 13 24.08 0.17 -1.88 0.99 BCDEF ------GMAL4047.o 1 5 3 NA 5 5.45 0.29 -3.21 1.61 BCDEF 4 0.9 0.20 -6.28 1.79 B GMAL4047.k 1 10 3 NA ------7 18.1 0.16 -1.77 1.42 AB GMAL4039.y 4 12 3 NA ------GMAL3610.c 3 12 1 S ------GMAL3635.i 1 4 1 R ------GMAL4309.g 3 12 NA S 7 56.94 0.20 -3.38 1.25 DEF ------PI633799 1 9 1 R ------GMAL4054.g 4 11 4 NA ------

Cont’d Washington 2013 Washington 2014 FB No. No. Genetic Collection FB Res._ Logit Mean Logit Mean Accession Res._ Shoots %SLB SE SE Shoots %SLB SE SE Cluster Site GH Inoc. %SLB Separation %SLB Separation Field Inoc. Inoc. PI657076 1 11 1 R 4 14.20 0.30 -1.60 1.74 ABCDEF 4 0.0 0.24 -5.93 2.16 AB GMAL3614.e 3 12 1 S ------PI633920 1 4 1 R ------GMAL4054.c 2 6 4 NA ------GMAL4054.t 1 9 4 NA ------PI657118 2 11 1 R 5 55.18 0.23 0.86 1.47 ABCDEF 2 18.0 0.28 -1.85 2.45 AB PI657117 2 11 1 NA 8 7.06 0.21 -2.84 1.32 BCDEF 6 8.6 0.19 -4.94 1.71 AB GMAL3688.j 4 12 1 R ------Jonathan ------8 12.72 0.19 -3.01 1.20 CDEF 9 5.8 0.15 -5.12 1.31 B PI657082 4 12 1 S ------GMAL4032.p 1 4 2 R ------3 12.6 0.25 -3.87 2.26 AB GMAL4211.b 3 12 4 NA ------GMAL4039.w 3 12 3 NA 3 5.83 0.29 -4.69 1.89 CDEF ------

91 GMAL4051.t NA 5 3 NA ------GMAL4011.u 2 10 2 R ------GMAL4032.m 3 12 2 R ------GMAL4054.q 2 3 4 NA ------PI657017 2 6 1 S 8 40.06 0.18 0.03 1.14 ABCDEF ------GMAL4290.f 3 11 5 S ------GMAL4011.x 1 5 2 R ------GMAL4054.m 3 12 4 NA 4 4.02 0.25 -3.76 1.59 CDEF 7 0.4 0.18 -5.79 1.56 B GMAL4035.t 1 9 3 NA ------PI657086 1 5 1 S 5 9.58 0.23 -4.93 1.46 DEF 3 -5.7 0.25 -4.05 2.26 AB PI657114 1 5 1 R 3 0.00 0.32 -4.85 1.92 CDEF ------GMAL4039.s 1 10 3 NA 7 0.22 0.20 -5.58 1.24 F 6 10.1 0.23 -4.39 2.05 AB GMAL4055.w 1 9 4 NA ------GMAL4032.r 1 4 2 S ------GMAL4028.k 2 11 2 S 6 8.60 0.21 -4.26 1.35 DEF 6 6.6 0.17 -5.11 1.54 AB GMAL4309.f 4 7 NA NA 3 0.00 0.32 -5.23 1.92 CDEF ------

Cont’d Washington 2013 Washington 2014 FB No. No. Genetic Collection FB Res._ Logit Mean Logit Mean Accession Res._ Shoots %SLB SE SE Shoots %SLB SE SE Cluster Site GH Inoc. %SLB Separation %SLB Separation Field Inoc. Inoc. GMAL4002.p 2 6 2 S ------GMAL3552.v 4 11 1 R ------5 2.6 0.23 -4.51 2.09 AB PI657101 1 6 1 R ------PI657041 1 9 1 R 5 0.00 0.29 -5.06 1.61 DEF 10 -5.4 0.15 -5.32 1.28 B GMAL3608.h 3 12 1 R ------PI657008 1 9 1 S 5 -3.90 0.25 -6.49 1.57 F ------GMAL3619.a 2 10 1 R ------9 22.7 0.15 -2.47 1.33 AB PI657056 1 3 1 R ------GMAL3683.k 2 11 1 R ------6 8.9 0.23 -5.07 2.05 AB GMAL4211.g 1 5 5 NA ------PI657049 3 11 1 S ------GMAL4177.d 1 10 4 NA ------GMAL3608.n 3 12 1 NA ------

92 PI657064 1 4 1 R ------PI657073 1 11 1 R ------GMAL4002.i 2 6 2 R 4 3.95 0.25 -4.42 1.59 CDEF ------GMAL4055.m 1 9 4 NA 7 37.75 0.20 -1.73 1.26 ABCDEF ------PI657009 2 6 1 R ------GMAL3685.c 1 11 1 R 2 6.82 0.35 -3.26 2.25 ABCDEF ------Empire ------GMAL3781.a 1 9 2 R ------PI633801 4 11 1 R ------GMAL4054.z 1 9 4 NA ------GMAL4177.b 2 6 4 R ------GMAL3688.c 2 11 1 R 6 7.77 0.21 -4.67 1.32 EF 6 -3.9 0.23 -4.28 2.05 AB GMAL3975.e 2 6 2 S 3 0.68 0.32 -4.62 1.92 BCDEF ------GMAL3683.e 2 11 1 R 6 85.06 0.19 -0.74 1.28 ABCDEF 8 -0.6 0.17 -5.41 1.54 B GMAL4047.h 1 9 3 NA 4 49.80 0.25 0.37 1.63 ABCDEF ------GMAL4053.k 4 11 3 NA ------3 0.7 0.25 -5.17 2.26 AB

Cont’d Washington 2013 Washington 2014 FB No. No. Genetic Collection FB Res._ Logit Mean Logit Mean Accession Res._ Shoots %SLB SE SE Shoots %SLB SE SE Cluster Site GH Inoc. %SLB Separation %SLB Separation Field Inoc. Inoc. GMAL3691.g 1 9 2 S ------Gala ------PI657110 1 9 1 NA 8 5.90 0.20 -3.70 1.21 DEF ------PI657102 2 6 1 R ------4 8.1 0.24 -4.94 2.16 AB Delicious ------GMAL4032.o 2 10 2 R ------PI657010 2 6 1 S ------GMAL4179.g 1 5 4 NA ------Splendour ------PI657058 1 4 1 R 6 0.50 0.21 -5.15 1.32 EF 5 7.3 0.23 -5.12 2.09 AB GMAL4082.d 1 6 4 NA ------GMAL4032.t 1 4 3 S ------6 0.7 0.23 -5.36 2.05 AB GMAL3975.c 1 9 2 R 4 0.00 0.30 -5.46 1.74 DEF 4 0.7 0.24 -6.32 2.16 AB

93 GMAL3989.c 2 6 2 R 4 14.69 0.25 -4.21 1.59 CDEF 4 7.3 0.24 -5.25 2.16 AB

Golden ------17 3.6 0.16 -5.12 1.45 B Delicious GMAL3691.i 1 5 2 R ------GMAL3691.c 1 9 2 R 6 -7.07 0.21 -5.57 1.36 F 5 7.3 0.23 -5.89 2.09 AB PI657109 1 10 1 R ------GMAL4198.d 1 5 4 NA 3 6.14 0.27 -4.02 1.81 BCDEF ------GMAL4211.c 1 4 1 NA ------PI657054 2 10 1 R ------GMAL4059.a 2 6 4 NA ------GMAL4051.o 1 5 3 NA ------PI633923 1 5 1 R ------GMAL4011.f 2 6 2 R ------PI657081 4 12 1 S 10 -4.52 0.18 -5.00 1.10 F 7 -8.4 0.22 -5.23 2.02 AB PI656998 2 6 1 R 3 1.04 0.32 -5.19 1.92 CDEF ------Robusta 5 ------

Cont’d Washington 2013 Washington 2014 FB No. No. Genetic Collection FB Res._ Logit Mean Logit Mean Accession Res._ Shoots %SLB SE SE Shoots %SLB SE SE Cluster Site GH Inoc. %SLB Separation %SLB Separation Field Inoc. Inoc. GMAL3616.o 3 12 1 S ------4 0.4 0.20 -5.50 1.73 AB GMAL4002.m 1 9 2 R ------GMAL4211.d 1 6 1 R ------5 -10.0 0.23 -5.39 2.09 AB GMAL4179.f 1 5 4 NA ------PI657097 1 9 1 S 10 107.90 0.18 2.43 1.09 ABCD 13 27.8 0.15 -1.47 1.30 AB GMAL4055.b 3 9 4 NA 3 62.66 0.28 -2.32 1.89 ABCDEF ------GMAL4039.L 2 6 3 NA ------4 5.0 0.20 -6.00 1.79 B GMAL4036.L 1 9 3 NA 12 53.98 0.17 0.80 1.02 ABCDEF 4 47.6 0.20 -1.02 1.73 AB PI657105 2 6 1 R 6 0.19 0.21 -2.21 1.36 BCDEF ------GMAL3637.d 1 4 1 R ------4 5.1 0.21 -5.08 1.86 AB GMAL4011.w 3 10 2 R ------PI657107 2 6 1 R 11 1.70 0.18 -4.56 1.07 EF 6 -0.2 0.17 -4.83 1.47 AB GMAL3623.i 2 3 1 R ------

94 GMAL4039.t 3 11 3 NA ------GMAL4039.d 1 9 3 NA 6 24.77 0.29 -1.45 1.62 ABCDEF ------GMAL4032.w 4 12 3 R 3 15.45 0.27 -2.63 1.81 ABCDEF ------PI633802 3 12 1 R 9 17.27 0.18 -2.81 1.13 CDEF ------GMAL4290.e 2 11 5 NA ------GMAL4068.b 2 6 2 NA ------GMAL3607.k 3 12 1 R ------GMAL4053.b 4 11 3 NA 9 11.53 0.19 -2.40 1.12 CDEF 11 15.6 0.16 -3.26 1.47 AB PI657089 1 9 1 R 6 10.44 0.21 -3.07 1.35 CDEF 24 10.3 0.12 -3.93 1.03 AB PI633922 3 12 1 R ------PI657067 1 5 1 S ------GMAL3629.m 1 4 1 R 3 0.00 0.27 -5.27 1.81 DEF ------GMAL3684.h 2 11 1 R ------5 -10.0 0.23 -6.21 2.09 AB GMAL4047.j 2 6 3 NA ------PI657120 2 11 1 NA ------GMAL3607.h 2 11 1 R ------

Cont’d Washington 2013 Washington 2014 FB No. No. Genetic Collection FB Res._ Logit Mean Logit Mean Accession Res._ Shoots %SLB SE SE Shoots %SLB SE SE Cluster Site GH Inoc. %SLB Separation %SLB Separation Field Inoc. Inoc. GMAL4011.i 2 6 2 R ------3 0.4 0.21 -4.89 1.88 AB PI657069 3 11 1 R 4 3.81 0.27 -3.81 1.70 BCDEF ------PI657021 2 6 1 R ------GMAL3544.j 4 11 1 R 8 26.63 0.21 -1.57 1.23 ABCDEF ------PI657112 1 5 1 NA ------8 2.2 0.20 -4.50 1.75 AB GMAL4051.p 1 5 3 NA ------GMAL4051.q 1 5 3 NA 7 6.88 0.21 -3.05 1.30 CDEF 8 4.7 0.16 -5.16 1.45 B GMAL4051.s 1 5 3 NA ------3 0.7 0.25 -5.27 2.26 AB GMAL3689.g 2 12 1 R ------3 0.7 0.25 -4.90 2.26 AB PI657088 1 5 1 R 5 2.70 0.29 -4.70 1.62 DEF 9 24.1 0.17 -1.86 1.50 AB PI657096 2 9 1 R ------GMAL4177.e 1 5 4 NA ------GMAL4051.m 1 5 3 NA 5 12.95 0.25 -3.79 1.51 CDEF 7 5.8 0.19 -4.45 1.68 AB

95 GMAL4039.c NA 6 3 NA ------6 1.0 0.18 -5.17 1.60 AB PI657093 1 9 1 R 3 7.51 0.32 -2.89 1.93 ABCDEF ------GMAL3689.a 3 12 1 R ------GMAL4039.n 2 6 3 NA ------GMAL4039.u 3 11 3 NA ------PI657072 2 11 1 R ------PI657001 2 6 1 S 4 0.00 0.30 -5.98 1.74 DEF 7 1.7 0.23 -5.02 2.02 AB GMAL3689.d 2 12 1 S ------GMAL4054.f 1 5 4 NA 3 49.44 0.26 -1.60 1.80 ABCDEF ------GMAL4304.a 3 12 5 NA ------GMAL3614.f 3 12 1 S 8 21.94 0.20 -3.13 1.23 CDEF 6 -10.0 0.23 -6.55 2.05 AB GMAL3682.d 1 4 1 S ------PI657111 1 4 1 R ------5 1.6 0.19 -5.28 1.66 AB GMAL3629.g 1 4 1 R 3 38.82 0.30 -0.80 1.91 ABCDEF ------GMAL4054.aa 2 6 1 S 3 76.39 0.32 3.17 1.92 ABCDEF ------GMAL4304.g 1 5 5 NA ------

CHAPTER FOUR: Consideration of cost and accuracy for advanced breeding trial designs in apple

Julia M. Harshman1, Kate M. Evans1, Craig M. Hardner2

1Department of Horticulture, Washington State University Tree Fruit Research and Extension

Center, 1100 North Western Avenue, Wenatchee, WA 98801, USA

2Queensland Alliance for Agriculture and Food Innovation, University of Queensland, St

Lucia, QLD 4072, Australia

Abstract

Trialling advanced candidates in tree fruit crops is expensive due to the long term nature of the planting. The proximity with which trait evaluations approximate the true trait value needs balanced with the cost of the program. Alternative trial designs of advanced apple candidates were modelled to investigate the effect of the number of locations, the number of years and the number of harvests per year on the cost and accuracy in an operational breeding program. Critical percentage difference, response to selection, and correlated response were used to examine changes in accuracy. The aim of this study was to model designs that would allow evaluation of additional candidates without sacrificing accuracy. For the traits evaluated, accuracy and response to selection was reduced by less than 5% for most trial designs. Risk management influences the decision to change trial design, and some designs had greater risk associated with them. Balancing cost and accuracy with risk yields valuable insight into advanced breeding trial design.

Introduction

New apple (Malus × domestica Borkh.) cultivars sustain consumer interest and increase industry profitability through improved returns or reduced costs relative to current varieties (Harker et al. 2003). The success of varieties released in recent decades, such as

‘Honeycrisp’, ‘Fuji’, and ‘Cripps Pink’, is largely due to their superior eating quality (Harker et al. 2003; Stebbins 1992). The aim of many apple breeding programs is improved fruit quality in order to produce successful new varieties.

Apple improvement programs typically operate a multi-stage selection program

(Brown 1975; Evans 2013; Kellerhals et al. 2009). The first stage generally involves evaluation of a large number of un-replicated seedlings. In subsequent stages, the reduced number of selected candidates are clonally propagated and planted in replicated trials which enable comparisons of genetic potential between candidate selections and current varieties for important characteristics. Multiple traits underlie fruit quality and thus, the decision to advance a candidate. Vegetative (or clonal) propagation allows both the additive and non- additive genetic variation to be targeted by selection. Identifying candidates superior to current cultivars is a function of the size of the selection population and the accuracy with which the available data predicts the genetic potential of a candidate selection (Falconer and

Mackay 1996). The design of a field trial impacts the accuracy with which traits are evaluated and the cost of a breeding program.

Accurate prediction of genetic potential in selection environments that are highly correlated with future commercial planting environments underpins the successful adoption of new varieties since it improves confidence in the genetic potential of the candidate selection (Cooper et al. 1993). In apple, these predictions are achieved by trialling clonal replicates over multiple years in multiple locations, usually with multiple blocks within

location and the accuracy is improved with increased replication. Maintaining replicated trials of clonal apple candidates is expensive (Weber 2008), and there are trade-offs between maximizing accuracy and minimizing cost to the program with the limited resources available to breeding programs. Several evaluation criteria can give insight into which trial designs most influence accuracy and those changes in accuracy can then be compared against changes in the breeding program cost.

Critical percentage difference (CPD) is a measure of accuracy that estimates the observed percentage difference needed to claim that a selection and a control are different with a confidence of α, similar to the least significant difference (LSD) statistic (Arief et al.

2015; Brennan et al. 1998; Cullis et al. 1996a, b; Patterson et al. 1977; Talbot 1997). CPD is a function of the percentage standard error of the estimate of the true mean difference recorded in the trial and the value from the standardized normal distribution that is exceeded with a probability of α.

Response to selection (RS) is the predicted gain from directional selection (Falconer and Mackay 1996). It can be used to evaluate trials of differing sizes (Arief et al. 2015) as variation in numbers of entries may affect both the selection intensity and accuracy of the predicted candidate mean. With clonal replicates, it can be used to explore how trial design changes alter the average trait value of a population from one stage to the next. RS is a function of selection intensity, accuracy of prediction of genetic potential in selection environments, and the correlation between genetic potential in selection environments and the genetic potential in the future commercial planting environments (Cooper et al. 1993).

Correlated response to selection (CRS) indicates the magnitude of directional gain that can be achieved in one trait when selection is applied to a second, correlated trait. This indirect selection can be more effective than direct selection when the second trait is highly

correlated and can be evaluated more quickly, with less cost, or with greater accuracy

(Falconer and Mackay 1996). Using instrumental traits to improve sensory traits is of interest because instrumental measures can have higher heritabilities (Hardner et al. submitted;

Oraguzie et al. 2009), are less expensive to measure, and eliminate problems with ‘taster- fatigue’ (Arnold 1983; Oraguzie et al. 2009). Gain from indirect selection is strongly dependent on the correlation between the two traits. In this study, CRS is used to evaluate the directional progress that can be made in a sensory fruit quality trait when selection is applied to the correlated instrumental fruit quality trait. As with RS, altering trial design factors changes the average trait value from phase two to phase three.

In this study, the Washington State University Apple Breeding Program (WABP) phase two field trials were used as a model to explore the influence of trial design on accuracy and cost. Under the current phase two design, fruit from candidate selections are harvested three times per year per location (in an attempt to harvest at optimum maturity) for a minimum of three years at three locations. The fruit are evaluated at harvest and after eight weeks in 2°C regular atmosphere (RA) storage (Evans 2013). The trial factors under consideration are the number of harvests per year, the number of years and the number of locations; the three criteria discussed above will be used to evaluate accuracy.

An in-depth analysis of the genetic architecture of appearance and quality traits evaluated in the WABP was undertaken for the two-month storage data used in this analysis as well as data from harvest assessments (Hardner et al. submitted). Substantial interactions between candidates and locations, years or harvests is lacking for most traits, indicating that a less intensive assessment design could be used to predict candidate performance (Hardner et al. submitted). Substantial candidate by storage duration interactions were lacking for all traits except firmness, indicating that selection could be made at either harvest or following

two months storage (Hardner et al. submitted). Short term storage was recommended and chosen for this analysis as commercially sold apples have been in storage, thus data from that regime better reflects the commercial process (Hardner et al. submitted).

The aim of this chapter is to validate methods for evaluating trial design accuracy and cost in tree fruit and horticultural crop breeding, using an operation apple breeding program as the model. Other trial design accuracy and cost goals could be investigated with the outlined methods, however this analysis will investigate alternative trial designs under which the total program cost would not increase in order to evaluate more candidates in each phase two trial.

Methods

Current WABP Phase 2 trial design. Phase two is the most intensive data collection phase of the WABP and consists of replicated trials planted with promising individuals from phase one (Hardner et al. 2015). Phase two plantings under the current design are sited at three locations distributed across the major growing regions each year. Five clonally propagated trees per candidate and several standard commercial cultivars are planted at each location. There is no replication within location as fruit harvested from the five clonal replicates are bulked into one sample. Plantings last no less than four years as most selections do not produce sufficient fruit to evaluate in the first year. Therefore, fruit from candidates can be considered to have been evaluated for three years.

Data collected includes yield efficiency, fruit quality at harvest and after eight weeks in 2°C regular atmosphere (RA) storage. Optimal harvest maturity is difficult to ascertain in new apple selections and as harvest maturity influences important fruit quality traits— storability, flavor and texture— selections are harvested in three consecutive weeks per harvest season. Maturity is estimated using a combination of the Cornell starch iodine index

100

(Blanpied and Silsby 1992), and appearance for the first harvest. At the third harvest, all of the remaining fruit are harvested (ie. “strip-picked”) for other uses in the breeding program

(not reported here). Fifteen fruit are sampled for each harvest; five fruits for instrumental analysis at harvest (not discussed here), and five fruit each for instrumental and sensory analysis after eight weeks in RA storage.

Sixteen ordinal traits are scored after storage (Hardner et al. submitted; Appendix 1).

Traits are evaluated by four experienced members of the breeding team, all of whom have received sensory analysis training. Each member tastes and scores the apple sample. The average of the four scores for each ordinal trait is then entered in the WABP database.

Sensory evaluations are based on the anticipation of the consumer’s sensory perceptions. It is assumed that the evaluators’ perception of taste matches that of the consumers.

Seven instrumental traits are also assessed after eight weeks in storage: SSC (soluble solids content, °Brix, using a digital refractometer [RX-5000α-Bev, ATAGO U.S.A, Inc.,

Bellevue, WA]), TA (titratable acidity, mg/L malic acid, using an auto-titrator [Metrohm®

815 Robotic USB Sample Processor XL, Metrohm USA Inc., Riverview, FL]), FRTD (fruit diameter, inches), FRTW (fruit weight, grams) and M1 (lb), M2 (lb) and CN (all assessed using a Mohr Digitest® texture analyzer [MDT-1, Mohr Test and Measurement LLC,

Richland, WA (Evans et al. 2010)]. Apples are individually assessed using the texture analyzer, and then an equatorial slice is used for starch pattern index analysis. The remaining halves of the fruit are cut into quarters and one quarter from the shoulder side of each of the five apples are pooled and juiced. Fresh juice is then measured for SSC and a tube of juice is frozen for later assessment of TA. TA is not typically measured in real time due to the large number of samples being analyzed and subsequent time constraints.

101

Cost Assessment of the WABP. Current trial design costs were summarized in a bioeconomic spreadsheet model to examine alternative designs effect on trial costs (Table 1).

The number of locations, number of years and number of harvests per year can be varied which changes the number of trees and fruit samples that are evaluated and therefore the associated field, consumable and labor costs. Locations were assumed to be equal distance apart and the per-site cost used was the average cost per site. Costs were divided into tree production, field establishment, field maintenance, and candidate assessment and are presented on a per candidate basis. Cost of evaluating an individual was expressed on a scale where 100 units represented the total cost of evaluation of a single candidate in the current trial. This analysis is for the life of a single phase two planting lasting four years. The total per candidate cost (100 units) multiplied by the number of candidates typically evaluated in each phase two trial (20) equals the total program cost (2,000 units) for the current design. To calculate the total number of individuals that could be evaluated under alternative designs, the total program cost was divided by the per candidate costs of each alternative design and rounded to the nearest whole number (Table 1). Design details are abbreviated so that 3 locations, 3 years, 3 harvests per year is denoted 3L/3Y/3H. Alterations from the current design will be in bold (ie. 2L/3Y/3H). Single factor-altered designs considered are

2L/3Y/3H, 3L/2Y/3H, 3L/3Y/2H and 3L/3Y/1H. Two factor-altered designs considered are

2L/3Y/2H, 2L/2Y/3H, and 3L/2Y/2H.

Statistical Methods. To obtain the variance components needed to estimate CPD, RS and CRS, a subset of data from 2004-2011 including only eight week storage assessment was reanalysed following the methodology outlined by Hardner et al. (submitted). With storage duration and relevant interaction terms removed, the final estimable model for individual trait observations presented was:

102

푦 = 퐿 ∗ (푌 + 퐴) ∗ 퐻 + 퐺⁄(푌 ∗ 퐻) + 퐿^푃⁄(푌 ∗ 퐻) + 푒

where L was location (referred to as F, Farm, in Hardner et al (submitted)), Y is year

(referred to as S, Season, in Hardner et al (submitted)), A is age, H is harvest, G is candidate, and P is plot. The symbol ^ denotes the interaction between terms, / denotes nesting of terms

(i.e. ABAAB/^ ) and * implies a full expansion of terms (i.e. ABABAB*^   ). The terms from the expansion of L* (Y + A)*H were treated as fixed and the rest as random.

To estimate variance components and fixed effects for fruit quality traits assessed after two months of storage, the general mixed linear random model, statistical methods and software were the same as those used by Hardner et al. (submitted).

For a balanced trial design where candidates (G) are harvested (H) h times at l

2 locations (L) in each y years (Y), the total variance of the predicted candidate effect (σ Ĝ) mean was given by:

휎2 휎2 휎2 휎2 휎2 휎2 휎2 휎2 = 퐺퐿 + 퐺푌 + 퐺퐻 + 퐺퐿푌 + 퐺퐿퐻 + 퐺푌퐻 + 퐺퐿푌퐻 Ĝ 푙 푦 ℎ 푙푦 푙ℎ 푦ℎ 푙푦ℎ

Plot variance was confounded with location variance as there are no plots within farm; if candidate × plot variance was assumed to be zero, then the total variance due to differences in trial locations was in the genotype × location variance component. Year variance was confounded with candidate age variance. If candidate age variance was assumed to be zero, then the total differences in the trial years was in the candidate × year variance component.

As described in Hardner et al. (submitted), variance components for the random factors defined in the mixed model were estimated by Restricted Maximum Likelihood (Patterson and Thompson 1971) with the software package ASReml (Gilmour et al. 2009). The Shapiro-

Wilk statistic (Shapiro and Wilk 1965) was calculated for the residuals to examine the assumption of normality. Wald statistics were used to test fixed effects (Kenward and Roger

103

1997). The likelihood ratio test (LRT) was used to test the significance of random terms

(Wilks 1938). A value of P=0.05 for significance testing.

Critical Percentage Difference. The critical percent difference (CPD) is the difference in the sample mean needed to reject the null hypothesis with a level of confidence of α was:

2 100 ∗ 2√휎퐺̂ 퐶푃퐷(훼) = 푍 휇 훼/2

2 where √휎퐺푥̂ was the standard error of the predicted candidate effect (described above) and the standard normal distribution value for α=0.05 was Zα/2 =1.96 (Patterson et al.

1977). The null hypothesis was that two entries have an equal true mean. The two-tailed hypothesis was chosen because selections are entered in the phase two trials to test their performance against standards, and prior to the results, there is uncertainty as to whether they are better or worse than the standards for a particular trait. CPD was presented for the current design. To allow for comparisons between traits and designs, the difference in CPD between an alternative design and the current design was presented. A positive change in CPD for an alternative design indicates that the alternative design was less accurate than the current design at predicting the candidate effect.

To determine if there was an interaction between altering one factor of design and altering two factors of design (ie. 2L/3Y/3H versus 2L/2Y/3H), the difference between the single factor altered design and the current design was calculated and those differences were added to get an expected CPD. The presence of an interaction was indicated by a difference between actual and expected CPD.

Response to Selection. Response to selection (RS) for a balanced design was given as:

104

휏퐺̂푥,퐺푥 푅푆푥 = 𝑖 2 2 √휎 퐺푥̂ + 휎퐺

where the selection intensity (i) equals 1.76 if the proportion of individuals selected is

0.1 (Falconer and Mackay 1996), τĜx,Gx is the covariance between the predicted candidate

2 effect and the true candidate effect, and √휎퐺푥̂ is the standard error of the predicted candidate effect. This equation was derived from that presented by Falconer and Mackay (1996;

Appendix 2). The selection intensity chosen represents the current selection intensity (2 candidates are advanced from a pool of 20) in the WABP. Change in response to selection is from either selection intensity or variation in the predicted candidate effect. By designating the selection intensity, response to selection reflects changes in variation of the predicted candidate effect due to changes in trial designs.

To determine if there was an interaction between altering one factor of design and altering two factors of design (ie. 2L/3Y/3H versus 2L/2Y/3H), the difference between the single factor altered design and the current design was calculated and those differences were added to get an expected RS. An interaction was indicated by a difference between actual and expected RS, as above for CPD.

Correlated Response. Correlated response to selection (CRS) was calculated for traits measured both instrumentally and organoleptically: CN v. CRISP; M1 v. HARD; SSC v.

SWEET; TA v. TART. The underlying assumption was that the WABP team’s sensory evaluation of traits approximates the average consumer’s assessment. To explore how selection for an instrumental trait (y) changed the response of a second, correlated sensory trait (x) under various trial designs, correlated response for a balanced design was calculated as:

105

푟퐺 퐺 휎2 휎2 푥 푦√ 퐺푥 Ĝ푦 퐶푅푆퐺푦 = 𝑖 휎2 √ Ĝ푦

2 where rGxGy was the genetic correlation between trait x and y, σ Gx is the variance of

2 the genetic effect of trait x, and σ Ĝy was the variance of the predicted candidate effect of trait y. The selection intensity (i) equals 1.76 when the proportion of individuals selected was 0.1 and 2.06 when the proportion is 0.05. This equation was derived from that presented by

Falconer and MacKay (1996; Appendix 2). Other parameters are assumed constant therefore change in correlated response are due to variation in selection intensity or variation in the predicted candidate effect. By choosing the selection intensity, changes in correlated response to selection are the effect of trial design changes on the variation in the predicted candidate effect of the sensory trait.

Genetic correlations estimated from a previous analysis of fruit quality traits after storage for entries in existing WABP field trials (Hardner et al. submitted) were used to estimate correlated response (Appendix 3). Correlated responses are in units of the traits targeted by selection, which are the ordinal sensory traits CRISP, HARD, SWEET, and

TART.

Results

Variance Components. Genetic effects accounted for 18 to 65% of the phenotypic variance for fruit quality traits and all of the residuals were normally distributed (Table 2).

The main effect of candidate was the largest source of phenotypic variance for all of the appearance traits except APPSUM, GCOL and RUSS, all of the instrumental traits except CN and SSC, but only TART of the sensory traits (Table 2). For those traits, residuals accounted for the largest source of variance. Candidate × harvest (G×H) interaction was zero for all

106

traits except OVERALL. Only a few significant interactions were detected for G×H

(OVERALL), G×Y×H (M2 and TCOL), and G×L×H (GCOL and TA). There was significant variation for G×L×Y for all traits except OVERALL. Variation due to G×L was significant for all traits except CN, M2 and TART, and there was significant variation due to G×Y for all except M1 and TART. G×L was smaller or equal to G×Y except for SIZE, M1 and TA.

Cost

Assessment is the most expensive component of the current design, accounting for 74.2 of the total 100 units (Table 1). Costs incurred during assessment include the staff time required for driving, 27 harvests (three locations harvested nine times over three years), data collection and data entry, as well as fuel, vehicle and consumable expenses. Driving costs are the most expensive component within assessment, accounting for 50.7 units. Data collection is the second most expensive component, accounting for 15.5 units. Sensory analysis is the most expensive component of data collection and the second most expensive individual component of the whole trial.

Changes in Costs. For the single-factor altered designs, the largest cost reduction was seen under the 2L/3Y/3H design. Decreased tree production, field establishment, field maintenance, and candidate assessment resulted in a reduction of 33.3 units to the per candidate costs (Table 1). Ten additional candidates could be evaluated for the same total program cost (Table 1). The reduction in per candidate costs was smallest under the

3L/3Y/1H design due to the tree production, field establishment, maintenance and driving costs remaining the same as the current design (Table 1). Only three additional candidates could be evaluated for the same total program cost.

Reducing two factors simultaneously resulted in a greater reduction in per candidate cost than reducing only one factor to two, except in the case of and 3L/2Y/2H which was 2.6

107

units more than 2L/3Y/3H. The 2L/2Y/3H design resulted in the greatest reduction in per candidate costs, by 50.8 units, due to the number strip-picks reduced to four (six for both the other two factor altered designs; Table 1). Twenty additional candidates could be evaluated for the same total program cost.

Effect on Accuracy

Changes in Accuracy from altering one design factor. Accuracy was decreased as measured by CPD when reducing any factor of the current design as there were less observational units; however, the reduction in accuracy was less than 5% for most traits

(Table 3). For all traits, the smallest decrease in accuracy was under the 3L/3Y/2H design.

The only decreases in accuracy greater than 5% were under the 3L/3Y/1H design for the traits: TCOL (6.2%), GCOL (5.4%), OVERALL (8.7%), and CN (7.4%).

Changes in Accuracy from altering two design factors. Altering two factors simultaneously resulted in greater reduction in accuracy than altering a single factor, though the reduction in accuracy was still less than 5% for most traits (Table 3).

For the appearance traits FRTWT and PCOL, CPD values were lower than expected and the same under both 3L/2Y/2H and 2L/3Y/2H. The decrease in accuracy under both

2L/3Y/2H and 2L/2Y/3H designs for SHAPE, RUSS, LENT, FRTDM, TCOL, GCOL, and

APPSUM were similarly small. The smallest decrease in accuracy for SIZE was under the

3L/2Y/2H design and the CPD value was lower than expected. The decrease in accuracy was greater than 5% for GCOL under 3L/2Y/2H and for TCOL under all three designs.

For the sensory traits CRISP, HARD, JUIC, AROM, TART and EQ, the smallest decrease in accuracy was under both 2L/3Y/2H and 2L/2Y/3H designs. Accuracy was equally reduced for all three designs for OVERALL (6%) and SWEET (2.7%).

108

For the instrumental traits, the decrease in accuracy for CN was greater than 5% for all three designs, with the smallest reduction under the 2L/3Y/2H design. The smallest decrease in accuracy for M1 and TA was under 3L/2Y/2H. Under both 2L/3Y/2H and 2L/2Y/3H, the reduction in accuracy for SSC was 1.3%.

Changes in Accuracy from altering three design factors. Decreases in accuracy were greater than 5% for most traits under 2L/2Y/2H; the greatest reduction in accuracy was for

CN (11.8) and OVERALL (10.1).

Interactions. CPD values for the two or three factor-altered designs were generally higher than expected based on CPD values for the single factor-altered designs, indicating an interaction. SIZE, PCOL and FRTWT are the exceptions, all having lower than expected

CPD values for the 3L/2Y/2H, 2L/2Y/3H and 2L/2Y/2H designs. The difference between the expected CPD value and the actual CPD value was less than 1% in most cases. CPD values were higher than the expected 2% for RUSS, LENT, and GCOL with 3L/2Y/2H, 2L/2Y/3H and 2L/2Y/2H.

CPD values were also higher than the expected 1% for CN, AROM, SHAPE, TCOL and OVERALL for 2L/2Y/2H, which indicates an interaction between the one factor alterations and the this design.

Effect on response to selection

Changes in Response to Selection from altering one design factors. For most traits, there was less than 5% reduction in response to selection under any of the single factor altered designs (Table 4). There was almost no reduction for any trait under the 3L/3Y/2H design. There was a 6% reduction in response under the 3L/3Y/1H design for JUIC. There was a similar reduction in response to selection for SWEET under 2L/3Y/3H, 3L/2Y/3H and

109

3L/3Y/1H design. There was a 13% reduction in response to selection for OVERALL under the 3L/3Y/1H design.

Changes in Response to Selection from altering two design factors. As seen for CPD, there was less than 5% reduction in response to selection for most traits under any of the two factor-altered designs (Table 4). RS values were greater than expected for many traits, which indicates an interaction between the one factor alterations and the two factor alterations.

However, the reduction in response for most traits was very small.

For the sensory traits, the reduction in RS was similar for all three designs for most traits. There was a measurable decrease in response to selection for all three two factor designs for AROM (6-7%) and SWEET (9-13%). There was a 6-7% decrease in RS for JUIC under both 3L/2Y/2H, and 2L/2Y/3H designs. The greatest decrease in response to selection for EQ (7%) and OVERALL (10%) was under 2L/2Y/3H. For the instrumental traits, there was a 6% decrease in response for CN under 2L/2Y/3H and for TA under all two factor- altered designs.

Changes in Response to Selection from altering three design factors. The largest reduction in response to selection for most traits was under the three factor altered design

(2L/2Y/2H), and RS values were higher than expected for most, indicating an interaction

(Table 4). For the appearance traits, the decrease in response was equal to expected for TCOL and smaller than expected for APPSUM. There was a reduction in RS of 19% for SWEET and of 13% for OVERALL. RS values were greater than expected for all of the sensory and instrumental traits, except the RS value for JUIC which was equal to expected.

Correlated Response

For the current design (3Y/3L/3H), direct selection on sensory CRISP increased the phase three trial average by 0.63 units on the ordinal scale, while indirect selection increased

110

the average by 0.5 units (Table 4). Indirect selection to improve CRISP was less effective than direct selection for all of the designs considered. Similar to CRISP, direct selection on

SWEET resulted in greater gains (0.32) than indirect selection on SSC (0.22) for all designs.

Indirect selection on M1 made slightly greater gains or equaled direct selection for all considered designs. Selecting indirectly on TA and directly on TART both resulted in a gain of 0.56 for the current design, 2L/3Y/3H and 3L/3Y/2H. Indirect selection using TA was slightly advantageous for the other designs.

Changing Intensity

Increasing selection intensity by reducing the proportion of selected candidates increased RS values for all traits (Table 4). Under the current design, where 10% of candidates (2 out of 20 candidates) are advanced to phase three, the cost per candidate is 100 units with a total program cost of 2,000 units (100 x 20; Table 1). The smallest reduction in accuracy and response to selection for most traits was under the 2L/3Y/2H design. If 5% of candidates are advanced (2 out of 40) with a per candidate cost of 62.2 units, then the total program cost is 2,488 units. For 1.5 times current total program cost, double the number of candidates can be evaluated. RS is highest at the 5% selection intensity for all traits. Thus, more progress could be made in the average trait values from phase two to phase three. For a total program cost of 1,990 units, 32 candidates could be evaluated; if only two are still advanced, then the proportion selected becomes 6% which means similar gain from selection could be made. CRS of all trait pairs increased; however, HARD and TART remained the only sensory traits that benefit more from indirect selection.

Discussion

There are inherent trade-offs between accuracy and cost in trial design efficiency.

Using the costs and data from an operational apple breeding program as a model, the

111

ramifications of reducing the design was explored in order to evaluate additional candidates.

The total cost of the program can be reduced if the design is reduced, but this leads to a decrease in accuracy and response to selection. Cost savings from a reduced trial design that allow evaluation of additional candidates warrant the decreases in accuracy and response to selection for the program we examined. This study demonstrates the applicability of these methods to tree fruit and horticultural crop breeding and the utility of the results in making informed trial design decisions.

Variance of fruit quality traits. More than half of the traits included in this analysis have residual variances as the largest single source of variance. Residual variances were greater than those reported by Hardner et al. (submitted) partially due to the removal of storage duration variance components. This could be due to a lack of consistency between observations not attributable to year, location, harvest, and their interactions, or the inherent variability among fruit for these traits (Hardner et al., submitted).

The summary traits APPSUM, EQ and OVERALL had relatively high residual variances, with OVERALL having the largest residual variance of all of the traits. Other factors influence the scoring decisions for summary traits, and those factors are not necessarily consistent between candidates. Consider two equally attractive apple samples evaluated by the breeding team, where one sample had large, bulbous stems that may increase water loss and the other had very open calyxes that would likely increase susceptibility to core rots. Both samples would receive the same lower APPSUM rating despite similar marks on the other ordinal appearance traits. OVERALL, in particular, is influenced by many unscored factors as this ‘trait’ denotes the selection decision. The three anchors for

OVERALL on the ordinal scale are ‘Reject’, ‘Re-evaluate, promising’, or ‘Advance.’ A fruit sample could be scored as ‘reject’ for a number of reasons not covered in any of the other 17

112

appearance and sensory traits or the post-harvest disorder incidence records (not reported here). The overall perception of a fresh-eaten apple is the interaction between the numerous traits measured by the WABP and therefore, the summary traits serve as a useful way to rate those interactions.

Performance of trial designs. Reducing one factor of the design resulted in smaller reductions in accuracy for both CPD and RS, but less reduction in cost. There was a greater reduction in accuracy when reducing two factors simultaneously, as well as greater reductions in cost. CPD and RS values that were greater or smaller than expected for the two- and three-factor altered designs indicates that there were interactions. These reflect the interactions between candidate and random effects (ie. G×L, G×Y, G×Y×L), highlighting the importance of analyzing the genetic architecture of traits evaluated in a breeding program prior to this type of analysis.

Reducing both harvests and locations (2L/3Y/2H) resulted in the smallest decrease in accuracy of the two factor-altered designs and would allow the program to evaluate twelve additional candidates for a similar total program cost. Interactions for G×H, G×H×Y, and

G×H×L were very small or non-existent for all of the traits evaluated, which may explain why the reduction in accuracy and gain for most traits was less than 5%. Decrease in accuracy and response to selection for some of the most important traits for advancement decisions (CRISP, JUIC, TART, EQ and OVERALL) were less than 5%. A lack of G×L in the WABP phase two trials (Hardner et al. IN PROGRESS) suggests that central Washington could be considered one selection environment. Removing a location resulted in a negligible decrease in accuracy and gain, but a sizeable savings of 33.3 units per candidate cost. Based on this analysis, there may be little value in the third location. 2L/2Y/3H was the other design with relatively small decreases in both accuracy and response to selection, and larger

113

reductions in cost than 2L/3Y/2H. G×Y and G×L×Y interactions were large for most traits.

Under the current design, G× age variance is confounded with G×Y variance. Our experience indicates that the initial crop(s) on young trees may not accurately represent a candidate selection’s fruit quality. Further studies could be conducted with the same genetic material planted over multiple years. Until then, the two years of fruit evaluations could be the third and fourth crop of the phase two trials, while still reducing the assessment costs by 41.3 units from the current design.

The presented accuracy and response to selection estimates assume that each trial is successful. Freezes that damage buds (fall), blossoms (spring), young fruitlets (spring) or hail damage on fruit (summer) can compromise an entire year’s data which would delay advancement decisions. In a crop such as apple, where time to release is already close to 20 years, each year of delay sacrifices speed and incurs considerable costs. The same risk management concerns for reducing years apply to reducing locations, especially if fruit assessment was delayed to the last two years. For that reason, reducing both locations and years (2L/2Y/3H) seems too high risk for the WABP.

Furthermore, stakeholder interests must be included. The grower is the ultimate consumer of new releases. If the third site is near a large proportion of interested growers or politically important, then removing that site may decrease their confidence and support in the program. All crop breeding programs are susceptible to pernicious weather events and serve their stakeholders to varying degrees, and thus must consider these contingencies.

Considerations for implementing new design in an operational breeding program.

Trial design accuracy must be balanced with managing risk, particularly in tree fruit crops where the investments in breeding, and by the grower in planting a new variety, are very high. An alternative option that further reduces risk of 2L/3Y/2H is to plant an incomplete

114

block, where each candidate was randomly planted at two out of three locations. This reduces propagation, field establishment and maintenance costs. Conversely, a complete block could be planted, but evaluated as an incomplete block each year. All three locations would be planted, but only two of the locations would be evaluated each year, saving much of the assessment costs associated with driving to harvest, harvesting, evaluating fruit and data entry. The third site acts as insurance in the event that one of the locations is compromised

(ie. freeze, hail, etc.) The cost for a single candidate in this scenario is 70.8 units. For a total program cost of 1,982 units, 8 additional candidates could be evaluated (28 total). CPD values for this design would be the same as the 2L/3Y/2H design. RS and CRS values would be between the 10% and 5% proportion selected values presented, but more complex methods would be needed to evaluate this quantitatively. Intensifying selection, or increasing the number of candidates evaluated without increasing the number of candidates advanced to the next phase, results in greater gains in trait values. The subsequent increase in RS values indicates that the overall quality of candidates advanced to phase three would likewise increase.

Utility of correlated response to selection in trial design considerations. Sensory evaluations tend to be expensive to measure, have lower heritabilities due to high variability

(Oraguzie et al. 2009) and suffer from ‘taste-fatigue’ (Arnold 1983), which was the motivation for investigating indirect selection using instrumental measures. Tartness (TART,

TA), sweetness (SWEET, SSC), crispness (CRISP, CN), and firmness (HARD, M1) are measured both organoleptically and instrumentally in the WABP. CRS indicated that more gain would be made for crispness and sweetness using direct selection of the sensory trait while tartness and firmness benefited from indirect selection using the instrumental measure.

The genetic correlation for both SWEET and SSC (0.57), and CRISP and CN (0.75) were

115

lower than for firmness or tartness (Appendix 4). Heritabilities were higher for TA, M1 and

CRISP than their paired trait and gained more from direct selection (Table 2). There is a higher heritability for SSC than SWEET (0.24, 0.21), however direct selection on SWEET made greater progress. This seeming discrepancy could be due to the low heritability of both measures of sweetness or the significant interactions between G×Y and G×L×Y observed for

SSC.

Sensory analysis is an expensive component of fruit assessment, and the results of this study suggest that organoleptic scoring of TART and HARD could be removed. However, both traits are scored with the other organoleptic traits so removing them would not decrease the costs significantly and subjective sensory analysis ultimately drives selection decisions

(Brown 1975; Hampson et al. 2000).

A similar argument could be made to remove the instrumental measurements CN and

SSC to reduce costs. CN is measured as part of a suite of variables on the Mohr® Digitest, so removing CN would not reduce costs of candidate assessment. However, measuring SSC adds to the time spent doing instrumental analysis and data collation. Nurserymen and growers are interested in SSC values at harvest and after storage for comparisons between cultivars and potential selections (i.e. Evans et al. 2012), as well as an indication of harvest maturity (Watkins 2003). SSC measurements could be restricted to phase three, as a way to reduce costs of candidate assessment.

RS and CRS assume that the type of selection for the trait is directional selection (ie. increasing or decreasing). Trial averages for several important traits such as TART and

HARD are within the desired range; directional progress for those traits would produce overly tart or low acid, and very firm or very soft apples, thus stabilizing selection is the goal.

116

RS ad CRS values may not be directly applicable for those traits, but are useful in understanding accuracy for either type of selection.

CRS values varied little despite changes in design for the four trait combinations presented, which was unexpected. While changes in the number of locations, years and harvests caused changes in the predicted candidate effect of the instrumental trait, the effect on the correlated response was small due to small G×L, G×Y, and G×H interactions. The larger denominator, due to the larger predicted entry effect values of the instrumental traits, also added to the small variation in CRS values.

Conclusion

The methods employed in this analysis offer a framework for other tree fruit and horticultural crop breeding programs to investigate their unique trial design accuracy and efficiency questions. Considerations specific to the WABP were outlined to demonstrate that trial design accuracy and cost must also be considered with specific breeding program’s needs, including risk management, stakeholder needs and characteristics of individual traits.

All breeding programs face challenges that would similarly inform their interpretation and the utility of the results.

Programs may want to examine improving accuracy while keeping the program cost static. In that case, a program might consider reducing some factors and increasing other factors. Additional methodologies could be joined with the cost analysis to investigate the effect of unbalanced trial designs (Piepho and Mohring 2007).

Previous analyses similar to this were undertaken in agronomic crops, where increasing yield was the main consideration (Butler et al. 2000; Cullis et al. 1996b; Patterson et al. 1977). For many of the target traits of agronomically important crops, directional selection is employed (Arief et al. 2015). Dessert apples are a unique commodity in that

117

multiple flavor and texture profiles are acceptable and expected by consumers (Bonany et al.

2014; Daillant-Spinnler et al. 1996), and thus the breeding targets are equally diverse. Apples may be an extreme example of this multi-targeted selection; however, the analysis yielded valuable results. The utility of the trial design efficiency analysis outlined in this study indicates that it would be equally, if not more, useful for breeders of other tree fruit and horticultural crops.

Author Contributions

JH, KE and CH designed the study. KE provided the data. JH analyzed the data with assistance from CH. JH drafted the majority of the manuscript. All authors read and approved of the final document.

Acknowledgements

This project was funded by Washington Tree Fruit Research Commission Projects

AP-11-103 and AP-13-106, and by AFRI NIFA Pre-doctoral Fellowship 123154-001. The authors wish to thank WABP grower cooperators for their interest, support and commitment to the WABP, as well as Dr. Bruce Barritt, Lisa Brutcher, Bonnie Schonberg and Nancy

Buchanan for patiently answering cost questions.

118

Literature Cited

Arief, V.N., DeLacy, I.H., Crossa, J., Payne, T., Singh, R., Braun, H-J., Tian, T., Basford, K.E., and Dieters, M.J. 2015. Evaluating testing strategies for plant breeding field trials: redesigning a CIMMYT international wheat nursery. Crop Sci. 55:164-177.

Arnold, G. 1983. A tasting procedure for assessing bitterness and astringency. Pages 109-114 in: Anonymous Sensory Quality in Foods and Beverages: Definition, Measurement and Control. Horwood Ltd., Florida.

Blanpied, G. D., and Silsby, K. J. 1992. Predicting Harvest Date Windows for Apples. Cornell Coop. Ext. Bul. 221.

Brennan, J. P., Kingwell, R. S., Thomson, F. M., and Cullis, B. R. 1998. Optimal strategies for regional cultivar testing. Aust. J. of Agric. and Resource Economics 42:209-226.

Bonany, J., Brugger, C., Buehler, A., Carbó, J., Codarin, S., Donati, F., and Schoorl, F. 2014. Preference mapping of apple varieties in Europe. Food Qual. Pref. 32:317-329.

Brown, A. G. 1975. Apples. Pages 3-37 in: Advances in Fruit Breeding. J. Janick and J. N. Moore, eds. Purdue University Press, Indiana.

Butler, D. G., Redden, R. J., Delacy, I. H., and Usher, T. 2000. Analysis of line × environment interactions for yield in navy beans. 1. Components of variance. Aust. J. Agric. Res. 51:597-605.

Cooper, M., Byth, D. E., Delacy, I. H., and Woodruff, D. R 1993. Predicting grain-yield in Australian environments using data from CIMMYT international wheat performance trials 1. Potential for exploiting correlated response to selection. Field Crops Res. 32:305-322.

Cullis, B. R., Thomson, F. M., Fisher, J. A., Gilmour, A. R., and Thompson, R. 1996a. The analysis of the NSW wheat variety database. 1. Modelling trial error variance. Theor. Appl. Genet. 92:21-27.

Cullis, B. R., Thomson, F. M., Fisher, J. A., Gilmour, A. R., and Thompson, R. 1996b. The

119

analysis of the NSW wheat variety database. 2. Variance component estimation. Theor. Appl. Genet. 92:28-39.

Daillant-Spinnler, B., MacFie, H. J. H., Beyts, P. K., and Hedderley, D. 1996. Relationships between perceived sensory properties and major preference directions of 12 varieties of apples from the southern hemisphere. Food Qual. Pref. 7:113-126.

Gilmour A. R., Gogel, B., Cullis, B. R., and Thompson, R. 2009. ASReml User Guide Release 3.0. VSN International Ltd., Hempstead.

Evans, K. 2013. Apple Breeding in the Pacific Northwest. Acta Hortic. 976:75-78.

Evans, K. M., Barritt, B. H., Konishi, B. S., Brutcher, L. J., and Ross, C. F. 2012. ‘WA 38’ apple. HortScience 47:1177-1179.

Evans, K., Brutcher, L., Konishi, B., and Barritt. B. 2010. Correlation of sensory analysis with physical textural data from a computerized penetrometer in the Washington State University apple breeding program. HortTechnology 20:1026-1029.

Falconer, D. S., and Mackay, T. F. C. 1996. Introduction to Quantitative Genetics. 4th ed. Longman Group Ltd., London.

Hampson, C. R., Quamme, H. A., Hall, J. W., MacDonald, R. A., King, M. C., and Cliff, M. A. 2000. Sensory evaluation as a selection tool in apple breeding. Euphytica 111:79- 90.

Hardner, C., Evans, K., Peace, C., and Brien, C. submitted. Genetic architecture of fruit quality traits in apple following short-term storage and implications for genetic breeding and selection. Tree Genet, Genomes.

Harker, F. R., Gunson, F. A., and Jaeger, S. R. 2003. The case for fruit quality: an interpretive review of consumer attitudes, and preferences for apples. Postharvest Biol. Technol. 28:333-347.

Kellerhals, M., Patocchi, A., Frey, J. E., Spuhler, M., and Duffy, B. 2009. Selection efficiency in apple breeding. Acta hortic. 814:177-183.

Kenward M. G., and Roger, J. H. 1997. Small sample inference for fixed effects from restricted maximum likelihood. Biometrics 53:983-997.

Munzel U., and Bandelow, B. 1998. The use of parametric vs. nonparametric tests in the statistical evaluation of rating scales. Pharmacopsychiatry 31:222-224.

120

Oraguzie, N., Alspach, P., Volz, R., Whitworth, C., Ranatunga, C., Weskett, R., and Harker, R. 2009. Postharvest assessment of fruit quality parameters in apple using both instruments and an expert panel. Postharvest Biol. and Technol. 52: 279-287.

Patterson H. D., and Thompson, R. 1971. Recovery of interblock information when block sizes are unequal. Biometrika 58:545-554.

Patterson, H. D., Silvey, V., Talbot, M., and Weatherup, S. T. C. 1977. Variability of yields of cereal varieties in UK trials. J. Agric. Sci. 89:239-245.

Piepho, H., and Mohring, J. 2007. Computing Heritability and Selection Response from Unbalanced Plant Breeding Trials. Genetics 177:1881-1888.

Shapiro S. S., and Wilk, M. B. 1965. An analysis of variance test for normality (complete samples). Biometrika 52:591-611.

Talbot, M. 1997. Resource allocation for selection systems. Pages 162-174 in: Statistical methods for plant variety evaluation, R.A. Kempton and P. N. Fox, eds. Chapman & Hall Ltd, London.

Watkins, C.B. 2003. Principles and Practices of Postharvest Handling and Stress. Pages 585- 614 in: Apples: Botany, Production and Uses, D.C. Ferree and I. Warrington, eds. CABI Publishing, Massachusetts.

Weber, M. 2008. Sustainable apple breeding needs sustainable marketing and management. 13th International Conference on Cultivation Technique and Phytopathological Problems in Organic Fruit-Growing, Weinsberg, Germany.

Wilks S. S. 1938. The large-sample distribution of the likelihood ratio for testing composite hypotheses. Ann. Math. Stat. 9:60-62.

121

Table 1. Costs of Phase 2 trial components for a single candidate standardized to the current design, total standardized Phase 2 trial cost and the total number of candidates that could be evaluated for the current total program cost for the current WABP Phase 2 trials and alternative

Phase 2 trial designs.

Current: Current: 3 Locations, 3 Years, 3 Harvests 2 Locations, 3 Years, 3 Harvests 3 Locations, 2 Years, 3 Harvests 3 Locations, 3 Years, 1 Harvest 3 Locations, 3 Years, 2 Harvest 3 Locations, 2 Years, 2 Harvests 2 Locations, 3 Years, 2 Harvests 2 Locations, 2 Years, 3 Harvests 2 Locations, 2 Years, 2 Harvests Nursery Production of Trees 7.0 4.7 7.0 7.0 7.0 7.0 4.7 4.7 4.7

Field establishment* 12.9 8.6 12.9 12.9 12.9 12.9 8.6 8.6 8.6

Field Maintenance† 5.7 3.8 4.3 5.7 5.7 4.3 3.8 2.9 2.9

Driving‡ 50.7 33.8 33.8 50.7 50.7 33.8 33.8 22.5 22.5

Harvest 6.5 4.3 4.3 4.2 5.3 3.6 3.6 2.9 2.4

Data Entry 1.6 1.1 1.1 0.5 1.1 0.7 0.7 0.7 0.5

Consumables** 0.1 0.1 0.1 0.0 0.1 0.0 0.0 0.0 0.0

Data Tree Diameter 0.4 0.2 0.2 0.4 0.4 0.2 0.2 0.2 0.2 Collection Sensory

Candidate Assessment 10.1 6.7 6.7 3.4 6.7 4.5 4.5 4.5 3.0 Analysis Instrumental 4.8 3.2 3.2 1.6 3.2 2.2 2.2 2.2 1.4 Analysis Total cost (units) per candidate 100.0 66.7 73.8 86.5 93.3 69.3 62.2 49.2 46.2

Total program cost (unit cost per 2000 1334 1476 1730 1866 1386 1244 984 924 candidate x 20 candidates) Total candidates that could be 20 30 27 23 21 29 32 40 43 evaluated for 2000 units *Includes fumigation, trellis materials, planting time, and labelling costs † Includes plot fees (trellis installation, pesticide application and irrigation), pesticides, pruning, thinning and orchard removal after 4 years ‡ Includes staff time, and per mile reimbursement for 12 weeks of the harvest season ** Includes labels printed for each sample

122

Table 2. Estimated phenotypic variance (v.P) and percentage of estimated variance components for individual random effects (G: candidate, H: harvest, L: Location, E: residual error) and interactions from the analysis of individual fruit quality traits assessed as part of the WABP

Phase 2 trials. Full details of traits given in text and in Table 1. Zero variance component indicates source of variation was not significant. Also shown is the repeatability for the fruit quality traits. The single largest source of variance for each trait is emboldened.

% Trait v.P G G.Y G.H G.Y.H G.L G.L.Y G.L.H E h2 Appearance APPSUM 0.38 38 7 0 0 4 6 0 45 0.38 FRTDM 0.08 51 5 0 0 4 11 0 29 0.51 FRTWT 3498 49 5 0 0 5 11 0 29 0.49 GCOL 0.33 38 8 0 0 1 7 3 42 0.38 LENT 0.58 52 6 0 0 2 5 0 35 0.52 PCOL 1.43 65 4 0 0 4 2 0 24 0.65 RUSS 0.96 37 8 0 0 2 8 0 45 0.37 SHAPE 0.91 38 8 0 0 5 0 0 48 0.39 SIZE 0.50 43 4 0 0 5 9 0 37 0.43 TCOL 0.45 59 3 0 2 2 6 0 29 0.59

Sensory AROM 0.46 26 7 0 0 3 16 0 48 0.26 CRISP 0.34 42 4 0 0 2 5 0 46 0.42 EQ 1.04 30 4 0 0 5 6 0 55 0.30 HARD 0.32 38 6 0 0 1 4 0 51 0.38 JUIC 0.26 27 8 0 0 2 6 0 57 0.27 OVERALL 0.23 18 4 2 0 4 0 0 72 0.18 SWEET 0.23 21 8 0 0 8 7 0 57 0.21 TART 0.24 46 0 0 4 0 8 0 42 0.46

Instrumental CN 4447 30 5 0 0 0 10 0 55 0.31 M1 6.42 61 0 0 0 2 16 0 22 0.61 M2 9.75 60 3 0 2 0 12 0 23 0.60 SSC 1.18 24 11 0 0 5 13 0 46 0.24 TA 0.02 57 3 0 0 4 2 2 32 0.57

123

Table 3. The critical percentage difference (CPD) required between sample means to reject the hypothesis that two candidates have the

same true mean with 95% confidence for traits assessed during Phase 2 trials by the WABP. Trait averages are presented for the current

design in the unit of the trait as is the CPD as a percentage. CPD presented for alternative designs were subtracted from the current

design to give the degree of change rather than absolute value.

% Current %- Alternative %

Current Trait Average

3 Harvests 3 Harvests 3 Harvests

Current: 3 Locations, 3 Years, 2 Locations, 3 Years, 3 Locations, 2 Years, 3 Locations, 3 Years, 3 Locations, 3 Years, 3 Locations, 2 Years, 2 Locations, 3 Years, 2 Locations, 2 Years, 2 Locations, 2 Years,

3 Harvests 1 Harvest 2 Harvest 2 Harvests 2 Harvests 2 Harvests

CRISP 3.18 10.5 1.7 2 3.6 1 3.3 3 3 5.6 HARD 3.54 9.6 1.3 2 3.3 0.9 3.2 2.5 2.5 5.1 AROM 2.9 17 2.6 3.3 3.9 1.1 4.6 4 4 8.1 124 JUIC 3.4 10.2 1.4 2.1 3.1 0.9 3.1 2.5 2.5 5

SWEET 3.28 11.5 1.8 1.8 2.7 0.7 2.7 2.7 2.7 5 TART 3.32 6.9 1.3 1.6 3.8 1.1 2.9 2.6 2.6 4.8 CN 195.5 20.8 3.2 4.7 7.4 2.1 7.2 5.9 8.5 11.8 M1 18.42 6.9 1.6 1.3 1.6 0.4 1.8 2.1 2.1 3.7 SSC 13.68 6.4 0.9 1.2 1.2 0.3 1.6 1.3 1.3 2.7 TA 0.64 12 2.1 1.7 3.7 1 3 3.4 3.4 5.5 EQ 5.78 11.7 2.1 1.9 3.6 1 3.2 3.3 3.3 5.9 OVERALL 1.7 19.2 3 3 8.7 2.5 6 6 6 10.1 SHAPE 2.98 22 2.9 3.7 5.7 1.5 5.7 4.9 4.9 9.5 SIZE 3.26 14 2.4 3.7 3.2 0.9 3.3 3.5 3.5 6.4 RUSS 3.82 17.3 2.2 1.9 4.3 1.2 5 3.8 3.8 8 LENT 3.04 14.8 1.9 1.4 3.8 1 4.2 3.3 3.3 6.8 FRTDM 2.94 6.2 1 1.1 1.2 0.3 1.5 1.4 1.4 2.7 FRTWT 209.54 18.6 3 3.9 3.3 0.9 4.1 4.1 4.1 7.8 TCOL 1.78 20 3.1 3.7 6.2 1.7 5.9 5.2 5.2 9.9 GCOL 2.04 18.7 2.3 2.9 5.4 1.5 5.5 4.3 4.3 8.8 PCOL 3.86 16.7 2.5 4.1 3.6 1 3.8 3.8 3.8 6.8 APPSUM 2.92 13.5 1.9 2.4 3.3 0.9 3.6 3.1 3.1 6.1

Table 4. Response to selection (RS) for traits assessed during Phase 2 trials by the WABP and correlated response to selection (CRS) for four

selected pairs of traits under the current and alternative trial designs. Selection intensity (SI) is 10%, unless noted. RS is in the unit of the trait

and CRS is in the unit of the sensory trait.

3 Harvests 3 Harvests 3 Harvests

Current: 3 Locations, 3 Years, Locations, 2 3 Years, 3 Locations, 2 Years, 3 Locations, 3 Years, 3 Locations, 3 Years, 3 Locations, 2 Years, Locations, 2 3 Years, Locations, 2 2 Years, Locations, 2 2 Years,

3 Harvests 3 Harvest 1 Harvest 2 Harvests 2 Harvests 2 Harvests 2 Locations, 2 Years 3 Harvests, 2 SI 5%

CRISP 0.63 0.62 0.62 0.61 0.63 0.61 0.62 0.61 0.60 0.72 HARD 0.58 0.57 0.57 0.56 0.58 0.56 0.56 0.56 0.54 0.66 AROM 0.54 0.52 0.52 0.51 0.53 0.51 0.51 0.50 0.48 0.60 JUIC 0.42 0.41 0.40 0.40 0.41 0.39 0.40 0.39 0.38 0.47 SWEET 0.32 0.31 0.31 0.30 0.32 0.30 0.30 0.29 0.28 0.35 TART 0.56 0.56 0.55 0.54 0.56 0.55 0.55 0.54 0.53 0.64 EQ 0.90 0.87 0.88 0.85 0.89 0.86 0.86 0.84 0.82 1.00 125 OVERALL 0.31 0.30 0.30 0.27 0.30 0.29 0.29 0.28 0.27 0.34 CN 59.68 58.37 57.72 56.47 58.83 56.57 57.19 55.97 54.42 67.03

CRCRISP 0.50 0.50 0.50 0.50 0.50 0.50 0.50 0.50 0.50 0.58 M1 3.39 3.35 3.36 3.35 3.38 3.34 3.34 3.3 3.28 3.91

CRHARD 0.59 0.59 0.59 0.59 0.59 0.59 0.59 0.59 0.59 0.69 SSC 0.81 0.78 0.77 0.77 0.8 0.75 0.76 0.73 0.72 0.90

CRSWEET 0.22 0.22 0.22 0.22 0.22 0.22 0.22 0.22 0.22 0.26 TA 0.18 0.18 0.18 0.17 0.18 0.17 0.17 0.17 0.17 0.2

CRTART 0.56 0.56 0.56 0.56 0.56 0.56 0.56 0.56 0.56 0.66 SHAPE 0.96 0.94 0.94 0.93 0.95 0.93 0.93 0.92 0.9 1.09 SIZE 0.77 0.76 0.76 0.75 0.77 0.75 0.75 0.74 0.73 0.88 RUSS 0.97 0.96 0.95 0.94 0.97 0.93 0.94 0.92 0.91 1.11 LENT 0.93 0.92 0.91 0.91 0.93 0.91 0.91 0.90 0.89 1.07 FRTDIAM 0.34 0.34 0.34 0.34 0.34 0.33 0.33 0.33 0.33 0.39 FRTWTG 69.01 67.79 67.79 67.69 68.68 67.32 67.32 66.31 65.65 78.91 TYPECOL 0.89 0.88 0.87 0.87 0.88 0.87 0.87 0.86 0.85 1.02 GRDCOL 0.58 0.58 0.57 0.56 0.58 0.56 0.57 0.56 0.55 0.66 PCOL 1.65 1.64 1.64 1.63 1.64 1.63 1.63 1.62 1.61 1.91 APPSUM 0.59 0.57 0.57 0.57 0.58 0.56 0.57 0.56 0.55 0.66

Appendix 1. Ordinal Trait Descriptions

Trait Abbreviation Trait Description Scale Description Ordinal Appearance Traits SHAPE Shape 10 increments, from flat to cylindrical SIZE Size 11 increments, from tiny to very large PCOL Extent of red color 10 increments, from 5% to 95% TCOL Red color type 6 increments, from blush to stripe GCOL Predominant background 6 increments, from green to yellow color RUSS Extent of russetting 11 increments, from severe to absent LENT Extent of lenticels 10 increments from large to absent APPSUM Appearance Summary 11 increments from ugly to beautiful Ordinal Sensory Traits HARD Hardness 11 increments from soft to very firm CRISP Crispness 10 increments from chewy to crisp JUIC Juiciness 11 increments from very dry to very juicy SWEET Sweetness 9 increments TART Tartness 11 increments AROM Aromatic flavor 12 increments from none to very fruity EQ Eating Quality Summary 15 increments from yuck to outstanding OVERALL Overall 4 increments from reject to advance

126

Appendix 2. Derivation of expression for predicting response to selection

Falconer presents response to selection as the multiplication of selection intensity (i), narrow-

2 sense heritability (h ) and standard error of the phenotype (σP).

2 푅 = 𝑖ℎ 휎푃

Heritability may be defined as the correlation between the selection clue and the true genetic value of the selection objective. For example, individual narrow sense heritability (h2) may be

2 휎퐴 written as 2 which is the correlation between the phenotype of an individual and the true 휎푃

2 휎퐴 휎퐴∗ 휎퐴 additive genetic value of the individual. 2 can be expanded to , so that response to 휎푃 휎푃∗ 휎푃

휎퐴∗휎퐴 selection becomes: 푅 = 𝑖 휎푃 휎푃∗휎푃

The numerator σP would cancel out a denominator σP. Hence, response to selection on the

휎2 phenotype to improve the additive genetic value can be written as: 푅 = 𝑖 퐴 휎푃

When referring the mean of a genotype rather than of an individual,  is used for phenotypic Gˆ

partitioning of the mean of a genotype, rather than σP. is the predicted candidate effect or the selection clue. Additive variance can be rewritten as covariance between the predicted candidate effect and the true clonal mean of the candidate, or selection objective. Therefore, response to

 ˆ selection was calculated as: Ri GG,  Gˆ

127

Appendix 3. Derivation of expression for predicting correlated response to selection

Falconer presents the equation for correlation response between two traits as:

퐶푅푆푦 = 𝑖ℎ푥푟푥푦휎퐺푦

휎퐺푥 푐표푣푥푦 The heritability of trait x is and the genetic correlation of trait x and y is 푟푥푦 = . 휎Ĝ푥 2 2 √ 휎퐺푥휎퐺푦

The square root of a variance turns into the standard error of the estimate. The equation then becomes:

휎퐺푥 푐표푣푥푦 퐶푅푆푦 = 𝑖 휎퐺푦 휎Ĝ푥 휎퐺푥휎퐺푦

The denominator of the genetic correlation is cancelled with the same terms in the numerator, so the equation becomes:

푐표푣퐺푥퐺푦 퐶푅푆푦 = 𝑖 휎Ĝ푥

Covariance of Gx and Gy can be rewritten as 푟퐺 퐺 휎2 휎2 푥 푦√ 퐺푥 Ĝ푦

The equation is thus:

푟퐺 퐺 휎2 휎2 푥 푦√ 퐺푥 Ĝ푦 퐶푅푆 = 𝑖 푦 휎 Ĝ푥

128

Appendix 4. Genetic Correlation matrix for apple fruit quality traits assessed following short term storage among (a) instrumental traits M1, M2 and CN, and sensory traits HARD, CRISP and JUIC, and (b) instrumental traits SSC and TAI, and sensory traits SWEET, AROM and

TART.

(a) M2 HARD CN CRISP JUIC

M1 0.86 0.96 0.31 0.17 0.01

M2 0.90 0.43 0.34 0.20

HARD 0.41 0.30 0.14

CN 0.75 0.71

CRISP 0.87

(b) AROM SWEET TA TART

SSC 0.56 0.57 0.12 0.25

AROM 0.84 0.05 0.25

SWEET -0.29 -0.06

TA 0.97

129

CHAPTER FIVE: Summary and Larger Impact

The wider impact and implications of the results and conclusions presented in this dissertation are seen in their use in the WSU apple breeding program:

Moldy Core Susceptibility. Moldy core is a minor postharvest disease compared to blue mold and grey mold, but is also important, particularly to consumers that find a molded core.

Furthermore, moldy core is more challenging to control than either of those diseases. Several studies have examined commercially important varieties for their susceptibility and disease cycle; however, breeders were making selection decisions against moldy core susceptibility on largely anecdotal evidence (K. Evans, personal communication), which potentially resulted in culling selections unnecessarily. This anecdotal evidence was supported by the results of this germplasm survey. Scored as part of a wider phenotypic survey on two diverse germplasm sets, open core/ open calyx individuals had higher incidence of moldy core than other core/ calyx combinations and an open core contributed to moldy core incidence more than an open calyx.

Two crosses with one susceptible parent (‘Pinova’, ‘Ginger Gold’) had increased number of progenies with moldy core susceptibility, indicating that moldy core may be heritable. Larger progenies would need to be developed and scored to confirm heritability.

Fire Blight Resistance. Fire blight is an increasingly important breeding target of the WABP.

As epidemics become more prevalent, the success of the cultivars released from this program will become more dependent on their resistance. It is important not to rely on only one source of fire blight resistance when incorporating resistance into a commercially viable cultivar, as was learned with scab resistance in apple, where Rvi6 (Vf) resistance, a major breeding target for

130

multiple programs, was overcome by the pathogen. Pyramiding resistances from different sources more likely leads to durable resistance. This is especially salient in a perennial crop such as apple, where plantings last upwards of 20 years, and having to replant an orchard to replace a now-susceptible variety would cause staggering economic losses to the grower.

Malus sieversii is an excellent potential source of fire blight resistance as its fruit quality is not as far removed from commercial acceptability as other sources of resistance. Of the 197 accessions screened, four showed no evidence of fire blight infection when inoculated or with natural occurrence. Eleven others developed slight fire blight infection, but seemed to have a substantial level of resistance. Fruit from all of these have been evaluated by the WABP at both fresh (after receiving them from WV) and after 2 months storage. All fruit samples had brix levels about 11° and the two largest fruit samples were between 75 and 80 grams. Three in total were given favorable remarks for use as parents and will likely be used in 2016 crosses.

This study and subsequent association mapping could result in identification of new genetic tests for MAPS or MASS. Studies like this also add value to the PGRU collections; using the collections and providing evidence of their utility to modern breeding efforts supports the preservation and management of wild germplasm related to important crop species.

Phase 2 Trial Design efficiency. Phase 2 is arguably the most important phase of the program as the majority of data is collected and intimate knowledge of the selections is gained. Furthermore, multiple research programs outside of the breeding program use the diverse genotypes to conduct research. Altering the design of Phase 2 is a decision to be made with careful deliberation. The methods and analysis outlined on the effect of design changes on the accuracy with which traits are evaluated and the costs of the program allow for a more informed decision to be made.

131

The design change suggested for the WABP, 3 locations planted but only 2 used for fruit evaluation each year, 3 years, 2 harvests per year, is being considered for the 2016 planting of

Phase 2. The program has not experienced a harvest year in which all three sites yielded valid data since this design alteration was suggested. One site had irrigation issues that led to drought conditions and a vole infestation in 2014 and in 2015, each site was compromised: early hail at all three, lack of water at one due to drought and water rights, and orchard maintenance issues at another. All three sites were still harvested in both years, however the outcomes of this analysis would encourage an earlier “abandon” decision for a particular compromised site in future years.

The methods and analysis are the first to be published on a horticultural tree fruit crop.

The goal was to outline methodology that other horticulture crop breeding programs could utilize; there has already been interest from two other programs: peach breeding at Clemson

University (K. Gasic, personal communication) and tart cherry breeding at Michigan State

University (A. Iezzoni, personal communication).

These three projects all function to improve accuracy or efficiency through quantification of different aspects of the Washington State University apple breeding program; the original hypotheses were all supported by the conclusions of the research:

Chapter 2: Increasing knowledge about the susceptibility to moldy core of currently available

germplasm will lead to more informed crossing decisions and ultimately, releases from

the program that have less susceptibility.

Chapter 3: Screening a large set of Malus sieversii accessions for fire blight resistance will lead

132

to more informed crossing decisions and ultimately, releases from the program that have

greater resistance.

Chapter 4: The accuracy with which traits are evaluated in advanced trials can be nominally

reduced in a way that allows the program to evaluate more selections in those trials for

the same total program cost, which increases the likelihood of releasing better apple

varieties.

133