Chapter 1: Introduction

INVESTIGATING THE IMPACTS OF ANTIBIOTICS AND ALTERNATIVES ON THE

SUSTAINABLE MANAGEMENT, DISTRIBUTION, AND SPREAD OF FIRE BLIGHT

A Dissertation

Presented to the Faculty of the Graduate School

of Cornell University

In Partial Fulfillment of the Requirements for the Degree of

Doctor of Philosophy

Anna Elizabeth Wallis

December 2020

INVESTIGATING THE IMPACTS OF ANTIBIOTICS AND ALTERNATIVES ON THE

SUSTAINABLE MANAGEMENT, DISTRIBUTION, AND SPREAD OF FIRE BLIGHT

Anna Elizabeth Wallis, Ph.D.

Cornell University 2020

Fire blight, caused by the bacteria Erwinia amylovora, is one of the most important diseases of apples. Despite over 200 years of intense investigation, it continues to cause devastating losses to growers worldwide. The most effective management tool in the Eastern US is the antibiotic streptomycin. However, antibiotic use in agriculture has come under scrutiny and there is a need to better understand the sustainability of streptomycin and alternative management solutions. The first goal of this research was to investigate impacts of streptomycin and alternatives on fire blight disease management and orchard health. Application of the plant growth regulator prohexadione-calcium pre-bloom was investigated as a novel alternative to antibiotics. Results indicated significant reduction in disease incidence to levels comparable with streptomycin, without compromising tree growth. Separate work investigated effects of streptomycin on endophytic bacterial communities in the canopy, communities potentially implicated in orchard health and antibiotic resistance. Minimal effects were detected on abundance and structure of communities, indicating the sustainability of current streptomycin programs. The second goal of this work was to describe distribution and spread of fire blight at multiple scales, with consideration for streptomycin resistance. First, we investigated risk of fire blight development and spread using mechanical thinning and pruning. Field trials indicated both practices could be used safely, and any risk mitigated by applying streptomycin after mechanical treatment. Next, distribution within orchard blocks was described for two fields over two years using spatial analyses. These case studies illustrated different mechanisms of introduction and spread. Finally, we investigated geographic distribution of E. amylovora strains and streptomycin resistance in apple production regions in the Northeastern US. Over 150 isolates were described in terms of streptomycin resistance phenotype and CRISPR profile. Several CRISPR profiles were widely distributed, while many others were only identified in one location. In a streptomycin resistant E. amylovora outbreak in Wayne County, all isolates exhibited CRISPR profiles matching the original SmR isolates discovered in NY, indicating the outbreak likely resulted from imperfect eradication. The work presented herein will inform recommendations for sustainable fire blight management, with immediate impacts to commercial production in NY

State and beyond.

BIOGRAPHICAL SKETCH

Anna Wallis was raised in Harford County, Maryland and attended Fallston High School, where she began her academic pursuit of biology and music. She attended University of

Maryland where she received a BS in Biology and a BA in Music, with a specialization in vocal performance. As an undergraduate she began working as a teaching assistant for the introduction to horticulture course in the Plant Science Department, where she fell in love with plant science and apple production. During this time and the year following her undergraduate work, she gained experience in plant science, entomology, and community agriculture, working in labs at the University of Maryland in College Park and the USDA in Beltsville, at the Green Farmacy

Garden, and at Eco City Farms. She then returned to school to pursue a Master’s Degree in Plant

Science with Dr. Chris Walsh, Professor of Horticulture, where she focused on lettuce and apple production in the Mid-Atlantic. After earning her degree, she worked for Cornell Cooperative

Extension as the Apple Specialist in the Champlain Valley of Eastern NY for three years, providing support to growers in through educational programming and applied research. She returned to school in 2017 to pursue a Ph.D. in Plant Pathology at Cornell University under the direction of Dr. Kerik Cox, where her work has focused on fire blight, a bacterial disease of apples caused by Erwinia amylovora.

iii

Dedicated to Colvin, my most loyal and patient companion,

relentless mischief-maker,

and welcome distraction.

ACKNOWLEDGEMENTS

I cannot thank my advisor Dr. Kerik Cox enough for how he has helped and inspired me.

Throughout my graduate work, he provided guidance, encouragement, and opportunity that has enabled my research and successful development as a professional in the fields of tree fruit production and plant pathology. I am continuously amazed by Kerik’s visionary research and teaching, and his tireless commitment to his work, students, and community. I have learned from his example what it takes to accomplish excellent research and cultivate an exciting, supportive lab environment, while maintaining a broad perspective and an all-important sense of humor.

I would also like to thank my committee members, Drs. Susan Brown, Lance Cadle-

Davidson, Andrea Ottesen, who have provided me with invaluable advice and compassionate support, in both professional and personal capacities. Their accomplishments, passion, and dedication are endless sources of inspiration on which I will model my future career.

My work would not be possible without the help of collaborators from the tree fruit community as well as the USDA and FDA, who facilitated this research by contributing their time, resources, and insights to these projects. In particular, Dr. Terry Bradshaw, who allowed me to conduct trials in his orchards at the University of Vermont Horticultural Research Station, the members of the commercial apple industry, Scott Palmer, Tom DeMarree, Kevin Bowman,

Chris Carballiera, Pete Ten Eyck, Kevin Bittner, and Bobby Brown, who allowed me to work in their orchards, and the Cornell Cooperative Extension Specialists and private consultants Dan

Donahue, Mike Basedow, Jim Eve, and Janet van Zoeren who collected numerous samples.

Thank you also to Padmini Ramachandran and Ganyu Gu who invited me into their labs and patiently taught me NGS and bioinformatics skills.

I would like to thank my family and friends for their love and support. I am incredibly fortunate to have so many people invested in celebrating my successes and sympathizing with my failures, no matter how great or small. Thanks Mom, for being my biggest fan and listening to countless practice presentations; thanks Dad, for always reminding me not to lose sight of my goals and who I am; and thanks Maggie, for reminding me to be kind, silly, and happy at a time when it has become easy to take everything too seriously. My labmates, Mei-Wah Choi, Katrin

Ayer, David Strickland, and Isabella Yannuzzi, have been incredibly supportive and helpful, and

I am so indebted to them for their help with my work, and general encouragement and positivity

(Type IV Fun, More Money, and many more inside jokes…); I hope I have been half as much of a positive influence on your work as you have been on mine.

Finally, thank you to the groups that supported this work financially, including the

Northeast Sustainable Agriculture Research and Education (NESARE) grant LNE19-385R, the

New York Farm Viability Institute (NYFVI) grant FVI 19 002, the Department of Agriculture &

Markets Apple Research & Development Program (ARDP) grants FY 2018-19 & 2017-18, agrichemical companies of Northeastern United States, and the Arthur Boller Research Fund grant 2019 and 2020.

TABLE OF CONTENTS

Page Front Matter Title Page ...... i Biographical Sketch ...... iii Dedication ...... iv Acknowledgements ...... v Table of Contents ...... vii List of Figures ...... ix List of Tables ...... xi

Chapter 1: Introduction ...... 1 Literature Cited ...... 8

Part I: Impacts of Antibiotics and Alternatives on Disease Management and Orchard Health ...... 11

Chapter 2: Management of Fire Blight Using Pre-bloom Applications of Prohexadione- calcium ...... 12 Abstract ...... 12 Introduction ...... 13 Materials and Methods ...... 19 Results ...... 24 Discussion ...... 30 Literature Cited ...... 34

Chapter 3: Endophytic Bacterial Communites in Apple Leaves are Minimally Impacted by Streptomycin Use for Fire Blight Management ...... 38 Abstract ...... 38 Introduction ...... 40 Materials and Methods ...... 44 Results ...... 52 Discussion ...... 59 Literature Cited ...... 65

Part II: Investigating Distribution and Spread of Fire Blight at Multiple Scales .....73

Chapter 4: Assessing and Minimizing the Development and Spread of Fire Blight Following Mechanical Thinning and Mechanical Pruning in Apple Orchards ...... 74 Abstract ...... 74 Introduction ...... 76 Materials and Methods ...... 80 Results ...... 87 Discussion ...... 102 Literature Cited ...... 108

vii

Chapter 5: Examining Spatial Distribution and Spread of Fire Blight in Apple Orchards: Two Case Studies ...... 112 Abstract ...... 112 Introduction ...... 113 General Methods ...... 115 Case Study 1 ...... 116 Case Study 2 ...... 119 Conclusions ...... 124 Literature Cited ...... 124

Chapter 6: Investigating the Distribution of E. amylovora Strains and Streptomycin Resistance in Apple Orchards in New York Using CRISPR Array Profiles: A Six-Year Follow Up...... 126 Abstract ...... 126 Introduction ...... 128 Materials and Methods ...... 132 Results ...... 139 Discussion ...... 151 Literature Cited ...... 159

Chapter 7: Conclusions ...... 164

Appendix ...... 167 Supplemental Tables and Figures for Chapter 6...... 167

viii

LIST OF FIGURES

Page

Figure 1.1. Overview of the life cycle of Erwinia amylovora, causal agent of fire blight of apples 3

Figure 1.2. Overview of fire blight management in commercial orchards 5

Figure 2.1. Hand cross-sections of apple fruitlet pedicels at 50 days post-treatment (40 days after full bloom) for trees treated with 0 g/L and 125 g/L prohexadione- calcium 22

Figure 2.2. Mean cell wall thicknesses (±1 SE) of fruitlet pedicels taken in 2018 and 2019 for Gala trees treated with prohexadione-calcium (trade name Apogee) at 20 days and 50 days after treatment 28

Figure 3.1. Alpha diversity (Shannon Diversity Index) values displayed as box plots for endophytic bacterial communities in 2018 and 2019, of leaf samples collected from apple trees treated with different fire blight management programs 55

Figure 3.2. Nonmetric multidimensional scaling (NMDS) ordination of Bray-Curtis dissimilarity measures in 2018 or 2019 of leaf samples collected from apple trees treated with different fire blight management programs 57

Figure 3.3. Relative abundance of the ten most dominant bacterial families in endophytic bacterial communities of leaf samples in 2018 and 2019, collected from apple trees treated with different fire blight management programs 58

Figure 4.1. Schematic of treatments used to evaluate the risk of fire blight development and spread with mechanical thinning and pruning in apple orchards. 83

Figure 4.2. Disease progress curves illustrating the relationship between mean incidence of blossom or shoot blight and distance from an inoculated point source within a row for ‘‘Idared’’ apple trees spaced at approximately 5 m, receiving mechanical thinning or pruning 101

Figure 5.1. Distribution of fire blight incidence within a vertical axis orchard block of mixed varieties on B.9 rootstock at Cornell AgriTech Research Orchards in Geneva, NY. 118

Figure 5.2. Distribution of fire blight incidence within a high density apple orchard block of mixed varieties on G.935 rootstock at Cornell AgriTech Research Orchards in Geneva, NY. 122

Figure 5.3. Erwinia amylovora strains recovered from three nearby orchard blocks at the Cornell Research Orchards in Geneva, NY in 2018 and 2019. 123

Figure 6.1. Spacer patterns of clustered regularly interspaced short palindromic repeat (CRISPR) regions CR1, CR2, and CR3 for E. amylovora isolates 142

Figure 6.2. Map of fire blight prevalence in counties in NY State from 2017 to 2020 145

Figure 6.3. Number or percent of clustered regularly interspaced short palindromic repeat (CRISPR) profiles identified for E. amylovora isolates 146

Figure 6.4. Phylogenetic tree of 19 E. amylovora isolates representing diverse geographic origin and host plant 150

LIST OF TABLES

Page

Table 2.1. Summary of pre- and post-bloom treatments evaluated for fire blight protection and horticultural impacts in 2016, 2017 and 2018 20

Table 2.2. Mean disease incidence and horticultural parameters of Gala trees treated with pre-bloom fire blight programs utilizing prohexadione calcium (Apogee) 26

Table 3.1. Fire blight disease management treatments applied to apple orchards in Geneva, NY 46

Table 3.2. Abundance of bacterial endophytes in apple leaves treated with different fire blight management programs for two orchards in Geneva, NY in 2018 and 2019 53

Table 4.1. Development of fire blight beyond the inoculated point source for apple trees receiving mechanical thinning or pruning 94

Table 4.2. Incidence of fire blight observed for trees receiving different mechanical thinning or pruning treatments, at 10 m distances from inoculated point sources 95

Table 4.3. Relative area under the disease progress curve (rAUDPC) for trees receiving mechanical thinning or pruning, from inoculated point sources 96

Table 4.4. The incidence of fire blight observed and predicted by two models (exponential decay and Gregory's power law), for trees receiving different mechanical thinning or pruning treatments, at 10 m distances from inoculated point sources 97

Table 4.5. Summary statistics used to evaluate models describing the development and spread of fire blight from inoculated point sources 99

Table 4.6. Characteristics of mechanical pruning practices in 13 commercial fruit farms located in the Lake Ontario Fruit Region, NY, US 108

Table 6.1. Area under the disease progress curve (AUDPC) for strains of E. amylovora evaluated for aggressiveness 139

Table 6.2. Clustered regularly interspaced short palindromic repeat (CRISPR) profiles of E. amylovora isolates collected from commercial apple orchards in NY from 2017 to 2020 142

Table 6.3. Characteristics of E. amylovora genomes sequenced in this study 149

xii

CHAPTER 1

INTRODUCTION

Apple production in NY State

Apples are one of the most important specialty crops worldwide, with 60-70 million metric tons produced annually (WAPA 2014). The United States is the second largest contributor to this market after China, producing an estimated 9-11 billion pounds of utilized production each year with a value of $3-3.5 billion (USDA NASS 2017). NY State is the second largest contributor to national production of apples, making up approximately 10% of production, with over 1,000 farms producing apples and over 50,000 acres planted, as of the 2012 Agricultural

Census (USDA NASS 2012). The industry provides over 10,000 direct agricultural jobs and another 7,500 indirect jobs (handling, distribution, and marketing) in NY State (NYAA 2012).

Clearly, apple production is a significant part of the NY economy, but with the nickname ‘The

Big Apple’ it is also a part of the state’s identity. It is essential that the apple industry remain vibrant for the welfare of the economy and the people of NY State and worldwide.

Fire Blight: Origin, Economic Significance, and Biology

Fire blight, caused by the bacterial pathogen Erwinia amylovora, is one of the most destructive diseases of apples worldwide. It is capable of devastating entire orchards in unforeseen epidemics and costs farmers millions of dollars in the United States and worldwide

(Norelli 2003). Originally identified in the Hudson Valley of New York in the 1790s, it was formally described by Thomas Burrill in 1880 (van der Zwet 2012). It was the first proven instance of a bacteria as the causal organism of a plant disease, shortly following the discovery of the germ-theory of disease and bacterial pathogens in human medicine. The disease has spread

to many apple production regions worldwide, with the exception of parts of South America,

Africa, and Asia (CABI ISC 2018; Norelli 2003; van der Zwet 2012). Economic losses in the

U.S. are estimated at over $100 million annually; most recently, fire blight outbreaks in the

Champlain Valley of NY in 2016 and Southwest Michigan in 2000 and 2005 impacted most commercial farms in the regions with total economic losses estimated at $9,500 per acre and $42 million across a region (Elizabeth Higgins personal communication 20 Sept. 2018, Norelli 2003).

These epidemics demonstrate that, despite over 200 years of study, there is still much that is poorly understood about the disease that impedes effective management in commercial systems.

E. amylovora is a gram-negative bacterium of Enterobacteriaceae, and classified as a hemibiotroph. Throughout its life cycle, the pathogen is capable of infecting and spreading to all tissues of the plant (Figure 1.1). The bacteria survive on living host tissue at the margin of cankers from old infection sites. In the spring under warm, humid conditions, cankers produce a sticky bacterial ooze; bacteria rapidly colonize the plant surface and are vectored to the stigmatic surface of the flower by insects and rain splash (van der Zwet 2012). The primary infection period occurs during bloom, when bacteria are washed into the flowers through floral nectaries, resulting in blossom blight (Schroth et al. 1974; van der Zwet 2012). A secondary form of infection is shoot blight, in which shoots become infected by ooze from cankers or blighted flowers via wounds caused by wind, hail, or insects (Norelli 2003; Schroth et al. 1974; van der

Zwet 2012).

Figure 1.1. Overview of the life cycle of Erwinia amylovora, causal agent of fire blight of apples. Bacteria overwinter at the margins of cankers formed from previous infections. Ooze in produced in the spring. Bacteria are transferred to blossoms where they preferentially colonize the stigmatic surface and are washed into the host via natural openings in the nectaries during wetting events. The bacteria may also enter and infect susceptible shoot tissue through wounds caused by wind whipping or hail. Bacteria may also travel systemically to the rootstock via the vascular tissue.

Fire Blight: Disease Management

The management paradigm for fire blight is well established and includes integration of of cultural, mechanical, and chemical tactics (Figure 1.2). Sanitation via pruning cankers during the dormant period of apple growth is a critical measure for reducing inoculum (Norelli 2003;

Schroth 1974; van der Zwet 2012). Additionally, removal of alternate hosts such as hawthorn and mountain ash is an effective way to reduce sources of inoculum. General antimicrobials such as copper and lime sulfur are applied at late dormant through green tip, to reduce inoculum on the surface of the plant. Host resistance, both of rootstocks (Norelli 2003; Russo et al. 2007) and

scion varieties (Brown 2012; Harshman et al. 2017; Norelli 2003; van der Zwet 2012) has also been employed to reduce orchard susceptibility and increase resilience of orchards to the disease.

The most effective form of commercial control for over 40 years has been the application of antibiotics during infection events. The predominant antibiotic used is the aminoglycoside streptomycin (Norelli 2003). Recently, kasugamycin, another aminoglycoside antibiotic, has also been registered in several states for fire blight management in orchards, in response to the emergence of streptomycin resistant E. amylovora (McGhee and Sundin 2011). The antibiotic oxytetracycline is also used for management, although less frequently because of its lower efficacy, likely due to bacteriostatic rather than bacteriocidal activity, shorter half-life, and sensitivity to UV radiation. To protect against blossom blight, antibiotics must be applied to open flowers (Norelli 2003; van der Zwet 2012). Follow-up antibiotic applications are used to protect against shoot blight throughout the season as necessary following trauma events as copper may result injury to fresh market crop. In the past 15 years, precise weather-based models (i.e.

CougarBlight, Maryblyt, and RIMpro) have been developed and widely adopted in decision aid support systems to predict infection events and provide recommendations for precise timing of management applications (Biggs and Turechek 2010; Carroll et al. 2017; Turechek and Biggs

2015). Streptomycin routinely provides greater than 90% reduction in blossom and shoot blight incidence in trials in the Northeastern US (Aldwinckle et al. 2002; Norelli 2003).

Figure 1.2. Overview of fire blight management in commercial orchards. Each image of a bud or flower cluster with blue background indicates a phenological stage of the apple host. Inoculum is removed during the dormant season by pruning out cankers formed from previous infections. Copper is applied in early spring to kill bacteria on the surface of the plant, from dormant through tight cluster. Antibiotics, biological materials, and plant defense inducing compounds are applied during infection events, primarily at bloom and as needed during the summer.

Streptomycin resistance in Erwinia amylovora

The use of antibiotics in agriculture has come under public scrutiny due to the potential for selection of resistance in pathogen populations and off-target bacteria (McManus et al. 2002).

Streptomycin resistance (SmR) has been identified in E. amylovora populations in major production regions in the United States. It was first reported in 1971 in California (Miller and

Scroth 1972) and has since been identified in most major production regions in the US, including

Washington and Oregon (Coyier and Covey 1975), Michigan (McGhee et al. 2011), and New

York (Russo et al. 2008; Tancos and Cox 2016). Additionally, antibiotics have been prohibited in organic production in the United States, following regulation by the National Organic

Standards Board in October 2014; this is consistent with organic regulations in other locations including Canada and European Union (Granatstein et al. 2013; Johnson and Temple 2013).

Alternative management strategies are necessary for the sustainable management of the disease.

There are two known determinants for streptomycin resistance: 1. presence of a point mutation in the rpsL gene coding for the ribosomal protein that is targeted by streptomycin

(Chiou and Jones 1995a) and 2. presence of a streptomycin-modifying enzyme encoded by the gene pair strA-strB (Chiou and Jones 1995b), typically present within the transposon Tn5393 on plasmids pEA29 and pEa34 (Förster et al. 2015; McGhee et al. 2011; McManus et al. 2002).

Both can be detected via DNA amplification. Resistant strains in western and eastern production regions nearly always have contrasting mechanisms, with the majority of strains in western production regions possessing the chromosomal mutation and strains in eastern production regions exhibiting the plasmid-mediated gene pair.

Alternative management programs for blossom and shoot blight have included biocontrols and plant regulators. Biological materials include biological control agents (BCAs) and biopesticides, defined as living organisms or naturally derived products respectively, applied to the plant or soil to control a disease or pest (Pal and Gardener 2011). Products used for fire blight management typically fall into three possible categories based on their mode of action.

Protectant materials in the form of competitive microorganisms include Pantoea agglomerans,

Pseudomonas fluorescens, and Aureobasidium pullulans (Norelli 2003; Sundin et al. 2009; van der Zwet 2012). Other protectants include products produced by microorganisms which have antibiotic properties, including Bacillus subtilus and B. amyloliquefaciens, marketed as Double

Nickel and Serenade (Norelli 2003; Sundin et al. 2009; van der Zwet 2012). There is also interest in products that elicit an ‘induced defense’ response in the host. Induced defense products

available for fire blight management include microorganisms such as Bacillus species, plant extracts including Reynoutria sachaliensis (Regalia, Marrone Bioinnovations), and synthetically produced products such as acibenzolar-S-methyl (Actigard, Syngenta) (Norelli 2003; Vallad and

Goodman 2004; van der Zwet 2012). These materials trigger a signaling cascade through two natural pathways in the host, the induced systemic resistance (ISR) or systemic acquired response (SAR) pathways, leading to induced resistance throughout the plant. Overall, increasing interest in biologicals is driving production of a multitude of new products and formulations.

Despite over 200 years of investigation, fire blight is still a major source of grower consternation and an active area of research (van der Zwet, T. 2012). Better understanding of pathogen biology and epidemiology will continue to inform more sustainable management of the disease.

Overarching Research Goal

The goal of this work has been to add to the scientific and commercial understanding of fire blight, a bacterial disease of pome fruits caused by Erwinia amylovora. The specific projects in this dissertation aim to provide a better understanding of the disease and pathogen at multiple scales (from the interaction with endophytic bacterial communities to the geographic distribution of E. amylovora strains), and to develop more sustainable management strategies, especially as related to development of antibiotic resistance.

Specific Research Objectives

Part I: Evaluate the impacts of antibiotics and alternatives on disease management and orchard health. Specifically:

1. Evaluate the impact of novel prohexadione-calcium programs on fire blight disease

occurrence and host performance in established orchards and young apple trees

2. Determine the impact of streptomycin applications for fire blight management on the

bacterial endophytic communities of the apple phyllosphere and potential impact on

development of streptomycin resistance

Part II: Investigate the distribution of fire blight at multiple scales. Specifically:

3. Determine the risk of fire blight incidence and spread when using mechanical thinning

and mechanical hedging

4. Describe the distribution and spread of E. amylovora within orchard blocks using spatial

analysis techniques

5. Utilize CRISPR strain characterization to describe the geographic distribution of strains

of E. amylovora across major apple production regions in NY

Literature Cited

Aldwinckle, H. S., Bhaskara Reddy, M. V., and Norelli, J. L. 2002. Evaluation of control of fire blight infections of apple blossoms and shoots with SAR inducers, biological agents, a growth regulator, copper compounds, and other materials. Acta Horticulturae. 590:325–331.

Biggs, A. R., and Turechek, W. W. 2010. Fire Blight of Apples and Pears: Epidemiological Concepts Comprising the Maryblyt Forecasting Program. Plant Health Progress. 11:23.

Brown, S. 2012. Apple. In Fruit Breeding, eds. Marisa Luisa Badenes and David H. Byrne. Boston, MA: Springer US, p. 329–367. Available at: https://doi.org/10.1007/978-1-4419-0763- 9_10.

CABI Centre for Agriculture and Bioscience International, Invasive Species Compendium. (2018). Erwinia amylovora (fireblight). [Datasheet 21908]. Retrieved from https://www.cabi.org/isc/datasheet/21908

Carroll, J., Weigle, T., Agnello, A., Reissig, H., Cox, K., Breth, D., et al. 2017. NEWA (Network for Environment and Weather Applications) Provides Fruit IPM and Production Tools from 400 Weather Stations. NY Fruit Quarterly. 25:19–24.

Chiou, C.-S. and Jones, A.L. 1995a. Molecular analysis of high-level streptomycin resistance in Erwina amylovora. Phytopathology. 85:324–328.

Chiou, C.-S., and Jones, A. L. 1995b. Expression and identification of the strA-strB gene pair from streptomycin-resistant Erwinia amylovora. Gene. 152:47–51.

Coyier, D. L. and Covey, R. P. 1975. Tolerance of Erwinia amylovora to streptomycin sulfate in Oregon and Washington. Plant Dis. Rep. 59:849-852.

Förster, H., McGhee, G. C., Sundin, G. W., and Adaskaveg, J. E. 2015. Characterization of Streptomycin Resistance in Isolates of Erwinia amylovora in California. Phytopathology. 105:1302–1310.

Granatstein, D., Johnson, K., Smith, T., and Elkins, R. Implementation of non-antibiotic programs for fire blight control in organic apple and pear in the western United States. Poster.

Harshman, J. M., Evans, K. M., Allen, H., Potts, R., Flamenco, J., Aldwinckle, H. S., et al. 2017. Fire Blight Resistance in Wild Accessions of Malus sieversii. Plant Disease. 101:1738–1745.

Johnson, K. B., and Temple, T. N. 2013. Evaluation of Strategies for Fire Blight Control in Organic Pome Fruit Without Antibiotics. Plant Disease. 97:402–409.

McGhee, G. C., Guasco, J., Bellomo, L. M., Blumer-Schuette, S. E., Shane, W. W., Irish-Brown, A., et al. 2011. Genetic Analysis of Streptomycin-Resistant (SmR) Strains of Erwinia amylovora Suggests that Dissemination of Two Genotypes Is Responsible for the Current Distribution of Sm R E. amylovora in Michigan. Phytopathology. 101:182–191.

McGhee, G. C., and Sundin, G. W. 2011. Evaluation of Kasugamycin for Fire Blight Management, Effect on Nontarget Bacteria, and Assessment of Kasugamycin Resistance Potential in Erwinia amylovora. Phytopathology. 101:192–204.

McManus, P. S., Stockwell, V. O., Sundin, G. W., and Jones, A. L. 2002. Antibiotic use in plant agriculture. Annual Review of Phytopathology. 40:443–465.

Miller, T.D. and Scroth, M.N. 1972. Monitoring the epiphytic population of Erwinia amylovora on pear with a selective medium. Phytopathology 62:1175-1182.

Norelli, J. L. 2003. Fire blight management in the twenty first century. Plant Disease. 87:756– 765.

NYAA New York Apple Association. 2012. Facts and Figures about New York State Apples. Retrieved from http://www.applesfromny.com/about/facts

Pal, K., and Gardener, B. 2011. Biological Control of Plant Pathogens. The Plant Health Instructor. DOI: 10.1094/PHI-A-2006-1117-02.

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 Disease. 92:714–718.

Russo, N. L., Robinson, T. L., and Fazio, G. 2007. Field Evaluation of 64 Apple Rootstocks for Orchard Performance and Fire Blight Resistance. 42:9.

Schroth, M. N., Thomson, S. V., Hildebrand, D. C., and Moller, W. J. 1974. Epidemiology and Control of Fire Blight. Annual Review of Phytopathology. 12:389–412.

Sundin, G. W., Werner, N. A., Yoder, K. S., and Aldwinckle, H. S. 2009. Field Evaluation of Biological Control of Fire Blight in the Eastern United States. Plant Disease. 93:386–394.

Tancos, K. A., and Cox, K. D. 2016. Exploring Diversity and Origins of Streptomycin-Resistant Erwinia amylovora Isolates in New York Through CRISPR Spacer Arrays. Plant Disease. 100:1307–1313.

Turechek, W. W. and Biggs, A. R. 2015. Maryblyt v. 7.1 for Windows: An Improved Fire Blight Forecasting Program for Apples and Pears. Plant Health Progress. 16:16–22.

USDA National Agriculture Statistics Service. 2012. Table 31. Fruits and Nuts: 2012 and 2007. 2012 Census of Agriculture – State Data. Retrieved from https://www.nass.usda.gov/Publications/AgCensus/2012/Full_Report/Volume_1,_Chapter_2_US _State_Level/st99_2_031_031.pdf

USDA National Agriculture Statistics Service. 2017. Commercial Apple Utilized Production United States: 2007-2016. Retrieved from https://www.nass.usda.gov/Charts_and_Maps/A_to_Z/in-apples.php

USDA Foreign Agricultural Service. 2018. Fresh Deciduous Fruit: World Markets and Trade (Apples, Grapes, and Pears). Retrieved from https://apps.fas.usda.gov/psdonline/circulars/fruit.pdf

Vallad, G. E., and Goodman, R. M. 2004. Systemic Acquired Resistance and Induced Systemic Resistance in Conventional Agriculture. Crop Science. 44:1920. van der Zwet, T. 2012. Fire Blight: History, Biology, and Management. St. Paul, MN: APS Press.

WAPA World Apple and Pear Association. 2014. Apple and Pear Production by Country and Year. Retrieved from http://www.wapa-association.org/asp/page_1.asp?doc_id=446

PART I:

IMPACTS OF ANTIBIOTICS AND ALTERNATIVES

ON DISEASE MANAGEMENT AND ORCHARD HEALTH

CHAPTER 2

MANAGEMENT OF FIRE BLIGHT USING PRE-BLOOM APPLICATION OF

PROHEXADIONE CALCIUM*

Abstract

Fire blight, a bacterial disease of rosaceous plants caused by Erwinia amylovora, is one of the most important diseases affecting commercial apple production worldwide. Antibiotics, applied at bloom to protect against blossom infection, are the most effective means of management, but raise concern due to the potential for antibiotic resistance in both the pathogen population and off-target organisms. In addition, most fire blight outbreaks in New York State often emerge in late June to July as shoot blight, calling into question the role of blossom infections and the antibiotic applications made to manage them. Prohexadione-calcium (PhCa) is a gibberellic acid inhibitor used post-bloom to control shoot vigor and to manage shoot blight.

However, the magnitude of shoot blight management is directly related to the suppression of shoot growth, which is undesirable, especially in young orchards during establishment years.

PhCa is believed to control shoot blight by thickening cell walls in cortical parenchyma, preventing invasion of host tissues by E. amylovora. We hypothesize that PhCa applied pre- bloom could similarly prevent invasion of blossom pedicels following infection leading to reduced disease incidence. We evaluated novel pre-bloom PhCa programs for their effects on disease management (blossom and shoot blight) as well as their horticultural impacts (shoot growth and yield) for three years in a mature ‘Gala’ orchard in NY. In all three years of the study, all PhCa programs resulted in less than 27% incidence in blossom blight and less than

13% incidence of shoot blight with minimal effect on tree growth and productivity. Inclusion of

a biopesticide during bloom further reduced the incidence of blossom blight. Using light microscopy, we found that cell walls in the cortical parenchyma of fruitlet pedicels on trees receiving pre-bloom PhCa applications were significantly thicker than those of untreated trees 40 days after full bloom and inoculation. Overall, we found that pre-bloom applications of PhCa had utility in reducing blossom blight and shoot blight with minimal impacts on tree growth and productivity. These pre-bloom programs would fit with standard production practices and may contribute toward the development of fire blight management programs without the use of antibiotics.

*Wallis, A. and Cox, K. 2020. Management of fire blight using pre-bloom application of prohexadione-calcium. Plant Disease. 104(4):1048-54.

Introduction

Fire blight, caused by the bacterial pathogen Erwinia amylovora, is one of the most destructive diseases of commercial apple production worldwide. Originally identified in the

Hudson Valley of New York in the early 1800s and formally described by Thomas Burrill in

1880, fire blight is still a major source of grower consternation and active area of research for nearly two centuries (van der Zwet, T. 2012). Fire blight has spread to nearly every apple production region worldwide, with the exception of parts of South America, Africa, and Asia

(van der Zwet, T. 2012), and is capable of destroying entire orchard blocks in a single season

(Norelli 2003; van der Zwet, T. 2012). Economic losses in the U.S. are estimated at over $100 million annually; most recently, fire blight outbreaks in the Champlain Valley of NY in 2016 and

Southwest Michigan in 2000 and 2005 impacted most commercial farms in the regions with total economic losses estimated at $9,500 per acre and $42 million across a region (Elizabeth Higgins personal communication; Norelli 2003). In recent years, fire blight has become particularly problematic due to the adoption of high-density orchards composed of smaller trees and popular cultivars, which are more susceptible to the disease and sustain the most damage from infection.

E. amylovora is a Gram-negative bacterium in the Enterobacteriaceae family and is classified as a hemibiotroph. The bacteria may persist in living host tissue at the margin of cankers from old infection sites. In the spring under warm, humid conditions, cankers produce a sticky bacterial ooze; bacteria rapidly colonize the plant surface and are transferred to the stigmatic surface of the flower by insects and rain splash. Primary infection occurs during bloom, when bacteria are washed into the flowers through nectaries in the hypanthium, invading the blossom and pedicel causing blight of the flowers in the flower cluster, leading to blossom blight

(van der Zwet, T. 2012). A secondary form of infection is shoot blight, in which shoots become

infected by bacterial ooze from cankers or blighted flowers entering wounds caused by wind, hail, or insects (van der Zwet, T. 2012).

For over 40 years, the most effective tool for the management of blossom blight has been the application of the aminoglycoside antibiotic streptomycin (Norelli 2003). More recently, kasugamycin, another aminoglycoside antibiotic, has also been registered in several states for fire blight management in orchards, in response to the emergence of streptomycin resistant E. amylovora (EPA registration number 66330-404; McGhee and Sundin 2011). Antibiotics are applied to open blossoms with guidance on application timing provided by one of several weather-based models (i.e. MaryBlyt, CougarBlight), which were developed to predict risk of infection events (Biggs and Turechek 2010; Carroll et al. 2017; Smith 1999; Turechek and Biggs

2015). Streptomycin routinely provides >90% reduction disease incidence in trials in the

Northeastern US (Aldwinckle et al. 2002; Norelli 2003). The use of antibiotics in agriculture has come under public scrutiny due to the potential for selection of streptomycin resistance in both pathogen populations and off-target bacteria (McManus et al. 2002). Streptomycin-resistant

(SmR) E. amylovora was first reported in 1971 in California (Miller and Scroth 1972) and has since been identified in most major production regions in the US, including Washington, Oregon,

Michigan, and New York (Loper et al. 1991; McManus and Jones 1994; McGhee et al. 2011;

Russo et al. 2008; Tancos and Cox 2016;). Additionally, antibiotic use has been prohibited in organic production in the United States; this is consistent with organic regulations in other locations including, Canada and European Union (Johnson and Temple 2013). In this regard, alternative management strategies are necessary for the sustainable management of fire blight.

Additionally, antibiotics may not be the most appropriate management tool for all phases of fire blight. Typically, the focus of commercial fire blight management programs is the

blossom blight phase, under the assumption that infected blossoms provide inoculum for secondary shoot blight infections observed during the summer. However, in NY, fire blight outbreaks are often reported in late June to July as shoot blight—rather than blossom blight which would occur earlier in the spring—as evidenced by recent statewide fire blight surveys in

NY (Tancos and Cox 2016). It is generally believed that shoot blight is more frequently reported in commercial orchards because symptoms are more easily identified, especially during periods of vigorous shoot growth. However, blossom blight is difficult to manage: it requires extremely precise management as the opportune application period occurs during a host phenological stage that may last only a few days, symptoms are less obvious and easily overlooked, and very low incidences of infected blossoms are required to produce inoculum for secondary infection. In this regard, should growers focus on managing both blossom and shoot blight earlier, prior to bloom and should materials other than antibiotics be investigated?

A plant growth regulator like prohexadione-calcium (PhCa) could be used to manage both blossom and shoot blight prior to infection. PhCa is a gibberellic acid inhibitor, originally registered for apple production in the United States in 1999 with the primary commercial use of canopy management and reduction of shoot vigor (Evans et al. 1999). Current recommendations for vegetative growth control are to apply 125 to 250mg/L for mature trees or 62.5 to 125mg/L for trees less than five years, at petal fall and again 14 to 21 days later, timing that corresponds with early season shoot growth (Evans et al. 1999). Early in development, it was discovered that

PhCa also was effective for managing fire blight (Breth et al. 1998; Jones et al. 1999; Yoder et al. 1999). Extensive investigations led to the discovery that PhCa applications were able to reduce shoot blight incidence from 40 to 90% when applied in a field setting with either natural or artificial bacterial inoculation; investigations usually evaluated one or two applications of 125

and 250mg/L, with a single high-rate application typically providing the best disease control

(Aldwinckle et al. 2000; Bazzi et al. 2003; Costa et al. 2001; Norelli and Miller 2004, 2006).

Application timing is critical for effective fire blight management, and coincides with application timing for vegetative growth control (Evans et al. 1999; Rademacher et al. 2006). Most studies have found that PhCa must be applied at least seven to ten days prior to infection or inoculation in order to reduce incidence of shoot blight; therefore, current recommendations are to make one or two applications of 125 or 250mg/L beginning at petal fall or when five to eight centimeters of shoot growth are observed (Aldwinckle et al. 2000; Costa et al. 2001; Schupp et al. 2002;

Yoder et al. 1999). Unsurprisingly, the magnitude of shoot blight management was directly related to the suppression of shoot growth, a problem for young orchards in which vigorous growth is encouraged as trees are pushed to fill canopy space. Therefore, prior research has concluded that PhCa’s negative impacts on tree growth prohibit its use for shoot blight control in young orchards, and it is primarily reserved for shoot blight control in mature orchards under high disease pressure (Aldwinckle et al. 2002; Costa et al. 2004; Norelli and Miller 2004, 2006).

Originally, the mechanism of action for shoot blight control was believed to be via the production of novel antibiotic compounds within the host, the result of an altered flavonoid biosynthetic pathway (Halbwirth et al. 2003; Rademacher et al. 2006; Roemmelt et al. 2003;

Spinelli et al. 2005). However, novel compounds did not show an appreciable effect on bacterial growth (Roemmelt et al. 2003) and transgenic apples with an altered flavonone pathway mimicking that of PhCa-treated plants did not produce adequate metabolic changes to reduce fire blight susceptibility (Flachowsky et al. 2012). Other proposed mechanisms of action included physiological resistance as well as morphological and histological changes (Bazzi et al. 2003).

More recent work has found reduced pathogen migration within the leaf xylem and parenchyma

tissue in vitro (Spinelli et al. 2006) and evidence for physical inhibition of bacteria within the host through the thickening of cell walls in the cortical parenchyma tissue to the point where E. amylovora’s Type III secretion system-associated pilus is prevented from delivering effectors

(McGrath et al. 2009).

Investigation of PhCa applied earlier in the season (pre-bloom) for both shoot blight and blossom blight has been limited due to label restrictions, the need for pre-infection applications, and the limited leaf surface area available before bloom (Evans et al. 1999; Rademacher and

Kober 2003). However, a few reports demonstrate distinct potential for PhCa-mediated blossom blight protection and early shoot blight protection, if applied pre-bloom (Buban et al. 2006;

Spinelli et al. 2006) or post-harvest leading to carry-over effects in the following season (Greene

2005). In addition, there is growing interest in the industry and extension to evaluate PhCa for control of blossom blight. Novel uses, including pre-bloom applications, could prevent invasion of the blossom tissues through the mechanism of cell wall thickening. Applying PhCa at lower rates for blossom blight control may allow for protection without inhibiting canopy growth by the end of the season. The paucity of academic research on the use of PhCa for blossom blight and early shoot blight management has led to increased reliance on extension publications, technical reports, and anecdotal reports from the industry to highlight research needs and drive hypotheses for academic research.

The objective of this work was to evaluate pre-bloom PhCa programs for fire blight management. Specifically, we evaluated PhCa applied at ‘pink,’ the phenological stage of apple preceding bloom, where the flowers are beginning to open and show a magenta coloration

(Meier 2001), for impacts on fire blight (blossom and shoot blight) as well as horticultural impacts on shoot growth and yield. We hypothesized that these PhCa programs would reduce

disease incidence of blossoms and shoots through the mechanism of cell wall thickening, without negatively impacting tree performance or productivity. Such management programs would fit with standard production practices for apple and may be a step towards the development of fire blight management programs without the use of antibiotics.

Materials and Methods

Evaluation of pre-bloom programs of prohexadione calcium for blossom blight and shoot blight management

To evaluate the effects of novel PhCa programs for fire blight management, field trials were conducted at Cornell AgriTech in a research orchard in Geneva, NY (42°87’70.19”N -

77°02’99.15”W) over three field seasons: 2016, 2017, and 2018. Experiments were performed in a planting of Gala on B.9 rootstock, planted in 2000 at 1.8 meter in-row spacing. A randomized complete block design with four single-tree replications was used to evaluate the management programs (Table 2.1). All applications were made using a Solo 475-B gas-powered mist blower

(Solo Incorporated, Newport News, VA) calibrated to deliver approximately 935.3L/ha, a standard volume for high-density apple plantings in the northeastern United States. To evaluate pre-bloom programs of PhCa for managing blossom blight, applications began at pink. In these programs Apogee (BASF Corporation, Research Triangle Park, NC) was applied at either 62.5 or

125mg/L, reflecting medium and high label rates. PhCa was either applied on its own at pink, or it was followed with a bloom application of the biopesticide Double Nickel LC (Certis,

Columbia, MD) was applied at 2.34 kg/ha to protect blossoms during this critical infection period. In these programs. In addition to the PhCa programs, two control programs (positive and negative) were included: i) untreated control plots were included to serve as a negative control

and indicator of overall disease pressure and ii) antibiotic programs were included as a positive control and industry standard, in which the antibiotics streptomycin and kasugamycin were applied at 40% bloom, 80% bloom, and petal fall. In these programs, FireWall 17WP

(AgroSource, Mountainside, NJ) and Kasumin 2L (Arysta LifeScience, Cary, NC) were applied at the label rates of 1.68 kg/ha and 5L/ha respectively, with the non-ionic surfactant Regulaid

(KALO, Overland Park, KS) applied at 3.18 L/ha.

Table 2.1. Summary of pre- and post-bloom treatments evaluated for fire blight protection and horticultural impacts in 2016, 2017 and 2018.

Treatment and Rate Time Active Ingredient Control Programs Untreated na na Bloom FireWall 17 1.68 kg/ha Bloom Streptomycin Bloom Kasumin 4.7L/ha Bloom Kasugamycin Pre-Bloom PhCa Programs Pink Apogee 62.5 mg/L Pink Prohexadione calcium Pink Apogee 62.5 mg/L Pink Prohexadione calcium + Bloom Double Nickel LC 2.34 kg/ha + Bloom + Bacillus amyloliquefaciens Pink Apogee 125 mg/L Pink Prohexadione calcium Pink Apogee 125 mg/L Pink Prohexadione calcium + Bloom Double Nickel LC 2.34 kg/ha + Bloom + Bacillus amyloliquefaciens

To ensure adequate disease pressure, trees were inoculated with E. amylovora strain

Ea273, a highly aggressive strain originally isolated from trees in New York State, routinely used in lab and field trials (Lee et al. 2010). Cultures were stored in 50% glycerol at -80°C prior to use, were streaked on Luria-Bertani (LB) agar and incubated at 28°C for 3 days. Cultures were then re-suspended in LB broth overnight while shaken at 28°C, and suspensions were diluted to

1x106 CFU ml-1 in phosphate buffered saline immediately prior to inoculation. Inoculum was applied using a hand-pumped Solo 475-B backpack sprayer (Solo Incorporated) at 80 to 90%

bloom. Inoculation was performed on May 11, May 10, and May 18 in 2016, 2017, and 2018, respectively. Weather conditions at the time of inoculation included average and high temperatures of 8.7°C and 11.9°C, 6.4°C and 11.5°C, and 10.1°C and 15.5°C in 2016, 2017, and

2018, respectively, with wind speed less than 9 km h-1 in all years. In all three years a rain event occurred within the three days following inoculation, in which 0.84, 0, and 1.5 cm of accumulation was recorded in respective years; leaf wetness ranged from 1 to 19 hours; and 10 to

32 hours of relative humidity >90% were recorded. Blighted clusters served as inoculum for shoot blight infections. To assess efficacy of pre-bloom programs for fire blight management, symptoms were assessed on blossom clusters and terminal shoots in June. The incidence of blossom or shoot blight was expressed as the number of blighted blossom clusters or terminal shoots out of 20 assessments for four replicate trees per treatment. Disease strikes were pruned out mid-summer following disease ratings to preserve trees.

Investigating PhCa-induced cell wall thickening as the mechanism of action providing blossom blight control

To investigate the mechanism of action of PhCa in reduction of blossom blight incidence in pre-bloom programs, we evaluated cell wall thicknesses in pedicels of developing flowers and fruitlets in 2018 and 2019. In each year, PhCa was applied at 0, 62.5, and 125 mg/L at pink as previously described. Measurements were then taken at two time points: i) at petal fall, approximately 14 to 20 days after PhCa treatments when treatment effects should have first been visible, and ii) approximately 40 days after full bloom (DAFB) or 50 days after PhCa treatments.

At each time point, king blossoms or fruitlets were randomly selected for each treatment replication, transported in plastic Ziploc bags to the lab (within 1 hour), held at 40°C, and

processed within 24 hours of collection. The pedicel was separated from the flower or fruitlet and cross-sectioned by hand at the mid-point. Thin cross sections (approximately one cell thick) were made under a dissection microscope by holding the pedicel steady between two microscope slides and making slices perpendicular to the stem with a double-edged razor blade (Personna,

Staunton, VA). Wet mounts were constructed, and prepared slides were imaged using a light microscope (Model BX50, Olympus, Webster, NY). Digital images were taken at 400x magnification using SPOT Imaging Software Version 5.3 (Figure 2.1; SPOT Imaging,

Diagnostic Instruments, Sterling Heights, MI). Fifty total cell wall measurements were made within the cortical parenchyma tissue directly below the epidermal tissue for each treatment replicate.

Figure 2.1. Hand cross-sections of apple fruitlet pedicels at 50 days post-treatment (40 days after full bloom) for trees treated with 0 g/L (A) and 125 g/L (B) prohexadione-calcium. Fifty (50) cell wall measurements in the cortical parenchyma tissue just below the epidermis were taken for each replication, as indicated by the arrow. Cell wall thickness in treated trees was significantly greater than cell wall thickness of untreated trees. Coloration reflects natural pigmentation of the pedicels at the time of sampling. Scale bars represent 50 μm.

Evaluation of pre-bloom programs of prohexadione calcium for horticultural impacts

To assess effects of pre-bloom PhCa programs on tree growth and productivity, shoot growth, fruit size, fruit number, and yield were evaluated each year. Shoot length and trunk circumference measurements were made at two points during the season: once in July at approximately terminal bud set, the point at which trees stop putting on significant vegetative growth, and once in October, following harvest when growth is concluded for the season. To evaluate shoot length, current season’s growth (bud scale scar to attachment point of terminal leaf) was measured for 20 randomly selected shoots in each treatment replication. Trunk circumference was measured for each tree at approximately 30 cm above the graft union and used to calculate trunk cross-sectional area (TCSA). Fruit measurements were taken once, at terminal bud set in July. For fruit size, fruit diameter was measured at the widest part of the fruit using digital calipers (General Tools & Instruments, New York, NY) for 20 selected fruit within each replicate. Potential yield (i.e. tree productivity) was represented as the number of fruit per cm2 TCSA, as specified by Stover, Wirth, and Robinson (2001).

Data Analysis

Disease incidence and host response data were subjected to analysis of variance

(ANOVA) for a randomized block design using Generalized Linear Mixed Models (GLIMMIX) procedure of SAS (version 9.4; SAS Institute Inc., Cary, NC). All percentage data was subjected to arcsine square root transformation prior to analysis. Cell wall thickness data were also evaluated using ANOVA for randomized block design using the GLIMMIX procedure of SAS

(version 9.4; SAS Institute Inc., Cary, NC). For models with significant fixed effects, differences

between treatments were determined using the LSMEANS procedure in SAS 9.4 at the α = 0.05 level of significance with an adjustment for Tukey’s HSD to control for family-wise error.

Results

Evaluation of pre-bloom programs of prohexadione calcium for blossom and shoot blight management

Blossom blight was observed in each year, following blossom inoculation, with highest incidence occurring in 2017. Blossom blight incidence for untreated control trees was 65±4%,

94±2%, and 55±3% in 2016, 2017, and 2018 respectively (Table 2.2). Shoot blight also developed following inoculation of blossoms in the each of the three years. The incidence of shoot blight in the untreated controls was 24±4%, 55±3%, and 28±3% in 2016, 2017, and 2018 respectively (Table 2.2). In all three years, there was a significant effect of treatment on the incidence of blossom blight (P < 0.0001) and shoot blight (P < 0.0001).

Both antibiotic programs provided excellent control with less than 6% blossom blight incidence, with the exception of kasugamycin in 2018 (Table 2.2). Blossom blight incidence for the streptomycin treatment was 3.9±1.8%, 0.9±0.9%, and 1.6±0.4% in 2016, 2017, and 2018 respectively. In kasugamycin programs blossom blight incidence was 2.0±0.4%, 3.3±3.1%, and

12.1±6.3% in the three years. Shoot blight incidence for streptomycin treatment was 0.0±0.0%,

2.4±1.9%, and 0.6±0.6% in 2016, 2017, and 2018 respectively. For kasugamycin programs shoot, blight incidence was 0.7±0.4%, 0.0±0.0%, and 3.2±1.9% in the three years.

There were no significant differences in the incidence of fire blight between any of the pre-bloom PhCa programs in any of the three years, with the incidence of blossom blight and shoot blight less than 27% and 13% respectively (P > 0.05, Table 2.2). Although not significant,

the program in which 125 mg/L PhCa was applied at pink and the biopesticide Double Nickel

LC (Certis) was applied at bloom tended to have lower incidences of blossom blight and shoot blight. All pre-bloom PhCa treatments had a lower incidence of blossom blight than the untreated controls. In 2016, the incidence of blossom blight in pre-bloom PhCa programs was higher than that for the two antibiotic controls, except for the program in which the 125mg/L rate of PhCa was applied at pink followed by the biopesticide Double Nickel LC (Certis) applied at bloom. By contrast, in 2017 and 2018, there were no significant differences in the incidence of blossom blight between pre-bloom PhCa programs and the antibiotic programs. Shoot blight in pre-bloom PhCa programs in 2017 and 2018 generally was higher than in the streptomycin program, but not in the kasugamycin program.

25 Table 2.2. Mean disease incidence and horticultural parameters of Gala trees treated with pre-bloom fire blight programs utilizing prohexadione calcium (Apogee) at the Cornell AgriTech research orchards in Geneva, NY in 2016

Blossom Blight Shoot Blight Shoot Length Shoot Length Fruit Fruit Size Yield Treatment (%)y (%)x July (mm)w Sept (mm)w Numberv (mm)u (Fruit/TCSA)t 2018 Untreated 65.00 az 23.90 a 308.51 a 330.39 ab 2.23 d 27.37 bcd 4.64 c Bloom FireWall 17 1.68 kg/ha 3.90 de 0.00 d 221.70 dc 316.80 abc 3.56 d 27.40 abcd 5.71 bc Bloom Kasumin 4.7L/ha 2.00 e 0.70 cd 277.00 abc 298.10 bc 3.23 bcd 28.15 ab 5.01 bc Pink Apogee 62.5mg/L & Untreated 25.60 b 6.30 bcd 265.00 abc 255.50 cd 2.30 d 26.40 cd 3.83 c Pink Apogee 62.5mg/L + Bloom Double Nickel LC 2.34 kg/ha 22.80 b 5.30 bc 258.30 abc 256.70 cd 4.60 ab 27.30 abcd 9.82 a Pink Apogee 125mg/L & Untreated 15.30 bc 3.90 bcd 260.60 abc 255.40 bc 3.72 bcd 27.79 abc 6.08 bc Pink Apogee 125mg/L + Bloom Double Nickel LC 2.34 kg/ha 12.50 bcd 5.40 bc 229.80 dc 179.60 ef 4.35 abc 25.85 d 8.45 ab P-Values <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 <.0001 2017 Untreated 94.38 a 55.04 a 207.93 d 239.88 d 0.89 c 16.74 a 1.53 d Bloom FireWall 17 1.68 kg/ha 0.94 e 2.37 c 292.08 bcd 293.25 ab 2.80 bc 19.28 a 5.68 abcd Bloom Kasumin 4.7L/ha 3.33 de 0.00 bc 308.32 ab 339.09 bcd 2.79 bc 20.91 a 4.71 abcd Pink Apogee 62.5mg/L & Untreated 26.92 bcde 11.51 b 278.92 abcd 321.42 abcd 4.73 a 18.58 a 7.47 a Pink Apogee 62.5mg/L + Bloom Double Nickel LC 2.34 kg/ha 10.68 cde 4.25 bc 289.50 abc 292.08 abcd 3.13 abc 15.87 a 5.12 abcd Pink Apogee 125mg/L & Untreated 13.21 bcde 12.85 b 259.61 abcd 277.72 bcd 3.49 ab 15.75 a 6.09 abc Pink Apogee 125mg/L + Bloom Double Nickel LC 2.34 kg/ha 6.44 cde 6.35 bc 277.30 abcd 281.13 bcd 2.45 bc 17.81 a 3.55 cd P-Values <.0001 <.0001 <.0001 <.0001 <.0001 0.5934 <.0001 2018 Untreated 55.15 a 27.77 a 205.97 d 205.97 c 3.37 a 25.53 a 5.42 a Bloom FireWall 17 1.68 kg/ha 1.60 b 0.60 b 311.79 a 318.00 a 4.10 a 25.75 a 6.15 a Bloom Kasumin 4.7L/ha 12.08 b 3.15 b 311.91 a 306.22 a 4.05 a 26.22 a 6.46 a Pink Apogee 62.5mg/L & Untreated 13.54 b 2.02 b 282.62 ab 282.62 a 3.97 a 26.29 a 6.10 a Pink Apogee 62.5mg/L + Bloom Double Nickel LC 2.34 kg/ha 14.38 b 4.71 b 268.56 abc 268.56 ab 3.77 a 26.32 a 3.72 a Pink Apogee 125mg/L & Untreated 15.00 b 5.22 b 227.97 cd 227.97 bc 3.43 a 26.02 a 5.22 a Pink Apogee 125mg/L + Bloom Double Nickel LC 2.34 kg/ha 8.75 b 3.21 b 243.27 bcd 243.27 abc 4.10 a 26.13 a 8.00 a P-Values <.0001 <.0001 <.0001 <.0001 0.9625 0.3234 0.2194

zValues within columns for a given year followed by the same letter are not significantly different (P < 0.05) according to the LSMEANS procedure in SAS 9.4 with an adjustment for Tukey’s HSD to control for family-wise error. yPercent incidence of blossom blight for 20 blossom clusters rated xPercent incidence of shoot blight for 20 terminal shoots rated wMean length of current season’s growth for 20 shoots vMean number of fruit per limb for 8 limbs uMean diameter of fruitlets measured at the largest point on the fruit for 20 fruit. Measurements taken mid-July. tTotal yield (extrapolated by multiplying mean fruit/shoot * 150 shoots) / TCSA (Trunk Cross

Sectional Area)

Evidence for PhCa-induced cell wall thickening as the mechanism of action providing blossom blight control

There was an effect of treatment program on cell wall thickness in the pedicel at the majority of sampling times (Figure 2.2). In 2018, trees treated with PhCa at pink did not have thicker cell walls than the untreated control trees at petal fall (e.g. shortly after the treatment, P =

0.0515), but trees treated with both 62.5 and 125 mg/L PhCa at 40 DAFB had thicker cell walls than untreated trees (P = 0.0006). In 2019, trees receiving both 62.5 and 125 mg/L PhCa at pink had thicker cell walls than untreated trees at petal fall (P < 0.0001) and 40 DAFB (P < 0.0001).

Figure 2.2. Mean cell wall thicknesses (±1 SE) of fruitlet pedicels taken in 2018 (A, B) and 2019 (C, D) for Gala trees treated with prohexadione-calcium (trade name Apogee) at 20 days (A, C) and 50 days (B, D) after treatment at the Cornell AgriTech Research Station in Geneva, NY. Within each graph, different letters above bars indicate significant differences between means based on Tukey HSD test (p<0.05).

Evaluation of pre-bloom programs of prohexadione calcium for horticultural impacts

Impact on shoot growth. A minimal impact was observed on shoot length for all programs across the three years, (P < 0.0001 for all three years) with reductions in growth being approximately 10 to 15 cm for both July and September measurements. In untreated control trees, growth tended to be lower because trees experience higher levels of shoot blight, therefore, the antibiotic treatments were used as controls for shoot blight comparisons. Pre-bloom PhCa programs rarely differed in shoot length from the antibiotic programs in either July or

September, with only two exceptions. In 2016, 125 mg/L PhCa followed by the biopesticide at bloom had shorter shoot growth than both antibiotic programs. In 2018, 125 mg/L PhCa without a bloom biopesticide application had shorter shoot growth than both antibiotics (P < 0.0001).

Impact on fruit number, fruit size, and yield. The number of fruit in 2016, 2017, and 2018

(P < 0.0001, P < 0.0001, and P = 0.9625) and size of fruit in 2016, 2017, and 2018 (P < 0.0001,

P = 0.5934, and P = 0.3243) was rarely affected by treatments and the few exceptions were inconsistent between years. In 2016, fruit number for 125mg/L PhCa followed by the biopesticide at bloom was higher than the untreated control trees and the streptomycin program.

However, in 2017, both 62.5 and 125 mg/L PhCa without the biopesticide at bloom had greater fruit number than the untreated control. In 2018 there were no differences in fruit number between any treatments. Yield (fruit cm2 TCSA-1) was minimally and inconsistently affected by treatments in 2016 (P < 0.0001), 2017 (P < 0.0001), or 2018 (P = 0.2200). In 2016, both 62.5 and

125 mg/L PhCa with the biopesticide at bloom had higher yield than the untreated control (P <

0.05); but in 2017 PhCa programs without the biopesticide had a higher yield than the untreated control (P < 0.0001). No differences were observed in yield between treatments in 2018 (P =

0.2194).

Discussion

Management of fire blight with antibiotics has become a tenuous topic and presently there is demand for alternative management programs. This is especially true for temperate production regions, where antibiotic alternative management programs developed for arid apple production regions (Johnson and Temple 2013) have been unable to reliably provide acceptable levels of control (Aldwinckle et al. 2002; Sundin et al. 2009). PhCa is currently recommended for shoot blight control in mature orchards after petal fall during the period of shoot elongation

(Aldwinckle et al. 2000, 2002; Norelli and Miller 2006, 2004; Yoder et al. 1999). There have also been cursory studies investigating potential for its use earlier in the season and at lower rates for control of blossom blight and pre-emptive management of shoot blight (Buban et al. 2006;

Greene 2005; Spinelli et al. 2006). In this research we tested novel management programs utilizing low rates of PhCa applied at earlier timings, to target blossom and subsequent shoot blight over three growing seasons in a mature planting of ‘Gala’ trees. These early PhCa programs resulted in less than 27% incidence of blossom blight and less than 13% incidence of shoot blight across all three years. Programs were similarly effective in seasons where weather conditions greatly favored fire blight development. While these programs did not provide a level of control equivalent with antibiotic programs, which typically had less than 5% incidence of blossom blight, they may have utility as part of a comprehensive antibiotic alternative management program as they do not mandate the same level of precise timing as bloom applications for antibiotics. It is also important to note that the PhCa programs were tested under experimental conditions with high levels inoculum to ensure infection, which are at levels much greater than would occur in the vast majority of commercial orchards. It is likely that the efficacy of these programs would be greater under the lower inoculum pressures of commercial orchards.

Although not always statistically significant, pre-bloom programs of PhCa made at the rate of 125mg/L and programs with the biopesticide Double Nickel LC (Certis) applied at bloom tended to have better control of blossom blight. This product contains the Bacillus amyloliquefaciens strain D747, which acts by producing antibiotic metabolites. As the most critical stage for infection by E. amylovora is during bloom, it would be an unnecessary risk for a commercial operation to abstain from control measures entirely during this period. Therefore, treatments were tested to include a biopesticide at bloom containing microbial metabolites with antibiotic activity, to inhibit the pathogen’s development on the floral surface as described in other work (Johnson and Temple 2013). Coupling the biopesticide with the PhCa mode of action appears to provide good control, while protecting the plant at the most crucial infection stages without reliance on antibiotics. Further study is warranted to explore the management implications of biopesticide use in fire blight control.

In all years and across all treatments, we observed much higher incidences of blossom blight, as compared to shoot blight. This is consistent with the established dogma that blossom infections are the more prevalent form of infection and that it is prudent to focus most intensive management programs around the blossom infection stage (Sundin et al. 2009; van der Zwet, T.

2012). Applying PhCa pre-bloom appeared to have utility in both blossom and shoot blight management, and therefore offers a desirable solution for commercial management in which elimination or minimization of antibiotics is desired. In contrast, multiple, and precisely-timed applications of antibiotics are time-consuming and expensive in terms of labor and wear on equipment and should be used judiciously as part of an integrated management plan.

The mechanism of action of PhCa has been described as a thickening of the cell walls in the cortical parenchyma of developing shoots, which provides a physical barrier for movement of

the pathogen between host cells (McGrath et al. 2009). We found evidence for this cell wall thickening in cross sections of fruitlet pedicels taken both at petal fall (14 to 20 days after treatment) and 40 DAFB (50 days after pre-bloom PhCa treatments). Unsurprisingly, a higher application rate consistently resulted in greater effect on cell wall thickness. A statistical difference was detected for trees treated with 125 mg/L PhCa in both years, while a statistical difference for 62.5 mg/L was only detected in 2019. In both years, trees treated with 125 mg/L

PhCa tended to have thicker cell walls than 62.5 mg/L.

The overall management implications of these findings may be affected by regional and seasonal conditions. In places in which the season is colder and growth slower (such as Geneva,

NY), the rate of cell wall thickening might be slower than in warmer regions in which trees grow more quickly (i.e. the Mid-Atlantic). Effects of warm weather on the speed of cell thickening could also be balanced by a greater aggressiveness of the pathogen in warmer climates. In regions with colder, slower-growing environments than the Northeastern United States, commercial applications of PhCa may need to start earlier than the pink phenological stage to achieve proper pedicel thickening prior to infection at bloom. Similar effects may be seen in orchard blocks with different nutrient management programs: trees receiving greater rates of nitrogen, for example, would be expected to experience greater shoot growth and therefore may be more impacted by PhCa programs. This would have implications for both disease incidence and horticultural considerations.

Given that the level of fire blight control is correlated with the suppression of growth, the use of PhCa has been limited to mature orchards, and excluded from young plantings being pushed to fill canopy space with vigorous canopy growth (Norelli and Miller 2006, 2004). Use of pre-bloom applications of PhCa in these experiments, rather than timing PhCa applications when

shoot growth averaged 5 to 8 cm of growth (i.e. post-bloom) as in the majority of previous work

(Norelli and Miller 2006, 2004), it appears that this practice is able to take advantage of disease suppression activity during critical infection windows at bloom, but allowing shoot growth to resume afterwards until terminal bud set, reducing the overall impact on tree growth. Certainly, in the present work, overall impact on shoot growth was minimal for pre-bloom PhCa programs.

In all three years, there was rarely a difference in shoot length and any differences observed were on the order of only a few centimeters. This magnitude of growth is arguably not horticulturally significant and may not have an impact on orchard establishment.

Aside from shoot growth, there is also a concern that using PhCa to manage fire blight may impact fruit set, size, and tree productivity. PhCa use can lead to the retention of more fruitlets, influencing the immediate seasons crop, return bloom, and fruit size and quality (Costa et al. 2004; Greene 1999, 2007, 2008). In this work, we observed only minimal effects on fruit number, fruit size, and yield for PhCa treatments. Any differences that were observed between the various pre-bloom treatments were inconsistent and sometimes contradictory between years, and typically were unrelated to a specific PhCa program. For example, fruit number and yield were higher for PhCa programs which included a biopesticide at bloom (Double Nickel LC,

Certis) in 2016, but were higher for PhCa programs not including the biopesticide at bloom in

2017 and 2018. We are aware of no reason or evidence that B. amyloliquefaciens D747 would influence fruit production or have activity as a phytohormone. Therefore, we believe that the observed differences were coincidental, and the minimal differences in fruit and yield observed here are reassuring that these programs may be used without negative consequences on yield response.

The results found here demonstrate potential for pre-bloom PhCa management programs for managing fire blight in commercial orchards. These programs provided control of fire blight with minimal impact to horticultural parameters. However, these programs must be evaluated in a variety of production settings (regions, cultivars, rootstocks, production systems, age of planting, etc.) in order to determine their efficacy in different climates and production systems.

Further, these programs should be tested in commercial orchards, both to evaluate effects in these systems and facilitate adoption by commercial growers.

Literature Cited

Aldwinckle, H., Robinson, T., Norelli, J., Momol, T., and Reddy, B. 2000. Controlling Shoot Blight with ‘Apogee.’ NY Fruit Quarterly. 8:4.

Bazzi, C., Messina, C., Tortoreto, L., Stefani, E., Bini, F., Brunelli, A., et al. 2003. Control of Pathogen Incidence in Pome Fruits and Other Horticultural Crop Plants with Prohexadione-Ca. European Journal of Horticultural Science. 68:108–114.

Biggs, A. R., and Turechek, W. W. 2010. Fire Blight of Apples and Pears: Epidemiological Concepts Comprising the Maryblyt Forecasting Program. Plant Health Progress. 11:23.

Breth, D. I., Momol, M. T., and Aldwinckle, H. S. 1998. Evaluation of BAS 125 for the control of fire blight in apple shoots. Fungicide and Nematicide Tests. 53:4.

Buban, T., Beszeda, E., Dorgai, L., Földes, L., Hudak, I., Dobranszki, J., et al. 2006. Erwinia amylovora infection of flowers and shoots in apple trees treated with prohexadione-ca. Acta Horticulturae. 704:271–276.

Costa, G., Andreotti, C., Bucchi, F., Sabatini, E., Bazzi, C., Malaguti, S., et al. 2001. Prohexadione-Ca (Apogee®): Growth Regulation and Reduced Fire Blight Incidence in Pear. HortScience. 36:931–933.

Costa, G., Sabatini, E., Spinelli, F., Andreotti, C., Bomben, C., and Vizzotto, G. 2004. Two years of application of prohexadione-ca on apple: Effect on vegetative and cropping performance, fruit quality, return bloom, and residual effect. Acta Horticulturae. 653:35–40.

Evans, J. R., Evans, R. R., Regusci, C. L., and Rademacher, W. 1999. Mode of action, metabolism, and uptake of BAS 125W, prohexadione-calcium. HortScience. 34:1200–1201.

Flachowsky, H., Halbwirth, H., Treutter, D., Richter, K., Hanke, M.-V., Szankowski, I., et al. 2012. Silencing of flavanone-3-hydroxylase in apple (Malus × domestica Borkh.) leads to accumulation of flavanones, but not to reduced fire blight susceptibility. Plant Physiology and Biochemistry. 51:18–25.

Greene, D. W. 1999. Tree Growth Management and Fruit Quality of Apple Trees Treated with Prohexadione-calcium (BAS 125). HortScience. 34:1209–1212.

Greene, D. W. 2005. Carryover Effects of Prohexadione-calcium on Apples. HortScience. 40:1340–1342.

Greene, D. W. 2007. The effect of prohexadione-calcium on fruit set and chemical thinning of apple trees. HortScience. 42:1361–1365.

Greene, D. W. 2008. The Effect of Repeat Annual Applications of Prohexadione–calcium on Fruit Set, Return Bloom, and Fruit Size of Apples. HortScience. 43:376–379.

Halbwirth, H., Fischer, T. C., Roemmelt, S., Spinelli, F., Schlangen, K., Peterek, S. Sabatini, E., Messina, C., Speakman, J., Andreotti, C., Rademacher, W., Bazzi, C., Costa, G., Treutter, D., Forkmann, G., and Stich, K. 2003. Induction of Antimicrobial 3-Deoxyflavonoids in Pome Fruit Trees Controls Fire Blight. Zeitschrift für Naturforschung 58c:765-770.

Johnson, K. B., and Temple, T. N. 2013. Evaluation of Strategies for Fire Blight Control in Organic Pome Fruit Without Antibiotics. Plant Disease. 97:402–409.

Jones, A. L., Fernanado, W. G. D., and Ehret, G. R. 1999. Controlling secondary spread of fire blight with prohexadione calcium. Phytopathology 89: S37.

Lee, S. A., Ngugi, H. K., Halbrendt, N. O., O’Keefe, G., Lehman, B., Travis, J. W., et al. 2010. Virulence Characteristics Accounting for Fire Blight Disease Severity in Apple Trees and Seedlings. Phytopathology. 100:539–550.

Loper, J., Henkels, M., Roberts, R., Grove, G., Willet, M., and Smith, T. 1991. Evaluation of Streptomycin, Oxytetracycline, and Copper Resistance of Erwinia amylovora Isolated from Pear Orchards in Washington State. Plant Disease. 75:287–290.

McGhee, G. C., Guasco, J., Bellomo, L. M., Blumer-Schuette, S. E., Shane, W. W., Irish-Brown, A., et al. 2011. Genetic Analysis of Streptomycin-Resistant (SmR) Strains of Erwinia amylovora Suggests that Dissemination of Two Genotypes Is Responsible for the Current Distribution of SmR E. amylovora in Michigan. Phytopathology. 101:182–191.

McGrath, M. J., Koczan, J. M., Kennelly, M. M., and Sundin, G. W. 2009. Evidence that prohexadione-calcium induces structural resistance to fire blight infection. Phytopathology. 99:591–596.

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, V. O., Sundin, G. W., and Jones, A. L. 2002. Antibiotic Use in Agriculture. Annual Review of Phytopathology. 40:443–465.

Meier, U., ed. 2001. Growth stages of mono-and dicotyledonous plants. BBCH Monograph. 2nd ed. Berlin, Wien: Blackwell.

Miller, T. D., and Scroth, M. N. 1972. Monitoring the Epiphytic Population of Erwinia amylovora on Pear with Selective Medium. Phytopathology. 62:1175-1182.

Norelli, J. L. 2003. Fire blight management in the twenty first century. Plant Disease. 87:756– 765.

Norelli, J. L., and Miller, S. S. 2004. Effect of prohexadione-calcium dose level on shoot growth and fire blight in young apple trees. Plant disease. 88:1099–1106.

Norelli, J. R., and Miller, S. S. 2006. Using prohexadione-calcium to control fire blight in young apple trees. In X International Workshop on Fire Blight, Acta Horticulturae. 704:217–224.

Rademacher, W. 2006. Prohexadione-Ca: modes of action of a multifunctional plant bioregulator for fruit trees. In X International Symposium on Plant Bioregulators in Fruit Production, Acta Horticulturae. 727:97–106.

Rademacher, W., and Kober, R. Efficient Use of Prohexadione-Ca in Pome Fruits. European Journal of Horticultural Science. 68(3)S:101–107.

Roemmelt, S., Treutter, D., Speakman, J. B., and Rademacher, W. 1999. Effects of Prohexadione-Ca on the Flavonoid Metabolism of Apple With Respect to Plant Resistance against Fire Blight, Acta Horticulturae. 489:359-363.

Roemmelt, S., Zimmermann, N., Rademacher, W., and Treutter, D. 2003. Formation of novel flavonoids in apple (Malus x domestica) treated with the 2-oxoglutarate-dependent dioxygenase inhibitor prohexadione-Ca. Phytochemistry 64:709–716.

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 Disease. 92:714–718.

Schupp, J., Rosenberger, D., Robinson, T., Aldwinckle, H., Norelli, J., and Porpiglia, P. 2002. Post-symptom sprays of prohexadione-calcium affect fire blight infection of “Gala” apple on susceptible or resistant rootstock. HortScience. 37:903–905.

Smith, T. J. 1999. Report on the development and use of Cougarblight 98C - A situation-specific fire blight risk assessment model for apple and pear. Acta Horticulturae. 489:429–436.

Spinelli, F., Andreotti, C., Sabatini, E., Costa, G., Spada, G., Ponti, L., et al. 2006. Chemical control of fire blight in pear: application of prohexadione-calcium, acibenzolar-s-methyl, and copper preparations in vitro and under field conditions. Acta Horticulturae. 704:233–238.

F. Spinelli, Vanneste, J. L., Marcazzan, G. L., Sabatini, A. G., and Costa, G. 2005. Effect of Prohexadione-Calcium on Nectar Composition of Pomaceous Flowers and on Bacterial Growth. Fruit Disease Management New Zealand Plant Protection. 58:106-111.

Stover, E., Wirth, F., and Robinson, T. 2001. A Method for Assessing the Relationship Between Cropload and Crop Value Following Fruit Thinning. HortScience. 36:157–161.

Sundin, G. W., Werner, N. A., Yoder, K. S., and Aldwinckle, H. S. 2009. Field Evaluation of Biological Control of Fire Blight in the Eastern United States. Plant Disease. 93:386–394.

Tancos, K. A., and Cox, K. D. 2016. Exploring Diversity and Origins of Streptomycin-Resistant Erwinia amylovora Isolates in New York Through CRISPR Spacer Arrays. Plant Disease. 100:1307–1313.

Turechek, W. W., and Biggs, A. R. 2015. Maryblyt v. 7.1 for Windows: An Improved Fire Blight Forecasting Program for Apples and Pears. Plant Health Progress. 16:16–22. van der Zwet, T., Orolaza-Halbrendt, N., and Zeller, W. 2012. Pages 5-11, 15-29, 155-159 and 333-347 in: Fire Blight History, Biology, and Management. American Phytopathological Society Press, St. Paul, MN.

Yoder, K. S., Miller, S. S., and Byers, R. E. 1999. Suppression of Fireblight in Apple Shoots by Prohexadione-calcium following Experimental and Natural Inoculation. HortScience. 34:1202– 1204.

CHAPTER 3

ENDOPHYTIC BACTERIAL COMMUNITIES IN APPLE LEAVES ARE MINIMALLY

IMPACTED BY STREPTOMYCIN USE FOR FIRE BLIGHT MANAGEMENT*

Abstract

Fire blight, caused by the bacteria Erwinia amylovora, is a devastating disease of apples routinely managed using the antibiotic streptomycin. Both E. amylovora and streptomycin interact with endophytic bacterial communities, which are increasingly recognized for their valuable roles in plant hosts. Altering these communities may have profound implications for plant health and selection for streptomycin resistance. In this study, we investigated the impact of streptomycin applications (1, 3, or 9 post-bloom sprays) on endophytic bacterial communities in apple leaves in two orchards, using both culture-dependent and independent methods.

Streptomycin programs were compared to untreated trees and two programs approved for organic production. Culture-dependent methods minimally impacted abundance of culturable endophytic bacteria, as indicated by colony forming units ml-1 on Luria-Bertani agar. Culture- independent methods, using 16S rDNA high-throughput sequencing, generally did not detect a significant difference in bacterial abundance or community composition between treatments. A greater difference was observed between the two orchards investigated, indicating a strong selection by locality. Communities were dominated by the families Pseudomonadaceae,

Enterobacteriaceae, Desulfovibrionaceae, Bacillaceae, and Burkholderiaceae, which are commonly described in other investigations of microbial communities of apple orchards. In addition, we found a high abundance of Amoebophiliaceae, a family dominated by arthropod parasites, which could indicate horizontal transfer of endosymbionts between insect and plant

hosts. Our results provide evidence that antibiotic applications in apple orchards have minimal effect on endophytic bacterial communities, adding to the growing body of evidence that current commercial practices are sustainable solutions for fire blight management.

*Wallis, A., Ramachandran, P., Reed, L., Gu, G., Ottesen, A.O., Cox, K. Impact of fire blight management programs on endophytic bacterial communities of apple canopy. Phytobiomes. In Review.

Introduction

Fire blight, caused by the bacterial pathogen Erwinia amylovora, is one of the most important diseases of apples and pears worldwide. It is routinely managed using antibiotics, primarily the aminoglycoside streptomycin, applied at bloom and post-bloom when infection events are predicted (Norelli 2003). However, antibiotic use in agriculture has increasingly come under scrutiny due to the potential for development of resistance in pathogen and off-target bacterial populations, especially in light of the increasing prevalence of antibiotic resistance in both agricultural and clinical settings (McManus et al. 2002; Stockwell and Duffy 2012). In some agricultural markets, antibiotics are already limited or prohibited. For example, antibiotics have been prohibited in organic production in the United States, following regulation by the

National Organic Standards Board in October 2014; this is consistent with organic regulations in other locations including Canada and European Union (Granatstein 2013; Johnson and Temple

2013).

Antibiotics are a critical tool for fire blight management in commercial apple production.

E. amylovora becomes active in the spring, and a bacterial ooze, composed of cells and exopolysaccharides, is produced in cankers formed from infections in previous seasons. Primary infection occurs when E. amylovora vectored from oozing cankers colonizes to the floral surface and is subsequently washed into the floral cup during a rain or heavy dew, entering the plant via natural openings in the nectaries (van der Zwet 2012). Subsequent infections occur when bacteria are transferred to and enter leaf tissue that has been damaged, which most commonly occurs with driving rain, piercing sucking insects, hail, and pruning cuts. Reducing inoculum by pruning out cankers and applying copper prior to spring growth are essential components of fire blight management. However, the utilization of these two practices is limited by the fact that copper

may result in severe damage to fruit finish when applied after bloom, and cankers are easily obscured making absolute detection and removal nearly impossible.

Therefore, application of antibiotics and biopesticides are applied at bloom when infection events are predicted based on weather conditions. Disease forecasting components of decision-aid systems, such as the Network for Environment and Weather Applications system

(NEWA, http://newa.cornell.edu/), are used to time management applications during bloom and as needed for the duration of the growing season. Since its introduction in 1955, streptomycin has been the most common and effective means of managing fire blight during bloom in apple production in the United States, east of the Mississippi River (Cox et al. 2013, Sundin et al.

2009). The antibiotics kasugamycin and oxytetracycline are also registered for fire blight control in apple orchards (McGhee et al. 2011a), but their use is limited due to cost and lower efficacy of these antibiotics, respectively. Other materials that may be used to mitigate fire blight infections include biopesticide products consisting of organisms and/or their metabolite that may kill or outcompete E. amylovora in its biological niche, and plant defense inducing compounds

(SAR/ISR products). However, these materials typically provide much lower levels of protection against infection during bloom (Aldwinckle et al. 2002; Sundin et al. 2009). Therefore, streptomycin remains the most important management tool for fire blight in commercial orchards.

Streptomycin resistant (SmR) E. amylovora has been documented in most regions where apples are produced, including California (Moller et al. 1972), Oregon and Washington (Coyier and Covey 1975), Israel (Manulis et al. 1998), New Zealand (Thomson et al. 1993), Michigan

(McGhee et al. 2011b), and New York (Russo et al. 2008). Repeated applications of streptomycin are believed to contribute to the development of SmR E. amylovora, but studies

evaluating selection for resistance in E. amylovora and microbial populations in an orchard environment have reported mixed results. In some cases, higher incidence of resistance genes strA (aph(3”)-Ib or orfH) and strB (aph(6)-Id or orfI) has been found in soils with a history of antibiotic use (Tolba et al. 2002). An increase in the proportion of culturable epiphytic populations of bacteria resistant to streptomycin was shown to increase after multiple consecutive applications of streptomycin, and decrease after multiple applications of kasugamycin (Tancos and Cox 2017). But in other work, proportions of antibiotic-resistant bacteria were not related to history of streptomycin use in orchard soil, or on leaves or flowers

(Overbeek et al. 2002; Shade et al. 2013a; Yashiro and McManus 2012). These results suggest that the extent streptomycin programs influence SmR in overall bacterial communities in the environment is still unclear. More importantly, little is known about the impacts to the bacterial community within the plant where E. amylovora primarily resides.

More broadly, antibiotic use may influence the microbial community colonizing the host plant (i.e. the microbiome). There is mounting evidence that the microbiome plays a dramatic role in plant phenotype, and offers services ranging from yield and growth promotion to health and protection (Hardoim et al 2015; Kandel et al. 2017; Ryan et al. 2008; Turner et al. 2013).

Application of antibiotics has the potential to influence these microbial communities by changing community composition (Lin et al. 2016), reducing microbial diversity (Liu et al. 2009), increasing antibiotic resistance (Ding et al. 2014), and causing disturbances in ecological functions (Cycoń et al. 2019). Research investigating microbial communities of apple systems has shown that these communities can be influenced by factors such as field treatments (i.e. organic vs. conventional management) (Ottesen et al. 2009), leaf or flower age (Afandhi et al.

2018; Shade et al. 2013b), and host genotype (i.e. rootstock) (Liu et al. 2018). However, other

research efforts investigating the impact of streptomycin on epiphytic communities in apple orchards found that the antibiotic had no measured effect on apple leaves or flowers (Shade et al.

2013a; Yashiro and McManus 2012; Yashiro et al. 2011). Therefore, the effects of streptomycin applications on the microbial communities in an orchard environment remains unclear.

While many studies investigating plant microbiomes have focused on the total microbial community, a more specific community deserving scientific investigation in the context of fire blight management may be endophytic bacterial communities of the orchard canopy. Endophytes are generally defined as microorganisms residing inside plant tissues for at least part of their life cycle without causing detrimental effects to their host (Brader et al. 2017; Hardoim 2015). The entire phytobiome (endophytes and epiphytes) is influenced by numerous factors including both environmental conditions such as temperature, moisture, and UV radiation, as well as host factors including host species, age or developmental stage, and plant health status (Dastogeer

2020; Lindlow and Brandl 2003). While epiphytes, or microorganisms colonizing the exterior surfaces of plants, are dramatically affected by environmental conditions, endophytes are protected by plant membranes and may be more resilient. In addition, they form close, complex relationships with their hosts without eliciting a defense response, although the mechanisms of these relationships are still poorly understood, and provide services to their hosts ranging from plant growth to pathogen protection (Dastogeer et al. 2020; Lindlow and Brandl 2003; Reinhold-

Hurek and Hurek 2011; Ryan 2007).

In the apple-fire blight pathosystem, endophytic communities may be especially important. E. amylovora is well known to travel and survive endophytically, colonizing the plant well ahead of symptom development (Rosen 1929; Momol 1998; Clarke et al. 1999; van der

Zwet 2012), whereas investigations of leaf and flower surfaces have found E. amylovora to be an

unsuccessful epiphyte (Bonn 1981; Manceau et al. 1990; Ockey and Thomson 2006; Tancos and

Cox 2017). Streptomycin, when applied to the canopy for fire blight management, has locally systemic activity (van der Zwet and Buskirk 1984). This has the benefit of providing protection against blossom and shoot blight infection up to 24 hours post infection event, but because of this local systemic activity, may also have detrimental effects to the entire endophytic community and, therefore, on plant health. The interaction between streptomycin, E. amylovora, and other apple foliar endophytes warrants more attention in order to understand the downstream effects on the apple host and orchard environment.

In this study, we investigated the cumulative effects of fire blight disease management programs on the endophytic bacterial communities of apple leaf tissue over two years. Multiple applications of streptomycin (1, 3, or 9 post-bloom applications) were compared to application of copper, application of a biopesticide, and untreated control. Both culture-dependent and independent methods were used to evaluate abundance, community composition, diversity, and streptomycin resistance. Results have the potential to improve our understanding of the impacts of antibiotics on the endophytic bacterial communities of the orchard canopy and inform more sustainable management recommendations for commercial producers.

Materials and Methods

Study Site

To evaluate the effect of fire blight disease management programs on endophytic bacterial communities, field trials were conducted at Cornell AgriTech in Geneva, NY

(42°87’70.19”N -77°02’99.15”W) over two field seasons in 2018 and 2019. Experiments were performed in a semi-dwarf planting of Gala on B.9 rootstock, planted in 2000 at 1.8 m in-row

spacing (Orchard A), and a high-density planting of Gala on G.935 rootstock planted in 2016 at 1 m in- row spacing (Orchard B). Trees were managed following commercial standards according to the Cornell Pest Management Guidelines (Agnello et al. 2019).

Experimental design, application of treatments, and sample collection

A randomized complete block design with four single-tree replications or a panel of four trees surrounded by buffer trees within and across rows was used to evaluate the management programs in the semi-dwarf (Orchard A) and high-density (Orchard B) orchards, respectively.

All applications were made using a Solo 475-B gas-powered mist blower (Solo Incorporated,

Newport News, VA) calibrated to deliver 935.3 L ha-1 of water, a standard volume for high- density apple plantings in the northeastern United States. Beginning at bloom, trees were subjected to one of three antibiotic programs (Table 3.1). Antibiotic programs included an application of streptomycin at 80% bloom, followed by 1, 3, or 9 post bloom (PB) applications of streptomycin on a weekly basis beginning at ‘petal fall’ phenological stage, approximately one week after bloom application (1PBStrep, 3PBStrep, and 9PBStrep). Such antibiotic programs were meant to represent a typical commercial program (1PBStrep) and two programs representing non-recommended practices, whereby a grower might make excessive streptomycin applications, attempting to avoid shoot blight in a young planting or nursery block (3PBStrep and 9PBStrep). The commercial product used was Firewall 17 (Nufarm, Morrisville, NC), at the label rate of 1.68 kg ha-1 of the product. These streptomycin programs were compared to two other bactericides commonly used to manage fire blight in organic commercial apple production.

These management programs included either an OMRI-listed formulation of Bacillus subtilis, a biopesticide, or an OMRI-listed fixed copper product, formulated as a soap, which both function

as general antimicrobials. Both were applied at 80% bloom and petal fall, to constitute one bloom and one post-bloom application. The commercial products used were Serenade Opti

(Bayer, Institute, WV) and Cueva (Certis, Wasco, CA), applied at the label rates of 1.40 kg ha-1 and 4.67 L ha-1, respectively (Treatments will be referred to Serenade and Cueva, respectively).

All treatments were compared to untreated control trees which received no chemical applications for fire blight management (Untreated). All treatments were applied to the same trees in subsequent years of the study in order to observe cumulative effects and interaction between treatments.

Table 3.1. Fire blight disease management treatments applied to apple orchards in Geneva, NY

Treatment Active Product (rate per ha) Timinga Rationale Ingredient Untreated None None None Control 1PBStrep Streptomycin FireWall 17 (1.68 kg) B, PF 1 post-bloom + Regulaid (1.4 L) application of streptomycin 3PBStrep Streptomycin FireWall 17 (1.68 kg) B, PF, 1C 3 post-bloom + Regulaid (1.4 L) applications of streptomycin 9PBStrep Streptomycin FireWall 17 (1.68 kg) B, PF, 1C, 2C 9 post-bloom + Regulaid (1.4 L) applications of streptomycin Serenade Bacillus subtilus Serenade Optimum (1.40 kg) B, PF Biopesticide strain QST713 + Regulaid (1.4 L) Cueva Copper Soap Cueva (4.67 L) B, PF General antimicrobial aB: Bloom, PF: Petal Fall, 1C: 1st cover (approximately 1 week after petal fall), 2C: 2nd cover (approximately 2 weeks after petal fall)

Sample collecting and processing

Foliar tissue samples were collected in late August 2018 and 2019, approximately two weeks after the final treatment date. Five mature leaves were randomly collected from each replication using gloved hands, placed in Ziploc bags, and transferred to the lab on ice. Samples were processed within 12 h. Using sterile forceps and scalpel, petioles were removed and leaf blades were surface sterilized as described by Afandhi et al. (2018) and Muresan (2017). Leaves were surface sterilized by gently washing with sterile water to remove dirt and debris, sequentially submerging in 5% sodium hypochlorite solution, 70% ethanol, and two rinses of sterile water for 30 s each. Removal of potentially contaminating epiphytic bacteria was verified by imprinting leaves on Luria-Bertani agar (LB) and incubating at 28°C for 7 days. Leaf blades were placed in extraction bags (BioReba, Reinach, CH) with 3 ml of extract buffer (160 g L-1 sodium chloride, 4 g L-1 potassium monophosphate, 23 g L-1 sodium phosphate dibasic, 4 g L-1 potassium chloride, 4 g L-1 sodium azide, 0.5 ml L-1 Tween 20, 20 g L-1 polyvinyl pyrrolidone, pH 7.4) according to Muresan (2017). Each sample was macerated for approximately 30 s using a Homex 6 semi-automated homogenizer (BioReba). A 1 mL aliquot of the maceration liquid from each sample was frozen for culture-independent methods. The rest of the maceration liquid was processed immediately for culture-dependent investigations.

Culture-dependent investigations of bacterial communities

Maceration liquid was serially diluted and spread on Crosse Goodman media (CG)

(Crosse and Goodman 1973) for identification of E. amylovora, and on LB agar for enumeration of total endophytic populations, at appropriate dilutions for visual colony forming unit (CFU) enumeration. Maceration fluid was also spread on CG and LB amended with 100 mg L-1

streptomycin for selection of SmR isolates. All media was amended with 10 mg L-1 cycloheximide in order to inhibit eukaryotic growth. Plates were incubated at 28°C for 48 h and the resulting colonies were enumerated to calculate total endophytic bacteria per sample. Growth on CG plates was observed using a dissecting microscope to identify E. amylovora colonies based on characteristic cratering morphology (Crosse and Goodman 1973; Tancos et al. 2016).

In order to characterize the most prevalent bacteria isolated, a subset of bacteria with unique colony morphologies were isolated and identified by sequencing the 16S rRNA gene using the primers 515f/806r, as previously described by Walters et al. (2016) and Wasimudin et al. (2019). PCR was performed in 25 μl volumes consisting of 12.5 μl EmeraldAmp GT PCR

Master Mix (Takara Bio USA, Mountain View, CA), 1 μl each of forward and reverse primer, 2

μl of bacterial suspension, and 8.5 μl nuclease-free H2O. Cycling parameters were 3 min at 94°C;

35 cycles of 45 s at 94°C, 60 s at 50°C, and 90 s at 72°C; followed by a final extension of 10 min at 72°C. Resulting PCR products were verified by gel electrophoresis and the presence of an approximately 290 bp amplicon (Walters et al. 2016; Wassimudin 2019). Amplified DNA was purified using ExoSAP-IT Express PCR cleanup reagent (ThermoFisher, Waltham, MA) and sequenced at the Cornell Biotechnical Resource center in Ithaca, NY using an ABI 3730xl capillary electrophoresis instrument (Applied Biosystems, Waltham, MA). Nucleotide sequences were analyzed using Basic Local Alignment Search Tool (BLAST; National Center for

Biotechnology Information (NCBI)) against the entire GenBank database, and identified to genus or species level.

Culture-independent investigation of bacterial communities

For extraction of bacterial DNA, the reserved 1 ml of maceration fluid was centrifuged for 10 min at 13,000 x g to pelletize bacterial cells, and extraction was carried out on the pellet using the DNeasy PowerLyzer PowerSoil Kit (QIAGEN, Germantown, MD) according to the manufacturer’s instructions. Extracted DNA samples were processed for 16S high-throughput sequencing according to the Earth Microbiome Project recommended protocols (Caporaso et al.

2011 and 2012), with minor adjustments to account for the high plastid content in these foliar samples. The 16S region was firstly amplified as described by Ikenaga et al. (2015) and

Lundberg et al. (2013) using the primer pair 63f/1492r (~1400 bp amplicon), including the addition of a pair of locked nucleic acid (LNA) clamps to block the amplification of off-target apple mitochondria and plastid sequences. PCR reactions were conducted in triplicate in 25 μl volumes consisting of 12 μl EmeraldAmp GT PCR Master Mix (Takara Bio USA), 1 μl each of forward and reverse primer, 3 μl each of forward and reverse LNA clamps (4 μM concentration),

3 μl of bacterial suspension, and 2 μl nuclease-free H2O. Cycling parameters were 3 min at 94°C;

35 cycles of 1 min at 94°C, 1 min at 70°C, 1 min at 54°C, and 2 mins at 72°C; followed by a final extension of 10 min at 72°C. A second PCR reaction was performed to amplify the V4 region of 16S rRNA gene using barcoded primers, 515f/806r, with Illumina adapters (Walters et al., 2016). PCR reactions were conducted in 25 μl volumes consisting of 10 μl Platinum Hot

Start PCR Master Mix (ThermoFisher), 2 μl of forward primer and reverse primers (5 µM), 2 μl of product from first PCR, and 9 μl nuclease-free H2O. Cycling parameters were 3 min at 94°C;

35 cycles of 45 s at 94°C, 60 s at 50°C, and 90 s at 72°C; followed by a final extension of 10 min at 72°C. The products of each PCR step were confirmed using gel electrophoresis. Triplicate

PCR products were pooled and aggregate samples were purified using Agencourt Ampure XP

bead (Beckman Coulter Inc., Brea, CA). Purified amplicons were quantified using Qubit 1X dsDNA HS Assay Kit (Invitrogen, Eugene, OR) and TapeStation 2200 (Agilent, Santa Clara,

CA), and pooled in equal concentration. Pooled DNA library was prepared as described previously (Gu et al. 2019) and sequenced on a MiSeq platform using 500-cycles v2 kits

(Illumina, San Diego, CA).

Following sequencing, 16S sequences were demultiplexed using deML (v 1.1.3). The demultiplexed sequences were then filtered to remove low quality sequences using Trimmomatic

(Bolger et al. 2014). Taxonomic abundance profiles and assignments were classified using

Kraken 2 classification tool (Wood and Salzberg 2014) and Bracken abundance estimator (Lu et al. 2017), with the SILVA database (Pruesse et al. 2007). Mitochondria and plastid sequences were removed using a Python script (extract_kraken_reads.py) which takes in a kraken-style output and kraken report and a taxonomy level to extract reads matching that level. Once the bacterial reads were extracted for each sample, they were subjected to Kraken 2 classification tool with SILVA database and Bracken abundance estimator. OTUs were defined by clustering with a 97% similarity threshold (3% divergence), OTUs with < 0.01% abundance were filtered out, and the remaining OTUs were classified to the family level to create an OTU table for downstream analyses.

Data analysis

Culture-dependent investigations. Within each field in each year, bacterial abundance was calculated for each treatment by determining the average CFU ml-1. The effect of treatment on bacterial abundance and effect on percentage of antibiotic resistant bacteria were determined by a mixed effects model using the lmer command in the lme4 package in R (Bates et al. 2015),

in which treatments were considered fixed effects and replications were considered random effects. All analyses were carried out in R version 3.6.3 (R Core Team 2020). Bacterial abundance values (CFU counts) were log transformed prior to analysis to normalize abundance values.

Culture-independent investigations. The dataset was analyzed in R version 3.6.3 (R Core

Team 2020) using the vegan package version 2.5-6 (Oksanen et al. 2019). Alpha diversity was evaluated using Shannon and Simpson diversity indices; normality of data was evaluated using

Shapiro-Wilks test, and differences were determined using Kruskal-Wallis non-parametric test and Dunn’s test for pairwise comparisons. To determine whether bacterial communities differed by field or treatment, OTU table was used to construct Bray-Curtis dissimilarity matrix that was used to evaluated beta diversity. To further investigate differences between field and treatment, nonmetric multidimensional scaling (NMDS) was conducted based on the Bray-Curtis dissimilarity matrix. NMDS analyses were implemented using the ‘metaNMDS’ function of the

R package (Oksanen 2019). Differences between sites or treatments were determined using a permutational multivariate analysis of variance (PERMANOVA), with the ‘adonis’ function.

Data was visualized using the ggplot2 program for R (Wickham 2016).

Results

Culture-dependent investigations of bacterial communities

Total endophytic bacterial abundance, as determined by CFU counts on LB agar, ranged from 0 to 2.16 log(CFU) ml-1 (Table 3.2). In Orchard A, bacterial abundance was not significantly affected by treatment in 2018 (P =0.1430) or 2019 (P = 0.0503). In Orchard B, in

2018, culture-dependent investigations of bacterial communities was not possible due to

insufficient number of CFUs (i. e. <2) recovered from leaf tissues in many treatment replicates on the lowest dilution plates. In 2019, in Orchard B, bacterial abundance was impacted by treatment (P = 0.0151), with lower CFU counts for trees that received 3PBStrep or 9PBStrep treatments (the two higher doses) than untreated trees, but other treatments were not significantly different (P > 0.05). E. amylovora was only detected in 2 samples across all treatments, fields, and years. These were both in 2019, in one sample from the Untreated treatment in Orchard A and one sample from the Cueva treatment in Orchard B. Therefore, abundance of E. amylovora was too low to draw meaningful conclusions about treatment effects. Similarly, bacterial CFUs were observed on media amended with streptomycin for only 6 total samples, and also too low in abundance to draw meaningful conclusions about treatment effects on streptomycin sensitivity in bacterial endophyte communities. These were in 2019, in one sample each from Untreated,

1PBStrep, 3PBStrep, and Serenade treatments from Orchard A, and the Untreated control and

Serenade treatments from Orchard B. Along these lines, none of the streptomycin treatments resulted in samples with a high number (>10) of SmR CFUs.

The most commonly recovered bacteria included Pseudomonas spp., Bacillus spp., and

Pantoea spp. All three were detected in at least one sample for each treatment. Most common were Pseudomonas spp., which were present in 64.6% of all samples (66.7% and 62.5% of samples from Orchard A and B, respectively), followed by Bacillus spp., in 35.4% of all samples

(33.3% and 37.5%) and Pantoea spp., in 27.1% of all samples (25.0% and 29.2%). Two other genera that were less frequently detected included Curtobacterium spp. and Frondihabitans spp. in 4.2% and 2.1% of all samples, respectively.

Table 3.2. Abundancez of bacterial endophytes in apple leaves treated with different fire blight management programs for two orchards in Geneva, NY in 2018 and 2019.

Orchard A Orchard B Treatment 2018 2019 2018 2019 Untreated 0.99 ± 0.70 ay 0.98 ± 0.07 a NAx 2.16 ± 0.39 b 1PBStrep 0.50 ± 0.67 a 0.78 ± 0.39 a NA 0.87 ± 0.34 ab 3PBStrep 0.35 ± 0.26 a 0.00 ± 0.00 a NA 0.45 ± 0.27 a 9PBStrep 0.35 ± 0.26 a 0.69 ± 0.30 a NA 0.26 ± 0.26 a Cueva 1.25 ± 1.18 a 0.56 ± 0.33 a NA 0.87 ± 0.34 ab Serenade 0.35 ± 0.35 a 0.91 ± 0.35 a NA 1.16 ± 0.30 ab p-value 0.1430 0.0503 NA 0.0151

zAbundance was measured as the log of colony forming units (CFU) ml-1 on Luria-Bertani media. yWithin a column, different letters indicate significantly different mean CFU ml-1 based on Tukey HSD test (P < 0.05). xNA: CFU counts were insufficient for enumeration (<2 CFU on lowest dilution plates).

Culture-independent investigations of bacterial communities

16S rDNA high-throughput sequence data was submitted to NCBI under the Bioproject

PRJNA663546 with SRA accession numbers SRR1270323-SRR12708369. After quality control, a total number of 4,332,792 bacterial reads were retained, ranging from 34,938 to 178,302 per sample. After mitochondrial and plastid sequences were filtered out, reads present in > 0.01 relative abundance in all samples were assigned to 37 and 59 families for 2018 and 2019 respectively.

Alpha diversity, as measured by Shannon diversity index (Figure 3.1), in 2018, was affected by treatment (P = 0.0054). Lower diversity was observed in trees receiving the

3PBStrep treatment than in Untreated trees (P = 0.0066), but was not affected by other

treatments (P > 0.05). In 2019, alpha diversity was unaffected by treatment (P = 0.7545).

Orchard A and B did not differ in alpha diversity in 2018 (P =0.0815), but, in 2019, there was greater diversity in Orchard A than Orchard B (P = 0.0005). Simpson diversity index analysis indicated similar trends. In 2018, diversity was influenced by treatment (P = 0.0478). Lower diversity was observed in trees receiving the 3BPStrep treatment than Serenade (P = 0.0476) and

Untreated trees (P = 0.0059). In 2019, diversity was not affected by treatment (P = 0.7734). In both 2018 and 2019, Simpson index indicated that there was higher diversity in Orchard A than

Orchard B (P = 0.0478 and 0.0014 in respective years). In 2018, a total of 37 families were detected, with 16 (42.1%) detected in all treatments in both orchards. In 2019, a total of 60 families were detected and 20 (33.3%) were detected in all treatments in both orchards. Families were seldom limited to a single treatment-orchard combination. In 2018, all families were found in at least two treatment-orchard combinations, and in 2019, 4 families (0.07%) were only detected in one treatment-orchard combination.

Figure 3.1. Alpha diversity (Shannon Diversity Index) values displayed as box plots for endophytic bacterial communities in 2018 (A) and 2019 (B), of leaf samples collected from apple trees treated with different fire blight management programs, in two different ‘Gala’ apple orchards (‘A’ and ‘B’) in 2018. 1PBStrep, 3PBStrep, or 9PBStrep: 1, 3, or 9 applications of streptomycin post-bloom; Cueva: a fixed copper bactericide, formulated as a soap, applied at bloom and petal fall; Serenade: a formulation of a strain of Bacillus subtilis, applied at bloom and petal fall.

NMDS ordination of Bray-Curtis dissimilarity matrices indicated strong separation between Orchard A and B in both 2018 and 2019, as determined by the permutational ANOVA using the adonis function in R (P < 0.001 in both years, Figure 3.2). In 2018, separation by treatment was detected (P = 0.044), and pairwise comparisons indicated the only significant difference was between 3PBStrep and Serenade treatments. In 2019, no separation between treatments was evident (P = 0.071).

The most abundant families detected in 2018 included Amoebophilaceae,

Desulfovibrionaceae, Bacillaceae, Enterobacteriaceae, Burkholderiaceae, and

Pseudomonadaceae, representing 61.0, 6.3, 5.7, 4.8, 4.8, and 4.1% of the total dataset, respectively (Figure 3.3A). In 2019, the most abundant families were similar and included

Amoebophilaceae, Enterobateriaceae, Pseudomanadaceae, and Desulfovibrionaceae, making up

33.4, 15.8, 11.5, and 11.1% of the total dataset (Figure 3.3B). Other families were present in

3.5% or less. The 10 most common families made up 94.9 and 86.3% of the total dataset in respective years, and were present in all samples.

Figure 3.2. Nonmetric multidimensional scaling (NMDS) ordination of Bray-Curtis dissimilarity measures in 2018 (A & C) or 2019 (B & D) of leaf samples collected from apple trees treated with different fire blight management programs, in two different ‘Gala’ apple orchards in 2018. Colored by treatment (A & B) or by orchard (C & D). Ellipses define the upper 95th percentile limit of the respective distribution.

Figure 3.3. Relative abundance of the ten most dominant bacterial families in endophytic bacterial communities of leaf samples in 2018 (A) and 2019 (B), collected from apple trees treated with different fire blight management programs, in two ‘Gala’ apple orchards (‘A’ and ‘B’) in 2018. Colors indicate bacterial families.

Discussion

Characterization of endophytic communities of apple phyllosphere

A growing understanding of the apple microbiome and the impacts of farming practices on microbial communities in agricultural landscapes has been facilitated by continuously improving and more available technologies in both culture-dependent and independent techniques. This has been especially important regarding the use of antibiotics in agriculture, which are invaluable tools for farmers, but have the potential to cause severe and lasting impacts to bacterial communities in the agricultural ecosystem. Numerous reports have scrutinized the use of antibiotics in agriculture, citing the potential for selecting antibiotic resistance in pathogen populations, the subsequent loss of invaluable crop protection tools, and potential off-target effects to the agroecosystem and other environments (McManus et al. 2002; Stockwell and Duffy

2012). Overuse of streptomycin is believed to contribute to the streptomycin resistance in E. amylovora populations, a problem that has severely limited the ability of the industry to control the pathogen in many production regions worldwide (McManus et al. 2002; Moller 1972;

Manulis et al. 1998; Thomson et al. 1993; McGhee and Sundin 2011b; Sundin and Wang 2018;

Tancos 2016). However, previous studies have found that streptomycin resistance genes are ubiquitous in both agricultural and pristine environments (Allen et al. 2010; Fitzpatrick and

Walsh 2016; Overbeek et al. 2002; Tolba 2002), and research investigating the impacts of antibiotics on the selection of resistance in microbial communities in soil, phyllosphere, and floral niches have found mixed results (Yashiro and McManus 2012; Shade 2013).

In this work, we investigated the effects of post bloom streptomycin applications for fire blight management on the endophytic bacterial communities of the apple canopy. While the

microbial communities of apple flowers have been studied extensively, the endophytic communities of foliar tissue have received much less attention. The flower has been the primary focus of many studies in the hopes of better understanding and developing controls for the blossom blight phase of the disease (Johnson and Stockwell 1998). Research in this area has led to a detailed understanding of the biology of the pathogen during a critical phase of the disease cycle (colonization of the stigmatic surface and host entry), and has elucidated many candidate microorganisms that may be used in biopesticides, including several that are now on the market

(Johnson and Stockwell 1998; Norelli et al. 2003; Pusey 2019). The internal movement of E. amylovora following host entry via the blossom or the shoot, is also well known and described, yet there has been little investigation of the endophytic communities of host plants as pertaining to the pathogen, nor have there been efforts to describe the effects of streptomycin use on the endophytic communities. This is surprising, given the growing appreciation for the influence of endophytes on their hosts (Berg et al. 2014; Bulgarelli et al. 2013; Chen et al. 2019; Hawkes and

Conner 2017; Kandel et al. 2017; Ryan 2008). Additionally, the possibility of genetic exchange between E. amylovora and other bacteria, as suggested by Chiou and Jones (1995), raises concerns about SmR development in pathogen populations in these environmental niches. In the present study we found that SmR recovery was rare and the frequency of streptomycin application had no bearing on the development of SmR endophytes in culture dependent or independent analysis. Similarly, despite a history of severe fire blight epidemics in both orchards, E. amylovora was rarely recovered from within leaf tissue. These two pieces of evidence suggest that shoot blight management programs involving frequent summer applications of streptomycin are unlikely to lead to selection of SmR endophyte community, and in turn, pose little risk for selection of SmR E. amylovora residing within trees.

In the present study, both culture-dependent and independent methods described similar communities of bacterial endophytes in apple foliage in both orchards. Members of the

Pseudomonadaceae family were most commonly isolated in culture and most abundant in 16S sequencing analyses. The common detection of Pseudomonas, Bacillus, Enterococcus, and

Pantoea in culture was consistent with community members found in most abundance in 16S high-throughput sequencing. Curtobacterium and Frondihabitans were also expected members of this community as they are common Gram-positive inhabitants of the soil and foliar environments, respectively. These results are consistent with other studies of the apple phyllosphere, in which the most common genera isolated or detected from foliar and floral tissues included Pseudomonas, Pantoea, Bacillus, Clavibacter, Curtobacterium, Erwina, and

Micrococcus (Tancos and Cox 2017, McGhee and Sundin 2011, Yashiro and McManus 2012,

Johnson and Stockwell 1998).

The high abundance of Amoebophiliaceae detected in all samples in both years was interesting and unexpected. A large part of this family is composed of the genus Candidatus

Cardinium, which are a group of Gram-negative bacteria commonly found as parasites within arthropods and nematodes (Bergey et al. 2015; Erban et al. 2020; Nakamura et al. 2009; de

Vienne 2016). Detection of these bacteria in such high abundance can be regarded as evidence for horizontal transfer of endosymbionts between insect and plant hosts. This is an actively growing area of investigation, with implications for the life cycles of both the bacterial parasites and their hosts (Chrostek et al 2017; Gonella et al. 2015). These bacteria would be undetected by culture-based methods, which underlines the importance of using sequencing technologies to characterize microbial communities. The high abundance of Desulfovibrionaceae was

unexpected in apple leaf samples. This family is composed of five described genera, which are desulfonating bacteria most typically found in anoxic marine environments (Kuever et al. 2015).

The high abundance of the remaining organisms was not unexpected. The

Burkholderiaceae family is composed of ecologically extremely diverse organisms, including environmental saprophytic organisms, phytopathogens, opportunistic pathogens, and mammalian pathogens (Coenye 2014, Garrity et al. 2015). Similarly, members of Microbacteriaceae,

Sphingomonadaceae, Microscillaceae, and Rhizobiaceae are diverse and commonly isolated from environmental samples including soil, water, and the plant phyllosphere and rhizosphere

(Alves et al. 2014; Bergey 2015; Evtushenko 2015; Glaeser and Kampfer 2014).

Beijerinckiaceae are metabolically diverse aerobic bacteria, which fix nitrogen (Marin and

Arahal 2014). Overall, the bacterial endophytic communities identified through both culture- dependent and independent methods were largely expected in the environmental niche we investigated, and commonly found in microbial communities described in other investigations of the apple microbiome.

Applications of streptomycin for fire blight management have minimal effects on endophytic bacterial communities

We found little evidence of post-bloom streptomycin applications impacting endophytic bacterial communities. In only a few comparisons, a greater number of antibiotic applications led to a small reduction in microbial abundance. Specifically, in only one orchard and in one year, trees receiving higher numbers of streptomycin applications (3PBStrep and 9PBStrep) had lower culturable bacteria than Untreated trees. Additionally, in 2018, bacterial abundance (alpha diversity) was lower for 3PBStrep than Untreated trees, and community composition (beta

diversity) was different for 3PBStrep than Serenade, an antimicrobial biopesticide. However, this was not true in 2019, and was not corroborated by results for the 9PBStrep treatment, for which diversity was not impacted. If an increasing number of streptomycin applications was responsible for this community shift, we should have expected to see a greater impact on trees receiving the 9PBStrep treatment, and greater impact in 2019 after two consecutive years of treatments. Based on these results, we found minimal evidence to support the notion that substantial decrease in endophytic bacterial communities can be caused by streptomycin programs. This was different than our previous work, in which increasing number of streptomycin applications resulted in rapid selection for SmR in epiphyte communities, with an increase in number of SmR Pseudomonads and reduction in the number of SmR P. agglomerans

(Tancos and Cox 2017). The difference in observations between our current and previous work is that the former study examined epiphytic bacterial communities, which would have more direct exposure to treatments and therefore be less resilient than the endophytic communities examined here.

There is also concern that antibiotic use may have greater impact on microbial communities than organic or biopesticide based programs. We found little evidence supporting this notion. Bacterial communities of trees receiving the Serenade, and OMRI-listed formulation of antimicrobial metabolites produced from fermentation of a strain of Bacillus subtilis¸ generally were not different than that with other streptomycin programs, with the one exception of 3PBStrep in 2018. Again, this trend was not consistent with the 9PBStrep treatment or the observations after the second consecutive year of treatment. Previous work examined the difference in microbial communities of orchards under conventional and organic management

(Ottesen et al. 2009). Streptomycin was used in both management programs, as this study was

carried out prior to the prohibition of antibiotics in organic agriculture in 2014 (Granatstein

2013). Therefore, differences in bacterial communities observed in this earlier study could be attributed to other management factors, such as fungicides, insecticides, fertilizers, or herbicides.

Moreover, the study by Ottesen et al. (2009) also examined only epiphytic communities, which would have more direct contact with orchard sprays and other inputs than endophytic communities. Our findings suggest that conventional fire blight management with antibiotics and biopesticide based programs are likely similar in their effects on endophytic bacterial communities. In addition, we observed very little change in endophytic bacterial communities, even when the number of streptomycin applications was far greater than the number typically applied under responsible commercial practices, demonstrating a very high level of resiliency of these communities.

Impact of orchard site on endophytic bacterial communities

Greater differences in microbial community composition and diversity were observed between orchards than among the management programs. This could be related to the resident bacterial communities in the environment at each site (soil, air, water). Orchards A and B were located in very close proximity (< 0.25 miles), indicating that these communities are extremely locality specific. Host factors, such as cultivar, rootstock, and tree history could play a large role in shaping the microbial communities. Liu et al. (2018) found that endophytic communities varied by different rootstock and scion combinations, suggesting a genotype-specific influence, and Steven et al. (2018) found evidence for cultivar and floral anatomy influencing the microbial communities. Additionally, studies of the floral microbiome provided strong evidence that this niche selected for specific communities in a successional manner, demonstrating the profound

effect hosts could play in shaping their microbial communities, even when challenged with E. amylovora (Cui et al. 2020; Shade 2013b). In our study, the same cultivar was investigated, but orchards included different rootstocks, year of establishment, and planting systems. Future work is warranted to investigate these host- and orchard-specific communities and the factors that shape them. Improved understanding of these dynamics could result in better recommendations for plant health and disease resistance.

In this work, we investigated the effects of streptomycin application for fire blight management on the endophytic bacterial community of the apple canopy. Using both culture- dependent and independent methods, we described communities consistent with other investigations of the apple tree tissues, dominated by Amoebophilaceae, Pseudomonadaceae,

Enterobacteriaceae, Desulfovibrionaceae, and Bacillaceae. Our results provided evidence that management programs had minimal effect on these communities, did not select for streptomycin resistance, and that communities were more greatly influenced by the orchard blocks than by antibiotic applications. This work adds to the growing body of evidence that current commercial practices are sustainable solutions for fire blight management.

Literature Cited

Afandhi, A., Choliq, F. A., Anggrilika W.S., H., and Tarno, H. 2018. Distribution of the Endophytic Fungi in Apple Leaves. AGRIVITA Journal of Agricultural Science. 40(1):91-100.

Agnello, A., Brown, B., Carroll, J., Cheng, L., Cox, K., Curtis, P., et al. 2019. Cornell Pest Management Guidelines for Commercial Tree Fruit Production. Ithaca, NY: PMEP Cornell University.

Allen, H. K., Donato, J., Wang, H. H., Cloud-Hansen, K. A., Davies, J., and Handelsman, J. 2010. Call of the wild: antibiotic resistance genes in natural environments. Nature Reviews Microbiology. 8:251–259.

Alves, L., de Souza, J., de Mello Varani A., de Macedo Lemos, E. 2014. The Family Rhizobiaceae in The Prokaryotes (ed. Rosenberg, E.). New York: Springer, pp.419-437.

Bates, D., Mächler, M., Bolker, B., and Walker, S. 2015. Fitting Linear Mixed-Effects Models Using lme4. J. Stat. Soft. 67(1):1-48.

Berg, G., Grube, M., Schloter, M., and Smalla, K. 2014. Unraveling the plant microbiome: looking back and future perspectives. Front. Microbiol. 5(148):1-7.

Bergey, D. and Holt, J. eds. 2015. Bergey’s Manual of Determinative Bacteriology. Philadelphia: Lippincott Williams & Wilkins. Print.

Bolger, A. M., Lohse, M., and Usadel, B. 2014. Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics. 30:2114–2120.

Bonn, W. G. 1981. Monitoring of epiphytic Erwinia amylovora and the incidence of fire blight of apple and pear in southwestern Ontario. Acta Hort. 117:31-36.

Brader, G., Compant, S., Vescio, K., Mitter, B., Trognitz, F., Ma, L.-J., et al. 2017. Ecology and Genomic Insights into Plant-Pathogenic and Plant-Nonpathogenic Endophytes. Annu. Rev. Phytopathol. 55:61–83.

Bulgarelli, D., Schlaeppi, K., Spaepen, S., van Themaat, E. V. L., and Schulze-Lefert, P. 2013. Structure and Functions of the Bacterial Microbiota of Plants. Annu. Rev. Plant Biol. 64:807– 838.

Caporaso, J. G., Lauber, C. L., Walters, W. A., Berg-Lyons, D., Huntley, J., Fierer, N., et al. 2012. Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. The ISME Journal. 6:1621–1624.

Caporaso, J. G., Lauber, C. L., Walters, W. A., Berg-Lyons, D., Lozupone, C. A., Turnbaugh, P. J., et al. 2011. Global patterns of 16S rRNA diversity at a depth of millions of sequences per sample. Proceedings of the National Academy of Sciences. 108:4516–4522.

Chen, Q.-L., Cui, H.-L., Su, J.-Q., Penuelas, J., and Zhu, Y.-G. 2019. Antibiotic Resistomes in Plant Microbiomes. Trends in Plant Science. 24:530–541.

Chiou, C.-S., and Jones, A. L. 1995. Expression and identification of the strA-strB gene pair from streptomycin-resistant Erwinia amylovora. Gene. 152:47–51.

Chrostek, E., Pelz-Stelinski, K., Hurst, G. D. D., and Hughes, G. L. 2017. Horizontal Transmission of Intracellular Insect Symbionts via Plants. Front. Microbiol. 8:2237.

Clark, G. G., Hickey, K. D., and Travis, J. W. 1991. Recovery of Erwinia amylovora from excised infected apple shoots and subsequent development of symptoms on pruning stubs in the orchard as influenced by pruning methods. [Abstract]. Phytopathology. 81:121.

Coenye, T. 2014. The Family Burkholderiaceae in The Prokaryotes 4th Edition (ed. Rosenberg, E.). New York: Springer, pp. 759-776.

Cox, K. D., Aldwinckle, H. S., Villani, S. M., and Bersocke, K. A. 2013. Evaluation of bactericidal programs for the management of fire blight on ‘Idared’ apples in NY, 2012. Plant Dis. Manage. Rep. 7:PF001.

Coyier, D. L. and Covey, R. P. 1975. Tolerance of Erwinia amylovora to streptomycin sulfate in Oregon and Washington. Plant Dis. Rep. 59:849-852.

Crosse, J., and Goodman, R. 1973. A selective medium for and a definitive colony characteristic of Erwinia amylovora. Phytopathology. 63:1425–1426.

Cui, Z., Huntley, R. B., Zeng, Q., and Steven, B. 2020. Temporal and spatial dynamics in the apple flower microbiome in the presence of the phytopathogen Erwinia amylovora. Microbiology. doi: https://doi.org/10.1101/2020.02.19.956078.

Cycoń, M., Mrozik, A., and Piotrowska-Seget, Z. 2019. Antibiotics in the Soil Environment— Degradation and Their Impact on Microbial Activity and Diversity. Front. Microbiol. 10:338.

Dastogeer, K. M. G., Tumpa, F. H., Sultana, A., Akter, M. A., and Chakraborty, A. 2020. Plant microbiome–an account of the factors that shape community composition and diversity. Current Plant Biology. 23:100161. de Vienne, D. M. 2016. Lifemap: Exploring the Entire Tree of Life. PLos Biol. 14(12):e2001624.

Ding, G.-C., Radl, V., Schloter-Hai, B., Jechalke, S., Heuer, H., Smalla, K., et al. 2014. Dynamics of Soil Bacterial Communities in Response to Repeated Application of Manure Containing Sulfadiazine ed. Gabriele Berg. PLoS ONE. 9:e92958.

Erban, T., Klimov, P., Molva, V., and Hubert, J. 2020. Whole genomic sequencing and sex- dependent abundance estimation of Cardinium sp., a common and hyperabundant bacterial endosymbiont of the American house dust mite, Dermatophagoides farinae. Exp Appl Acarol. 80:363–380.

Evtushenko, L.I. 2015. Microbacteriaceae. In Bergey's Manual of Systematics of Archaea and Bacteria (eds M.E. Trujillo, S. Dedysh, P. DeVos, B. Hedlund, P. Kämpfer, F.A. Rainey and W.B. Whitman). doi:10.1002/9781118960608.fbm00035

Fitzpatrick, D., and Walsh, F. 2016. Antibiotic resistance genes across a wide variety of metagenomes ed. Kornelia Smalla. FEMS Microbiology Ecology. 92:fiv168.

Garrity, G.M., Bell, J.A. and Lilburn, T. 2015. Burkholderiaceae fam. nov. In Bergey's Manual of Systematics of Archaea and Bacteria (eds M.E. Trujillo, S. Dedysh, P. DeVos, B. Hedlund, P. Kämpfer, F.A. Rainey and W.B. Whitman). doi:10.1002/9781118960608.fbm00181

Glaeser, S. P. and Kӓmpfer, P. 2014. The Family Burkholderiaceae in The Prokaryotes 4th Edition (ed. Rosenberg, E.). New York: Springer, pp. 641-707.

Gonella, E., Pajoro, M., Marzorati, M., Crotti, E., Mandrioli, M., Pontini, M., et al. 2015. Plant- mediated interspecific horizontal transmission of an intracellular symbiont in insects. Sci Rep. 5:15811.

Granatstein, D., Johnson, K., Smith, T., and Elkins, R. Implementation of non-antibiotic programs for fire blight control in organic apple and pear in the western United States. Poster.

Gu, G., Yin, H.-B., Ottesen, A., Bolten, S., Patel, J., Rideout, S., et al. 2019. Microbiomes in Ground Water and Alternative Irrigation Water, and Spinach Microbiomes Impacted by Irrigation with Different Types of Water. Phytobiomes Journal. 3:137-147.

Hardoim, P. R., van Overbeek, L. S., Berg, G., Pirttilä, A. M., Compant, S., Campisano, A., et al. 2015. The Hidden World within Plants: Ecological and Evolutionary Considerations for Defining Functioning of Microbial Endophytes. Microbiology and Molecular Biology Reviews. 79:293- 320.

Hawkes, C. V., and Connor, E. W. 2017. Translating Phytobiomes from Theory to Practice: Ecological and Evolutionary Considerations. Phytobiomes Journal. 1:57-69.

Ikenaga, M., and Sakai, M. 2014. Application of Locked Nucleic Acid (LNA) Oligonucleotide– PCR Clamping Technique to Selectively PCR Amplify the SSU rRNA Genes of Bacteria in Investigating the Plant-Associated Community Structures. Microbes and Environments. 29:286- 295.

Ikenaga, M., Tabuchi, M., Oyama, T., Akagi, I., and Sakai, M. 2015. Development of LNA oligonucleotide–PCR clamping technique in investigating the community structures of plant- associated bacteria. Bioscience, Biotechnology, and Biochemistry. 79:1556-1566.

Kuever, J., Rainey, F.A. and Widdel, F. 2015. Desulfovibrionaceae fam. nov. In Bergey's Manual of Systematics of Archaea and Bacteria (eds M.E. Trujillo, S. Dedysh, P. DeVos, B. Hedlund, P. Kämpfer, F.A. Rainey and W.B. Whitman). doi:10.1002/9781118960608.fbm00199

Johnson, K. B., and Temple, T. N. 2013. Evaluation of Strategies for Fire Blight Control in Organic Pome Fruit Without Antibiotics. Plant Disease. 97:402-409.

Liu F., Ying G. G., Tao R., Zhao J. L., Yang J. F., Zhao L. F. 2009. Effects of six selected antibiotics on plant growth and soil microbial and enzymatic activities. Environ. Pollut. 157:1636-1642.

Kandel, S., Joubert, P., and Doty, S. 2017. Bacterial Endophyte Colonization and Distribution within Plants. Microorganisms. 5:77.

Lin, H., Jin, D., Freitag, T. E., Sun, W., Yu, Q., Fu, J., et al. 2016. A compositional shift in the soil microbiome induced by tetracycline, sulfamonomethoxine and ciprofloxacin entering a plant-soil system. Environmental Pollution. 212:440-448.

Lindow, S. E., and Brandl, M. T. 2003. Microbiology of the Phyllosphere. Applied and Environmental Microbiology. 69:1875-1883.

Liu, J., Abdelfattah, A., Norelli, J., Burchard, E., Schena, L., Droby, S., et al. 2018. Apple endophytic microbiota of different rootstock/scion combinations suggests a genotype-specific influence. Microbiome. 6(18):1-11.

Lu, J., Breitwieser, F. P., Thielen, P., and Salzberg, S. L. 2017. Bracken: estimating species abundance in metagenomics data. PeerJ Computer Science. 3:e104.

Lundberg, D. S., Yourstone, S., Mieczkowski, P., Jones, C. D., and Dangl, J. L. 2013. Practical innovations for high-throughput amplicon sequencing. Nature Methods. 10:999-1002.

Manceau, C., Lalande, J., Lachaud, G., Chartier, R., and Paulin, J. 1990. Acta Hort. 273:73-82.

Manulis, S., Zutra, D., Kleitman, F., Dror, O., David, I., Zilberstaine, M., et al. 1998. Distribution of streptomycin-resistant strains of Erwinia amylovora in Israel and occurrence of blossom blight in the autumn. Phytoparasitica. 26:223-230.

Marin, I. and Arahal, D. 2014. The Family Rhizobiaceae in The Prokaryotes (ed. Rosenberg, E.). New York: Springer, pp. 115-133.

McGhee, G. C., Guasco, J., Bellomo, L. M., Blumer-Schuette, S. E., Shane, W. W., Irish-Brown, A., et al. 2011. Genetic Analysis of Streptomycin-Resistant (SmR) Strains of Erwinia amylovora Suggests that Dissemination of Two Genotypes Is Responsible for the Current Distribution of SmR E. amylovora in Michigan. Phytopathology. 101:182–191.

McManus, P. S., Stockwell, V. O., Sundin, G. W., and Jones, A. L. 2002. Antibiotic Use in Agriculture. Annual Review of Phytopathology. 40:443-465.

Moller, W. J., Beutel, J. A., Reil, W. O., and Zoller, B. G. 1972. Fire blight resistance to streptomycin in California. In Abstracts of Papers Accepted for Presentation at the Sixty-Fourth Annual Meeting of The American Phytopathological Society, Mexico City, Mexico: Phytopathology, p. 779.

Momol, M. T., Norelli, J. L., Piccioni, D. E., Momol, E. A., Gustafson, H. L., Cummins, J. N., et al. 1998. Internal Movement of Erwinia amylovora Through Symptomless Apple Scion Tissues into the Rootstock. Plant Disease. 82:646-650.

Muresan, L. E. Cultivable Bacterial and Fungal Endophytes from Apple Tissues and Their Potential for Biological Control of Venturia inaequalis. Master’s thesis: The University of Guelph.

Nakamura, Y., Kawai, S., Yukuhiro, F., Ito, S., Gotoh, T., Kisimoto, R., et al. 2009. Prevalence of Cardinium Bacteria in Planthoppers and Spider Mites and Taxonomic Revision of “Candidatus Cardinium hertigii” Based on Detection of a New Cardinium Group from Biting Midges. AEM. 75:6757-6763.

Norelli, J. L. 2003. Fire blight management in the twenty first century. Plant Disease. 87:756- 765.

Ockey, S. C., and Thomson, S. V. 2006. Influence of rain on the transient populations of Erwinia amylovora on leaf surfaces. Acta Hortic. 704:113-120.

Oksanen, J., Blanchet, F. G., Friendly, M., Kindt, R., Legendre, P., McGlinn, P., Minchin, P. R., O'Hara, R. B., Simpson, G. L., Solymos, P., Stevens, M. H., Szoecs, E., and Wagner, H. 2019. vegan: Community Ecology Package. R package version 2.5-6. https://CRAN.R- project.org/package=vegan.

Ottesen, A. R., White, J. R., Skaltsas, D. N., Newell, M. J., and Walsh, C. S. 2009. Impact of Organic and Conventional Management on the Phyllosphere Microbial Ecology of an Apple Crop. Journal of Food Protection. 72:2321-2325.

Overbeek, L. S., Wellington, E. M. H., Egan, S., Smalla, K., Heuer, H., Collard, J.-M., et al. 2002. Prevalence of streptomycin-resistance genes in bacterial populations in European habitats. FEMS Microbiology Ecology. 42:277-288.

Pruesse, E., Quast, C., Knittel, K., Fuchs, B. M., Ludwig, W., Peplies, J., et al. 2007. SILVA: a comprehensive online resource for quality checked and aligned ribosomal RNA sequence data compatible with ARB. Nucleic Acids Research. 35:7188-7196.

Reinhold-Hurek, B., and Hurek, T. 2011. Living inside plants: bacterial endophytes. Current Opinion in Plant Biology. 14:435-443.

Rosen, H. R. 1929. The life history of the fire blight pathogen, Bacilus amylovorus, as related to the means of overwintering and dissemination. Ark. Agric. Exp. Stn. Bull. 244.

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 Disease. 92:714-718.

Ryan, R. P., Germaine, K., Franks, A., Ryan, D. J., and Dowling, D. N. 2008. Bacterial endophytes: recent developments and applications. FEMS Microbiology Letters. 278:1-9.

Shade, A., Klimowicz, A. K., Spear, R. N., Linske, M., Donato, J. J., Hogan, C. S., et al. 2013. Streptomycin Application Has No Detectable Effect on Bacterial Community Structure in Apple Orchard Soil. Applied and Environmental Microbiology. 79:6617-6625.

Shade, A., McManus, P. S., and Handelsman, J. 2013. Unexpected Diversity during Community Succession in the Apple Flower Microbiome. mBio. 4:10.1128/mBio.00602-12.

Steven, B., Huntley, R. B., and Zeng, Q. 2018. The Influence of Flower Anatomy and Apple Cultivar on the Apple Flower Phytobiome. Phytobiomes Journal. 2:171-179.

Stockwell, V. O., and Duffy, B. 2012. Use of antibiotics in plant agriculture. Scientific and Technical Review of the Office International des Epizooties (Paris). 31:199–210.

Sundin, G. W., and Wang, N. 2018. Antibiotic Resistance in Plant-Pathogenic Bacteria. Annu. Rev. Phytopathol. 56:161-180.

Sundin, G. W., Werner, N. A., Yoder, K. S., and Aldwinckle, H. S. 2009. Field Evaluation of Biological Control of Fire Blight in the Eastern United States. Plant Disease. 93:386-394.

Tancos, K. A., and Cox, K. D. 2017. Effects of Consecutive Streptomycin and Kasugamycin Applications on Epiphytic Bacteria in the Apple Phyllosphere. Plant Disease. 101:158-164.

Tancos, K. A., and Cox, K. D. 2016. Exploring Diversity and Origins of Streptomycin-Resistant Erwinia amylovora Isolates in New York Through CRISPR Spacer Arrays. Plant Disease. 100:1307-1313.

Thomson, S. V., Gouk, S. C., Vanneste, J. L., Hale, C. N., and Clark, R. G. 1993. The presence of streptomycin resistant strains of Erwinia amylovora in New Zealand. Acta Horticulturae. 338:223-230.

Tolba, S., Egan, S., Kallifidas, D., and Wellington, E. M. H. 2002. Distribution of streptomycin resistance and biosynthesis genes in streptomycetes recovered from different soil sites. FEMS Microbiology Ecology. 42:269-276.

Turner, T. R., James, E. K., and Poole, P. S. 2013. The plant microbiome. Genome Biol. 14:209. van der Zwet, T. ed. 2012. Chapter 10: The Disease Cycle of Fire Blight in Fire Blight: History, Biology, and Management. St. Paul, MN: APS Press. van der Zwet, T. and Buskirk, P. D. 1984. Detection of endophytic and epiphytic Erwinia amylovora in various pears and apple tissues. Acta. Hort. 151:69-77.

Walters, W., Hyde, E. R., Berg-Lyons, D., Ackermann, G., Humphrey, G., Parada, A., et al. 2016. Improved Bacterial 16S rRNA Gene (V4 and V4-5) and Fungal Internal Transcribed Spacer Marker Gene Primers for Microbial Community Surveys ed. Holly Bik. mSystems. 1:e00009-15, /msys/1/1/e00009-15.atom.

Wasimuddin, Schlaeppi, K., Ronchi, F., Leib, S. L., Erb, M., and Ramette, A. 2019. Evaluation of primer pairs for microbiome profiling across a food chain from soils to humans within the One Health framework. Microbiology. doi: https://doi.org/10.1101/2020.02.19.956078.

Wickham. H. 2016. ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag: New York.

Wood, D. E., and Salzberg, S. L. 2014. Kraken: ultrafast metagenomic sequence classification using exact alignments. Genome Biol. 15:R46.

Yashiro, E., and McManus, P. S. 2012. Effect of Streptomycin Treatment on Bacterial Community Structure in the Apple Phyllosphere ed. Jack Anthony Gilbert. PLoS ONE. 7:e37131.

Yashiro, E., Spear, R. N., and McManus, P. S. 2011. Culture-dependent and culture-independent assessment of bacteria in the apple phyllosphere: Apple phyllosphere bacteria. Journal of Applied Microbiology. 110:1284-1296.

PART II:

INVESTIGATING DISTRIBUTION AND SPREAD OF FIRE BLIGHT

AT MULTIPLE SCALES

CHAPTER 4

ASSESSING AND MINIMIZING THE DEVELOPMENT AND SPREAD OF FIRE BLIGHT

FOLLOWING MECHANICAL THINNING AND MECHANICAL PRUNING

IN APPLE ORCHARDS*

Abstract

The adoption of mechanical thinning and pruning in commercial apple orchards has largely been limited by the risk of development and spread of fire blight. This devastating disease, caused by the bacterial pathogen Erwinia amylovora, may be transmitted by mechanical injury such as pruning, especially under warm, moist conditions conducive to bacterial growth, infection, and disease development. However, risk may be mitigated by avoiding highest risk times and applying a bactericide, such as streptomycin, following mechanical thinning or pruning. In ‘Gala’ and ‘Idared’ orchards, we evaluated the risk of fire blight development and spread following mechanical thinning early in bloom (20% bloom), when seasonal temperatures are cooler and there are few open flowers available for infection. In both orchards, we also evaluated the spread and development of fire blight by mechanical pruning in July and in

August, before and after terminal bud set when shoot growth is slowed and less susceptible to infection. We also assessed the potential efficacy of a streptomycin or Bacillus subtilis biopesticide application following mechanical thinning and pruning to mitigate the spread of fire blight. In the ‘Gala’ orchard, disease never developed beyond the inoculated tree following thinning or pruning, which was unexpected for this highly susceptible cultivar. In the ‘Idared’ orchard, incidence of blossom or shoot blight from the point source, represented as relative area under the disease progress curve (rAUDPC) was rarely different for trees that received

mechanical thinning or mechanical pruning compared to untreated trees, and was frequently eliminated or reduced when the antibiotic streptomycin or the B. subtilis biopesticide was applied within 24 hrs of mechanical thinning or pruning. For both thinning and pruning, incidence of fire blight dropped off quickly beyond the inoculated tree in the ‘Idared’ orchard and generally was not observed in trees beyond 10-15 m from the inoculated point source or predicted beyond 10 m by exponential and power law models fit to the disease progress curves. The results of this work demonstrate the low risk for fire blight development and spread by mechanical thinning and pruning when practiced under low-risk conditions—early in bloom for mechanical thinning, and after terminal bud set (in August) for mechanical pruning—especially when paired with a subsequent bactericide application. This study demonstrates the safe use of mechanical thinning and pruning in commercial apple production, corroborated by anecdotal evidence from apple growers in Western New York State.

*Wallis, A. and Cox, K. 2020. Assessing and minimizing the risk of fire blight following mechanical thinning and mechanical pruning in apple orchards. Plant Disease. First Look available online: https://doi.org/10.1094/PDIS-06-20-1324-RE.

Introduction

The adoption of high-density apple orchards in recent years has been driven by the potential for these systems to dramatically increase yields and fruit quality (Lordan,

Francescatto, et al. 2018; Lordan et al. 2019; Robinson et al. 2007). However, modern orchards are limited in their ability to realize higher return on investment by their high labor requirement, greater upfront investment, and the availability of practices and materials to complete highly time-sensitive tasks (He and Schupp 2018; Lordan et al. 2018b, 2019; Robinson et al. 2007).

Therefore, the success of modern orchards is dependent on incorporating labor-efficient practices, such as mechanization.

One area of interest for mechanization in modern apple production is thinning. Fruit thinning is essential to managing fruit size and yield. It is typically accomplished by applying chemicals from bloom to 25 mm in fruit size, followed by extensive hand-thinning. Chemical thinners include lime sulfur or plant hormones such as carbaryl (a carbamate insecticide), benzyl adenine (BA; a synthetic cytokinin), or naphtaleneacetic acid (NAA; a synthetic auxin) alone or in combination (Cline et al. 2019; Greene and Costa 2013; Petracek et al. 2003; Sazo et al.

2016). However, chemical thinning is problematic due to frequently erratic results related to weather conditions, cultivar, phytotoxicity for bloom applications of caustic materials in humid climates, and restrictions on many of these products related to concerns about food safety and environmental protection (Cline et al. 2019; Green and Costa 2013; Lordan 2018a). Carbaryl was completely banned in Europe in 2008; in the U.S., its use as an insecticide was recently removed from the label and its use exemption for fruit thinning is likely to expire in coming years (He and

Schupp 2018). Additional products currently registered for thinning include 1-

aminocyclopropane carboxylic acid, metamitron, and abscisic acid, but their efficacy and availability is still limited (Gonzalez et al. 2019; Lordan et al. 2018a).

Aside from chemical options, there is growing interest in using mechanical string thinners in the production of apple and other tree fruits. These machines have been used successfully in peach production, in which chemical thinning options are extremely limited

(Asteggiano et al. 2015; Baugher et al. 2010; Schupp and Baugher 2011), and more recently in apple production (He and Schupp 2018; Kon et al. 2013; Solomakhin and Blanke 2010) to reduce labor, improve crop load and quality, and act as a more environmentally friendly solution to common chemical thinners.

Another area of interest for mechanization is pruning to shape tree architecture and optimize crop load. In apple production, canopy management is practiced at two growth stages annually, during tree dormancy in late winter and during fruit sizing in mid-July to August, and is essential to maintaining tree size and architecture, vegetative and reproductive balance, and crop quality (He and Schupp 2018). Pruning has the highest labor requirements of orchard management practices, aside from harvest (He and Schupp 2018; Robinson et al. 2007). High density systems have the unique potential to be maintained as or transitioned to fruiting wall systems using mechanical pruning, with benefits including improved light interception, crop quality, and labor efficiency (He and Schupp 2018; Robinson et al. 2007, 2013; Sansavini 1978;

Sazo 2018).

A significant barrier to the adoption of mechanical treatments is the potential to spread fire blight (Kon et al. 2013; Ngugi and Schupp 2009). This disease, caused by the bacterial pathogen Erwinia amylovora, is one of the most economically devastating diseases of apple worldwide. The highest risk for infection is during warm, wet weather, conducive to pathogen

growth (van der Zwet, T. 2012), which commonly occurs in spring and early summer. Infections are initiated by primary inoculum in the form of bacterial ooze produced in the spring from overwintering cankers, which is vectored to open flowers by wind and rain. Bacteria preferentially colonize stigmatic surfaces and are washed into the plant via natural openings in the nectaries during a wetting event. An infection leads to a complete necrotic blight of the flower cluster and pedicles, referred to as blossom blight. Bacteria form cankers and infect blossoms, and are later vectored to young developing shoots by wind-driven rain, insects, and equipment. The bacteria infect the shoot tissue through wounds caused by mechanical injury, leading to a spreading necrotic blight referred to as shoot blight. Shoot blight infection can lead to a systemic infection of the entire tree as bacteria travel through the vascular system toward the actively growing parts of the plant and can lead to complete tree mortality. These systemic infections can produce additional ooze and facilitate rapid spread within an orchard leading to severe losses within a season and enormous impact for the life of the orchard. As mentioned, a critical condition for infection is that the pathogen requires a natural opening or damaged tissue to enter the host. The use of machinery at bloom or during terminal shoot elongation could lead to wounding and greatly increase the risk of fire blight. Mechanical thinning and pruning introduce a considerable number of wounds in the canopy during the period of bloom through shoot elongation, raising obvious concerns about the potential for spreading fire blight.

Few studies have evaluated the implications of mechanical thinning and pruning on fire blight spread in orchards. Ngugi and Schupp (2009) found that using a mechanical string thinner on healthy trees immediately after using it on inoculated trees resulted in high levels of blossom blight. However, in this study, conditions were especially conducive to fire blight infection, including artificially high levels of inoculum, greater than 100 times that used in typical field

research trials (Cox et al. 2016; Outwater and Sundin 2016; Yoder and Kowalski 2016) and likely much higher than typically present in grower orchards. In addition, trials were conducted during full bloom, with continuous precipitation and condensation under warm temperatures

(>18°C) prior to and following the thinning procedure. Moreover, inoculated trees were exhibiting early signs of fire blight, conditions under which growers would not typically apply mechanical thinning. The unrealistic conditions of this study and the lack of any other studies in this area suggest further study is needed.

Several steps may be taken by commercial growers to mitigate the risk of fire blight spread when using mechanical thinning and pruning equipment. Thinning may be conducted earlier in bloom when seasonal temperatures are cooler and there are few open blossoms available for infection by E. amylovora. In addition, an application of streptomycin after thinning was suggested by Ngugi and Schupp (2009) as a possible means for reducing risk. Anecdotal reports from growers in NY indicate that mechanical thinning and pruning is being used cautiously in commercial production without any perceived increased risk of fire blight infection, and an application of streptomycin following mechanical treatment is often being used to reduce possible risk (Mario Miranda-Sazo, personal communication). However, the risk of fire blight development and spread following mechanical thinning in early bloom with applications of streptomycin has not been evaluated in controlled experiments.

The objectives of this work were to determine the risk of fire blight development and spread following 1) mechanical string thinning and 2) mechanical pruning under field conditions reflecting the practices of commercial apple growers. Mechanical thinning and pruning were evaluated both with and without bactericide application following treatments. We hypothesized that mechanical thinning and pruning would increase risk of both blossom and shoot blight in

terms of overall disease development and spread along an orchard row, but that bactericide applications would reduce or eliminate this risk. Establishing the level of risk for fire blight infection associated with mechanical thinning and pruning practices and identifying practices effective for mitigating any apparent risk would enable informed recommendations of these valuable labor- and cost-saving practices in commercial orchards and for commercial orchards, especially productions adopting high-density systems.

Materials and Methods

Orchard Sites

Field trials were conducted at Cornell AgriTech in research orchards in Geneva, NY

(42°87’70.19”N -77°02’99.15”W) in 2017 and 2018. Experiments were performed at two orchards, both consisting of semi-dwarf apples planted at approximately 5 m in-row spacing, established in 2000. Tree canopies were approximately conical in shape with two tiers of rigid permanent scaffolds approximately 2 m in diameter. Cultivars used in the experiments were

‘Gala’ at the first orchard and ‘Idared’ at the second orchard, both on B.9 rootstock. Trees were maintained according to commercial standards (Agnello et al. 2019).

Establishing point sources and experimental design

In order to better understand the risk of fire blight infection and spread with mechanical thinning and pruning in apple orchards, we established trials using mechanical thinning or pruning in replicated rows. Replications consisted of tree rows with a minimum of 11 trees and minimum total length of 50 m. Distance was measured from the center of the first tree to the center of each tree in the row, which was approximately every five meters. The first tree in each

row was inoculated at king bloom (approximately 20% blossoms on trees open) to serve as a point source of inoculum and reflect the presence of a recent quiescent infection at the time of the commercial mechanical thinning window of early bloom. Inoculation was performed on 1

May and 15 May in 2017 and 2018, respectively, one day (approximately 24 hours) prior to mechanical thinning. Inoculations were carried out using E. amylovora strain Ea273, a highly aggressive strain originally isolated from trees in New York State and routinely used in lab and field trials (Cox et al. 2016). Cultures were stored in 50% glycerol at -80°C prior to use, streaked on Luria-Bertani (LB) agar, and incubated at 28°C for 3 days. Cultures were then re-suspended in LB broth overnight while shaken at 28°C, and suspensions were diluted to 1x106 CFU ml-1 in phosphate buffered saline immediately prior to inoculation. Inoculum was applied using a hand- pumped Solo 475-B backpack sprayer (Solo Incorporated, Newport, VA) at king bloom.

Evaluating the risk of fire blight development and spread following mechanical thinning

In order to assess the risk of fire blight infection and spread associated with mechanical thinning, mechanical string thinning was performed using a Darwin mechanical string thinner

(Darwin PT-250; Fruit-Tec, Deggenhauserertal, Germany). Thinning was conducted at king bloom (20% bloom) to represent thinning timing recommended for commercial apple growers, on 2 May and 16 May in 2018 and 2019, respectively. The machine consisted of a tractor- mounted frame with a 10 ft tall vertical spindle in the center of the frame. Attached to the spindle were 54 steel plates securing a total of 216 plastic cords, each measuring 60 cm in length. Speed of the clockwise rotating spindle was adjusted with a hydraulic motor. Forward tractor speed was maintained at 8.0 km h-1 and spindle speed at 220rpm. A detailed description of the machine can be found in Schupp et al. (2008). Treatments included 1) mechanical string thinning with no

subsequent disease control, 2) mechanical string thinning followed by an application of the antibiotic streptomycin, 3) mechanical string thinning followed by an application of a Bacillus subtilis biopesticide, a program appropriate for organic growers, and 4) untreated trees, which were inoculated but were not mechanically thinned nor did they receive an application of a bactericidal product, to serve as a control for natural disease development and spread in the plots.

Treatment scheme is diagrammed in Figure 4.1. Each treatment was replicated in three rows at the ‘Idared’ orchard, except for treatment 4 (untreated trees), which was replicated twice. At the

‘Gala’ orchard, treatments were replicated in two rows. These bactericides were applied approximately 24 hours after mechanical thinning, to represent the post-thinning management that could be practiced by commercial apple growers to reduce populations of E. amylovora and subsequent risk of fire blight development. Commercial products were applied at the full field rates for apples as indicated by the labels: FireWall 17 WP (AgroSource, Mountainside, NJ) was applied at 1.68 kg ha-1 to represent the streptomycin treatment, and Serenade Opti (Bayer Crop

Science, Institute, WV) was applied at 1.40 kg ha-1 to represent the biopesticide treatment. Both bactericides were applied with the non-ionic surfactant Regulaid (KALO, Overland Park, KS) at

3.18 L ha-1, using a Solo 451 gas-powered mist blower (Solo Incorporated, Newport News, VA) calibrated to deliver approximately 935.3 L ha-1.

Figure 4.1. Schematic of treatments used to evaluate the risk of fire blight development and spread with mechanical thinning and pruning in apple orchards. Treatments were applied to tree rows of semi-dwarf ‘Gala’ and ‘Idared’ apple trees at 5 m within-row spacing, a minimum of 50 m or 11 trees in length. 1. At king bloom (20% bloom) the fire tree in the row was inoculated with Erwinia amylovora strain Ea273 at 106 CFU ml- to serve as a point source. 2. Approximately 24 h following inoculation, rows were subject to mechanical thinning, pruning, or neither for blossom blight; for shoot blight, natural inoculum was used in the form of infected blossoms in the first (inoculated) tree in each row. 3. Approximately 24 h later, rows received a treatment of Firewall 17 (streptomycin; ‘Strep’), Serenade Opt (Bacillus subtilis Biopesticide; ‘Bio’), or neither.

Evaluating the risk of fire blight development and spread following mechanical pruning

In order to assess the risk of fire blight development and spread by mechanical pruning, mechanical pruning was performed using a Lagasse Orchard Hedger (LaGasse Machine and

Fabrication, Lyons, NY) mounted to a John Deere 5525 tractor, equipped with a 3.35 m nominal cutter bar capable of hydraulic tilting and lateral movement. The bar was equipped with 44 knives 7.6 cm in length, vee-shaped with tapered and serrated edges, bolted edge-to-edge onto the knife bar, and capable of moving 7.6 cm in both directions along bar. Tractor speed was maintained at 8 k hr-1 and the bar was vertically positioned (90 degree angle) approximately 51-

61 cm from trees. Two pruning trials were conducted in each season at times suggested by Sazo

(2018) for thinning of smaller and larger fruited cultivars: once in mid-July (16 July in both years), during vegetative tree growth, and again in mid-August (7 and 8 August in 2017 and

2018, respectively), following terminal bud set after most vegetative growth had concluded for

the season. The early and late pruning trial dates reflect the approximate times mechanical pruning is typically practiced commercially, for small and large fruited cultivars, respectively.

Treatments included 1) pruning without subsequent chemical control, 2) pruning followed by application of the antibiotic streptomycin, 3) pruning followed by a Bacillus subtilis biopesticide, and 4) untreated trees, which were not mechanical pruned nor did they receive an application of a bactericidal product, to serve as a control for natural disease development and spread in the plots. All treatments were applied as described in ‘Objective 1. Thinning’ above. Infections from blossom inoculation were used as inoculum for shoot blight and trees were not artificially inoculated prior to mechanical pruning. Shoot blight strikes from thinning trials were pruned out of all but the inoculated trees (first tree in the row) prior to pruning in order to prevent confounding shoot blight data from thinning and pruning trials. In point source trees, strikes were removed so that 10 strikes were present in each tree, in order to equalize inoculum in each row.

Environmental conditions were monitored using field units located adjacent to the study sites, approximately 50 m from the orchards, and weather data sets accessed using the Network for

Environment and Weather Applications (NEWA, http://newa.cornell.edu/; Carroll et al. 2017).

Weather data included temperature (daily mean, high, and low), precipitation, leaf wetness, and relatively humidity (number of hours greater than 90%).

Disease Assessments and Data Analysis

Disease development and spread was evaluated in terms of incidence of blossom and shoot blight for thinning trials and shoot blight for pruning trials. The incidence of blossom blight and shoot blight symptoms was assessed for every tree in the replicated treatment rows as soon as symptoms were reliably detectable. Both blossom blight and shoot blight were assessed

for thinning trials. Shoot blight symptom assessments were conducted following each pruning trial (mid-July or early August treatment). For each tree in the treatment row, the incidence of blossom blight was assessed by determining the number of blighted blossoms out of five (5) in a cluster with 20 cluster assessments for each tree. Shoot blight was assessed for each tree in the treatment rows, expressed as the percentage of terminal shoots with discoloration or ooze out of the total number of shoots on a tree for thinning trials or out of the total number of shoots that were cut by the hedger for pruning trials.

To determine the overall risk of fire blight development with mechanical thinning and pruning, the area under the disease progress curve (AUDPC) was determined for each treatment replication in each year, as described by Madden et al. (2007). In addition, these models were produced in order to make predictions about the incidence and spread of the disease from point sources of inoculum. The AUDPC was estimated as follows:

푛−1 푦 + 푦 AUDPC = ∑ ( 푖 푖−1) (푡 − 푡 ) 2 푖+1 푖 푖 in which n is the number of assessment times or distances, y is the disease intensity, and t is the time or distance from the original measurement. In the present study, n was between 10 and 11 trees, based on how many trees were in each row, y was the intensity of disease (percent incidence), and t was the distance from the point source. Because rows differed slightly in number of trees (tree loss following development of fire blight) and row length (+/- 3 m), relative

AUDPC (rAUDPC) was calculated by dividing by the total distance for each row. Mean rAUDPCs in each trial (blossom blight and shoot blight for thinning trials; shoot blight in July and shoot blight in August for pruning trials) were subjected to one-way analysis of variance

(ANOVA) for a randomized block design using the aov package (Chambers 1992). Means were

separated using the Tukey HSD test at the α = 0.05 level of significance. All data analysis was conducted using R ver. 3.6.3 (R Core Team, 2020).

In order to determine patterns and make predictions of fire blight spread by mechanical thinning or pruning from a point source, two models were fit to disease incidence data for thinning and pruning trials in each year. The first model, Gregory’s power law model (Bock

2005; Gregory 1968) can be described as follows:

y = a*x(b) and the second model, an exponential decay model or Kiyosawa and Shiyomi model (Kiyosawa and Shiyomi 1972; Mundt 1999), is described as follows:

y = a*e(-b*x) in which, for both models, y is the intensity of disease at a given distance from the point source, a is the intensity of disease at the point source, b is the slope of the gradient or the rate parameter, e is the base of natural logarithms, and x is a distance from the point source.

Models were fit to disease intensity in the form of incidence of blossom or shoot blight

(y) as a function of distance in meters from the inoculated first tree of each row which served as point source (x) of inoculum. Models were fit to the average disease incidences at each distance for the replications of a given treatment, resulting in a total of four curves in each year (blossom blight and shoot blight for thinning trials, shoot blight in July and shoot blight in August for pruning trials). This was done because fitting a curve for each replication of each treatment in each experiment greatly complicated interpretation of the model fit and comparison of slope values. Coefficients of determination, standard errors of the estimate, significances of slopes, and residual patterns were used to evaluate the model fit. All model fitting was conducted using

SigmaPlot software (version 11; SPSS Science Inc., Chicago, IL).

Results

Conditions and overall disease incidence

At the time of inoculation and mechanical thinning average temperatures were 16.4°C and 16.2°C, and high temperatures were 21.1 and 20.8°C in 2017 and 2018, respectively, with wind speed less than 9 km h-1. In both years at least one rain event occurred within the three days following inoculation (May 1 to 4 2017; and May 15 to 18 2018), in which 2.64 and 1.47 cm of rain accumulated, and 19 and 14 hours of leaf wetness and 12 to 31 hours of relative humidity

>90% occurred in respective years. Weather conditions at the time of mechanical pruning in both years were similar. The average and maximum temperatures the day of and three days following mechanical pruning ranged from 20.3 to 22.2°C and 26.6 to 33.2°C, respectively. With the exception of August pruning in 2017, a rain event occurred during the four days following pruning with 0.9-1.1 cm of accumulation. In the four days, following pruning in each year, between 8-19 h of leaf wetness and 13-33 h of relative humidity over 90% occurred. These conditions were more than sufficient for infection and to facilitate disease development in each trial.

At the ‘Gala’ orchard, symptoms only developed in untreated trees (trees that were inoculated, but neither received mechanical thinning or pruning, nor an application of streptomycin or a biopesticide). In these trees, the incidence of blossom blight and shoot blight ranged from 31.0 to 35.0% and 14.9 to 20.2% incidence, respectively in 2017, and ranged from

22.0 to 24.0% and 8.6 to 13.4% incidence, respectively in 2018. Symptoms of blossom blight and shoot blight never developed beyond the inoculated tree, rendering rAUDPC analyses and outcomes of model fitting between treatments no different than these observations (data not

shown). Similarly, in pruning trials conducted in July, shoot blight developed only in untreated trees (neither pruning nor bactericides), but not beyond the inoculated tree in these rows. The incidence of shoot blight in July trials ranged from 6.6 to 10.5% and 3.2% to 7.7% incidence in

2017 and 2018, respectively. Shoot blight did not develop in any trees that were mechanically pruned in August in either year. As a result, further data analyses were conducted only on data collected from the ‘Idared’ orchard.

Risk of fire blight infection and spread following mechanical thinning

In the ‘Idared’ orchard, disease symptoms developed in all inoculated trees, but for some years and tree rows, disease symptoms did not develop beyond this inoculated point source

(Tables 4.1 and 4.2). In thinning trials in 2017, all replicates of the mechanical thinning treatment developed both blossom and shoot blight beyond the inoculated trees. However, in other treatments (untreated trees or mechanical thinning followed by streptomycin or the biopesticide), symptoms only developed beyond the inoculated tree in two tree rows: one row of untreated trees and one row of trees that received a streptomycin application after thinning. In

2018, however, both blossom and shoot blight developed beyond the first tree in all tree rows except for one row receiving streptomycin after thinning.

Overall development of fire blight, expressed as relative area under the disease progress curve (rAUDPC), was frequently affected by thinning treatments (P < 0.05, Table 4.3). Blossom blight and shoot blight was often highest for trees that were thinned but did not receive subsequent application of streptomycin or the biopesticide, as compared to all other treatments.

The rAUDPC for thinned trees without subsequent bactericide and rAUDPC for untreated trees was rarely statistically different, although untreated trees tended to have slightly lower overall

disease development. Blossom and shoot blight were almost always significantly lower in trees receiving either streptomycin or the biopesticide application than trees thinned alone. rAUDPC tended to be the lowest in trees that received a streptomycin application following thinning.

Disease spread usually only extended a short distance beyond inoculated trees, with symptoms of blossom and shoot blight typically observed in trees adjacent to the inoculated tree and the subsequent 2-3 trees, or approximately 10-15 m beyond the point source of inoculum

(Tables 4.2 and 4.3). Therefore, we compared disease development at 10 m from the inoculated point source, as an indicator of the likelihood of disease spread. At 10m, blossom and shoot blight ranged from 0 to 31.3 and 0 to 11.0% incidence, in 2017 and 2018, respectively. Similar to overall disease development (rAUDPC), the highest incidence was observed for trees that were thinned without subsequent streptomycin or biopesticide application, with 20.0 and 1.9% incidence of blossom blight and 31.1 and 11.0% incidence in 2017 (P = 0.0004 and 0.0314) and

2018 (P = 0.0167 and 0.0810), respectively. For trees that were thinned followed by a streptomycin application, incidence of blossom and shoot blight was always lowest, ranging from 0 to 1.2%.

Occasionally, disease symptoms were not observed in the tree adjacent to the inoculated point source but were observed in the last 1-3 trees or 10-15 m from the end of a row. This was unlikely due to spread from the inoculated point source, and more likely due to an alternate source of inoculum. There was an adjacent, unmanaged orchard of ‘Idared’ that was devastated by fire blight in 2015 and remained for the 2017 and 2018 seasons. Trees in this orchard had oozing cankers and produced shoots that developed fire blight in 2017 and 2018. This orchard’s proximity to the study orchard may have allowed this to serve as a source of inoculum at the far end of the orchard rows.

Models were used to further describe the spread of disease with thinning treatments. The fit of the Gregory power law model and exponential decay model was similar for all treatments in both years, as indicated by similar values for coefficient of variance (R2 value) and standard error of the estimate (SEE) for the two models fit to a given data set (Table 4.5 and Figure 4.2).

Model fit was significant at P < 0.05 in all cases. Coefficients of variance indicated excellent fit, with values always greater than 0.75, and greater than or equal to 0.9 in all except 3 cases. In very few instances, the coefficient of variance indicated a perfect or near perfect relationship (R2

= 1), reflecting that disease did not develop beyond the inoculated tree (for example, blossom blight incidence for the untreated trees in the 2017 thinning trial). For the power law model the rate parameters or slopes of the gradients (b) were always more negative for mechanical thinning followed by a streptomycin application, indicating sharper decline in the curve and shortest distance for spread of disease from the inoculated tree. The only exceptions were blossom blight and shoot blight in the 2017 thinning trial, in which the rate parameter for thinning followed by the biopesticide was higher than thinning followed by streptomycin. Rate parameter values for mechanical thinning alone or untreated trees were similar in all instances (almost always within

0.1). Similarly, the rate parameter for the exponential models were almost always greatest for mechanical thinning followed by streptomycin. This was another indication of a sharper decline in the curve and the least amount of spread, as measured in distance from the inoculated tree.

Model predictions were close to observed incidence values (Table 4.5, Figure 4.2). Both typically over-predicted severity at the inoculated tree (distance = 0), but at 10 m both models deviated by 10% disease severity or less. Overall, predicted blossom blight severity at 10 m ranged from 0 to 36% incidence, with highest values in trees that were thinned with no subsequent antibiotic or biopesticide application (0 to 18.2% incidence) and untreated trees (0 to

36.6%). Lower incidence was predicted at 10m for thinning followed by streptomycin application (0 to 3.1%) and thinning followed by biopesticide application (0 to 15%). Both models occasionally over-predicted how far the disease progressed down a row, but predicted values were often low. For example, for thinning followed by an application of streptomycin or the biopesticide, incidence at 50 m was less than 1% except for one case. Additionally, this was likely an artifact of the alternate source of inoculum that may have been the source of blossom and shoot blight symptoms at the end of rows in the study orchard.

Risk of fire blight infection and spread following mechanical pruning

In the ‘Idared’ orchard, patterns of disease development and spread for mechanical pruning trials were similar to those observed in the mechanical thinning trials. Disease symptoms developed in all inoculated trees (Table 4.1). In both July and August pruning trials in both years, shoot blight never developed beyond the first tree in tree rows receiving a streptomycin application after pruning. Symptoms developed beyond the inoculated point source for all other treatments.

Overall disease development, expressed as rAUDPC, was again frequently affected by pruning treatments (P < 0 .05, Table 4.2). For both July and August trials, rAUDPC was highest for trees that were thinned without a subsequent bactericidal application, ranging from 9.5 to

22.0% shoot blight incidence. In untreated trees, rAUDPC was similar, ranging from 7.5 to

16.1% incidence. rAUDPC was significantly lower for trees that received an application of streptomycin (ranging from 0.5-3.1) or the biopesticide (ranging from 2.0-15.1) following mechanical pruning as compared to thinning alone.

Also similar to mechanical thinning trials, disease spread usually only extended a short distance beyond inoculated trees, with shoot blight symptoms typically observed in trees adjacent to the inoculated tree and the subsequent 2-3 trees, or approximately 10-15 m beyond the point source of inoculum (Tables 4.2 and 4.3). Symptoms developed furthest in the row for trees that were pruned and did not have a subsequent application of streptomycin or the biopesticide: for July trials, shoot blight incidence was still relatively high at 30m with 36.3 and

18.4 % incidence in 2017 and 2018, respectively. For August trials, in trees that were pruned without subsequent streptomycin or biopesticide application disease did not extend as far into the row: disease incidence at 30 m was 1.4 and 2.2% incidence in 2017 and 2018, respectively. In comparison, untreated trees had similarly high incidence close to inoculated trees, but symptoms were not observed beyond 10 m from the point source. As previously stated, in trees that received a streptomycin application after pruning, symptoms were not observed beyond the inoculated tree.

Models were again used to describe the spread of disease with pruning treatments. The fit of both the Gregory power law model and exponential decay model was again similar for all treatments for both July and August trials in both years, as indicated by similar values for coefficient of variance (R2 value) and standard error of the estimate (SEE) for the two models fit to a given data set (Table 4.5 and Figure 4.2). Model fit was significant at P < 0.05 in all cases with only three exceptions, all of which were for the power law model fitting to progress curves in the July trials. These were for the untreated trees in both 2017 and 2018, and for the mechanical pruning treatment with no subsequent application of streptomycin or the biopesticide and untreated trees in the 2017 trial. Coefficients of variance were somewhat lower in July trials with values ranging from 0.51 to 1.0, than for August trials, which were very high, ranging from

0.79 to 1.0. In July trials, coefficients of variance were usually slightly higher for the exponential model than the Gregory power law model. Rate parameter (b) values again indicated a steep decline in disease incidence from the inoculated point source for all treatments. For Gregory power law models, the rate parameter values were much more negative for pruning followed by streptomycin than for other treatments in both July and August, indicative of the lack of spread from the inoculated trees. Pruning followed by the biopesticide was always slightly more negative than the other two treatments. A similar pattern was true for the rate parameter values for exponential models.

Model predictions were also close to observed incidence values (Table 4.4, Figure 4.2).

Consistent with observed values, predicted disease incidence was highest for pruning without subsequent antibiotic or biopesticide application in July trials, with Gregory power law models predicting 23.0 and 16.4% incidence and exponential models predicting 20.4 and 12.7% incidence in 2017 and 2018, respectively. For this treatment, both models predicted disease beyond the 50 m row at 4.2-20% incidence. In models for August trials, predictions for disease incidence at 30 m for pruning without subsequent streptomycin or biopesticide application were lower, ranging from 2.3 to 7.4% incidence. Model predictions for untreated trees were lower than mechanical pruning without subsequent application of streptomycin or the biopesticide, with predicted values at 30m ranging from 4.3 to 9.3% incidence in July and 1.2 to 5.5% incidence in

August. Consistent with observations, shoot blight was not predicted beyond the inoculated tree for pruning followed by streptomycin. For pruning followed by the biopesticide, low levels of shoot blight were predicted to the end of the row by Gregory power law models, but by 20 m values were low: 8.8% or less in July and 2.5% or less August. Similarly, exponential models

predicted shoot blight at 20m to be 9% or less in July and 1.4% or less in August, indicating the biopesticide helped prevent the disease from spreading a significant distance.

Table 4.1. Development of fire blight beyond the inoculated point sourcez for apple trees receiving mechanical thinning or pruning.

Thinning Pruning 2017 2018 2017 2018 Treatmenty Row BB SB BB SB SJ SA SJ SA Thinning or Pruning 1 + x + + + + + + + Thinning or Pruning 2 + + + + + + + + Thinning or Pruning 3 + + + + + + + + T/P+Streptomycin 1 - - + + - - - - T/P+Streptomycin 2 - - + + - - - - T/P+Streptomycin 3 + + - + - - - - T/P+Bio 1 - - + + + + + + T/P+Bio 2 - - + + + + + + T/P+Bio 3 - - + + + + + + No T/P 1 - - + + + + + + No T/P 2 - + + + + + + +

zExperiments were conducted in a planting of ‘Idared’ trees at Cornell AgriTech research orchards in Geneva, NY. The first tree in each row was inoculated with E. amylovora at 20% bloom to serve as a point source. yTreatments include mechanical thinning (T) or pruning (P), followed by an application of Firewall 17 (streptomycin) or Serenade Opti (Bacillus subtilis biopesticide; 'Bio') within 24 h, or no mechanical thinning or pruning. Mechanical thinning or pruning was applied from the inoculated point source down the length of an entire row of apple trees which as a minimum of 50 m in length. xGrey cells with '+' indicate instances in which fire blight symptoms (BB: blossom blight, SB: shoot blight, SJ: shoot blight evaluated mid-July, SA: shoot blight evaluated mid-August) were observed beyond the inoculated point source.

Table 4.2. Incidence of fire blight observed for trees receiving different mechanical thinning or pruning treatments, at 10 m distancesz from inoculated point sourcesy.

Thinning Blossom Blight Shoot Blight

Treatmentx 0 10 20 30 40 50 0 10 20 30 40 50

2017 Thinning 50.6 20.0 0.0 0.8 1.1 0.0 42.3 1.9 0.0 1.2 0.6 0.0 Thinning + 13.4 0.0 0.4 0.0 0.0 0.0 6.5 0.0 1.0 0.0 0.0 0.0 Strepw

Thinning + Biow 19.4 0.0 0.0 1.3 7.4 0.0 13.6 0.0 0.0 0.0 4.5 0.0

No Thinning 41.6 0.0 0.0 0.0 0.0 0.0 33.8 0.0 0.0 0.0 0.0 0.0

2018 Thinning 93.6 31.3 11.7 11.7 1.7 1.7 67.5 11.0 11.9 5.3 0.0 0.9

Thinning + Strep 35.8 0.8 5.0 0.0 0.8 0.8 36.0 1.2 0.9 0.0 0.9 0.9

Thinning + Bio 95.0 15.0 5.0 2.1 0.0 0.0 50.4 5.4 13.5 5.1 0.8 0.0

No Thinning 92.5 15.0 . 7.5 0.0 10.0 65.3 9.1 . 5.0 12.5 21.1

Pruning Blossom Blight Shoot Blight

Treatment 0 10 20 30 40 50 0 10 20 30 40 50

2017 Pruning 37.0 57.4 35.6 36.3 3.7 0.0 33.3 18.5 6.9 1.4 0.0 0.0

Pruning + Strep 47.1 0.0 0.0 0.0 0.0 0.0 10.6 0.0 0.0 0.0 0.0 0.0

Pruning + Bio 55.6 3.7 2.8 3.2 3.7 0.0 16.7 5.9 5.6 1.9 2.8 0.0

No Pruning 33.3 33.3 0.0 0.0 11.1 0.0 47.2 4.2 0.0 0.0 0.0 0.0

2018 Pruning 56.5 47.2 22.6 18.4 5.0 0.0 45.9 10.9 8.3 2.2 0.0 0.0

Pruning + Strep 25.7 0.0 0.0 0.0 0.0 0.0 15.3 0.0 0.0 0.0 0.0 1.4

Pruning + Bio 38.3 5.0 0.0 1.9 2.2 0.0 20.2 2.2 0.0 0.0 0.0 0.0

No Pruning 30.0 38.9 0.0 0.0 11.1 0.0 50.3 13.9 0.0 0.0 6.7 0.0

zData were taken at ~5 m intervals which corresponded to the presence of an apple tree. Data is only shown for 10 m intervals to simplify presentation. yThe first tree in each row was inoculated with E. amylovora at 20% bloom to serve as a point source. xTreatments include mechanical thinning or pruning followed by an application of Firewall 17 (streptomycin) or Serenade Opti (Bacillus subtilis sp.) within 24 h, or no mechanical thinning or pruning. Mechanical thinning and pruning was applied from the point source down the length of an entire row of apple trees which was a minimum of 50 m in length. wStrep refers to streptomycin; Bio refers to the Bacillus subtilis biopesticide Serenade Opti uColor gradient indicates disease incidence, with darker color indicating higher incidence.

Table 4.3. Relative area under the disease progress curve (rAUDPC)z for trees receiving mechanical thinning or pruning, from inoculated point sourcesy.

Thinning 2017 2018 Treatmentx Blossom Blightw Shoot Blight Blossom Blight Shoot Blight Thinning 8.1 ± 0.8 bv 3.4 ± 0.7 b 20.0 ± 5.4 b 14.4 ± 5.5 a Thinning+Strep 0.9 ± 0.3 a 0.3 ± 0.1 a 3.1 ± 1.0 a 3.4 ± 1.0 a Thinning+Bio 1.7 ± 0.5 a 1.2 ± 0.5 ab 9.6 ± 0.8 ab 8.0 ± 0.6 a No Thinning 2.6 ± 1.2 a 2.6 ± 1.3 ab 19.8 ± 1.4 b 16.5 ± 1.6 a P value 0.0004 0.0314 0.0167 0.0810

Pruning 2017 2018 Treatment July August July August Pruning 31.8 ± 5.5 b 10.0 ± 0.7 b 22.0 ± 2.7 b 9.5 ± 0.3 b Pruning+Strep 3.1 ± 0.5 a 0.5 ± 0.0 a 1.5 ± 0.2 a 0.8 ± 0.1 a Pruning+Bio 15.1 ± 5.1 ab 3.2 ± 0.7 a 6.3 ± 1.2 a 2.0 ± 0.4 a No Pruning 16.1 ± 4.8 ab 7.2 ± 0.9 b 15.4 ± 1.3 b 10.1 ± 1.3 b P value 0.0130 <0.0001 0.0002 <0.0001

zrAUDPC values were calculated from the incidence of blossom and shoot blight collected at ~5 m intervals, which corresponded to the presence of an apple tree, up to a minimum distance of 50 m from the point source. Values followed by the same letter indicate a lack of significant differences. yThe first tree in each row was inoculated with E. amylovora at 20% bloom to serve as a point source. xTreatments included mechanical thinning or pruning, followed by an application of Firewall 17 (streptomycin; 'Strep') or Sernade Opti (Bacillus subtilis biopesticide; 'Bio') within 24 h, or no mechanical thinning or pruning. Mechanical thinning and pruning was applied from the inoculated point source down the length of an entire row of apple trees, which was a minimum of 50 m in length. wDisease severity was evaluated for thinning trials by rating percent incidence blossom blight (BB) or shoot blight (SB); pruning trials were conducted twice in each year, once in July and again in August, and each was rated for shoot blight (SJ = shoot blight July trial, SA = shoot blight August trial). Trials were repeated in 2017 and 2018. vWithin a column, values followed by different letters indicate significantly different means.

zData were taken at ~5 m intervals which corresponded to the presence of an apple tree. Data is only shown for 10 m intervals to simplify presentation. yThe first tree in each row was inoculated with E. amylovora at 20% bloom to serve as a point source. xTreatments include mechanical thinning or pruning followed by an application of Firewall 17 (streptomycin; 'Strep') or Serenade Opti (Bacillus subtilis biopesticide; 'Bio') within 24 h, or no mechanical thinning or pruning. Mechanical thinning and pruning was applied from the point source down the length of an entire row of apple trees which was a minimum of 50 m in length. wObserved and predicted values are reported for 5 m distances from inoculated point source. For power models, distance 'x=0' was calculated using 'x=0.01.' vOB = Observed; PL = 'Gregory's' Power Law Model; EX = Exponential Decay Model uModels predicting >100% incidence capped at 100%.

Table 4.5. Summary statistics used to evaluate models describing the development and spread of fire blight from inoculated point sourcesz.

zThe first tree in each row was inoculated with E. amyovora at 20% bloom to serve as a point source. yTreatments include mechanical thinning or pruning, followed by an application of Firewall 17 (streptomycin; 'Strep') or Serenade Opti (Bacillus subtilis biopesticide; 'Bio') within 24 h, or no mechanical thinning or pruning. Mechanical thinning or pruning was applied from the inoculated point source down the length of an entire row of apple trees which was a minimum of 50 m in length. xModels include the exponential model: y = a*x^(b), and Gregory's power law model: y = a*e^(- b*x). wThe cells highlighted in light gray indicate relationships that were not significant (P > 0.05). vCoefficients of determination; instances of a perfect or near perfect relationship (R2 = 1) reflect that disease symptoms were not observed beyond inoculated trees. uStandard error of the estimate SEE = sqrt(sigma(y-y')^2/n) tRate parameter or slope of gradient

100

Figure 4.2. Disease progress curves illustrating the relationship between mean incidence of disease incidence and distance from an inoculated point source within a row for 'Idared' apple trees space at approximately 5 m. A to D correspond to blossom blight (A,B) or shoot blight (C,D) following mechanical thinning in 2017 and 2018 respectively. E to G correspond to shoot blight following mechanical pruning in July (E,F) or August (G,H) in 2017 and 2018, respectively. The first tree in a row was inoculate with E. amylovora strain Ea273 at 2x106 CFU ml-1 to serve as a point source of inoculum. Approximately 24 hours after inoculation tree rows were subject to mechanical thinning, pruning, or neither. Tree rows were treated with Firewall 17 (streptomycin), Serenade Opti (Bacillus subtilis biopesticide) 24 h later, or left untreated. Treatments included thinning or pruning alone (dark circles), thinning or pruning followed by an application of streptomycin (light circles) or the Bacillus subtilis biopesticide (upright dark triangles) at approximately 24 h, or no mechanical treatment or bactericidal treatment (downward light triangles). Disease intensity gradient was described by two models, which were similarly well fit to the data: exponential decay models for Thinning (black solid curves), No Thinning (grey solid curves), Thinning + Bio (black dashed curves), or Thinning + Strep (grey dashed curves), as well as the Gregory power law model (not shown). Model fit is described in Table 4.5.

Discussion

Mechanical thinning and pruning have largely been limited in commercial orchards due to concern over the potential risk of spreading fire blight and economic considerations. However, in the present study we found little evidence of increased risk of mechanical thinning or pruning spreading fire blight at the proposed timings for commercial use in NY apple production.

Further, we found that applying streptomycin or a bactericidal biopesticide within 24 hours after mechanical thinning or pruning contributed to mitigating some of the potential risk in ‘Gala’ and

‘Idared’ apples.

Concern about spreading fire blight via mechanical thinning or pruning is justified given that E. amylovora is capable of surviving on a variety of surfaces, including soils and cloth for multiple days, or pruning shears and rubber boots for up to 24 hours (Choi et al. 2019). It is well known that the pathogen cannot penetrate uncompromised host tissues, but numerous studies have demonstrated that mechanical damage can contribute to fire blight infection, with bacteria typically introduced into shoot tissue by insects with piercing-sucking mouth parts, wind-driven rain, or pruning cuts (Kleinhempel et al. 1987; van der Zwet, T. 2012). Because pruning shears are an effective means of transmitting the bacteria, management recommendations have long included sanitizing equipment between pruning cuts with 70% ethanol or bleach when removing strikes (Hasler 1996; Keil 1979; Norelli 2003; van der Zwet, T. 2012). On the other hand, most of the studies conducted on sanitation of pruning equipment used extremely high levels of inoculum, and lower risk may be associated with transmission of bacteria by equipment in a commercial orchard setting. Mechanical thinning and pruning introduce numerous wounds and can actively distribute foliar tissue in the tree canopy, causing concern for dissemination of the pathogen.

We are aware of only one study to date directly evaluating the risk of spreading fire blight with mechanical thinning, and none evaluating risk with mechanical pruning. In this work, investigators found in both potted-tree and field experiments that the incidence and severity of shoot blight in trees adjacent to inoculated trees was considerably higher for trees that were treated with a mechanical string thinner than those that had not received mechanical thinning

(Ngugi and Schupp 2009). However, as acknowledged by the researchers, experimental conditions were intentionally produced to be optimal for promoting infection and unlikely to be present in a commercial system during recommended thinning timings. In particular, moisture levels were kept artificially high as foliage was thoroughly moistened immediately following mechanical thinning, and in the potted tree experiment, foliage was moistened at least once per day, including immediately before and after the mechanical thinning or pruning. Additionally, the concentration of inoculum introduced at bloom was 2x108 CFU ml-1, over 100 fold higher than what is used in other research trials to achieve high levels of disease incidence (Cox et al.

2016; Outwater and Sundin 2016; Yoder and Kowalski 2016), and much higher than would be expected in a commercial block. This undoubtedly resulted in high levels of blossom blight infection. In addition, researchers acknowledged that early symptoms of disease were visible at the time of mechanical thinning. In a commercial setting, priority would be to intentionally avoid mechanical thinning or pruning in orchards where symptoms are present or disease levels are known to be high, with efforts focused instead on managing the disease.

In the present study, we found that the risk of fire blight infection and spread was only marginally increased by mechanical thinning or pruning when these practices were performed during periods of low risk for disease development. In our trial at the ‘Gala’ orchard, we saw no development of blossom or shoot blight at all beyond the inoculated trees. This was unexpected

103

given that ‘Gala’ is a highly susceptible variety and that this orchard receives more than adequate irrigation and fertilization to promote ample vegetative growth especially vulnerable to fire blight infection. Moreover, this block has been used routinely for fire blight trials in the past in which we reliably achieve high levels of both blossom and shoot blight (Cox et al. 2016).

Similarly, in the trials conducted at the ‘Idared’ orchard, another orchard routinely used for fire blight trials in previous years, fire blight often did not progress beyond the inoculated trees for any of the treatments. Despite high levels of artificial inoculum, the overall disease development

(rAUDPC) were comparable to the untreated trees, representing potential for spread under natural conditions, and overall very low for all treatments. In addition, models predicted that disease symptoms would decline quickly as distance from the inoculated tree increased, and seldom predicted symptoms beyond the first 2-3 trees or 10-15 m in the row. A slightly higher risk of spread was indicated for pruning in July as compared to August, in which model predictions were higher for the spread of disease further into the row for trees that were mechanically pruned, as indicated by higher predictions to 30 m. However, this risk was completely mitigated by including a subsequent application of a bactericide.

There was a notable disparity in disease incidence between the two orchard blocks. This cannot be attributed to environmental conditions, because inoculation and treatments were always applied at the same time and in the same methods for each block. Similarly, orchard blocks are located in close proximity to each other, were planted in the same year, and are managed almost identically. Therefore, the difference was most likely related to the inherent susceptibility of the two cultivars. While both ‘Gala’ and ‘Idared’ are consistently rated a susceptible to fire blight, there is a range of susceptibility in ‘Gala’ reported from different sources (Khan, 2020). Moreover, the ‘Idared’ orchard block used in this work routinely has

104

higher fire blight incidence than the ‘Gala’ orchard used, so it was not completely unexpected that the disparity existed in this trial. These results provide evidence that cultivar and other horticultural characteristics are important contributors to fire blight susceptibility.

One of the main factors that may contribute to disease development and spread is environmental conditions. Temperature and moisture are incredibly important for the development of populations of E. amylovora of and the subsequent development of fire blight, with optimal temperatures for infection generally between 21-27°C, and any surface moisture

(including high atmospheric humidity) being especially conducive to fire blight outbreaks

(Ockey and Thomson 2006; van der Zwet, T. 2012). In all years of our study, temperatures and relative humidity levels following mechanical thinning were adequate for pathogen growth and subsequent wetting events were conducive to infection. However, thinning was performed in early bloom (20% blossoms open) rather than full bloom (75-80% blossoms open). Conducting mechanical thinning earlier in bloom, as demonstrated in the present study, could reduce or eliminate the risk of infections because fewer growing degree hours would have accumulated to allow population growth of E. amylovora on stigmatic surfaces, and fewer blossoms would be open to serve as infection courts. In fact, using the online Network for Environment and Weather

Applications system (NEWA, http://newa.cornell.edu/), the disease forecasting model,

Cougarblight (Smith 1996), predicted low or moderate risk for blossom blight in 2017 and 2018, respectively during the four days following inoculations.

Selective timing also appears to be an effective strategy for mitigating risk of fire blight development and spread for mechanical pruning. Young tissue and vigorous growth are especially susceptible to shoot blight (Norelli 2003), therefore, conducting pruning in July or in

August when trees are approaching the end of growth signaled by terminal bud set would reduce

105

risk. In the present study, we found that risk was somewhat lower for August pruning as compared to July, corroborating this idea. This is a strategy that would be easily employed in commercial settings. The timing of mechanical thinning and pruning during bloom in the

Northeastern United States is flexible, especially as compared to traditional chemical thinning

(Robinson et al. 2016; Sazo et al. 2016), and growers may therefore select days when fire blight infection risk is low. Forecasting models and tools already in use by commercial farmers, such as the NEWA weather system, can be used to project periods of high risk so that mechanical thinning or pruning can be avoided at these times (Carroll et al. 2017). Moreover, Sazo et al.

(2016) recommended mechanical thinning at king bloom (20%) for optimal bloom thinning.

Application of a bactericidal product following mechanical thinning or pruning also appears to be effective at reducing risk of fire blight development and spread. Applying streptomycin following mechanical injury to the canopy, such as by wind storms and hail, is a widely accepted management practice for reducing risk of shoot blight (Ockey and Thomson

2006; van der Zwet, T. 2012). In the present study, applying streptomycin was effective for reducing and effectively eliminating the spread of fire blight down a row from the inoculated point source. This is evidenced by much lower rAUDPC values for trees receiving a streptomycin application after mechanical thinning or pruning and by the rate parameters in both exponential and power law models fit to the data, indicating steeper decline in disease with distance from the inoculated tree for such thinning programs. Similarly, the biopesticide was effective at reducing risk, although at a lower magnitude.

There is anecdotal evidence that these recommended practices are already being used successfully by commercial NY apple growers to enable mechanized practices in commercial orchards (Table 4.6). A total of 17 commercial farms in Western NY currently practice

106

mechanical pruning. Earliest adoption was observed in the early 2000s (Table 4.6), but there is variation in parameters including machine type, timing (dormant, summer, post-harvest), cultivars, and use of streptomycin following mechanical pruning. Machines may use a single- or double- sickle bar, tractors and equipment may move at different speeds, and thinner or hedger bars may operate at various angles and distances from the canopy. These variations in equipment and practices translate to various speed of strings or blades, amount of contact with trees and tissue damage, and ultimately the potential to spread fire blight. Even so, a small minority of the farmers interviewed have reported any difficulty in managing fire blight in relation to mechanical pruning. Survey results indicate streptomycin is already routinely being used by growers as a method for reducing risk of fire blight infection following mechanical practices.

The research presented here supports their method of control. Clearly, mechanized thinning and pruning are valuable time- and labor-saving practices that the commercial industry is already adopting and growers are adopting practices in this study to reduce the risk of spreading fire blight.

107

Table 4.6. Characteristics of mechanical pruning practices in 13 commercial fruit farms located in the Lake Ontario Fruit Region, NY, USA.

Adoption of Mechanical pruning Fire blight (FB) Machine type Main Timings Year Strep Related FB (sickle bar) started Use issues/strikes Single- Double- Dormant Summer Post- Orchard harvest 1 x x x x 2011 no no 2 x x x 2012 no no 3 x x x 2014 no no 4 x x x 2015 no no 5 x x x 2018 no no 6 x x x 2018 no no 7 x x 2018 no no 8 x x x x x 2016 no no 9 x x 2017 yes minor issues 10 x x x x 2018 no no 11 x x x 2016 no no 12 x x x 2016 no no 13 x x x x 2008 no no

In conclusion, we found little evidence of increased risk of fire blight infection or spread following the use of mechanical thinning or pruning. Application of streptomycin or the biopesticide B. subtilis biopesticide within 24 h of mechanical thinning or pruning effectively reduced risk to tolerable levels. Many growers appear to already be employing these practices, but many factors including environmental conditions, timing, machine parameters, and cultivars, warrant further investigation to provide more specific recommendations. Overall, our results support the safe use of mechanized thinning and pruning under field conditions in commercial orchards.

Literature Cited

Agnello, A., Brown, B., Carroll, J., Cheng, L., Cox, K., Curtis, P., et al. 2019. Cornell Pest Management Guidelines for Commercial Tree Fruit Production. Ithaca, NY: PMEP Cornell University.

108

Asteggiano, L., Giordani, L., Bevilacqua, A., Vittone, G., Pellegrino, S., and Costa, G. 2015. Bloom mechanical thinning improves fruit quality and reduces production costs in peach. Acta Hortic. 1084:389–394.

Baugher, T. A., Ellis, K., Remcheck, J., Lesser, K., Schupp, J., Winzeler, E., et al. 2010. Mechanical String Thinner Reduces Crop Load at Variable Stages of Bloom Development of Peach and Nectarine Trees. Hort. Sci. 45:1327–1331.

Bock, C. H., Parker, P. E., and Gottwald, T. R. 2005. Effect of Simulated Wind-Driven Rain on Duration and Distance of Dispersal of Xanthomonas axonopodis pv. citri from Canker-Infected Citrus Trees. Plant Disease. 89:71–80.

Carroll, J., Weigle, T., Agnello, A., Reissig, H., Cox, K., Breth, D., Robinson, T., Loeb, G., Wilcox, W., Eggleston, K., DeGaetano, A., Olmstead, D., Grant, J., Gibbons, J., and Petzoldt, C. 2017. NEWA (Network for Environment and Weather Applications) Provides Fruit IPM and Production Tools from 400 Weather Stations. Fruit Q. 25(1):19-24.

Chambers, J. M., Freeny, A and Heiberger, R. M. (1992) Analysis of variance; designed experiments. Chapter 5 of Statistical Models in S eds J. M. Chambers and T. J. Hastie, Wadsworth & Brooks/Cole.

Choi, H. J., Kim, Y. J., Lim, Y.-J., and Park, D. H. 2019. Survival of Erwinia amylovora on Surfaces of Materials Used in Orchards. 식물병연구. 25:89–93.

Cline, J. A., Bakker, C. J., and Gunter, A. 2019. Response of Royal Gala apples to multiple applications of chemical thinners and the dynamics of fruitlet drop ed. Christian Willenborg. Can. J. Plant Sci. 99:1–11.

Cox, K.D., Ayer, K.M., Kuehne, S.A., and Tancos, K.A. 2016. Evaluation of bactericide programs for the management of fire blight on 'Gala' apples in NY, 2016. Plant Dis. Manag. Rep. 11:PF003. Online Publication. doi: 10.1094/PDMR11.

Gonzalez, L., Torres, E., Carbó, J., Alegre, S., Bonany, J., Àvila, G., et al. 2019. Effect of different application rates of metamitron as fruitlet chemical thinner on thinning efficacy and fluorescence inhibition in Gala and Fuji apple. Plant Growth Regul. 89:259–271.

Greene, D., and Costa, G. 2013. Fruit thinning in pome- and stone-fruit: state of the art. Acta Hortic. 998:93–102.

Gregory, P. H. 1968. Interpreting Plant Disease Dispersal Gradients. Annu. Rev. Phytopathol. 6:189–212.

Hasler, T. 1996. Disinfection of fire blight contaminated tools. Acta Horticulturae. 411:369–371.

He, L., and Schupp, J. 2018. Sensing and Automation in Pruning of Apple Trees: A Review. Agronomy. 211:1-18.

109

Keil, H. L. 1979. Fire Blight: A Bacterial Disease of Rosaceous Plants.

Kiyosawa, S., and Shiyomi, M. 1972. A Theoretical Evaluation of the Effect of Mixing Resistant Variety with Susceptible Variety for Controlling Plant Diseases. Ann. Phytopath. Soc. Japan. 38:41–51.

Khan, A. (2020). “Disease Susceptibility of Common Apples.” Database of Apple Diseases. Cornell University Blog Service. https://blogs.cornell.edu/applevarietydatabase/disease- susceptibility-of-common-apples/

Kleinhempel, H., Nachtigall, M., Ficke, W., and Ehrig, F. 1987. Disinfection of pruning shears for the prevention of the fire blight transmission. Acta Horticulturae. 217:211–218.

Kon, T. M., Schupp, J. R., Winzeler, H. E., and Marini, R. P. 2013. Influence of Mechanical String Thinning Treatments on Vegetative and Reproductive Tissues, Fruit Set, Yield, and Fruit Quality of ‘Gala’ Apple. Hort. Sci. 48:40–46.

Lordan, J., Alins, G., Àvila, G., Torres, E., Carbó, J., Bonany, J., et al. 2018. Screening of eco- friendly thinning agents and adjusting mechanical thinning on ‘Gala’, ‘Golden Delicious’ and ‘Fuji’ apple trees. Scientia Horticulturae. 239:141–155.

Lordan, J., Francescatto, P., Dominguez, L. I., and Robinson, T. L. 2018. Long-term effects of tree density and tree shape on apple orchard performance, a 20 year study—Part 1, agronomic analysis. Scientia Horticulturae. 238:303–317.

Lordan, J., Gomez, M., Francescatto, P., and Robinson, T. L. 2019. Long-term effects of tree density and tree shape on apple orchard performance, a 20 year study – part 2, economic analysis. Scientia Horticulturae. 244:435–444.

Madden, L. V., Hughes, G., and van den Bosch, F. 2007. The Study of Plant Disease Epidemics. St. Paul, MN: The American Phytopathological Society, APS Press.

Mundt, C. C., Ahmed, H. U., Finckh, M. R., Nieva, L. P., and Alfonso, R. F. 1999. Primary Disease Gradients of Bacterial Blight of Rice. Phytopathology. 89:64–67.

Ngugi, H. K., and Schupp, J. R. 2009. Evaluation of the Risk of Spreading Fire Blight in Apple Orchards with a Mechanical String Blossom Thinner. Hort. Sci. 44:862–865.

Norelli, J. L. 2003. Fire blight management in the twenty first century. Plant Disease. 87:756– 765.

Ockey, S. C., and Thomson, S. V. 2006. Influence of rain on the transient populations of Erwinia amylovora on leaf surfaces. Acta Hortic. 704:113–120.

Outwater, C.O. and Sundin, G.W. 2016. Evaluation of Cueva, Nu-Cop HB, Nu-Cop 30HB, Oxidate, and Actigard for resistance management and control of fire blight on Jonathan apples, 2016. Plant Dis. Manag. Rep. 11:PF002. Online Publication. doi: 10.1094/PDMR11.

110

Petracek, P. D., Silverman, F. P., and Greene, D. W. 2003. A History of Commercial Plant Growth Regulators in Apple Production. HortScience. 38:937–942.

R Core Team. 2020. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Robinson, T. L., DeMarree, A. M., and Hoying, S. A. 2007. An economic comparison of five high density apple planting systems. Acta Horticulturae. 732:481–489.

Robinson, T., Hoying, S., Sazo, M. M., DeMarree, A., and Dominguez, L. 2013. A Vision for Apple Orchard Systems of the Future. Fruit Quarterly. 21:11-16.

Robinson, T., Lakso, A., Greene, D., Reginato, G., and Rufato, A. 2016. Managing fruit abscission in apple. Acta Hort. 1119:1-14.

Sansavini, S. 1978. Mechanical pruning of fruit trees. Acta Horticulturae. 65:183–197.

Sazo, M. M. 2018. New Advances to Narrower Canopy Systems: Transitioning from 3-D to 2-D Canopies or Fruiting Walls – Part 3. Fruit Quarterly. 26:31-37.

Sazo, M. M., Francescatto, P., Sanahuja, J. L., and Robinson, T. L. 2016. Mechanical Blossom Thinning Followed by 6-BA Shows Promise as an Alternative to Thinning Without Carbaryl. Fruit Quarterly. 24:29-36.

Schupp, J. R., and Baugher, T. A. 2011. Peach Blossom String Thinner Performance Improved with Selective Pruning. HortScience. 46:1486–1492.

Schupp, J. R., Baugher, T. A., Miller, S. S., Harsh, R. M., and Lesser, K. M. 2008. Mechanical Thinning of Peach and Apple Trees Reduces Labor Input and Increases Fruit Size. HortTech. 18:660–670.

Smith T. J., 1996. A risk assessment model for fire blight of apple and pear. Acta Hort. 411:97- 100.

Solomakhin, A. A., and Blanke, M. M. 2010. Mechanical flower thinning improves the fruit quality of apples. J. Sci. Food Agric. 90:735-41. van der Zwet, T. 2012. Fire Blight: History, Biology, and Management. pp. 183-191. St. Paul, MN: APS Press.

Yoder K., and Kowalski, A.L. 2016. Suppression of fire blight on ‘Idared’ apple blossoms, 2016. Plant Dis. Manag. Rep. 11:PF023. Online Publication. doi: 10.1094/PDMR11.

111

CHAPTER 5

EXAMINING SPATIAL DISTRIBUTION AND SPREAD OF FIRE BLIGHT IN APPLE

ORCHARDS: TWO CASE STUDIES*

Abstract

Fire blight, caused by the bacteria Erwinia amylovora, is an incredibly destructive disease of apples, capable of spreading rapidly through an orchard block. The pathogen is endemic to many apple production regions worldwide, but it is often introduced into newly planted sites on infested host material, while locally it is typically vectored by wind driven rain, hail, and insects.

Here we presented two case studies of orchard blocks infected with fire blight at the Cornell

AgriTech Research orchards in which CRISPR pattern characterization was used to identify strains present in the blocks and spatial analyses were used to describe the distribution and spread of the disease over the course of two years. Results indicated two very different sources of introduction (a nearby block or planting material) and patterns of spread (from a corner of the field or from focal points within the block). Describing the distribution and spread of fire blight within and between orchard blocks has the potential to improve our understanding of the disease movement, inform appropriate management recommendations, and facilitate traceback efforts.

*Wallis, A. and Cox, K. 2020. Examining spatial distribution and spread of fire blight in apple orchards: two case studies. Plant Health Progress. Melhus Symposium Proceedings. In Review.

112

Introduction

Fire blight, caused by the bacteria Erwinia amylovora, is one of the most serious and destructive diseases of apples. It can spread rapidly within an orchard, costing farmers considerable financial losses worldwide annually (Norelli 2003). Describing the distribution and movement of the pathogen between and within orchards is integral to understanding how the pathogen spreads and developing sustainable management strategies.

E. amylovora requires living host tissue for survival making the most probable mode of dissemination between geographic regions infested plant material such as budwood or nursery stock (van der Zwet 2012). Locally, the pathogen may be introduced and spread over shorter distances and time spans by wind and rain splash, insect vectors, birds, and humans (Miller 1929; van der Zwet 2012). In previous research, spread in a simulated nursery environment was associated with previous storms and distribution of outbreaks was typically within a row or forming tight clusters, while wind-driven rain related to spread across rows and greater distances

(McManus 1994). However, conditions are often erratic and unpredictable, and further characterization of introduction and spread is warranted. Additionally, multiple strains may exist in the same orchard, making identification of introduction events and related patterns of spread within an orchard block difficult.

Recently, the ability to differentiate E. amylovora strains has become possible by sequencing clustered regularly interspersed short palindromic repeat (CRISPR) regions. CRISPR regions, which are present in approximately 40% of bacteria and 90% of archaea, are created as the organism interacts with foreign DNA (i.e. a virus) and retains a portion of the DNA as a

‘spacer’ for future recognition. Because spacers are unique and accumulated in chronological order these regions effectively act as a fingerprint for strain identification, facilitating the study

113

of epidemiology, tracebacks, and strain characterization (McGhee and Sundin 2012; Rezzonico et al. 2011; Tancos and Cox 2016). In E. amylovora, three CRISPR regions (CR1, CR2, and

CR3) have been described, and reference libraries have been created containing over 50 unique

CRISPR patterns. The resulting CRISPR array patterns have been used to determine distribution of E. amylovora strains and investigate outbreaks of streptomycin resistance in New York and the Northeastern United States (McGhee and Sundin 2012; Tancos and Cox 2016; Tancos et al.

2016).

The purpose of this work was to describe two case studies of fire blight infection and spread within an orchard block, and possible introduction events. CRISPR pattern characterization was used to identify pathogen strain(s) present in the orchard blocks and nearby locations. Subsequent spatial analyses were used to determine the distribution and spread of strains within a block. The results of this work have the potential to inform more sustainable fire blight management recommendations by improving our understanding of the movement of E. amylovora strains within orchard blocks and informing the most likely methods of introduction into infected orchards.

114

General Methods

In order to investigate the distribution and spread of fire blight within orchards, two orchard blocks at the Cornell AgriTech research orchards in Geneva, NY(42°87’70.19”N -

77°02’99.15”W) were evaluated over the course of two years. Each individual tree in a block was rated for fire blight incidence, as presence or absence of shoot blight. A sample was collected from each infected (symptomatic) tree and the pathogen was isolated in pure culture, as described by Tancos et al. (2016). The CRISPR pattern was identified for each isolate using methods described by McGhee and Sundin (2011) and Tancos and Cox (2016).

The resulting data were subjected to spatial analyses to characterize the distribution of a strain within the given block. Ordinary runs analysis was used to determine whether the pattern of disease incidence was random or aggregated within individual rows (Gibbons 1976; Madden et al. 1982). A run was defined as one or more trees with the same incidence rating (presence or absence). Significant aggregation (non-random distribution) was indicated by a Z-statistic, ZU ≤

= -1.64 (P ≤ 0.05), which occurred if the number of runs was significantly different than expected for a random distribution. Data were also subjected to Spatial Analysis by Distance

IndicEs (SADIE), which determines the extent of non-randomness of data mapped in two dimensions by calculating the distance or 'number of moves' from the observed distribution to a completely regular distribution (Li et al. 2012; Perry 1995; Perry et al. 1999; Xu and Madden

2005). The Index of aggregation (Ia) was calculated as Ia = Dr/Ea, in which Dr is the distance to regularity and Ea is the expected distance to regularity for 100 randomization results. Significant aggregation (non-random distribution) was indicated by an Ia value of > 1. SADIE analysis was performed using the ‘sadie’ function in the ‘epiphy’ package (Gigot 2018) in R version 3.6.3 (R

Core Team 2020).

115

Case Study 1: Orchard A

The first case study was carried out in 2019 and 2020 in a mixed block of ‘Cortland’,

‘Jersey Mac’, ‘Gingergold’, and ‘Golden Delicious’ on B.9 rootstock (Figure 5.1). Trees were planted in 2008 at 2.4 m within row and 3 m between row spacing, and trained to vertical axis system. The block was composed of six rows, each 39 trees in length. In 2019, the adjacent block of apples, which was located directly north (20 m between blocks) and oriented with rows parallel to the study block, was inoculated at king bloom with E. amylovora (strain Ea273) at 106

CFU ml-1 using a hand-pumped Solo 475-B backpack sprayer (Solo Incorporated, Newport,

VA). In 2019 and 2020, each individual tree in the study block was rated for disease incidence

(presence or absence) as soon as shoot blight symptoms were reliably detectable. Samples were collected at the time of rating, and processed as described above. Shoot strikes were pruned out in both years at least 30 cm beyond visible symptoms, as soon as data collection was complete.

Prior to 2019, the study block had not presented symptoms of fire blight. However, in

2019 and 2020, 38.5% and 39.7% of trees exhibited fire blight symptoms, respectively (Figure

5.1). CRISPR pattern characterization of E. amylovora isolates revealed that only one strain was present, which was Ea273, the strain used to inoculate the adjacent block. Ordinary runs analysis indicated there was not significant aggregation in any row during either year (ZU > -1.64, P >

0.05), with the exception of the single row closest to the adjacent inoculated orchard block in

2019. SADIE indicated there was significant aggregation in the plot in both 2019 (Figure 5.1 B,

Ia = 1.77, Pa < 0.0001) and 2020 (Figure 5.1 C, Ia = 1.27, Pa < 0.0001).

Only one strain of E. amylovora was recovered from this orchard block, matching the strain used to inoculate the adjacent parallel block, providing strong evidence that the pathogen was introduced from the adjacent block. Furthermore, transfer between fields would have been in

116

the direction of prevailing winds. This is consistent with other studies investigating fire blight movement within an orchard environment, in which dissemination by wind and rain has been observed over distances from 1 m (Bauske 1967) to 100 m (Billing and Berrie 2002) depending on environmental conditions. One-dimensional spatial analysis (ordinary runs) did not find evidence for aggregation within rows, indicating that there was not aggregation within individual rows. Rows were oriented east to west, or perpendicular to the presumed source of inoculum, therefore it is expected that aggregation would not be detected within rows. In contrast, two- dimensional spatial analysis (SADIE) found strong evidence for aggregation within the block.

Clusters were evident in the northeast corner of the field in both years. We expected aggregation at the northern part of the field due to the source of inoculum. One explanation for aggregation in the eastern portion of the field is that the east side of the block was adjacent to a wood line, which may have provided environmental conditions more conducive to disease development

(slower drying time and higher humidity), whereas the west side is an open drive lane that would have provided less conducive conditions. Our results describing the introduction and spread of fire blight within an orchard block provide evidence that planting location and surrounding environment are important to consider in a fire blight management strategy. For instance, sustainable management strategies may include planting orchards or susceptible varieties in a location with less wooded area adjacent to the block, in order to promote conditions less conducive to disease development.

117

Figure 5.1. Distribution of fire blight incidence within a vertical axis orchard block of mixed varieties on B.9 rootstock at Cornell AgriTech Research Orchards in Geneva, NY. Distance to the adjacent infected orchard was approximately 10 m. (A) Incidence in 2019 and 2020. Individual cells represent a single tree. Cell colors indicate year in which tree was affected: 2019 (grey), 2020 (black), or both years (striped): 2019&2020). (B) and (C) Spatial Analysis of Distance IndicEs (SADIE) plots for 2019 and 2020, respectively. Plots indicate extent of clustering, with larger, red circles indicating greater clustering

118

Case Study 2: Orchard B

The second case study was carried out in 2018 and 2019 in a high density planting of mixed varieties on G.935 rootstock. Trees were planted in 2016 at 1 m within row and 3 m between row spacing and trained to a tall spindle system. The block was composed of nine rows, each 104 trees in length. Trees in this block were not inoculated. Two nearby blocks

(approximately 100 m and 500 m away) were routinely used for fire blight research and exhibited high levels of fire blight symptoms (Cox et al. 2016; Cox et al. 2017). Similar to

Orchard A in Case Study 1, each individual tree in the study block was rated for disease incidence (presence or absence) in both 2018 and 2019, as soon as shoot blight symptoms were reliably detectable. Samples were collected at the time of rating, and processed as described above. Shoot strikes were pruned out in both years at least 30 cm beyond visible symptoms, as soon as data collection was complete.

In the year of planting and the following season (2016 and 2017), the block experienced very low levels of fire blight; data was not collected on incidence or distribution at that time. In

2018 and 2019, 3.2% and 3.6% of trees exhibited shoot blight, respectively (Figure 5.2).

CRISPR pattern characterization of E. amylovora isolates detected only one strain in 2018, with

CRISPR pattern matching strain NY17.2 (CRISPR pattern 41:23:38), a well-studied streptomycin resistant strain present in commercial orchards in NY and the Northeastern United

States (Tancos and Cox 2016). In 2019, two CRISPR patterns were identified: the majority matched strain NY17.2, but four samples matched strain Ea273 (4:21:38), the strain used routinely to inoculate nearby research blocks for E. amylovora trials. Spatial analyses were carried out using only the data for samples matching NY17.2, the dominant strain in the block.

Ordinary runs analysis indicated there was significant aggregation in rows 1, 6, and 9 in 2018

119

(ZU = -4.26, -10.78, and -1.94, P < 0.05), and in rows 1, 2, 4, and 8 in 2019 (ZU = -2.64, -5.06, -

3.69, and -6.37, P < 0.005). SADIE indicated there was significant aggregation in the plot in both

2018 (Figure 5.2 B, Ia = 1.49, Pa = 0.01) and 2019 (Figure 5.2 C, Ia = 2.74, Pa < 0.0001).

Our CRISPR strain characterization provided evidence of two strains of E. amylovora in this block, indicating that there were likely two separate introductions. To further investigate the nearby blocks as sources of inoculum (Figure 5.3), several samples were randomly collected within each of these blocks and processed as described above to identify the E. amylovora strain(s) present. Only Ea273 was present in the block 500 m away; in the block 100 m away, three strains, with patterns matching Ea273, NY17.2, and an additional previously undescribed strain, were identified. The detection of Ea273 and NY17.2 in both the study block the nearby blocks provides evidence for the possible transfer of inoculum between blocks, although it does not determine the original source of inoculum. Alternatively, the same strains may have been introduced to each block independently. While these results do not provide conclusive information about the source of inoculum, they illustrate the utility of strain characterization in detecting the number of introduction events and sources of inoculum, which may enable traceback efforts in future investigations.

Aggregation of disease incidence in the block was detected in both years, with the extent of aggregation greater in the second year. Ordinary runs analysis indicated clustering in more rows in the second year and SADIE indicated more significant clustering (lower P value) in the second year. The increasing aggregation in the second year may indicate that the pathogen was introduced on plant material obtained from the nursery. The distribution of disease incidence appeared to be originally aggregated in multiple places throughout the block, rather near a specific location in the field such as in Case Study 1: Orchard A in which clustering was focused

120

in the northeast corner of the field. This was followed by greater aggregation of disease incidence in 2019, which could be explained by the pathogen spreading from infected trees to healthy trees nearby.

121

Figure 5.2. Distribution of fire blight incidence within a high density apple orchard block of mixed varieties on G.935 rootstock at Cornell AgriTech Research Orchards in Geneva, NY. (A) Incidence in 2018 and 2019. Individual cells represent a single tree. Cell color indicates year and strain: strain NY17.2 in 2018 (grey), strain NY17.2 in 2019 (black), or strain Ea273 in 2019 (black and white check). (B) and (C) Spatial Analysis of Distance IndicEs (SADIE) plots for 2019 and 2020, respectively. Plots indicate extent of clustering, with larger, red circles indicating greater clustering. 122

Figure 5.3. Erwinia amylovora strains recovered from three nearby orchard blocks at the Cornell Research Orchards in Geneva, NY in 2018 and 2019. Distance between the study block (Orchard B) and the nearby orchards are indicated by lines with double arrows. Both adjacent orchard blocks were routinely inoculated with E. amylovora strain Ea273 for fire blight trials. Strains were identified using CRISPR pattern profiling; CRISPR patterns are indicated in parenthesis.

123

Conclusions

In this work, we presented two case studies of fire blight spread within orchard blocks.

Using CRISPR pattern characterization we were able to determine the number of strains present in an infected block and hypothesize the number of introduction events. Spatial distribution analyses facilitated the description of two very different patterns of aggregation within the two study blocks, which were used to make inferences about how the pathogen was introduced into the block and describe how the pathogen spread over the course of two years. These methods have the potential to be used in future outbreaks to improve our understanding of the distribution and spread of fire blight within blocks, inform appropriate management recommendations, and facilitate traceback efforts.

Literature Cited

Bauske, R. 1971. Wind dissemination of waterborne Erwinia amylovora from Pyrus to Pyracantha and Cotoneaster. Phytopath, 61: 741-2.

Billing, E., and Berrie, A. M. 2002. A re-examination of fire blight epidemiology in England. Acta Horticulturae. 590:61–67.

Cox, K.D., Villani, S.M, and Tancos, K. 2016. Evaluation of bactericide and chemical regulator programs for the management of fire blight on 'Idared' apples in NY, 2015. Plant Dis. Manag. Rep. PF014.

Cox, K.D., Ayer, K., and Kuehne, S.A. 2017. Evaluation of bactericide programs for the management of fire blight on ‘Gala’ apples in NY, 2016. Plant Dis. Manag. Rep. PF003.

Gibbons, J.D., 1976. Nonparametric Methods for Quantitative Analysis. Holt, Rinehart and Winston, New York 463 pp.

Gigot, C. 2018. epiphy: Analysis of Plant Disease Epidemics. R package version 0.3.4. https://CRAN.R-project.org/package=epiphy

Li, B., Madden, L. V., and Xu, X. 2012. Spatial analysis by distance indices: an alternative local clustering index for studying spatial patterns: A new SADIE clustering index. Methods in Ecology and Evolution. 3:368–377.

124

Madden, L., Louie, R., Abt, J., and Knoke, J. 1982. Evaluation of tests for randomness of effected plants. Ecology and Epidemiology. 72(2): 195-198.

McGhee, G. C., and Sundin, G. W. 2012. Erwinia amylovora CRISPR Elements Provide New Tools for Evaluating Strain Diversity and for Microbial Source Tracking ed. Ching-Hong Yang. PLoS ONE. 7:e41706.

McManus, P. S., and Jones, A. L. 1994. Role of Wind-Driven Rain, Aerosols, and COntaminated Budwood in Incidence and Spatial Pattern of Fire Blight in an Apple Nursery. Plant Disease. 78:1059–1066.

Norelli, J. L. 2003. Fire blight management in the twenty first century. Plant Disease. 87:756– 765.

Perry, J. N. 1995. Spatial Analysis by Distance Indices. The Journal of Animal Ecology. 64:303.

Perry, J. N., Winder, L., Holland, J. M., and Alston, R. D. 1999. Red-blue plots for detecting clusters in count data. Ecol Letters. 2:106–113.

R Core Team 2020. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/.

Rezzonico, F., Smits, T. H. M., and Duffy, B. 2011. Diversity, Evolution, and Functionality of Clustered Regularly Interspaced Short Palindromic Repeat (CRISPR) Regions in the Fire Blight Pathogen Erwinia amylovora. Applied and Environmental Microbiology. 77:3819–3829.

Tancos, K. A., and Cox, K. D. 2016. Exploring Diversity and Origins of Streptomycin-Resistant Erwinia amylovora Isolates in New York Through CRISPR Spacer Arrays. Plant Disease. 100:1307–1313. van der Zwet, T. 2012. Fire Blight: History, Biology, and Management. St. Paul, MN: APS Press.

Xu, X., and Madden, L. V. 2005. Interrelationships Among SADIE Indices for Characterizing Spatial Patterns of Organisms. Phytopathology. 95:874–883.

125

CHAPTER 6

INVESTIGATING THE DISTRIBUTION OF STRAINS OF ERWINIA AMYLOVORA

AND STEPTOMYCIN RESISTANCE IN APPLE ORCHARDS IN NEW YORK USING

CRISPR PROFILES: A SIX-YEAR FOLLOW-UP

Abstract

Fire blight, caused by the bacteria Erwinia amylovora, is one of the most important diseases of apple. The antibiotic streptomycin is routinely used in the commercial apple industries of NY and New England to manage the disease. In 2002 and again from 2011 to 2014 outbreaks of streptomycin resistance (SmR) were reported. Motivated by new grower reports of control failures, we conducted a follow-up investigation of the distribution of SmR and E. amylovora strains for major apple production regions of NY over the last six years.

Characterization of clustered regularly interspaced short palindromic repeat (CRISPR) profiles revealed that several ‘cosmopolitan’ strains were widely prevalent across regions, while other

‘resident’ strains were confined to one location. In addition, we uncovered novel CRISPR profile diversity in all investigated regions. SmR E. amylovora was detected only in a small area spanning two counties from 2017 to 2020, and always associated with one CRISPR profile

(41:23:38), which matched the profile of SmR E. amylovora strains discovered in 2002. This suggests the original SmR E. amylovora was never fully eradicated and went undetected due to several seasons of low disease pressure in this region. Investigation of several representative isolates under controlled greenhouse conditions indicated significant differences in aggressiveness on ‘Gala’ apples. Potential implications of strain differences include the propensity of strains to become distributed across wide geographic regions and associated

126

resistance management practices. Results from this work will directly influence sustainable fire blight management recommendations for commercial apple industries in NY State and other regions.

* Wallis, A. Yannuzzi, I. M., Spafford, J., Choi, M., Ramachandran, P., Timme, R., Pettengill, J. B., Cagle, R., Ottesen, A., and Cox, K. 2020. Investigating the distribution of strains of E. amylovora and streptomycin resistance in apple orchards in New York using CRISPR profiles: a six-year follow-up. Plant Disease. In preparation.

127

Introduction

Fire blight, caused by the bacterial pathogen Erwinia amylovora, is a devastating disease of apples (Malus x domestica) and is now present in many apple producing regions worldwide.

Annual losses in the United States alone have been estimated at over $100 million annually

(Norelli 2003). E. amylovora originated in Northeastern North America, with fire blight first reported in the late 1790s in the Hudson Valley of NY and subsequently detected in west coast production regions of North America (Bonn and van der Zwet 2000). Only relatively recently had the pathogen been spread to other regions outside North America, including New Zealand in the 1910s, Europe in the late 1950s, and the Middle East in the 1960s (Bonn and van der Zwet

2000; Rezzonico et al. 2011). In many other countries, E. amylovora remains a quarantine pathogen with strict eradication requirements and has yet to be reported in several countries with considerable apple industries, including Australia, Brazil, South Africa, and the entirety of South

America. In the last decade alone, there have been new reports of E. amylovora in Korea in 2015

(Park 2017) and in Kazakhstan, in the center of origin of modern apples, in 2010 (Djaimurzina et al. 2014). These events illustrate the importance of understanding and preventing the spread of the pathogen by detecting, tracing, and managing outbreaks efficiently.

It has been particularly difficult to investigate the distribution of and conduct trace-back studies on E. amylovora due to low genetic variability observed between strains. Genomic comparisons have reported over 99.90% identity between nucleotide sequence (Smits et al. 2010;

Mann et al. 2012; Zheng et al. 2018). However, CRISPR loci have been identified as significant sources of genomic variability and have been used successfully to differentiate strains of E. amylovora (Rezzonico et al. 2011; McGhee and Sundin 2012; Tancos and Cox 2016). CRISPR loci, which are widespread in both archaea and bacteria, function as an adaptive immune system,

128

where by the bacterium conserves pieces of foreign DNA (i.e. bacteriophages) for future recognition (Mojica 2005). These conserved pieces alternate with repeating segments approximately 20 base pairs in length. In E. amylovora, there are three CRISPR loci (CR1, CR2, and CR3), each of which may be described by the ‘pattern’ of spacers present; a ‘CRISPR profile’ of a given isolate may be determined as the combination of the three patterns written

‘CR1:CR2:CR3.’ A comparison of 37 strains from around the world found 18 CRISPR array profiles, which clustered into three groups (Rezzonico et al. 2011). Strains segregated by geographic location: either isolates from United States, or isolates Europe, New Zealand, or the

Middle East. These researchers found evidence that E. amylovora originated in the Eastern

United States, was subsequently introduced to the Western United States production regions, and later a single founder strain initiated global distribution (Rezzonico et al. 2011). Later studies by

McGhee and Sundin (2012) characterized 85 strains of global distribution, detected 28 CRISPR array profiles, and found evidence corroborating the findings of Rezzonico et al. (2011). CRISPR profiling has also been used on a local level to investigate the distribution of E. amylovora within

Michigan (McGhee and Sundin 2012) and New York (Tancos and Cox 2016), documenting the spread of the pathogen within commercial apple industries.

In commercial orchards, fire blight is managed using a combination of cultural, mechanical, and chemical management tactics. The bacteria overwinter on the margins of cankers, which result from prior infections. Therefore, removing inoculum through dormant pruning and early season applications of copper is essential (van der Zwet management 2012).

Ooze produced by cankers in the spring is transferred to blossoms, where bacteria preferentially colonizes the stigmatic surface. The bacteria are washed into the floral cup during wetting events, such as rain or heavy dew, and enter the plant via the natural openings in the nectaries,

129

leading to ‘‘blossom blight” (van der Zwet 2012). At this point in the infection cycle, control is achieved by applying antibiotics or biopesticides to kill, outcompete, or prevent growth of the bacteria on the stigmatic surfaces of flowers (van der Zwet 2012; Johnson et al. 2009). Since its introduction in 1955, Streptomycin, an aminoglycoside antibiotic, has been the most effective means of controlling fire blight in apple production regions east of the Mississippi River (Cox et al. 2013; McManus et al. 2002; Russo et al. 2008; Sundin and Ehret 2010). Other antibiotics available for fire blight control in the United States include oxytetracycline, which exhibits bacteriostatic activity, and kasugamcyin, another aminoglycoside antibiotic recently registered for fire blight management, which are both used in places where antimicrobial resistance has occurred (McGhee and Sundin 2010).

Development of streptomycin resistance (SmR) in populations of E. amylovora has prompted great concern and become the subject of numerous scientific investigations. SmR E. amylovora was first reported in California in 1971 (Moller et al 1972), and shortly after in

Oregon and Washington (Coyier and Covey 1975). It has since been identified in numerous places throughout the world including Missouri (Shaffer and Goodman, 1985), Egypt (El-

Goorani and El-Kasheir 1989), New Zealand (Thomson et al. 1993), Israel (Manulis et al. 1998),

Lebanon (Saad et al. 2000), British Colombia (Sholberg et al. 2001), Michigan (McGhee et al.

2011), and Mexico (Ponce de Le´on Door et al. 2013).

In New York, SmR E. amylovora was first discovered in 2002. Based on the mechanism of resistance and CRISPR profiling, it was determined that the most likely source of introduction was on plant material from a nursery outside of NY (Russo et al. 2008). Infected trees were promptly removed and annual surveys did not identify SmR E. amylovora isolates from 2004 to

2006, so the strain was believed to be successfully eradicated. However, following grower

130

reports of failure to control fire blight with streptomycin, resistance was detected again in 2011.

Surveys conducted from 2011 to 2014 examining the distribution of SmR E. amylovora in the major apple production regions of the state (Tancos et al. 2016), followed by systematic comparison of CRISPR array profiles of streptomycin resistant and sensitive isolates (Tancos and Cox 2016), revealed that resistance was primarily confined to one county in Western NY, and that most SmR isolates exhibited the same CRISPR array profile as the original 2002 isolates. However, six other CRISPR array profiles with SmR phenotype were identified, and for five of these profiles, isolates were also detected with streptomycin sensitive phenotypes. These observations provide evidence for separate introductions and/or development of SmR E. amylovora within the state.

There are two known mechanisms of streptomycin resistance in E. amylovora. The first is the gene pair strA and strB, which code for an aminoglycoside transferase (Chiou and Jones

1995a). This is the primary form of resistance in eastern North America and the only detected mechanism of resistance in NY isolates (Forster et al. 2015; McManus and Jones 1994; McGhee and Sundin 2011; Tancos et al. 2016). The other mechanism of resistance is a point mutation in the rpsL gene, coding for S12 ribosomal protein, which alters the target for streptomycin binding

(Chiou and Jones 1995b). This was the original mechanism of resistance found in western North

America (Moller et al. 1981).

Stakeholders in Western NY have continued to voice concern about the development of streptomycin resistance and to report cases where they believe streptomycin has failed to manage fire blight. Therefore, sampling and testing are conducted annually to identify whether the pathogen is present and determine streptomycin sensitivity status for individual outbreaks. The purpose of this work was to provide a follow-up to the previous investigations of the distribution

131

of E. amylovora strains, as determined by CRISPR array profile characterization (Tancos and

Cox 2016), and streptomycin resistance (Tancos et al 2016) in NY State. From 2015 to 2020 extensive survey efforts were conducted in the three major production regions of the state, as well as other nearby apple production regions in the Northeastern United States. Results from this work may continue to improve our understanding of the distribution and spread of E. amylovora in NY and the northeastern United States, and inform management recommendations for commercial apple orchards, especially as related to preventing the development of streptomycin resistance.

Materials and Methods

Collection of fire blight samples and isolation of E. amylovora

Samples were collected and processed using approximately the same methods described in the original work by Tancos et al. (2016). Fire blight samples were collected from 2015 to

2020 from outbreaks identified in commercial and research orchards. Similar to study efforts from 2011 to 2014, sample collection was accomplished as a large collaborative effort between

Cornell AgriTech Scientists, Cornell Cooperative Extension, private consultants, and commercial growers. Fire blight sampling efforts were advertised in publications distributed widely by the industry, and included solicitation and instruction for sample collection. Samples were delivered to Cornell AgriTech in the most efficient means available to preserve the pathogen populations in the tissue collected, either by overnight mail or in person on ice, when possible. Once samples were received by the lab at Cornell AgriTech, they were stored in refrigerated rooms (4°C) until processing, which took place within one week.

132

From each sample, several 2-3 cm sections of infected tissue, including the margin of the lesion, were removed using sterilized scalpel and pruners. Tissue sections were surface sterilized by immersing in 10% bleach solution for 10 min with agitation, followed by a 1 min rinse with sterile deionized water. Several 1 cm sections of cambium tissue were dissected from shoot sections using sterile scalpel and forceps and plated immediately on Crosse Goodman medium

(CG; Crosse and Goodman 1973). Petri plates were incubated at 28°C for approximately 48 h.

Bacterial growth was identified as E. amylovora based on characteristic cratering morphology, and pure culture isolates were obtained by re-streaking an individual colony on CG and incubating for approximately 48 h. DNA was extracted by suspending an individual colony in

200 µl sterile deionized water and vortexing for 1 min, Samples were stored at -20°C until further molecular tests were performed.

Presence of E. amylovora was confirmed using polymerase chain reaction (PCR), to amplify the plasmid pEA29. This plasmid is ubiquitous in, and unique to, E. amylovora worldwide (McGhee and Jones 2000; Tancos 2016). PCR reactions were carried out as described by Tancos et al (2016). In brief, 25 µl reactions included 2 µl of template DNA, 12.5 µl

EmeraldAmp® GT PCR Master Mix (Takara Bio Inc.), 1 µl each of the forward and reverse primers (AJ75 and AJ76), and 8.5 µl H2O. Cycling parameters included 5 min at 94°C; 35 cycles of 94°C for 30 s, 52°C for 30 s, and 72°C for 30 s; and 7 min at 72°C. PCR products were visualized using a 0.5% agarose gel. An 800 bp band indicated the presence of pEA29.

Determination of CRISPR profiles

Strain identity of each isolate was determined by characterizing CRISPR profiles, as previously described by McGhee and Sundin (2012) and Tancos and Cox (2016). Isolates with

133

the same CRISPR profile were considered the same strain. Spacer patterns for each CR region were determined using PCR and Sanger sequencing. PCR Reactions were performed in 25 µl volumes including 2 µl of template DNA, 12.5 µl EmeraldAmp® GT PCR Master Mix (Takara

Bio Inc.), 1 µl each of the forward and reverse primers (CR1-F1 and CR1-R0 for CR1, CR2-F1 and CR2-R1 for CR2, and CR3-F1 and CR3-R1 for CR3), and 8.5 µl H2O (Appendix:

Supplementary Table S6.1). Cycling parameters were 5 min at 94°C; 40 cycles of 94°C for 30 s, annealing for 30 s (58°C for CR1 and CR2, 55°C for CR3), and 72°C for 4 min; and 7 min at

72°C. PCR products were visualized using a 0.5% agarose gel. PCR products were purified using

ExoSAP-IT PCR Product Cleanup Kit (Applied Biosystems) and sequenced at the Cornell

Biotechnical Resource Center in Ithaca, NY using Sanger sequencing. CRISPR spacer profiles were determined using CLC Main Workbench (v 20, CLC Bio, Qiagen) to assemble fragments into complete CR1, CR2, and CR3 regions. Fragments were aligned with reference sequences of known CRISPR array patterns, created based on a library of known spacer sequences and patterns provided by Tancos and Cox (2016). Finally, spacer pattern was determined for the

CR1, CR2, and CR3 regions in each isolate. Then, CRISPR profile was defined as the combination of CR1, CR2, and CR3 spacer patterns, written ‘CR1:CR2:CR3.’

Characterization of streptomycin resistance in E. amylovora isolates

Sensitivity to streptomycin was determined by phenotyping as previously described by

Tancos et al. (2016) and Tancos and Cox (2016). Initial evaluation of streptomycin sensitivity was conducted by assessing growth on CG medium amended with 100 µg ml-1 streptomycin.

Individual colonies were streaked on medium with and without the streptomycin amendment and

134

incubated at 28°C for approximately 48 h; growth on the streptomycin amended media indicated a streptomycin resistant phenotype.

Streptomycin resistance was confirmed, and the mechanism of resistance identified, using

PCR amplification of genes conferring streptomycin resistance. First, SmR isolates were screened for the presence of strA and strB genes. PCR was performed in 25 µl volumes, as previously described by Tancos et al. (2016), and included 2 µl of template DNA, 12.5 µl

EmeraldAmp GT PCR Master Mix (Takara Bio Inc.), 1 µl each of the forward and reverse primers (AJ75 and AJ76), and 8.5 µl H2O. Cycling parameters were 5 min at 94°C; 35 cycles of

94°C for 30 s, annealing for 30 s (56°C for strA and 53°C for strB), and 72°C for 30 s; and 7 min at 72°C. PCR products were visualized using a 0.5% agarose gel. A 406 or 403 bp band indicated presence of the strA or strB genes, respectively.

If isolates were identified with SmR phenotype but the strA and strB genes were not present, the rpsL gene was investigated as a possible source of resistance. PCR amplification was performed in 25 µl volumes as previously described by Tancos et al. (2016) and Russo et al.

(2008), and included 2 µl of template DNA, 12.5 µl EmeraldAmp GT PCR Master Mix (Takara

Bio Inc.), 1 µl each of the forward and reverse primers (rpsL212-F and rpsL212-R), and 8.5 µl

H2O. Cycling parameters were 5 min at 94°C; 35 cycles of 94°C for 30 s, 53 for 30 s, and 72°C for 30 s; and 7 min at 72°C. PCR products were visualized using a 0.5% agarose gel. PCR products were purified using ExoSAP-IT PCR Product Cleanup Kit (Applied Biosystems) and sequenced at the Cornell Biotechnical Resource Center in Ithaca, NY using Sanger sequencing.

Sequences were examined for the point mutation conferring SmS or SmR phenotypes using CLC

Main Workbench (v 20, CLC Bio, Qiagen).

135

Whole genome sequencing of select E. amylovora isolates

In order to further investigate a subset of isolates whose CRISPR regions were unsuccessfully amplified using the methods above, whole genomes were sequenced. Sequencing libraries were prepared using Nextera Flex prep kit (Illumina, CA) according to the manufacturer’s instructions. The prepared Flex libraries were sequenced using Illumina Miseq

(Illumina, CA) on Illumina V2 kit (250 x 250) following the manufacturer recommended protocol. The fastqs were quality filtered and CRISPR regions were annotated manually in CLC

Main Workbench by identifying spacers and repeat regions based on a library of known spacer sequences and patterns provided by Tancos and Cox (2016). WGS was explored for anomalies that would prevent amplification and primer target sites were determined to be absent. New primers were designed to amplify the CR1 region using Primer3 primer design tool (Untergasser et al. 2012).

Draft genome assemblies were created from the raw sequence data using skesa v2.2 with default settings (Souvorov et al. 2018). The program kSNP3, with k=19, was then used to detect nucleotide differences among the isolates (Gardner et al. 2015). kSNP3 is a k-mer based variant

(single nucleotide polymorphism; SNP) detection method where variants are inferred from the assemblies. We discuss the results from the core SNP matrix (i.e., no missing data where only positions for which every isolate had a nucleotide state are used). The program FastTree was used to infer a phylogeny based on the core SNP matrix (Price et al. 2018). Two strains originating from non-apple hosts were included as outgroups: Ea646 collected from raspberry and Ea472 from hawthorne.

136

Evaluation of the aggressiveness of select strains of E. amylovora

In order to investigate possible differences in aggressiveness between strains of the pathogen, eight isolates of E. amylovora representing different strains (as determined by CRISPR patterns) from the major apple production regions of NY were evaluated in a controlled greenhouse experiment using potted apple trees at Cornell AgriTech in Geneva, NY. Strain selections represented the most prevalent CRISPR patterns of E. amylovora isolates obtained from samples in NY State as well as several control strains commonly used lab and field experiments (Table 6.1). One-year-old ¼” caliper ‘Gala’ trees on G.41 rootstock were purchased from Wafler Nursery (Wolcott, NY). Trees were grown in 18.9 L pots in Lambert LM-16 Bark

Mix (Griffin Greenhouse Supplies, Auburn, NY), in a greenhouse with conditions maintained at

22.2°C/18.3°C daytime/nighttime temperatures with 12 h of light and 12 h of dark. At the time of potting, trees were headed to approximately 30 cm above the graft union in order to promote vigorous shoot growth from several lower buds, and trees were thinned to 4-5 healthy shoots after two weeks. Trees were grown for four weeks, until shoots were approximately 20 cm in length. Isolates of selected strains, stored in 50% glycerol at -80°C prior to use, were streaked on

CG agar and incubated at 28°C for 48 h. Cultures were then re-suspended in LB broth overnight while shaken at 28°C, and suspensions were diluted to 1 x 108 CFU ml-1 in phosphate buffered saline immediately prior to inoculation using a SmartSpec Plus Spectrophotometer (Bio-Rad

Laboratories, Hercules, CA, USA). To facilitate conditions conducive to disease development, greenhouse conditions were adjusted 24 h prior to inoculation and for the duration of the experiment to 26.7°C/21.1°C daytime/nighttime temperatures and mist risers were installed and set to run for 5 min every 6 h. Ten replicate shoots for each treatment (strains), arranged in a completely randomized design, were inoculated as described by Singh et al. (2020). In short,

137

sterile scissors were dipped into the bacterial suspension, or sterile deionized water for untreated control trees, and used to bisect the youngest unfolded leaf across the midrib. Shoot lengths were measured at inoculation and the length of necrotic lesions were measured 2, 3, 4, 5, 6, and 7 d after inoculation using calipers. Percent lesion length was calculated as lesion length divided by total length of the shoot at inoculation multiplied by 100.

In order to determine the overall aggressiveness of different strains on potted trees, as indicated by overall disease development over time, AUDPC was calculated for percent lesion length as described by Madden et al. (2007). AUDPC was estimated as follows:

푛−1 푦 + 푦 AUDPC = ∑ ( 푖 푖−1) (푡 − 푡 ) 2 푖+1 푖 푖 in which n is the number of assessment times, y is disease intensity, and t is the time or distance from the original measurement. In the present study, n was equal to six (the number of evaluations following inoculation, y was evaluated as percent lesion length, and t was the number of days post-inoculation (2-7 d). Mean AUDPCs were subjected to one-way analysis of variance

(ANOVA) for a completely randomized design using the ‘aov’ command (Chambers 1992).

Means were separated using the Tukey HSD test at the α = 0.05 level of significance. All data analysis was conducted using R ver. 3.6.3 (R Core Team, 2020).

138

Table 6.1. Area under the disease progress curve (AUDPC)z for strains of E. amylovora evaluated for aggressiveness on potted 'Gala' trees on G.41 rootstock in a greenhouse setting at Cornell AgriTech in Geneva, NY.

CRISPR Sm Isolate profile phenotypey AUDPCx control na na 0 ± 0 a 273 4:21:38 SmS 0.93 ± 0.1 bcd 4001a unknown SmS 1.05 ± 0.05 cd 17.2 41:23:38 SmR 0.73 ± 0.11 bcd 88-100 unknown SmR 0.55 ± 0.13 b 1210 4:27:38 SmS 1.12 ± 0.06 d 1133 2:22:38 SmS 0.63 ± 0.1 bc 1101 53:27:38 SmS 0.98 ± 0.13 bcd 1116 4:23:38 SmS 0.77 ± 0.08 bcd P-value <0.0001

zTo evaluate strain aggressiveness, the newest unfolded leaf was bisected with scissors dipped in a bacterial suspension of a given isolate at 108 CFU ml-1. Length of necrotic lesion was measured each day after inoculation for 7 days and percent lesion length was calculated as lesion length divided by total shoot length multiplied by 100. yStreptomycin sensitivity phenotype. SmS: Sensitive, SmR: Resistant. xAverage AUDPC ± 1 standard error of 10 replicate inoculations per isolate. Different letters indicate significantly different means.

Results

Collection of fire blight samples and isolation of E. amylovora

From 2015 to 2020, total of 461 samples were collected, and E. amylovora was successfully isolated from 327, as confirmed by colony morphology and PCR amplification of pEA29. E. amylovora isolates represented 75 commercial and research farms in the major production regions of NY State (Western NY (WNY, 226), the Champlain Valley (CV, 45), and the Hudson Valley (HV, 34)), as well as Vermont (VT, 13), and Massachusetts (MA, 9). Within

139

NY, samples represented 18 counties; a majority of the samples came from WNY, with high sample collections several counties including Clinton (34), Niagara (22), Orleans (33), Ontario

(59), Orange (22), and Wayne (89). Sample collection was greater in these counties, representative of the greater number of orchards present with more land in apple production, greater fire blight pressure due to environmental conditions conducive for disease development in these years, and greater motivation by stakeholders for testing due to concern for streptomycin resistance. The cultivars from which isolates were most commonly obtained included ‘Gala’,

‘McIntosh’, ‘Fuji’, ‘Ida Red’, and ‘EverCrisp’, as well as several cider apple cultivars. Six isolates were obtained from commercial plantings of Bartlett pears.

In 2015 and 2016, samples were processed exclusively for the purpose of assessing streptomycin sensitivity, as a routine surveillance effort. A total of 158 samples were received in these years, and E. amylovora was isolated from 89. From 2017 to 2020, following renewed concern about streptomycin resistance detection and its distribution, the volume of samples collected increased and evaluations expanded to include both streptomycin sensitivity monitoring and strain tracking by CRISPR profile evaluation. In these years, a total of 303 samples were received and E. amylovora was isolated from 238.

Determination of CRISPR profiles

Of the 238 E. amylovora isolates obtained from 2017 to 2020, CRISPR profiles were determined for 164 isolates (Table 6.2). A total of 22 CRISPR profiles, defined as the combination of the CR1, CR2, and CR3 spacer patterns for a given isolate (written

‘CR1:CR2:CR3’), were detected. Across all CRISPR profiles, a total of 12 patterns for CR1, 14 for CR2, and 2 for CR3 were detected (Figure 6.1).

140

These 22 CRISPR profiles include several new spacers, patterns, and profiles were discovered. A total of 28 new spacers were discovered; spacers were considered new if they differed from previously described spacers by more than 4 base pairs. The two new spacers 800 and 801 were found together, located together at the 5’ terminus of the CR2 region. These spacers resulted in the new pattern 61, which was present in two isolates, 1214 and 1247. New identified spacers 802-809 were found in isolates in which the CRISPR arrays were unable to be fully sequenced. Spacers 810-827 were found in 4 isolates obtained from HV samples, and resulted in new patterns for both CR1 (pattern 67 and 68) and CR2 (69 and 70) in these isolates.

In addition to the new patterns resulting from these new spacers, several new patterns for CR1 and CR2 were discovered containing novel orders of previously described spacers. In addition to the new patterns resulting from new spacers, described above. For CR1 two new patterns were discovered due to novel spacer orders. Pattern 59 was found in two isolates collected from two nearby farms in WNY, and Pattern 60 was found in four isolates collected from the same orchard block. For CR2, three additional patterns were found with novel spacer orders (62, 63, and 64), each of which was found in one location.

141

Table 6.2. Clustered regularly interspaced short palindromic repeat (CRISPR) profiles of E. amylovora isolates collected from commercial apple orchards in NY from 2017 to 2020.

CRISPR Number of Profile Countiesz Regionsy Years Isolates 2:22:38 Ulster (2) HV (7) 2020 (7) 7 Columbia (4) Orange (1) 4:21:38 Wayne (6) WNY (21) 2017 (1) 26 Ontario (15) VT (4) 2018 (9) Chittenden (4) CV (1) 2019 (10) Albany (1) 2020 (6) 4:22:38 Tompkins (5) WNY (5) 2018 (2) 7 Clinton (2) CV (2) 2020 (5) 4:23:38 Chittenden (3) VT (3) 2019 (5) 5 Orange (2) HV (2) 4:24:38 Chittenden (1) VT (1) 2020 (1) 1 4:27:28 Chittenden (3) WNY (15) 2018 (2) 26 Monroe (2) CV (7) 2019 (7) Niagara (2) VT (3) 2020 (17) Norfolk (1) MA (1) Orleans (4) Wayne (7) Clinton (6) Saratoga (1) 4:57:38 Wayne (1) WNY (1) 2020 (1) 1 4:62:38* Wayne (2) WNY (2) 2020 (2) 2 4:63:38* Worcester (1) MA (1) 2020 (1) 1 5:27:38 Worcester (1) MA (1) 2020 (1) 1 40:27:38 Worcester (2) MA (2) 2020 (2) 2 40:65:38* Wayne (1) WNY (1) 2020 (1) 1 41:23:38 Ontario (40) WNY (64) 2020 (43) 64 Wayne (24) 2019 (9) 2018 (8) 2017 (4) 41:64:38* Ontario (1) WNY (1) 2018 (1) 1 43:27:38* Clinton (1) CV (1) 2020 (1) 1 53:27:38 Wayne (2) WNY (2) 2019 (2) 2 59:24:38* Niagara (2) WNY (2) 2020 (2) 2 60:22:38* Ulster (4) HV (4) 2020 (4) 4 67:69:38* Orange (1) WNY (1) 2018 (1) 1 68:70:38* Orange (3) HV (3) 2018 (3) 3 71:66:38* Clinton (1) CV (1) 2020 (1) 1 4:61:39* Niagara (2) WNY (2) 2020 (2) 2

*An asterisk indicates new CRISPR profile discovered in this work. zCounties, Regions, or Years in which the isolate was identified. Number of isolates is indicated in parentheses. yMajor production regions of apples in NY State and Northeastern US. CV: Champlain Valley of NY, HV: Hudson Valley of NY, MA: Massachusetts, VT: Vermont, WNY: Western NY.

142

Figure 6.1. Spacer patterns of clustered regularly interspaced short palindromic repeat (CRISPR) regions CR1, CR2, and CR3 for E. amylovora isolates used in this study, collected from 2015 through 2020 from commercial apple orchards. Spacers are represented by boxes, uniquely numbered at top of each column. White boxes indicate absence of a spacer in a given pattern. Spacer numbers that are shaded (teal) are newly discovered in the present study. Spacers are arranged in patterns, depicted in individual rows. Patterns are uniquely numbered at the left hand side of each row. Text color of pattern number indicates who and when it was identified: black, McGhee and Sundin (2012); red, Tancos and Cox (2016); teal, this study. GAP indicates no spacer present between two adjacent repeat segments. Adapted from Tancos and Cox (2016).

Overall, eleven new CRISPR profiles were discovered; sequences of CR1, CR2, and CR3 regions of isolates with novel patterns were submitted to the National Center for Biotechnology

Information database (NCBI) and given a reference accession number (Appendix:

Supplementary Table S6.2). Similar to the criteria for identifying new spacers, a profile was considered new if it exhibited one of the following features: novel spacers present, a new pattern of spacers within a CRISPR region, or a previously undescribed combination of CR1, CR2, and

CR3 patterns. Ten of these were the result of the new patterns described above, resulting from new spacers and/or novel spacer patterns (4:62:38, 4:63:38, 40:65:38, 41:64:38, 59:24:38,

60:22:38, 67:68:38, 68:70:38, 71:66:38, and 4:61:39). The one additional new profile (43:27:38), which was found in one isolate (1365), was a new combination of previously described CR1,

CR2, and CR3 patterns. The new CR1, CR2, and CR3 patterns, CRISPR profiles, and spacer sequences were compiled with those previously described by Tancos and Cox (2016) and

McGhee and Sundin (2011) into a master reference collection for future work (Appendix:

Supplemental Figure S6.1 and Supplemental Table S6.3). At the time of this work, a cumulative total of 44 profiles have been described, including 801 spacers organized into 25 CR1, 18 CR2, and 2 CR3 patterns.

The majority of isolates were from 2020 (97), followed by 2019 (33), 2018 (26), and

2017 (5). These isolates represented 18 counties in NY, with most from WNY (116), and fewer from CV (15) and HV (17), as well as VT (11) and MA (5) (Figure 6.2 and 6.3). Only 3 of 22 distinct profiles (13.6%) were identified in more htan 7 samples and were obtained from a broad geographic area. The most prevalent was 41:23:38, which was identified in 64 isolates collected from 14 farms, followed by 4:21:38 (26 isolates from 6 farms), and 4:27:38 (26 isolates from 15 farms) (Table 6.2).

Figure 6.2. Map of fire blight prevalence in counties in NY State from 2017 to 2020. Darker shade indicates a greater number. A. Relative number of Erwinia amylovora isolates obtained by county. Shades are in increments of 10 isolates. B. Relative number of clustered regularly interspaced short palindromic repeat (CRISPR) profiles identified by county. Shades are in increments of 2 profiles. C. Presence of streptomycin resistant (SmR) and sensitive (SmS) E. amylovora detected within a county, indicated as presence of SmS only (blue/lighter shade), or both (purple/darker shade).

145

Figure 6.3. Number (A and C) or Percent (B and D) of clustered regularly interspaced short palindromic repeat (CRISPR) profiles identified for E. amylovora isolates collected from major apple production regions in the Northeastern US (A and B) or Counties in NY State (C and D) from 2017 to 2020. Different colors indicate different profiles.

146

Characterization of streptomycin resistance in E. amylovora isolates.

Streptomycin resistance was detected in 70 isolates, as determined by growth on CG media amended with streptomycin. In all these isolates, the mechanism of resistance was the strA/B gene pair, which was determined by successful PCR amplification of these genes. All these isolates exhibited CRISPR profile 41:23:38, and this profile was always associated with

SmR phenotype. Of the SmR isolates, 40 were from Ontario County, 26 from Wayne County, 3 from Orleans County, and 1 from Clinton County. In Ontario County, the SmR isolates were from only two orchards, representing 7 non-contiguous orchard blocks, while the SmR isolates from Wayne County were from 11 orchards, representing 21 non-contiguous blocks. The SmR isolates in Orleans and Clinton Counties were collected in 2015 from 3 farms; it was not identified in samples from the same farms in subsequent years.

The distribution of SmR isolates was followed over the course of 2017 to 2020. In 2017 and 2018 it was identified only in one orchard in Ontario County. In 2018, it was identified in 9 isolates from 3 orchards, representing 6 separate orchard blocks. In 2020, it was identified in 43 isolates from 13 farms, representing 32 separate orchard blocks.

Whole genome sequencing of select E. amylovora isolates

Whole genome assemblies were created for 19 strains collected in this study and previous work (Table 6.3) representing diverse geographic locations of collection. Sequences of these genomes were submitted to NCBI and given a reference accession number (Table 3). Based on the SNP matrix and phylogeny inferred from it, the two outgroup strains originating from non- apple hosts (Ea646 collected from raspberry and Ea472 from hawthorn) separated from other isolates by 29,793 and 1,566 SNPs respectively (Figure 6.4). The remaining 17 isolates clustered

147

into three groups, segregating primarily by geographic origin. The first group included two isolates collected from Washington State (WA1 and 88-100) and differed by 95 SNPs; the second group included the four isolates collected from the Hudson Valley of NY (982, 983, 984, and 985) and varied by 47 SNPs. The third group consisted of the remaining 11 isolates, which were collected from the Champlain Valley of NY (967 and 965), Western NY (272, 273, 249,

254, 300-2a, 685b, 974, and NY17.2) and Ontario (4001a); which varied by 91 SNPs.

148 Table 6.3. Characteristics of E. amylovora genomes sequenced in this study

Alternate CRISPR Accession Isolatea Name Regionb Country Year Host profile Smc Mechanismd Groupe Number Reference Ea646 646 Quebec Canada .f Raspberry . . . Outgroup 1 . Singh and Khan 2019 Ea472 472 WNY US . Hawthorne . . . Outgroup 2 . Singh and Khan 2019 WA1 EaWA2018a WA US 2018 Pear . . . WA . Singh and Khan 2019 88-100 88-100 WA US 1988 Apple . SmR rpsL WA Loper and Henkels 1991 982 HV18-1 HV US 2018 Apple 67:69:38 SmS na HV This study 985 HV18-2 HV US 2018 Apple 68:79:38 SmS na HV This study 983 HV18-10 HV US 2018 Pear 68:79:38 SmS na HV This study 984 HV18-11 HV US 2018 Apple 68:79:38 SmS na HV This study 967 CV18-E1 CV US 2019 Apple 4:27:38 SmS na WNY/CV This study 974 IR2-5 WNY US 2019 Apple 41:64:38 SmS na WNY/CV This study 965 CV18-CM1 CV US 2019 Apple 4:22:38 SmS na WNY/CV This study 301 301 WNY US 2012 Apple 41:23:38 SmS na WNY/CV Tancos and Cox 2016 272 272 WNY US 2012 Apple 51:27:38 SmS na WNY/CV Tancos and Cox 2016 NY17.2 NY17.2 WNY US 2002 Apple 41:23:38 SmR strA/B WNY/CV Russo et al. 2008 4001a Ea266 Ontario Canada 1980 Apple 4:21:38 SmS na WNY/CV Norelli et al. 1984 254 254 WNY US 2012 Apple 41:23:38 SmR strA/B WNY/CV Tancos and Cox 2016 685b 685b WNY US 2014 Apple 4:27:38 SmS na WNY/CV Tancos and Cox 2016 300-2a 300-2a WNY US 2012 Apple 41:23:38 SmR strA/B WNY/CV Tancos and Cox 2016 249 249 WNY US 2012 Apple 41:23:38 SmS na WNY/CV Tancos and Cox 2016 Ea273 273 WNY US 1980 Apple 4:21:38 SmS na WNY/CV Norelli et al. 1984 aName used in this study. bRegion from which isolate was obtained. CV: Champlain Valley of NY; HV: Hudson Valley of NY; WA: Washington; WNY: Western NY cStreptomycin resistance phenotype, as indicated by growth on media amended with 100ppm streptomcyin sulfate. SmS: sensitive; SmR: resistant. dMechanism of streptomycin resistance. rpsL: mutation in the rpsL gene; strA/B: presence of this gene pair. eCluster within phylogenetic tree. fIndicates unknown.

Figure 6.4. Phylogenetic tree of 19 E. amylovora isolates representing diverse geographic origin and host plant. Isolates are colored by geographic origin and host, Yellow: Hudson Valley of NY; Red: Champlain Valley of NY; Pink: Western NY; Blue: Washington State; Green: Non-apple host. Isolate name reflects sample number, isolate name, year collected, and geographic origin.

Evaluation of the aggressiveness of select strains of E. amylovora

Disease symptoms developed on all potted trees inoculated with E. amylovora indicating that all strains were virulent on ‘Gala’ apples. None of the control trees inoculated with sterile deionized water became infected. There was a significant effect of isolate (treatments) on disease determined by AUDPC (P < 0.0001, Table 6.1). For all isolates evaluated, AUDPC was greater than the controls inoculated with sterile water. The most aggressive of these isolates was 1210

(CRISPR Pattern 4:27:38), which led to a larger AUDPC than 88-100 (unknown CRISPR

Pattern) and 1133 (2:22:38). Isolate 4001a (unknown CRISPR Pattern) was the second most aggressive, which led a larger AUDPC than 88-100. Isolates with intermediate aggressiveness included NY17.2 (41:23:38), 1116 (4:23:38), 273 (4:21:38), and 1101 (53:27:38).

Discussion

Distribution of strains of E. amylovora

Over the last six years, we identified many isolates of E. amylovora with known or novel

CRISPR profiles from fire blight outbreaks in NY. Prior to this work, a total of 36 CRISPR profiles had been identified (McGhee and Sundin 2012, Tancos and Cox 2016). A majority of the isolates in the present study had one of the previously described profiles. However, of the 22 profiles we identified in this study, half were newly described. Additionally, the CRISPR profile of several of the isolates collected could not be identified. The current methods and primers used in CRISPR profile identification were unsuccessful at amplifying and sequencing the CR1 and

CR2 regions, but successfully amplified CR3, indicating that these isolates have novel sequences in the CR1 and CR2 CRISPR regions. Due to the discovery of new diversity in all of the

geographic regions we investigated, it is likely that even more strain diversity exists within E. amylovora populations in the Northeastern US.

CRISPR profiles varied greatly in the extent of their geographic distribution. A few

‘cosmopolitan’ CRISPR patterns (4:21:38, 4:27:38, and 41:23:38) were present in all or most of the sampled regions. Previous investigations have also identified these profiles in numerous samples across wide geographic regions. It has been suggested that the most likely explanation for the pervasiveness of these isolates is that these particular strains became established at central locations for the distribution of plant material (i.e. nurseries) and then disseminated to commercial orchards in diverse geographic locations via plant material (McGhee and Sundin

2012; Tancos et al. 2016). In addition to being the result of coincidental circumstances, these strains may also possess adaptations that make them predisposed to greater geographic distribution. For instance, more resilient strains capable of surviving in suboptimal environmental conditions (i.e. water availability, pH or salinity fluctuations, temperatures) could survive longer in adverse conditions; more aggressive strains would be able to establish infections more quickly; and strains having a broader host range would have greater opportunity for survival and movement. Such investigations could be conducted in follow-up studies. In contrast, many ‘resident’ CRISPR profiles were only present in one or two isolates collected from a single orchard site. This included five in Western NY (4:57:38, 4:62:38, 40:65:38,

53:27:38, and 4:61:39), two in the Hudson Valley (60:22:38, and 67:69:38), three in the

Champlain Valley (41:64:38, 43:27:38, and 71:66:38), and three in MA (4:63:38, 5:27:38, and

40:27:38). Often, these profiles were newly described. The HV, CV, and MA are areas in which less sampling and CRISPR profile investigation has taken place in the past. In these areas, we detected a high proportion of new profiles, as well as unidentified isolates, indicating novel

152

isolates and high diversity within each given region. Most likely, these isolates have been geographically isolated, without either the opportunity or the ability to be transported over great geographic distance. In all, each production region that we investigated appears to have its own unique strains.

The novel profiles and strain diversity we identified in the Northeastern US, and in NY in particular, were consistent with our expectation. The Hudson Valley is considered the center of origin for E. amylovora, and therefore likely a center of diversity for this pathogen (van der Zwet

2012). In addition, the bacteria require living host tissue to be distributed over larger geographic distances and time spans. This has been facilitated primarily by human propagation of cultivars of Malus x domestica, which has only occurred relatively recently in evolutionary history (van der Zwet 2012). New CRISPR profiles were identified in isolates from almost all the major apple production regions from which we received samples, indicating there is likely even more diversity in pathogen populations across the Northeastern US, especially from production regions that we have not yet sampled.

The whole genome analysis of 19 select isolates and our large-scale CRISPR profile investigations led to similar conclusions, consistent with those in other works. Overall, we detected low genetic diversity between the samples, as indicated by SNPs. Similar results have been found in other, more comprehensive and detailed analyses of E. amylovora genomes, which have typically detected over 99% similarity between isolates (Smits et al. 2010; Mann et al.

2012; Zheng et al. 2018). In our work, CRISPR profiles tended to cluster by geographic origin: those originating in Washington (WA), the Hudson Valley of NY (HV), or the Champlain Valley and Western NY (CV/WNY). Less than 100 SNPs were detected within clusters, but groups separated by over 1,000 SNPs. This provides further justification for our observations from the

153

CRISPR analyses. There appears to be geographic separation of E. amylovora strains and/or unique strains within a given region. A larger separation was observed between hosts, with the two isolates collected from non-apple hosts separating by nearly 300-fold the number of SNPs.

Other studies have also found that E. amylovora isolates from Spiroideae or Rosacea hosts are the most genetically distinct (McGhee and Sundin 2012; Rezzonico et al 2011; Singh et al.

2019). Moreover, the isolates for which we conducted whole genome analysis, and all the isolates included in the CRISPR profile investigations, were collected from commercial apple orchards. Exploration of CRISPR profiles of isolates from alternate and wild hosts would likely yield even more diversity. However, finding wild hosts with fire blight infection often proves to be difficult and it may likely be the monoculture condition that promotes high level of conspicuous disease. Overall, we believe CRISPR profile investigations provide an economical way to identify isolate diversity consistent with a full genome analysis, and could act as a starting point for identifying isolates suitable for more thorough investigations.

One particularly interesting finding was the new spacer pair 800 and 801. These spacers were found in two isolates (1214 and 1247) collected from nearby locations. The resulting new pattern (61) matched pattern 29 except for the addition of these new spacers, which were located adjacent to each other at the 5’ end of the pattern. This is interesting because spacers are accumulated chronologically with the newest at the 5’ end (McGhee and Sundin 2012), so the

800 and 801 spacer pair may be an indication of evolutionary change in these particular isolates as they interacted with novel foreign DNA, as compared to isolates with pattern 29. In this way it may be possible to use CRISPR patterns to describe evolutionary relationships of isolates.

Also of interest was the identification of two isolates with CR3 pattern 39 (1237 and

1369), both from apple hosts. To date, only two patterns have been described for CR3 (38 and

154

39). In previous work, pattern 38 has always been associated with isolates obtained from apple, pear, and quince hosts, while pattern 39 has been associated with isolates obtained from Rubus hosts (McGhee and Sundin 2012). This finding may provide evidence for isolate crossover between hosts. Further investigation is warranted to better understand the extent of pathogenicity and host range for these particular strains and others.

Distribution of streptomycin resistant isolates of E. amylovora

The new discovery of streptomycin resistant E. amylovora again in 2018 caused concern within the apple industry in New York and was one of the main motivations for this work.

Unlike apple producers in other regions, such as Washington, Oregon, and California, those in

NY State and the Northeastern US have maintained streptomycin as the primary management tool for fire blight. Following the outbreaks in 2002 and 2011-14, diligent efforts to identify and control SmR E. amylovora led to what was believed to be successful eradication (Tancos et al.

2016). Unfortunately, our investigations detected SmR E. amylovora on several commercial apple orchards in NY State from 2016 to 2020, indicating it is still a threat to the NY State industry. These farms were confined to two counties (Wayne and Ontario), the same counties in which the earlier SmR outbreaks occurred. In the present study, only one CRISPR profile

(41:23:38) was associated with SmR. Consistent with the previous investigations in NY, this

CRISPR profile was always associated with SmR (never SmS). In addition, this is the same profile that was identified for the original SmR isolates collected in 2002 and a large majority of the SmR isolates collected in the survey efforts from 2011 to 2014. These findings provide strong evidence that the SmR E. amylovora originally introduced into NY State was never fully eradicated. Most likely, it went undetected during years in which fire blight pressure was low,

155

due to environmental factors not conducive to infection events (i.e. cool temperatures during bloom), and growers were able to control the disease satisfactorily under these conditions with other management practices. In 2018 and 2019, only a few isolated SmR were uncovered, enough to raise concern, but not great enough to affect industry production. However, in 2020 overall fire blight pressure was extremely high, due to very warm weather over a prolonged period during bloom. As a result, much more fire blight was reported throughout the state and control failures were more conspicuous. Overall, relatively few SmR isolates were identified and

SmR was only found in a very small geographic area. This provides confidence that much of industry is practicing effective and sustainable resistance management in regards to fire blight, such as making only necessary, well-timed applications of streptomycin and rotating with other antibiotics like kasugamycin and biopesticides. However, the intensified detection of SmR in

2020, a year with higher fire blight pressure, has renewed the importance and urgency of continuing to survey for SmR and practicing resistance management. Fortunately, our routine fire blight screening efforts provided rapid detection of SmR; identification and feedback to growers typically occurred within 2 weeks, providing enough time for growers to respond within the season with appropriate management actions. In addition, other management tools are available and better understood, including the antibiotic kasugamycin (McGhee and Sundin

2011) and the plant growth regulator prohexadione-calcium (Wallis and Cox 2019). Further research and education on antibiotic alternatives will be necessary to protect streptomycin as an invaluable management tool and provide control options where SmR is present.

It is important to note that two of the locations where SmR E. amylovora was detected were research institutions: the USDA Apple Germplasm Repository and Cornell AgriTech

Research Orchards, both of which are located in Geneva, NY. While these are not commercial

156

operations, it is arguably more important that SmR E. amylovora was detected and appropriate control measures were implemented at these locations. The collections and research conducted at these locations informs all aspects of apple production both in NY State and worldwide; protecting these assets is critical to the global industry (McCandless 1995; Robinson 2002). In addition, these are central locations for transfer of knowledge as well as plant material. This unfortunate discovery has provided a unique opportunity to emphasize the importance of continued screening and education about fire blight management and streptomycin resistance.

Strain aggressiveness and implications

In a subset of E. amylovora isolates representing the most prevalent strains (as indicated by CRISPR profiles) across commercial apple production regions in New York State, we found that all were virulent on ‘Gala’ but had differences in aggressiveness. We observed significant differences in aggressiveness among these isolates, with aggressiveness defined as percent lesion length developing on inoculated shoots of potted trees under greenhouse conditions. Of the eight strains evaluated, strains 4001a and 273, two well-studied strains originally isolated from Simcoe

Ontario Canada and Wayne County NY, respectively (Norelli and Aldwinckle 1986), were included as reference controls. In the present study, both 4001a and 273 were found to be highly aggressive, with 4001a the more aggressive of the two. This agrees with other studies, which have characterized aggressiveness of numerous strains of E. amylovora using immature fruit, detached flowers, and shoot inoculation of greenhouse plant assays (Cabrefiga and Montesinos

2005; Khan et al. 2018; Lee et al. 2010; Norelli and Aldwinckle 1986). Differences in aggressiveness have been attributed to growth rates, amylovoran production, and biofilm formation, and the resulting differences in the time to symptom development (initiation of the

157

disease progress curve) and rate of disease progression (Cabrefiga and Montesinos 2005; Lee et al. 2010; Norelli and Aldwinckle 1984).

The most aggressive isolate in our study was 1210 with CRISPR pattern 4:27:38. This strain is widely distributed in Northeastern North America and is one the most prevalent in NY fire blight surveys between 2011 and 2020. The wide distribution of this CRISPR profile could reflect the ability of the strain to colonize hosts more efficiently and spread more rapidly. The least aggressive strain in our experiment was 88-100, a streptomycin-resistant isolate originating in WA. Resistance in this isolate is conferred by the point mutation in the rpsL gene, resulting in a structural change in the binding site of streptomycin. Lower aggressiveness in this isolate may indicate a tradeoff in fitness (Agrios 2005). This strain has not been identified in the

Northeastern US; the narrow distribution of the strain may be attributed to its relatively low aggressiveness. The other isolate exhibiting an SmR phenotype included in our work was

NY17.2 with CRISPR profile 41:23:38. Streptomycin resistance in this isolate is conferred by the presence of the gene pair strA and strB, which code for an aminoglycoside transferase. In surveys from 2011-16, this was the most common CRISPR pattern identified for SmR isolates, and from 2018-20, this was the only profile for SmR isolates. NY17.2 was found to be moderately to highly aggressive in our study, indicating no tradeoff in aggressiveness and fitness. Neither of the SmR isolates exhibit higher aggressiveness than the other most common

CRISPR patterns distributed in NY State, indicating no need to assign different threat levels to

SmR isolates in extension communications aside from the streptomycin resistance. As discussed above, the differences we observed in aggressiveness between isolates may partially explain why some strains become ‘cosmopolitan’ while others remain ‘residents’ of a certain geographic area.

158

Further exploration of pathogen aggressiveness needs to be explored, especially as related to isolates originating in different geographic regions and with different CRISPR profiles.

In summary, we used CRISPR profiling of isolates of E. amylovora collected from commercial apple orchards to describe strain distribution across the major apple production regions in NY and the Northeastern US. Greater diversity was uncovered than previously described and it is likely that more diversity exists in this region. The distribution of the CRISPR profiles across several major apple production regions indicated that some strains tend to be

‘cosmopolitan’ having a wide geographic distribution or ‘residents’ of a specific farm or region.

Through our sampling efforts, we identified streptomycin resistant E. amylovora, previously believed to be eradicated from NY State, and provided a timely description of the distribution and most likely method of introduction. Results from this work have improved our understanding of the distribution of E. amylovora strains across the Northeastern US, with implications for strain diversity, aggressiveness, and associated management solutions.

Literature Cited

Aćimović, S. G. 2015. Control of fire blight (Erwinia amylovora) on apple trees with trunk- injected plant resistance inducers and antibiotics and assessment of induction of pathogenesis- related protein genes. Frontiers in Plant Science. 6:1–10.

Agrios, G. 2005. Chapter four: Genetics of Plant Disease pages 125-176 in: Plant Disease 5th ed. Elsevier Academic Press, New York.

Bonn, W. G. and van der Zwet, T. 2000. Breeding for resistance to fire blight. Pages 253-273 in: Fire Blight: The Disease and Its Causative Agent. J.L. Vanneste, ed. CAB International, Wallingford, UK.

Cabrefiga, J., and Montesinos, E. 2005. Analysis of Aggressiveness of Erwinia amylovora Using Disease-Dose and Time Relationships. Phytopathology. 95:1430–1437.

Chambers, J. M., Freeny, A and Heiberger, R. M. 1992. Analysis of variance; designed experiments. Chapter 5 of Statistical Models in S eds J. M. Chambers and T. J. Hastie, Wadsworth & Brooks/Cole. 159

Chiou, C.-S., and Jones, A. L. 1995a. Expression and identification of the strA-strB gene pair from streptomycin-resistant Erwinia amylovora. Gene. 152:47–51.

Chiou, C., and Jones, A. 1995b. Molecular analysis of high-level streptomycin resistance in Erwinia amylovora. Phytopathology. 85:324–328.

Coyier, D. L. and Covey, R. P. 1975. Tolerance of Erwinia amylovora to streptomycin sulfate in Oregon and Washington. Plant Dis. Rep. 59:849-852.

Crosse, J., and Goodman, R. 1973. A selective medium for and a definitive colony characteristic of Erwinia amylovora. Phytopathology. 63:1425–1426.

Djaimurzina, A., Umiralieva, Z., Zharmukhamedova, G., Born, Y., Bühlmann, A., and Rezzonico, F. 2014. Detection of the causative agent of fire blight - Erwinia amylovora (Burrill) Winslow et al. - in the southeast of Kazakhstan. Acta Hortic. 1056:129–132.

El-Goorani, M. A., and El-Kasheir, H. M. 1989. Distribution of streptomycin resistant strains of Erwinia amylovora and control of fire blight disease in Egypt. J. Phytopathol. 124:137-142.

Förster, H., McGhee, G., Sundin, G., Adaskevag, J. 2015. Characterization of Streptomycin Resistance in Isolates of Erwinia amylovora in California. Phytopathology. 105(1):1302-1310.

Gardner, S. N., Slezak, T., and Hall, B. G. 2015. kSNP3.0: SNP detection and phylogenetic analysis of genomes without genome alignment or reference genome. Bioinformatics. 31(17):2877-2878.

Johnson, K. B., Sawyer, T. L., Stockwell, V. O., and Temple, T. N. 2009. Implications of Pathogenesis by Erwinia amylovora on Rosaceous Stigmas to Biological Control of Fire Blight. 99:128–138.

Johnson, K. B., and Temple, T. N. 2013. Evaluation of Strategies for Fire Blight Control in Organic Pome Fruit Without Antibiotics. Plant Disease. 97:402–409.

Khan, A., Desnoues, E., and Clark, M. 2018. Bacterial Strain Affects Cultivar Response to Fire Blight in Apples. NY Fruit Quarterly. 26:15–20.

Madden, L. V., Hughes, G., and van den Bosch, F. 2007. The Study of Plant Disease Epidemics. St. Paul, MN: The American Phytopathological Society, APS Press.

160

Mann, R. A., Smits, T. H. M., Bühlmann, A., Blom, J., Goesmann, A., Frey, J. E., et al. 2013. Comparative Genomics of 12 Strains of Erwinia amylovora Identifies a Pan-Genome with a Large Conserved Core ed. Jesús Murillo. PLoS ONE. 8:e55644.

Manulis, S., Zutra, D., Kleitman, F., Dror, O., David, I., Zilberstaine, M., et al. 1998. Distribution of streptomycin-resistant strainsof Erwinia amylovora in Israel and occurrence of blossom blight in the autumn. Phytoparasitica. 26:223–230.

McCandless, L. 1995. New York State Agricultural Experiment Station 115 Years of Service to New York Food and Agriculture. https://ecommons.cornell.edu/handle/1813/17319

McGhee, G. C., and Jones, A. L. 2000. Complete Nucleotide Sequence of Ubiquitous Plasmid pEA29 fromErwinia amylovora Strain Ea88: Gene Organization and Intraspecies Variation. Appl. Environ. Microbiol. 66:4897–4907.

McGhee, G. C., and Sundin, G. W. 2012. Erwinia amylovora CRISPR Elements Provide New Tools for Evaluating Strain Diversity and for Microbial Source Tracking ed. Ching-Hong Yang. PLoS ONE. 7:e41706.

McGrath, M. J., Koczan, J. M., Kennelly, M. M., and Sundin, G. W. 2009. Evidence that prohexadione-calcium induces structural resistance to fire blight infection. Phytopathology. 99:591–596.

McManus, P. S., Stockwell, V. O., Sundin, G. W., and Jones, A. L. 2002. Antibiotic Use in Agriculture. Annual Review of Phytopathology. 40:443–465.

Mojica, F., Díez-Villaseñior, C., García-Martínez, J., and Soria, E. 2005. Intervening Sequences of Regularly Spaced Prokaryotic Repeats Derive from Foreign DNA Elements. J. Mol. Evo. 60:174-182.

Moller, W. J., Beutel, J. A., Reil, W. O., and Zoller, B. G. 1972. Fireblight resistance to streptomycin in California. In Abstracts of Papers Accepted for Presentation at the Sixty-Fourth Annual Meeting of The American Phytopathological Society, Mexico City, Mexico: Phytopathology, p. 779.

Moller, W. J., Scroth, M. N., and Thomson, S. V. 1981. The Scenario of Fire Blight and Streptomycin Resistance. Plant Disease. 65:563–568.

161

Norelli, J. L. 2003. Fire blight management in the twenty first century. Plant Disease. 87:756– 765.

Norelli, J. and Aldwinckle, H. 1986. Differntial susceptibility of Malus spp. cultivars Robusta 5, Novole, and Ottawa 523 to Erwinia amylovora. Plant Disease. 70(11): 1017-1019.

Norelli, J., Aldwinckle, H., and Beer, S. 1984. Differential host x pathogen interactions among cultivars of apple and strains of Erwinia amylovora. Phytopathology. 74(2):136-139.

Park, D. H. 2017. Current status of fire blight caused by Erwinia amylovora and action for its management in Korea. Journal of Plant Pathology. 99:59-63.

Ponce de León Door, A., Romo Chacón, A., and Acosta Muñiz, C. 2013. Detection of streptomycin resistance in Erwinia amylovora strains isolated from apple orchards in Chihuahua, Mexico. Eur. J. Plant Pathol. 137:223-229.

Price, M. R., Dehal, P. S., and Arkin, A. P. 2010. FastTree 2 – Approximately Maximum- Likelihood Trees for Large Alignments. PlosOne. 5(3): e9490.doi:10.1371/journal.pone.0009490.

R Core Team. 2020. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/.

Robinson, T. 2002. The New York State Apple Research and Development Program: Ten Years of Successful Research Support for the Apple Industry. Fruit Quarterly. 10(2):3-17.

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 Disease. 92:714–718.

Saad, A. T., Hanna, L., and Choueiri, E. 2000. Evaluation of streptomycin and oxytetracycline resistance of Erwinia amylovora populations in Lebanon. (Abstr.) Phytopathology 90(suppl.):S68.

Shaffer, W. H., and Goodman, R. N. 1985. Appearance of streptomycin-resistant Erwinia amylovora in Missouri apple orchards. (Abstr.) Phytopathology 75:1281.

Sholberg, P. L., Bedford, K. E., Haag, P., and Randall, P. 2001. Survey of Erwinia amylovora isolates from British Columbia for resistance to bactericides and virulence on apple. Canadian Journal of Plant Pathology. 23:60–67.

Singh, J., and Khan, A. 2019. Distinct patterns of natural selection determine sub-population structure in the fire blight pathogen, Erwinia amylovora. Sci Rep. 9:14017.

162

Smits, T. H. M., Rezzonico, F., Kamber, T., Blom, J., Goesmann, A., Frey, J. E., et al. 2010. Complete Genome Sequence of the Fire Blight Pathogen Erwinia amylovora CFBP 1430 and Comparison to Other Erwinia spp. MPMI. 23:384–393.

Souvorov, A., Agarwala, R., and Lipman, D. L. 2018. SKESA: strategic k-mer extension for scrupulous assemblies. Genome Biology. 19:153.

Sundin, G. and Ehret, G. 2010. Fire blight control on Jonathan apple with antibiotics in 2009. Online publication. Plant Dis. Manage. Rep. 4:PF006.

Tancos, K. A., and Cox, K. D. 2016. Exploring Diversity and Origins of Streptomycin-Resistant Erwinia amylovora Isolates in New York Through CRISPR Spacer Arrays. Plant Disease. 100:1307–1313.

Tancos, K. A., Villani, S., Kuehne, S., Borejsza-Wysocka, E., Breth, D., Carol, J., et al. 2016. Prevalence of Streptomycin-Resistant Erwinia amylovora in New York Apple Orchards. Plant Disease. 100:802–809.

Thomson, S. V., Gouk, S. C., Vanneste, J. L., Hale, C. N., and Clark, R. G. 1993. The presence of streptomycin resistant strains of Erwinia amylovora in New Zealand. Acta Horticulturae. 338:223–230.

Untergasser A, Cutcutache I, Koressaar T, Ye J, Faircloth BC, Remm M and Rozen SG. Primer3--new capabilities and interfaces. Nucleic Acids Res. 2012 Aug 1:40(15):e115. van der Zwet, T. 2012. Fire Blight: History, Biology, and Management. St. Paul, MN: APS Press.

Wallis, A. E., and Cox, K. D. 2020. Management of Fire Blight Using Pre-bloom Application of Prohexadione-Calcium. Plant Disease. 104:1048–1054.

Zeng, Q., Cui, Z., Wang, J., Childs, K. L., Sundin, G. W., Cooley, D. R., et al. 2018. Comparative genomics of Spiraeoideae-infecting Erwinia amylovora strains provides novel insight to genetic diversity and identifies the genetic basis of a low-virulence strain: Comparative genomics of Erwinia amylovora. Molecular Plant Pathology. 19:1652–1666.

163

CONCLUSIONS

The overarching goal of this work has been to add to the scientific and commercial understanding of fire blight, a bacterial disease of pome fruits caused by Erwinia amylovora, in order to facilitate more sustainable management practices for the commercial industry.

In Part I, we evaluated the impacts of antibiotics, the current industry standard for fire blight control in the Northeastern US, and alternative strategies on disease management and orchard health. First, we evaluated the effects of a novel management strategy, applying the plant growth regulator prohexadione-calcium (PhCa) prior to bloom, on disease management (blossom and shoot blight) and horticultural impacts (shoot growth and yield) in field trials at Cornell

AgriTech. Disease reduction with this program was comparable to streptomycin applications, with minimal effect on tree growth and productivity, and may offer as an alternative to conventional management practices. However, the effectiveness of this treatment may depend on many variables, and further investigations in various geographic regions, cultivars, and planting systems are necessary. To this point, we have initiated trials evaluating the efficacy of this strategy at research orchards in two different production regions under disease pressure. In addition, we have six field trials at commercial orchards evaluating the horticultural impact on young high-density orchards during establishment. Results thus far are promising, with no detected effect on growth or productivity after two consecutive years of treatment. These pre- bloom programs would fit with standard production practices and may contribute toward the development of fire blight management programs without the use of antibiotics.

Next, we evaluated the impact of streptomycin applications for fire blight management on endophytic bacterial communites of the phyllosphere. These communities have the potential

164

to impact orchard health and harbor antibiotic resistance genes. Overall, results indicated minimal impacts on these communities. A greater difference was observed between the two orchards investigated, indicating a strong selection by locality. Interestingly, we found a high abundance of Amoebophiliaceae, a family dominated by arthropod parasites, which could indicate horizontal transfer of endosymbionts between insect and plant hosts. This may be an interesting avenue for future research, regarding apple and insect interactions with microbial communities. Consistent with the growing body of work evaluating the effects of streptomycin in an orchard environment, we provided evidence that such applications have minimal effect on endophytic bacterial communities, and that commercial practices are sustainable solutions for fire blight management.

In Part II of this work, we investigated the distribution and spread of the pathogen at multiple scales. First, we evaluated the risk of spreading the disease using mechanical thinning and mechanical pruning in two orchards. Contrary to previous work, we found that these practices could be used safely, without drastically increasing fire blight incidence or spread.

Development of fire blight beyond the inoculated point source was rare, and any risk was mitigated by applying streptomycin after mechanical treatment. Timing in these trials was important, with thinning occurring early in bloom (king bloom) and pruning occurring after terminal budset in August. These times correspond with prudent horticultural practices. Our results provide promising evidence for the safe use of invaluable labor and time saving practices for commercial use. In fact, in our survey of growers in Western NY, we found that many have already adopted mechanical thinning and pruning, without experiencing fire blight problems. We hope that this work will help inform and enable use of these practices for the industry.

165

Next we investigated spread of fire blight within two orchard blocks using spatial analyses and CRISPR profile characterization. Our results indicated two very different sources of introduction (a nearby block or planting material) and patterns of spread (from a corner of the field or from focal points within the block). Describing the distribution and spread of fire blight within and between orchard blocks has the potential to improve our understanding of disease movement, inform appropriate management recommendations, and facilitate traceback efforts.

Finally, we utilized CRISPR profile characterization to describe the geographic distribution of strains of E. amylovora and streptomycin resistance across major production regions in NY and nearby states. This work was a six-year follow-up to previous work that was conducted following streptomycin resistant outbreaks in 2002 and again in 2011-2014. We found that several ‘cosmopolitan’ strains were widely prevalent across regions, while other ‘resident’ strains were confined to one location. In addition, we uncovered novel CRISPR profile diversity in all investigated regions. SmR E. amylovora was detected only in a small area spanning two counties from 2017 to 2020, and always associated with one CRISPR profile (41:23:38), which matched the profile of SmR E. amylovora strains discovered in 2002. This suggests the original

SmR E. amylovora was never fully eradicated and went undetected due to several seasons of low disease pressure in this region. Overall, we found CRISPR profile characterization to be an effective tool for conducting strain tracking and traceback investigations. In the future, improving efficiency of CRISPR pattern identification and development of a central database and tracking system would be invaluable for facilitating such efforts.

In conclusion, the work herein has improved both the scientific understanding of fire blight and the commercial management of the disease. We hope that it has built upon the great body of work that has been dedicated to developing more sustainable management strategies.

166 APPENDIX

SUPPLEMENTAL FIGURES AND TABLES FOR CHAPTER 6

Supplemental Table S6.1. Primers used in this study to confirm presence of E. amylovora, identify streptomycin resistance mechanism, and amplify CRISPR regions.

Target Purpose Primer Sequence (5'-3') Source pEA29 confirm presence of E. amylovora AJ75 CGTATTCACGGCTTCGCAGAT McManus and Jones 1995 AJ76 AACCGCCAGGATAGTCGCATA McManus and Jones 1995 strA evaluate SmR mechanism strA406-F TGACTGGTTGCCTGTCAGAG Russo et al. 2008 strA406-R CGGTAAGAAGTCGGGATTGA Russo et al. 2008 strB evaluate SmR mechanism strB403-F ATCGCTTTGCAGCTTTGTTT Russo et al. 2008 strB403-R CGTTGCTCCTCTTCTCCATC Russo et al. 2008 rpsL evaluate SmR mechanism rpsL212-F CGTACGCAAAGTTGCAAAAA Russo et al. 2008 rpsL212-R GGATCAGGATCACGGAGTGT Russo et al. 2008 CR1 Amplify and Sequence CR1 region CR1-F1 CGCCGCCACGCTGCCATTT McGhee and Sundin 2012 CR1-F2 ATAAACCGCAAGCGATCAACCTGT McGhee and Sundin 2012 CR1-F3 GCTTATACAACTGACAAAATCGTG McGhee and Sundin 2012 C1-R0 TCCAGCGCCTGTAAAGCGGC McGhee and Sundin 2012 C1Uni-F AGCCGAYTTTYCCYGYTTTRAG McGhee and Sundin 2012 CR1-R2 ATGGTATCCAGCGCCTGTAA This study CR2 Amplify and Sequence CR2 region CR2-F1 GCGGCCAACAGATGCGGAAAG McGhee and Sundin 2012 CR2-F2 GTCTGGCGCAAAAACTGGAG McGhee and Sundin 2012 CR2-F3 CCGCCCTTCTGGTGTTTTGA McGhee and Sundin 2012 CR2-R1 TGCGGGGAACACTCGACATCTAAT McGhee and Sundin 2012 CR3 Amplify and Sequence CR3 region CR3-F1 TTTTCGCCGGGTAACAGG McGhee and Sundin 2012 CR3-R1 ATGAGAAGCCCGTGAAGCAAAGTA McGhee and Sundin 2012

Supplemental Table S6.2. Accession numbers for CR1, CR2, and CR3 regions of E. amylovora isolates identified in this study with new CRISPR profiles.

Isolate CR1 CR2 CR3 974 MW221391 MW221392 MW221393 982 MW221394 MW221395 MW221396 983 MW221397 MW221398 MW221399 984 MW221400 MW221401 MW221402 985 MW221403 MW221404 MW221405 1214 MW221406 MW221407 MW221408 1217 MW221409 MW221410 MW221411 1219a MW221412 MW221413 MW221414 1247 MW221415 MW221416 MW221417 1279 MW221418 MW221419 MW221420 1280 MW221421 MW221422 MW221423 1282 MW221424 MW221425 MW221426 1284 MW221427 MW221428 MW221429 1302 MW221430 MW221431 MW221432 1353 MW221433 MW221434 MW221435 1354 MW221436 MW221437 MW221438 1365 MW221439 MW221440 MW221441 1369 MW221442 MW221443 MW221444 1375 MW221445 MW221446 MW221447

Supplemental Table S6.3. Spacer CRISPRtionary (page 1 of 6)

Number Sequence Number Sequence 1 TTATTCATGAGCCTTTTTATCTTCGCGGCATG 61 GACGCTCAAATCAGTGGCGGCGAAACCCGACA 2 GTAAATAGCAAAATGATAAATAATTTATCAAT 62 CCAGAGGGGATTTAGCAAACGTCATTTCTGAC 3 CTATGCAGAAGCGGAGGGCGGCGAGTGATGGA 63 TCATCTGCGGGTCGGGTAGGCTGCTTACGGGT 4 AGCATCTCGGGAACTGTGTTTTTTGTATAAAA 64 CAGCTATTCCCCGCATCGGTCAGTACTGCGCT 5 AAGATGCTTTGACATTAATTATCTCCATAAAA 65 GCCAATGGATTCAGGATTGGAGCCAGAATTTA 6 CAAGCGATCAACCTGTTTTTCAGTAGGTTTAA 66 AAAAAAAGCCTAAAGCTCGAAAGAATAAAAAT 7 GATTGCGCATGAGCACTGAAATTGTTCACAGC 67 ATGATGGCGCTGATAGTTTTATTAGATGTCGA 9 ACAAAAGACAACACCCCCCTTACCCCCCCACG 68 TAAATGGTTGTCCGTTCTTGGCGCAGACGGCT 10 CAGGTATTTCGGATAGCCGGTTGTCTCGGCGG 69 GCTACTACGTGTACGCACAGCCGCTGGCCAGT 11 ACTGAAATTTAAAATCACCGCTAACCCGCCAG 70 TGGCCCACAATGGTAAAACCGGCGGCTTTCCA 12 GGCGATGAGGGAGTACGCGGAGCGGCAGGGTA 88 GATTTGCGTGAACTGCATATCCGCCGGGCAGC 13 AAAAGCCAACCGCCCGCCCGTAATAAACCTGA 89 ATGGACAAAGGGCGAATGCTCCACATGATGGA 14 GTTGCAGAGACTTAAAGATCGTCTGCTAGTTA 90 CCGCTTTCGGATGAGACGGTACGTGAGTTAAC 15 TAAAGGAGCATGCTTATACAACTGACAAAATC 91 ATTCGGCATTTTTATCCCTCACAACATGCGTT 16 AGATTTGGCGGAAATGTCGGCGGAGATGCCCC 92 ACCCTGCTGCTCTGAATGATGTGTGTCAGTGA 17 AAATGTCCTGTGGCTCGGCCCGATGCTGCAAT 93 GTCGTGGGGCGAAAGGCACAGACCGGATGCTT 18 GAGATCATTCTCATCCCTCATGTTTTCCAGGA 94 TGAACGAGCGCGAGGCACCTGACGGTGCAGAC 19 ATTGTAAAATCCTCTCCGCCAAATTTGATTAC 95 GTGGTGACCTCTTTGTGGGCGGCGTTTCTGGG 20 AAACTCTCGCATACATGGACGGAATTTAACGA 96 GTTCACTTAGATTCTCAATCGTTGCAATCATG 21 ACGATTTGCCTGAAACCTCAACGAAGTTCGAC 97 GTTCTTTAATAATGTTTATGGCCTTTAGCGTA 22 CTGATGGCGTCACGAGCCATACGGAATGTGAC 98 CGTACCCTGCAGACGGCAAATAACTACACGAA 23 CAAAAATTTGCGCATGTCATCTATCTTTTTTT 99 GATTTCTCCCCGCATGCGGGCTGCACTCAGAA 24 CCCTCGGGGAGGGCTTTGCGTTGTTACTCAGA 100 GTTTCGGCGAGCGCGTATATGGTCACGTTCAC 25 GTTACGTTGAATGTATCGTTGGATGTGATTAA 101 ATGGCCCGCTAAATGTTGACATGTCTGGTCGG 26 TACATCGAACAATGCCAATTGTTGACGTTCTT 102 TCTTTATGACGCAAAGCAAGCCTCAGCCGAAC 27 CCGCGAAAATCCGCAGTGAGCTGGCAATGAGC 103 CGCGAGTACCCATCCATCCCCGCAGAGGCATT 28 GCTGTCTATCTGGGCTGCCTCTATCCAGCAAT 104 GCAGGGCCGGTTTACGTTGCGCAATCGGAGAT 29 ACTTCGGTGAGAATGTCGAATTGCCACCAGAT 105 GCGGCGAAGAGACCGGAGCATGGGCTGTTGAA 30 TTGAATCAGAGTCTTTCAGGGACGATGTTTTC 106 CCCATTTCCCTGATTTTCCTGGGATTATTTCT 31 TGAAGCAGCCAGAATCCCATCCGGCCTTTATC 107 CGGGCATTAGCGGCTTTGAAACGAGAACTGGA 32 GCTTTTGTACCCTTTACAGTCAACGTACTGCT 108 CACGATCACACTGTCAGCTAGATTTTATGATG 33 AGATTGAGATCTTATCAACGGACTCTGACGCC 109 AATATCTAGTGTTTATGCGTGCCTTTTCTGGC 34 TTTCTTCACACACAACGGTGAGGGCATTGTCT 110 CAACTGAGCAACTCCGCTTTGCCCCATACCAA 35 TGATAAAGTAACGTTCCGAATGGCGCGTGATG 111 TTTGTCCCAGATTTCCAGCGCGTCGAACTGAC 36 CCATTTTATGACAGTCTGGCGCAAAAACTGGA 112 CGGAAAAGTGCGACAGAGAATTAGCTCCACTC 37 GAGATGCACTGGATATACCGACTCCTCACTGA 113 AAATCCAGATTCTCCGCCACCCAAGAAGCGGT 38 GCTCGGGGGGACATGAGCTTGTACAAAACAGC 114 CAGGCCAGCGACTACCCCACGCAGAAGCGCTA 39 TTTTTAGCAGCGTGACAGTTATGGAGCCGCTC 115 TGCGTCAATTTTGCACGCATGTTCCGTTAAGT 40 ATTATAAGGATCACTTGCTAGGGCATTATATA 116 TTTTTTAATTGCGTGCTGCGTTGCAGGGCATG 41 ACTTGACTGTTTATGCAGTGGTTGTATTTCTT 117 GCCAGATGCTTATTATCTGGCATCCGTGGCCC 42 ATCGGAACGACTTAGATTAGCGTCCTTGCACAT 118 CTGCGCCACTTTATGCTCCATGACCTCCAGAG 43 TAACCGCAACCATCGCCGCGATAAATCCACTG 119 TCGTCCAGATAATGGGAACACCGCGCAGGACG 44 GTTGATACGGCTGATTACAATAAAATGTCACT 120 ATCAGTCAGAATTTGAAAAAAACCCCGCCATT 45 TAAAAAATGCCGCCATCGAATCAGCAAAATCG 121 AATGCAGGTGCGCTGAAGAGAATTCATAATCA 46 CTGCGGAGCGTCAAACGGGCGTTAACTCTCGA 122 GGAGTTTATCTGAATGGTCTCGCGCAAAACTC 47 CCCTTCTGGTGTTTTGATTCTCCTAGGTGATT 123 GGGATCCGCTATTCTTTACCCTACAGTATACA 48 AATGGACGAGATTTCACAGAAAATATCTGTTC 124 CTGATGCCGCTGAACCTTCCGGGTGGTTTCC 49 CAGATGAGGCTGCAAATTCCAGGCACTTTTTA 125 ATGTAGGCCACACTTTAATTTTTGGGCCAAAT 50 GATGGTCGTACCGATGTTTGCGAAAGATTCGC 126 TAAAATGGCTTAGCGCTGCACATGAACGTGCT 51 ACGGTCAGATGGTGGCGCTGGTTGCGCTGGCA 127 AGCTGGGCGTAAACCTGCAAAGCGAATTTGAA 52 CTGTTTATGAAAAATGCCAACAAACAGGAAGC 128 GCTTCAATTAACGCGGCTGGCTCCTTTGAGAA 53 ATTTTTCAGGAACGGGCCGACACGAAAATTTAT 129 GGACATTAATCACCGTCCTGAATTATGCCGAA 54 ATATTTACTAGCATTTCCCCATGCTGTATCAC 130 CTGCTATGCGCTGGTCAGAGGGGGAATAAGGA 55 CTGGAGCATGAGACGAAATCGGGGGTAGTGCT 131 GCAGTCTTTCAGAACTCCAAACGCCTTGGTAA 56 CCGGTTCAGGTTTGATAGGTTCTGCCTAACTC 132 ACCGGACTAACGGCAATAGCATCCGCCACCGC 57 TTCGCATACGACAATCTCCCGGCACTGATTAA 133 CCAGCTCAGCCGTGAGCAATCAATCCTCATTG 58 AGGGTGACGCAACGATTGTTGCAATTCCTAAC 134 AGCCAGAGCTAATTACAAACTTCTGCCGGCTA 59 CACCAGTGTGTACATTCCAGACTCAGAAACCAC 135 CAGCAGCAGAGAATCGAAATAACGGGATTGGA 60 CCTCGAGGTGTTCTAAGCACTCCGGGGCTTTT 136 CAATCCAAACCATACGCCTGCTGCATATGTTC

169

Supplemental Table S6.3. (Page 2 of 6)

Number Sequence Number Sequence 137 CGTAGATGAAAGCTTTATTCATGATGGGACGT 197 CGTTTAAGAGCTGTAACAAAAGCCAGGATGGA 138 AAAAAGGATTGCGCATGAGCACTGAAATTGTT 198 CTGTCTATGCCAACGAATACCTGCCGCAGGTT 139 GCATTGGAGTTCTTCCGTTCACATGAGCTGGT 199 AACCATAGTTCCTATAATTATCTTTCTGTTGT 140 CGTCGCGGCGTGGACGTCAGGGTGGTAGTGGA 200 ACCATCAGTTCCCTAATAATGATATCGGTTAT 141 TATTTCTGAGCGATGCTATTCGACTGCTGAC 201 GCGGCGATTTAAAACCGGGCGCAATATACAAA 142 TTTCAGGAGGCATTCAAAATGACCGCTGGGAT 202 CCTTGGCTGGAGCCGTGACAGGGCAGCTGACG 143 CCGCATGTTCTTTTAACATGGCTTTAATATTA 203 ACATTTGTTTGGTTATACTTCTTATCGTGATG 144 CGGCACACCCGCCTGAACTAACCAGCTCGCCC 204 ATGAGGTGCGTTTTCTCCGGGGTTGTGAAACC 145 GCAGCTGGCAATGATGACGAATGAGGATAGAC 205 ACGTGTCGATCGCTGATGATATCTCATCGTTC 146 GGGGAATAAAGATATTAATCAGGATTTGCGTG 206 CATCACTCTAGACTCGTCAACTGGCAACCCCC 147 TGCAGGCGCTGCAGGAGGGGGAGAACGTCACC 207 TCACCGAGCAATGGGATAAGTGCGGCGTCGAT 148 CTGAAAATGAGTGGTGGGCAGCCGGGAGTTAG 208 TGTTGGGGTGCTCCCCCCGCCCCCGCCGCCTG 149 TCATAGTATTCCATCACTACGACACCCGTTAC 209 CCAGCTGAGTCAGACTACATCTGCAAATCCTG 150 AATGCCAGTACGAGCGCGCCTTAAATATCTGG 210 TACTGCGACAGCGTAACGTTAGCCGGTAACAA 151 TTGGCCACCAACAACAACGCATGGCATGTTTA 211 TGACTGAACCGTGTAACCCGCAGCGAGGCCAG 152 CCAGAGGACAACTTTCCGCCTTTTCTCTGCGT 212 ATATTTCAGTGGCGGGCTGCCATATGAAACAA 153 AGGTGGCGGAATACGGCAACGAGTGGATAACT 213 TTGGGGTGAGTAAATACGACTCTGTTAGAGCC 154 GTCGGGGTATATACGGTACTGGACAAATATTT 214 ATTTGGGGCGCAGTGGTTTACTGGCTTCCGGC 155 ACGCATCACCCGAACGTCTCTTCCGTCCACTT 215 GAGAGCAACAACAAACTCCGACCGGAGAAACC 156 CAGGTTGTAGTATGCTCATCCGATGATCCATT 216 TCATTTTTATAGCTAGAATTTTCCTTTGCTTC 157 GGGCCTTTACGACACGCCAGAGGCTTTCCCGT 217 TTCCTGCTGTGATTGCAGCAAAACCTAAATTT 159 CACCGGCAGTAGCTGGCGATGCTGTATTTGCA 218 CTAATCATAGGTGTCGCTGCAGTCGTTGGGAT 160 TTAACGGGGCCCACTAAAACTGTATAGAGTCT 222 TGTATCTGTCAGTGTGAGAATCTGCGCGTTACT 161 TAACTGCTGCCAATGACCGCAGCCATAATTAA 223 CTCTGCATATCCATTGTGTCTTTGTCTGACAT 162 CCAAATGCCGGTAGCCCATATAATGATTCCTT 224 GAACCTGCTTGCGGTCAAATACACGCTTCCAG 163 ACATAAACATGCTTAACTGGTATCTGGCCTGT 237 CCAGTTCGACGCGCTGGAAATCTGGGACAAAC 164 TTAATCTGATAATTTTGATCACATCTTAACCA 238 AAGGCAGCAGGCTGATTGCATACCCTGATGGC 165 CATAAAATCCATCAATTCTATTTATTACGAAC 239 CATCCCCAGCAGCCCACGCAGCAGCCCACGCA 166 ATAAATGAGGCGCTGCCAGAGCACCGTGATGG 240 CGGTCACCGTCGAGGCAGCATATAGGGCGAAT 167 GTTTCAGGAGTCTGGAGGGCAAGCCGATGCCC 241 CCTTATGAGTCGCAGTAATTCTGCCATCTCTG 168 AAAGGGGTTGTCGGAATGATTCTGGTCGATTA 242 AAATAAGAAACCAAACGAAGGACCCAGAAAGT 169 TCACCATCCTGTTCGATTAAGCAAAAATACTC 243 CTGCGACATCAGAACCGGAGCAGAAGCGCCGC 170 GGCGGGGGCAGGCAGTGCACAGGAAAATGACA 244 CGTAATAAGCGTTCCCACTGCCACGCCGAGAA 171 CGGATCACCTCTTCGCCCAGCCAGCTGTTCAG 245 AAACCAAATCTGTTTATCTTCTTCGCCACTCA 172 CATCTTGCACTAATGAAAGCAATGATGAGGGC 246 AGCGCCGGGGCGAGTAACGCGCAGATTCTCAC 173 ACAGGTGCTCTAAACACACGTCCGGGTGCCGG 247 GTTATTAGCAATAAGGGCATGCAGGGCAGTGG 174 GAAAAATTATGTATCGCGTATATCTGCGCCAC 248 CCAATTAACGGAGGTGCTTATGACAGTAGTTA 175 TAGCCAGGTTAACACTCACAAAAATATCACTC 249 TCGCTAAGCAGCTCGCCGATGATGGATATATT 176 CTGCATAAGTTCTGCCATCTCTGACCATGAAA 250 GAAGCGTTTACAGGATCGGTAAATCACGGTTT 177 TATTGACCTAGAGTTTGAGCAACGTAAGATTT 251 AGAGCGGGAACTCTTTTGCAGCTTAGTGCGAA 178 TGAATCAGCATACTTATCAAAGAACGGCGTCT 252 CACGGCCCCCGTACTATTGAATCCACCCGTGT 179 CAACTAATGACCTCGTTTATATGATATTCGGCC 253 CGGACAGTTTTCTTGCTCAATGTCAGTTTTCA 180 GAAGGAGACGAGGAACAGGATTAGGTATACAT 254 CCCAAAGGAACAACGAGAGCTATCCAAACCAA 181 AAGCAGAAGTGACAGTAACGCTCTGTTTGACT 255 TTGAGCACGGTAACCCTCGCCACCCGGCAGCT 182 ATAAATTCAGGATAAAATATGAATGAAAAAAT 256 CCAGAAAATGAAAGGGCAGCAACGGCGCAGGAG 183 TAGCAAGGGCTTATTACTCTCACAATCCACGA 257 CAGAGCGGGTCAGATTCAGGGATGTATGTCA 184 TCACCACTGGCCAGCGGATCCCGGGGACCGGT 258 GATGCCATTGCACWKGCYAAAKCGCAGTATGR 185 GTGCTGATGGAAACCATCGAGAAGGCCGGGGC 259 ATTTCAKGCTTGCCGCCCTCGCCTACCCGGTA 186 TACACTGCAGGCAGTACGGGAATGGTTAACCC 260 TACCGAGACAATTTCTCTGCTGTTTCAAGAGA 187 CCATTCCCGAGCCATCGTCGCGACATCCAGGC 261 GGCTTTGCGTTGTTACTCAATATGACGAACTA 188 AAAGAAACTACCCCCAACCCCCTAACGGGTGT 262 CTGATTTTAGCGTTAAATATTTTCCTGAATTA 189 CGACTATTCGGCGAACTGAAAGAGCCAACCGT 263 CGGAAACGACATATTATATGTACGTTCGCACA 190 ACGCCGCCATTCGGATCCGGGGAGGATTCGGA 264 CGAAGGAAATTGCGGGAAACAGCGAGCAGGCC 191 AATGCCTCATACCTGGCGGCCAGCCTTCAGCG 265 GCTCGCGCTGGCAATGTCTGCTCGAGATATAT 192 TACCGCCAGTGGGCCATTGCATCCGGCATGGC 266 AGCTGAATTTCCCCGATGCCGCTGGAATTAAA 193 TAACTACGTCAGCGGTGTCCAGCATGACGGAA 267 TCTATTTCACGTGCCAGCCGGGGATGTGATCG 194 GCAGCAGGTATTTATTGATAGCGCGATTGCAG 268 ATACTCCCTCTCAAGGATTGCGGCTTCTGCTT 195 TGGTTTCTGCCACTACCCACCCTATCTAATGA 269 AGCCCAGGGGGCAATTTCCGGGCTGGAAAATG 196 CTCGCCATCAGTTTTTCACGCTGGCCGGGGGG 270 CCGACAAATCAGCGCTGGGGAATTCCTCGTCG 170

Supplemental Table S6.3 (page 3 of 6)

Number Sequence Number Sequence 271 CGGAAACTCCGCATCCATGCCGAGGTGTTCAT 332 CAACCAGTTTCGTTAGTTGTTTCTGGGATTAA 272 GCCAGCAGTACCGCACCGCAAAGTTAAACGTT 333 GTTGAAAATTGACCATTACAAGAAATTTAAAA 273 GGCAGGAGCGGTGGGACATTCGTAACCAGATT 334 TCGCGGCAAAGCGCCTTTGCCGCTGTATGGC 274 TCCGCGCACGAATCCGCAAAGGGGGCGTTGAG 335 GTGTTCTTCATCTGTATCCACTTCACCCACGC 275 ACAATATTACAAAAGACAAGAAAGACAATAAT 336 CAGGGTGTGTGGCTCACGCGCTTTCATCACGT 276 TCTCCCGCAGCAAAACAGCAATCGAAAAACCC 337 CATAACGCTTCTTTTCTTTGGTTCTTCTGTCA 277 ATTCTTTCACATTCGACTATCGAATTTTGCAG 338 TTTTACGAATATGTCTTGTATCCCGGCTCTGG 278 GTCATATTCAAGGCGAAGTGATGTCGGCCAGT 339 GGAAAAATTATCGAGCTGGCAAATAATAAAGG 279 CCACACGTGCATTGTGGTCCAGGATTAGCCGG 340 GATAATGCGGAGTTTTCGGGCACGCTAGAGGC 280 TAGGTTAGGTGGGGGTGGGGGTTATACAATCA 341 GTGCAGAGATGCCGATATTGACAGATCCAATT 281 AATTAATGCGGATCGATTTTATTAGTTCATCT 342 CGCGATTGAAAGGGCTGTGGTTTATCGTGATG 282 CTAATCCTGTTCGAAATTATCAACACACGGTA 343 TTCCCATACCTTCACGGCCCGACACTTTGCGA 283 TGGTCATCGAGGATGGTTTATTACTCGCAAGA 344 CGCTTAATCAGCTCAACGCAGCGAAGGACGCT 284 CTCGAGTCCGGCAAACTCCAGCAGTTCGCGGA 345 CGCAATTTTTAGACAATGCAGAATTTTATTTT 285 TCATTACCCGCACAGCCAGATAGGGCAAAGAA 346 GATAGGAGCAAGGCAACGCAAAATCCCGTGAA 286 CCCGAGCACGCGCCAGTAATTTATATTACGGA 347 AATGTTTGTGTTAATTGGCTTTTCTGCCTCAA 287 ATGATTGCAACAATGAAAGACAAAGACGAAAG 348 TTATAAAAGCGTTTATTTGTAATAATCGTAAT 288 CAACAAGCGCAGTGATTGCCTGTGAGTCGAGA 349 AGCCTGAATTGCCGGAGATATCTGATGTTGAC 289 CCATCGCCTGAAAAGTACGATGTTATCGATCA 350 CGCTGAAATGTGAATCAGTGCGGTATTTAGCC 290 GATTCAATAAATGTTCGTATATTCCAAATCAA 351 CTCAATCAGGGGACTGATTCCGCGTTAATAAC 291 ATCGTCTGCTAGTTATCAGTGACAATCAAAAC 352 ATAATAAAAGGCCGCGCATAGCGACCTGTTG 292 GACAATAATACCCGGTAGGTGTGGGGTGATGGT 353 TCGGGTATAGTTCATCTCCCATTTTCCAATTA 293 CCGTTGCAGATGACCAGCTTGGAAATGCAATG 354 CTTCGAGATTAGAATATGACTCAGTACCATCTA 296 CGAGCGATCAACCTGCCTTTCGGCACATTTAA 355 GGCACCGCCAATCTTCTGGCCCCAGTGATTAT 297 ATTATCATAACCCTTACATCTCAATAGTATTAAA 356 GAGATAGCTAACATACTTAACTTATCGCCTAA 298 AGATTCAGATCCTATCAGTCGGACTCTGTCACA 357 GTGATAAGCGTACCCTGCAGACGGCAAATAAC 299 TTTTTATCAGCGTGACAGTTATGGAGCCGCTC 358 CCAGCTCCAGTGCTACCGCAACGCCAGATTCA 300 CTGTTTTCGCTGCTATTGAATCCGGGAGTGAA 359 AGTAATTAATGATTCTGAAATCTCTCTTAATA 301 GTCTTGTGTCTGACGATTTCACCATCGCGGCG 360 TTTGGTGACAAAGGATGGGTGCGCGAGGAAAT 302 ACTGCGTGATCACGTAATCGCAGGGGGCATAA 361 AACGTAATCAGTGGGCTGATATGCACGGTCTG 303 CGAGGCGTGACGCAGCTCGCCAGCGCAAATAG 362 CCGCAGTGGTGTTGTGGTTCTTAGCAGGCCGA 304 GACGAGCCATGCCTTGCAGAAAGTTTTATTCG 363 GGTGAGTGTTTAATACTTCCCCTTTGGAGGCA 305 TTCGGGGGAACGGATATCATTACAGCCCGAGC 364 TTCCCGCGTTTCGTGGGATACCCCTTTTATGC 306 TTTACGTTTGCGTTAACAGTAAGCTCTGCAAC 365 CGAACCTATGCGGCATTAGCGGATGCCGCCGG 307 ATCGCACCCCACTGATTGAAGAGCAGCACACT 366 GTGTTCCGCCTGGCTGGGAGAATGTGGATGGC 308 TAGCAATAAATTCGATAGACGCTGATTTGCGT 367 AACTTACGGTGGCCTGGCTTGAACTTCCGGAC 309 TGGCTTCAATTGCGGTCGGGTGTGATGCATCA 368 CCGGGCGAATGCGGGGATTGCAGGGGCAATGG 310 GCCAACGTTCACTGTCATTTAGCCACGCTTCCG 369 GTATTTATAAACGATTCATTAATACGCTTATA 311 GGCTAAATGTCTGAATTCAACGTTCTACAAAA 370 CAGCAACCGGTGGTATTTTTCCGGGTCGGTTA 312 CGATGGGAAAATACTAAAAACTGGAATTGCTG 375 CTGTGTCAGGTTTCGCCACTTCTGACAGATTC 313 ACTCGCTTAGGCTCTGAGTGCCGTTAAGAAGC 378 ATTACTGGTAGTAACTCAGCAGTATTCGGTAC 314 GAAAATTCTAGCTATAAAAATGACTTAACAGA 381 GGTACCTGAAGCACTTCGGAGAATACCCCATG 315 CAATGATGAGGGCAGCAGAACGAGACTGGAAA 382 GCGCGACTGTGTGGTCTCATACCCTCACCAAC 316 TCGTCGGCGCTTTTCTTTGGTGTCAACTTCGT 383 CACTTCGGAGAATACCCCATGTTTATTTTTAT 317 AACGGCTCCCATATCACCCCAGTCTCGTGCCA 384 TTACCACCTGTGACGCGTCACAGTTATTTGCC 318 ACAGCAGAGTTAACACAAAACATACGCAATAC 385 CAGGGGTGAGTTCGAAGCGTGGTGGTCTATAA 319 GTTTTCATCAATATGGATACACGATCACTCAC 386 ACGCTAGGTCTCTGGGATTCTGTCGCCTCGCC 320 TTTTCAAGCCGGAGCTAATGCAACAGAATCTC 387 TGAGTACGCTTCCATAGTTACCTCATTAGATA 321 CMCCGAAAAAMCAATTGTCGTTAGCATGAAGT 388 GTACATAGCCAGCATGCGTGTTCGTTCCTCGT 322 ATACCTTTACGGGTACATCTTGTTTTCCGATT 389 ACGCCATGTACCGCGAAGGGTACCTGAAGCAC 323 TTCGCAGTACACGGTTTATCAGACAGTGTTCG 390 AAAGAACGCAAAAAAGCAATGCAGATAGAGAC 324 CATTTTTTTGATGTGACTGCAAAAAATTGTGT 391 CTATCGTTGATGAGGGCGCTGTGTTTGAAACG 325 CGAAGAGAGCTACCCTGAAACCTCATTGAGAG 392 CGTTTACGTCACGCCGATTTTATGGCAAAGGC 326 CAAAAATTTATTCAGCTGTTAGCAGGATGAAT 393 CAGATCGTAACCGTCCGTCACCGGATAGGGGC 327 CCAGATGTGGACCTGAACTCTGGTAGTCACCA 394 CCTGCACAATTTTGCACAAATAATTTGATGCT 328 TTTTTGTACTGTGAGAAAACGCTGTCAGAGGA 395 TTGCTGGTCCAGAAAAATTCATCAAAATAGAG 329 CCAGTTAGGATGTCTCTTTCATGATTTATATA 396 ATAGGAATGACAACGGGAACGCCTGAGCAAAG 330 CCCGAGCCCATGTGCTAGTGCCGGTAAGAAAA 397 GTAATATGCGAGACATGTCCGCGGTGTTTACT 331 GCGTCCTGGCCATCGAGTGCGTATATCCATGT 398 CATGGGTAAGCAATAAATCCACGTCGGTTTGC

171

Supplemental Table S6.3 (page 4 of 6)

Number Sequence Number Sequence 399 GTGGCTGACTCAGAGTTTGCGAGCATATCTCC 461 AAATCTCTGAGGCTCTCGGCAATAAGACCGTT 400 AGACCGGAGCATGGGCCGTTGAAATTCTTGCC 462 CAAATCTGGCGGCGTGCTGATGGACGAGGCTT 401 GCGACCCGGTCGCCGTGTCAAAAGGCCGAATT 463 GCGGGGAGCACGTTGCTGTGACACATGACCGC 402 CTGCAGCCGCCCACTAAAACGGGGATCAGCAT 464 GATTCAGTCAGGAAAATGCGCATCGGCTTGTT 403 ACGCGGGTTACAGTGACCTCGGCGCAGCAATG 465 GGGATGTCAAAAGAGGCCATAGCTATTGACGG 404 TGGTTGGGCGATGTGAATGGACGTTTGAGCCG 466 CCACATTGCCAGTTTTGGCAGGTTTTTTGAGT 405 ACATGCAAAAGCGCCACTGGCAGCAGCCAATG 467 TGGTCAAAGGCTCGTGGGTGTGGGCAATCCCT 406 GAGTTCCCCGCACCACAGCGCAGCAAAAGCGG 468 CCTGGCCGTTCTGGACCGGCGATCGCTGGAAT 407 ATATAGACATTGTCGGCTTTACTTTTCCATGT 469 ATTGCCGCCTCTAACATGAGCAAGAGCGAGTT 408 CTCAGTGATTCTTTCTTCCAGATTCGGGTCAA 470 ATGCNCATGATTATGATGAATTGGCAAATCAA 409 CTCTATGCTATCGACCACGGCGGGTTGAAAAG 471 TTGCNTGGATAAACCCGGAGGATAGCAAATCA 410 ACAACTGTTTAAAAAAATCAAAAACTCCGAAC 472 ATTGNACATGAACTCCGGGGCTTTGTTTTTTG 411 TAGCCAGCGCAACGAAAATTGACGTTGGAAAC 473 TTATAACGCCAAAAGGTGCAGAGATGCCAATA 412 TGGGTATTGTTAGCCGCCCACTTGTGAATAGT 474 TCTTTAAATCGGCGCTGGGCATGACAGAGATA 413 AGGAAATTAGAGCTATCCAAGATTGAAGGGGA 475 CTTAAACGTTTGTCCCCTTCTGTCTACCTCTG 414 AATGGCCACAGTTCGGCGGCGTGCTGGCGCTG 476 CGTGGGATTCGATAACGTCGCGTTGGCAAAAT 415 GAATTCTTCATTAGCGAGAGGACATACGTATG 477 CATCATGGTGGAAATAGAGGAGTGAAATCATG 416 CCCAGAAAAAACAATTGTCGTTAGCATGAAAT 478 GTAGAAATGCCGCTGTACAGCAGGTTATTGCG 417 TTTTCATTATGGAGACAATACGATCTTCTCCA 479 TTTTCCCCCATTGCCGCTTCAACCATCTGGGG 418 TTTTTCGGGCATATATAAAAAATAGAGTTAAA 480 CCGAAATCAGTCGGTAAGGGGTTTACCTTGCT 419 GTTCGCAAGAGTTGATGCTCTTTTGAACCGCG 481 TTTCGATATCCTCGTTTGGATTATCATTGTCT 420 GGAAAAAATACAGTTGGATTGCAACTAATCAG 482 CCACAAAATTCAGGGTTACTATCAATTTAACA 421 GCTTCACGTTCCCGCCGAATTTTGCAGCTGCG 483 TTGGAAAGAGGCGCGGAATAGAATGGAGTGGG 422 CGGTCAATTTTCCTGCTCAATGTCAGTTTTCA 484 GCTTCCTCGTCAGAAAGTGAATAAAGCAAAAA 423 CCGCCAGCTCCCAGACCTACACCATTGCGCAT 485 ACACCAGGCGTCCACTGAACAAAACCCGGCGC 424 CAATTGATGATGTTGTAATTCGCCTCCATGC 486 CCTCTGATCGATACCGGGCAGTATCGACGGGC 425 CCATAAACACGCATTATTACTTTGTCGTTAGT 487 TACCAACTAGTACGGGTGATGCAGATCCGGTT 427 GCTCTAGACCGATTCCTGTCATGGGCAGAGAT 488 TGTGTCCCTATGGGATGTGTTGCTTCAGATTC 428 TCCATAACGCTGGTTACTCCCCCAAATATTTT 489 GCGGGTCATCTTCAGCTGATGACGCTGATACT 430 CCACAGGATTCAGGGTTACTATCAATTTAACA 490 TAGAGGTATATCGAGCATCTATGGAAATCTTC 431 AATCCACCATTGAGGAATATTCCTGGCCGTAT 491 GTATCGCTGCRCAGTGACCAGTTGTGCTTTTT 432 TGTGCAACTCAATCAGCAAAACAGAGGCTGTC 492 GCTGAAATCAATAACATCTCCTGGATTTACAG 433 CGGTGATGGAGGAGGAGACCAAAAATCAGGGC 493 ATGTCGAACCTGTCGGCGTATCAACTGGGCTG 434 GCGTTCCCCCTCCAGGACGGCTGGGAACCACG 494 CTTTGCCAGGCCTGTGTTTGCCCTGGGAATTT 435 CTGTTCAGCCTTTTTGATGGCTTTTTTCTCTT 495 GAACCATAGAGCCGGGAGATCTAATTTTCGTA 436 TGAGTTTGACGCTTTTGCAGACCATCACCAGG 496 TTGGGCGAGATTACTACCCTTTGGAGGGTTGA 437 GCTGCTTCCAGCTCAGGCCAGATGTTGGTCCA 497 CATAGCGTCTGAAGCGCTCAAAACTGTTAGGC 438 ATATCGGTCAGACCGTCCGCAAGCGTTTCTCT 498 AAGSTCGCCAGCATTACTGGTTTCTCTCAGGC 439 CAATTGATCGGCTATCAGGCAGAGATCGCCAG 499 gGAGCAAAAGGTATTTTATTGATATGCTCTTG 440 TTAGCGCCCTTCTAAGCCGTAGGTCACAGGTT 500 CTCGACACTGATATTAAGCCAGTCGTTATGCA 441 TCGACCCCGCCTCTAACGCCCCGCCAGATTTT 501 GGAGCGTGTGCTGAAATAAAACGATGGGTTTA 442 GCCAACGGAAAATGTTGCTACTCGATCTTTGT 502 STCGTCAGCTGGGGTTTACGACAGAKGTTTGC 443 TTGTCAGTTGCAAACATCAACAGCTCAATTAC 503 CTTGTTATTACGGAATTCTTTCTCTAAATCTT 444 TCCACCAAGAGGAAATTAGAGCTATCCAAGAT 504 GAGGTAGACAGCAGAATACAGGCGTAAAAAAA 446 GACTCTTAACTCATTCTGCCGCTTTCACGGTC 505 CTATTCATAAATAACATACGAGGTTTCTATGA 447 AAAATGTTATAAGCCGTATTCCTTGCGGAACA 506 CAcTGTTCgTCGAgCGtTAATGTAGTAGCAAC 448 GGTTGAATACCTTCAGGGATTTAGAGCTACCA 507 CTCGCCCGAATTATAAATCAGAAGGTGGATCC 449 ATAGTGCTCGACGTTAACGCGTGCGCCGTACA 508 GTACACCTCTCACCCGGTACGACTAATTGCTG 450 GCATTGACATGAAAAATTGTGCTAGCAGTTTT 509 CTGACAGTAACGGAAATCTGACAGTTTCCGGC 451 TTTCCCCGCCCGGTCAGATACAAAGCGGCGAC 510 TGTTCTCCGTCGAAAAACGACGCCAGGATCAG 452 GCTTCTGTTGCGGTCAGGCTGGTTTTCGATGT 511 AGCAGTACAACATGTTCGGCTCGGCAATTAAT 453 CGGTAAAGTTTACAGAATCTGTAAATATTACA 512 GGGATTCAGCGGGTAAGCGCGAGCGTTAACAG 454 GCGATTGCCACCGCCAACCGACTTCAAGAGAT 513 TGCCGGAGCAAGAGACCCCCACAATGGACTCT 455 AAAATAAATTCAAATCAAATAGTTTTATGAGC 514 CATTCCTGAAATTGAGCAGGCAGATAGCTTAG 456 CAGGTGAACAGGCGCAGCTTGTGAGGAACGGA 515 GGCAGAACATGCAAACTTATGACCGCGCAACC 457 AGGTTTAAATGGCTACATATAGTTTTATGGAT 516 ACGTAACACTTATCTTGCTAATGTCGCTGAAT 458 CTGCTCTACCCTGTATGTATAATCAGTATAGC 517 AGGAGCCTAATGATTACGGCGAGTATATCGAT 459 CTATACCAAACACATAAAAACGTTGTGCCTCT 518 CAGACCAAACCATACCCCTGCTGAATATGTCC 460 TAACGGGAAACCCCATGGCAGAAACATTTACA 519 CGTAATCTGATGGCTGATAGCTGGGGTGTTGGT

172

Supplemental Table S6.3 (page 5 of 6)

Number Sequence Number Sequence 520 CCTGTCTGATGGCCTTCAATCAGTTACCTAAC 579 ATTGGTCGCATCGGTGGATCTTCCACAGAAGA 521 TGAGATTGATGCAATATCGGTCATTGTGTCAC 580 GCGCTGGAGTCGTACAGAAAAGAGTGGGACGA 522 CATCCATCGGCAGTTGCGGCTGATATATTTGA 581 GTGTAGGTCTGGATCATCAGGTCACTGCCCTC 523 GTTATGTATGGTTTGCCAGTCTCTTAAATCCA 582 CTGGAAGATGATATTCAGGCCATCACGCAGGG 524 CAGATTGAATCACCACTTTCCGGGAGGCTGGCA 583 ATGGTGTCGCAAAGTGTCGGGCCGTGAAGGTA 525 CCTGCGTGAGTCAGCGAGTCAAAAGGGGCTTT 584 CGGCGCTGTCTATCCCCTGCGCGATCCCCTTT 526 GTTCCCGGCGCTTCGCCACTCTGCCACCACTC 585 ACCCAGCTAGGGAATGTAGGTCTAGTCATGAA 527 TCGGCTATGCCGTGGGGTGAATGATGGAATTT 586 GACGCTATAAGCCAGGCAAAAAGCCAGTATGA 528 GCAAAAAAACATGACTGCATATCTGTATCGGA 587 GGAACCATATTCTTCATAGCCTAGCTTTTCAT 529 TCCAGAAGGGCATAACCTGCATTAACCCGCGA 588 CGGGCGAGTTTTTCAGGATGCGTCTCATTGAA 530 AAGCAAACCAGAATGGGGCTAAGAGTGCGGCA 589 CAATTTGTATACAGTCGAGAACGTGCCTTATG 531 GGCTGCAACCCGACCGCAATTTTGTAGCGCAT 590 AACCTTTTACCATTGTCATAACAATATCTTCA 532 TTCACAGACTGCTTTCTGCTTGCGGCTATATT 591 CCAGGGGAGTTTCTTATTTTCCGGAGAGAATG 533 CAGGAACATTCAATTGACGCTGAAATGGCGGA 592 CGTATTGATCAGATACTTACACTTGGCCAGAT 534 CGGTCGATGCATACGGCACTGTGCCTACGTGG 593 TCATCGCTGATACCTGTTCAGTAGTTAAAACA 535 CGCTAAAGAGGGTGCTGCGACATTCCAGGCGC 594 GCCAGTGCTATCAGCAGGAGATACGGAACCCC 536 CCGCAGAGTTTCGCCGATATGCGCGATAGAGC 595 ATGCAACGTCAGCCCAGTCAAAAGAAATAGGA 537 CATATAAGAGCGGAAAAGCAGTGGTAGATGGT 596 CCAATGACGCCGCAAAACTGCGCTCTATTTCA 538 CCGTACCTGCGCCGGTAACGGTGGGTAACTTT 597 ATCTAGGTTTGCCGGTTTCAGGATCATCAGGA 539 AAATTTGATGATTTCCTGAAAGCGAAACTTTA 598 GTAGGGAAGAATAAAAATAGTTATGATGGAAA 540 ACGCAGTCGCAGCTGATTATGGGCGGCACGGC 599 TTAATTTTATATTCATTTCTGCAGTCTCCAGC 541 TTGGAACGTCTGAACGCCACGCGCACGCAGGA 600 CCAGGATAGTCCGTTTTGATACCGTTAATCAG 542 CGGAACATCCGCAGATGCACGTAACCGACCCG 601 CGAAAAACATTGCACAAACCGAGCAGGTGGGG 543 CCCTGAGGAAATCAGCAGTGGTCTGATGAACC 602 GTATTTATTTTTGCTCTACTTGCCGGGATGTC 544 CCCGTCAACAGGACAATCAATATCTGGAGCAG 603 CCGGTTAACGTGCTTAAGGGCATCACTGAGAT 545 CGATGCCATTAGTGCAGACAGTAGGATTAGTG 604 ATATCGCATTGATTTTATGAAATTATCTAGTG 546 GGAGGTTCCTGGTGGGTGGCGAAAATGCAGAA 605 AGCGGGACAGGTATATCAGATCAACTCTGGCA 547 CTGGATAAATGCGTAATTAGAGCGCTGTGAGG 606 TCCGTTCATTGGACAACAGAACGCTTCAACAT 548 GGGTTGTCGGAATGATTCTGGTCGATTACCTA 607 GAGGCATGTAGTGCGTTCGTGTCCTGAACAAAT 549 CCAGGGCAATACGTTATAGTCCAGCCCTGTCA 608 GCTCATCTGCGCACGATGAAAGTCAGAGGGTA 550 CCTGCCAAAGAGAAAAAAAGAGCAACGCAGTTC 609 GGGTTAAAAATGAACAAATCAACGCTTTTCAC 551 AATCCCGTAAAGAACTTGAATTATCAAATCTG 610 CAAAATCGCCGCATAGCAAACCGCTCTTTAAA 552 ATTGCATTCATATCAAAGGTTGCAGAGGATC 611 CAACAGTACGCGGGGTGCTGTTTATTCTGTCC 553 CGAGTATGGCAGGACTTCGCAGAGGGTAGCGG 612 CGATCAATGACTTTATGTTAGCCAGGGAAAAT 554 ACGAATATCCTTAGGCATAGCCTGCCATCCAA 613 CCAACCAGCCGGCATTCAGTAGCCAGAATGCT 555 CTCGGCAGCTGGGGTTTACTACTGAGGTTTGC 614 CGGCGGCTCACAGCCCCTACGGACGTCTGTTT 556 ATTTCGGGCGCGATAGCCCGGAATACTCTGCC 615 TTCACGCGCTCTTGTCTGCCACTCGTCAAGAC 557 TGATTCATTGCGCTAATCAGCGCCGTTTTTTC 616 CCATCGCAGATGTGGCAATCAACAACAGGCGA 558 AAAAATGTACCGGCGGGCAAAGCGGCAACTGG 617 GCAGAGCGAGGCGAGTCTGTTTACTTCACACA 559 GGATGAAATTTGATGCCTGGGTGAAAATGATA 618 GGGCATGCGCAGAGATTAAACGCTGGGTTTAT 560 CTGGTACAGGATGAGATGTACCCAATCAAAGA 619 GCCCGCAGGCAGGACGAGGAACGAACACGCAT 561 GACGCGCCCGCATTGCATCACTAGCCGCTCAG 620 GTTGTCCTGTGGCTTGCTAAATTTGGATTCTC 562 ACTGGCGCTGCGTCCGTTCGGCTGGCGTGAAG 621 ATATAAAATGAATGAATGGATCAGGGCCTATG 563 TTATTCGATACTCGTTCATGTTTTTATCAACC 622 GTCAATTATGGACGTACAGGTAGCAGCTACCC 564 TGTCGAGCGAGATAGAATCTGTTAGCTGGCAC 623 ACCAAAGGCGACCATCAGGTAGAACTGGCCTT 565 CTCTGGAAACTCACTTTCGGCATGTATGCGAC 624 AGCTCGAATGCGACGCATCAACACTGAATAAA 566 TATGTTAATTTCCCCACGTTCGTACTCATCGC 625 TGCTGAATACCGGTAAGCCTGGCTCTATCACC 567 TCATCGCGCCGGTAATTCGTCATCTCATGAGG 626 GGAGAACGGCGGTCATATTCCGCTAAACTGCA 568 AATTGAGCTTTCAACGATTGCAGTCCCCGCGA 627 CCGGAGACAGTGTTGAAGCCGCAGTCACTCTA 569 CGTGATTTTTCAATAGGAGCAATGGATAGGGC 628 AGTGCTGACTGTGTTAGTAGACAACCGGTCTT 570 TAATGGGAACACCGCGCAGGACGGATGTAAAT 629 CTTCGAGATTAGAATATGACTCAGTACCACTA 571 TGAATATCCAGCCCTTTCTGCGCGACAACCTC 700 CGATGCCGCTGAACCCTTCCGGGTGGTTTCCC 572 CCGAAAACCTGTCCCGTCTGCAAATCTGAATT 701 TTTGGATCTAAGCCAGCCCAGCAATCGAAGGT 573 GGGGCCACGGTAGATTTATTCGATACTCGTTC 702 GGCAATGAACTACTGGGGAGTCGGTGGTAGAA 574 GAGTTGCAGGGCGAAATTAAGCAACGTCACAA 703 ACAATTAACGATGGATGGCCGCGAGAGCGGCC 575 ATCGTACTGTCAAACGGCGCAGAGCTGCATTT 704 GGTAGGAATTAACGCGTATTGCCTTTACTAAT 576 AGGTCGACGTTGAACTGAAGGGCGGTGACAAG 706 CGACCAGAAAACGTCCTGTGAGATCCCGAGCA 577 GCATTTATTTTTGCACTACTTGCCGGGATGTC 707 AATCACAGACTGCTGTAACGGGATTTCTGTTT 578 AGAATGGCTAGTTTGATTGAATATTCTACCTCG 708 ATCAGGGGAATCAAATCAAAAAAACACTCTTA

173

Supplemental Table S6.3 (page 6 of 6)

Number Sequence Number Sequence 709 TGGTATTGACTTTCGATCTGGAGCGAATGGGG 768 CGATTTTGAAATAACACCATCAATCCAATCCC 710 TAATAACAATAATAATAAAAACACTACGTCAG 769 AAATAGGAAAGCTGGCTGATAAACTTTTTGAA 711 GATGGCACTGAAAATACATTTGGCGTGAACAT 770 CACGCGCTTCACCTGGTCCGGTGTAATGTCCT 712 ACATGGTTCTCATAGCACAATCGCTTATAGCG 771 TCGAAATGGCGACAGCAGAAGAGAAGGATCGG 713 GCGTGCTCGTACTGGCATTCTTTAGTTGTGAA 772 CCTTAGGCGAGCTGGGATCACTGATTTCAGAT 714 TGGCGTCAAAAAGCCACTAAATGCGAGTGGAT 773 CTACCATCTGGCCTTATAATTCAATGGGTTAGT 715 ATTCTTGCTGCATGTAAGCCATATATTCTGCG 774 GCCGGATTATACATTACTGGCATTTGAGCCAG 716 TTTCCTGTTTCACCGTTAACTGCCTTAGCGTT 775 CCTTGTCAGCCTACGAGATGGCAAGGTTTATC 717 ATCAGTAGCCGTGAAGGTGACTATATGGAGGG 776 ACTGTCGCAACAAGGAGAAATGCAAATGAGTA 718 GTATCGCTGCGTAGTAACCAGTTGTGCTTTTT 778 CCGGCAACGTTCTACTCCAGTCCGCACCAGGT 719 TACGAATTTGGAACGACAACGGGGTTACTGAT 779 CAAACTGCACTGAGTTCGTTTCCTCACTGCTC 720 AACAGCCAGCTTAAAAACCATGTTGCTGATC 780 CATCTGGCCTTTGCTAATTCCCCGAAACCGCTA 721 ATGATGGGGCTGCGCTGAACGCAACAGATCCG 781 TCCGCCGTTTACGTTTCTGCTAATTTCGACAT 722 GAAAACCACTATGTGTATCACCGTCACCGAAA 782 GAACTCCCGGGCCTTGGGTAATTGATGGTGGC 723 CCTAAAAGGAAAGTGTTACGTAAAAATAAAGA 783 CGTGATTTTCAATAGGAGCAATGGAAAGGGC 724 GGCCCTCAACCGGACAGTCAATATCTGGAGCA 784 AAATCTCACCAGTGGAGACTCCAGCGTCGAAA 725 GGGATTTATTTTTGGACTGGATGTGCTTTAGAGT 785 AACGTCTGGGTATCTGGCGCACATCATGAAAG 726 CGCGATTTTGAGACAATGCAGAATTTTATTTT 786 TTGGAAAAAAGGATATCAACCAGTTTCTCAGT 727 GAGAAATGCAAATGAGCAATGTCACTAATCTT 787 GTGATATGAGCGTCCTTGAGGATATGGGCGTA 728 CCAAAAACATTGCACAAACCGAGCAGGTGGGG 788 GCGATCTTATGGCGTTTGGAGAAAAAATAATC 729 ACAGCTAAACAGGTAAGGGGAATGCTGGATTT 789 CGGATCTTACGATGGTTCTAATGACTCGGTCA 730 GCAACGAATTAGCCCCACTCAACGTGCAATGG 790 CGCGCAAAACATATCCTCATCGCTGATTTATC 731 CGTACTCGTACTGGCATTCTTTAATTGTGAAA 791 AAAAAAACCGGATTCCTCACGCTATTTGCGCG 732 AACACCTTATCAAGGAGGGCATTGGTGCTGTG 792 ACGCGAATACAAAGCAGCACGGCAGGGGGGAG 733 CAAACCAGAATGGGGCTAAAAGCGCGGCAGTT 793 TCGTTCCGTAAATGGTGTCTATCGTGGCAGTG 734 CGTTGGCCATTTCTTCTGTGGAAGATCCACCG 794 TTTTTTAATTGCGTGTTGAGTCGCGGGGCAGA 735 TTCATATGCGCGTAAATATTGCCTCAACGGAC 800 TACACCGCCAGTTCGTGCCGGCACACTACCGG 736 GAAAGGCGATTACATGCTAATCGGGCAATCTA 801 TTTGCCGGCCTTGGCCGGCTTCTCGTCCAGAC 737 GTTAGTAATTTGGCCCACACAGCTATCTGCAA 802 CAGGAAACTGGAATTGATATTGAATGGCGAACG 738 TCTTGACTTAGCAGCATCAATAACGGCCTTAC 803 TTTTATTCGAACTCTGCACGAATTCGATGTTA 739 CCGTGTTGTGTTGAGCTGCAAGCGCGCCGTAA 804 TGGCTGGTAAATTTCAACGATTAAAAAGTGGC 740 GCCGCATTAAAGCCGCCGCCGCCGGAACCCCC 805 GCTGTGTGGTGCCTGCATAATTACATCTGTCA 741 GTGGCAGCGGCATTTGGGAAGGATCACAAAGA 806 ACTGGCGCGGCAAGGGGAAGGGCAAATGAGTA 742 GTTTTCGGGCTGATCAGGGGAATGGTGAATTT 807 TTCATTAGTCATTGTATATGCCTTTTTGGGTT 743 CAAATTATTGCTATTCATGGGCTGTCTAATAA 808 CGCGATTGACGTGATGAATCAGCTAAACAGGC 744 GTTTAGTGTAAATGATGATTCCAGACTATCAA 809 AACACCACCGCCATTTTATCTGTCGAATGCCC 745 GTTGCAACGGCTTTTGGGAAGGATCACAAAGA 810 CTGGGTAAATACGGTCTGGTTGCTGCCCGCCA 746 CAGCTCATTAGCGATACGGGTATAACCATTAT 811 ACCGCGAACAGCTCAATCTTCGTTATCAGCAG 747 CACGTCTTTTTTCGACAGCACACGGAGCGCGT 812 CAATATCTTTTGTTTGCGTATAGATATCCTGC 748 AAAGGGGAATAGCTAATTTAGCTCCGGTTTTC 813 GTCAGGGAACTGACCAAAACCAGCTTTGAAAT 749 CCAACATAGCTACATATATCGCGGCTTCACTA 814 TCCTCCTGCAGGCGGAGCATGAGGATAAAGTC 750 GCCCGGTGTGTCGTCGATGTAGAGCATGTCTA 815 TTATCACATTGATTTTATGAAATTATTTAATG 751 TAACGCGCATGACTTCATTTTGTATGAATTAC 816 AGCGGGACAGGTATATCGCATTAACACTGGCA 752 ACCTGGTGGGTAGTAGATATCGCTGGCTGCAA 817 TCAGTTCATTGGGCAACAGAACGCATCAACAT 753 ACCATTTGATCATATCGTTCAGTTTGTCCCAT 818 GAGGTATGTACTGCATCGTGCCCTGAACAAAT 754 CGTGGCGAGTGGACGAGCTGGGCGACCTGCCC 819 AGCTCAAATGCTACGCATCAGCACTGATTAAA 755 ATTTTACATTCCCACCGAATTTTGCAGCTGCG 820 TGTCTCCAAGTTAGCAATTTAGTCTCTAAACC 756 GATGGGGTACAGCTGCTTCCGTTCAGCCAGAT 821 TCATATGCGATTTTCAGCCTTTTAAGACCTTC 757 CGGGGTAATCTGTGTTCATCATCACGGCGCAG 822 TACGTCAGCGGCATTATAAGGATTTACACGGA 758 CATCTCTATAAATTGCGAGATATGAAAAATGA 823 CCGGCACCAAATGCTTATCACTGCCAGCGGCG 759 TTAAATATCGAACCGAAAATAATTTTTATAAA 824 ATTTCTGGATTAACACTCAGGCGGCTCATGAC 760 CTGTACACCGTTGACGGCGGCACCTCCCACCA 825 GTCATTCATACGGGCTATACCGTCCTCCAGAT 761 GCGCATCTGCGCACGATGAAAGTTAAAGGCTA 826 GTCAGCAGCTGTAATAATGCGCACCTTTTCCC 762 TTTTTGAGCTATTGCGGATGATTGAGAGGCTG 827 CCTTCAGGCCGCTGGATAGATGCTAACCCACC 763 CACAGGAGCTGCATGACAAGCTAAAGTCTGAC 764 TAGCTTATAGCGTTTTGTCGACTTAAGCAGAT 765 CTAGAGAAGTCCACCAAGAGGAAATTAGAGCT 766 GTGTGTTTTTTGAGCTACGGCAAATGGATTCA 767 CCCACGCAGCAGCCCGCGCAGCATCCCTCGCA

174 Supplemental Figure S6.1 (page 1 of 2)

Pattern 1 2 3 4 5 6 7 7 9 10 11 12 13 14 68 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 1 2 3 4 5 6 7

Pattern 88 89 90 10 91 92 93 94 95 13 21 22 23 254 28 29 30 31 32 33 34 35 8

Pattern 1 2 3 4 5 6 7 7 9 10 11 12 13 14 68 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 9

Pattern 258 259 260 261 119 262 263 264 265 312 313 314 315 316 317 293 276 266 267 268 269 270 271 272 273 274 275 277 35 10

Pattern 580 581 582 583 328 584 585 488 586 600 601 602 603 604 604 606 607 338 339 341 342 343 344 334 335 336 337 364 365 369 370 378 357 358 359 360 360 625 361 10 94 95 11 12 13 14 68 345 346 347 348 352 626 627 628 629 11

Pattern 446 447 448 449 487 596 599 608 609 610 611 612 613 614 615 616 617 618 619 493 620 621 24 622 27 254 623 11 *

Pattern 327 255 328 329 330 331 332 333 338 339 340 341 342 343 344 334 335 336 337 364 365 366 148 367 368 368 369 370 378 364 355 356 357 358 359 360 625 361 362 363 9 10 94 95 11 12 13 14 68 15 345 346 347 348 349 350 12 13 14 15

Pattern 351 352 353 626 627 354 375 446 447 448 449 487 488 489 490 592 593 594 595 596 597 598 599 608 609 610 611 612 613 614 615 616 617 618 619 493 620 621 24 GAP 622 27 254 623 12 13 14 15

Pattern 414 415 93 416 417 418 419 410 412 413 425 426 427 428 173 482 483 484 485 486 450 320 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 16

Pattern 389 390 391 392 393 394 395 396 431 432 433 434 435 436 437 438 284 499 545 546 399 547 548 549 550 551 552 569 570 571 572 573 574 575 422 576 577 578 579 553 554 555 556 557 558 488 559 560 561 562 563 506 507 508 509 510 511 17

Pattern 405 406 407 408 409 410 411 412 413 425 427 428 173 430 483 484 485 486 450 320 530 531 451 452 455 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 472 474 475 476 477 478 479 480 481 CR1 18

Pattern 381 382 383 384 385 7 386 387 388 420 421 422 423 424 500 501 502 503 504 505 533 534 535 536 537 538 539 540 541 542 543 544 511 19

Pattern 397 98 399 400 401 402 403 404 439 440 441 442 443 444 422 491 492 493 494 495 496 512 513 514 515 516 517 518 519 520 521 564 565 566 567 568 532 522 382 523 524 525 526 527 528 529 511 20

Pattern 1 2 3 4 5 6 7 7 9 10 11 12 13 14 68 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 40 41 42 43

Pattern 327 255 328 329 330 331 332 333 338 339 340 341 342 343 344 334 335 336 337 364 365 366 148 367 368 369 370 378 364 355 356 357 358 359 360 625 361 362 363 9 10 94 95 11 12 13 14 68 15 345 346 347 348 349 350 351 44

Pattern 352 353 626 627 354 375 446 447 448 449 487 488 489 490 592 593 594 595 596 597 598 599 608 609 610 611 612 613 614 615 616 617 618 619 493 620 621 24 GAP 622 27 254 623 44

Pattern 1 2 3 4 5 6 7 7 9 10 11 12 13 14 68 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 47 50 51 52 53

Pattern 1 2 3 4 5 6 7 7 9 10 11 12 13 14 68 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 59 60

Pattern 580 581 810 811 582 602 603 812 813 814 815 816 817 818 338 341 342 343 344 335 336 337 364 365 366 148 367 369 370 378 357 358 359 360 360 625 361 10 94 95 11 12 13 14 68 15 345 346 347 348 352 626 629 375 447 67 68

Pattern 448 449 487 596 599 608 609 610 611 612 613 614 615 616 617 618 619 493 620 621 24 GAP 622 27 254 623 819 820 821 67 . 68 x .

Pattern 1 2 3 4 5 6 7 7 9 10 11 12 13 14 68 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 71

Figure S6.1 (page 2 of 2)

Pattern 36 37 38 39 40 41 42 43 44 45 46 47 48 69 49 50 51 52 53 54 55 56 57 58 58 59 60 61 62 63 64 70 65 66 67 21 22 23 24 25

Pattern 587 588 589 590 591 48 102 69 49 50 51 52 53 54 55 56 57 103 104 58 59 60 61 62 63 64 70 65 66 67 26

Pattern 36 37 38 39 40 41 42 43 44 45 46 47 48 69 49 50 51 52 53 54 55 56 57 58 58 59 60 61 62 63 64 70 65 66 67 27 28 29

Pattern 111 112 113 114 69 115 116 117 118 119 120 121 122 123 124 159 160 161 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 162 67 30

Pattern 318 300 301 302 303 304 305 118 121 122 123 124 159 160 161 125 319 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 162 67 31

Pattern 96 97 98 99 278 279 108 108 110 187 188 189 190 191 192 193 255 215 186 100 41 42 43 44 45 46 47 101 48 102 69 49 50 51 52 54 55 56 57 103 104 58 59 60 61 62 63 66 67 32

Patern 320 321 322 323 324 325 326 203 204 205 206 207 208 209 210 211 212 213 214 143 144 145 146 147 148 149 150 151 152 153 171 67 33

Pattern 105 106 107 108 109 110 187 188 189 190 191 192 193 255 215 186 100 41 42 43 44 45 46 47 101 48 69 49 50 51 52 53 54 55 56 57 103 104 58 59 60 61 62 63 64 70 65 66 67

CR 2 CR 34

Pattern 163 164 165 166 167 168 169 170 238 239 2 240 7 241 242 243 244 129 245 246 247 248 249 250 251 252 237 253 154 155 156 148 149 157 150 151 152 153 171 67 35

Pattern 172 173 174 175 176 177 178 179 256 280 281 282 283 284 285 286 287 288 289 290 291 292 224 145 180 181 182 183 184 185 69 257 36

Pattern 194 195 196 197 198 199 200 201 202 216 170 217 218 43 44 101 48 102 222 223 69 49 50 51 52 53 54 55 56 57 104 58 59 60 61 62 63 64 70 65 66 67 37

Pattern 36 37 38 39 40 41 42 43 44 45 46 47 48 69 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 70 65 66 67 55 56 57 58

Pattern 800 801 36 37 38 39 40 41 42 43 44 45 46 47 48 69 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 63 64 70 65 66 67 61 62 63 64 65 66

Pattern 589 591 822 827 823 824 825 826 107 108 109 110 187 188 189 190 191 193 255 215 186 100 41 42 43 44 45 46 47 101 48 102 69 49 50 51 52 53 54 55 56 57 103 104 58 59 60 61 62 63 64 70 65 66 67 69 70

Pattern 306 307 308 309 310 311 38 CR CR 3 39

Figure S6.1. Master reference collection of spacer patterns of clustered regularly interspaced short palindromic repeat (CRISPR) regions CR1, CR2, and CR3 for E. amylovora isolates from this study, as well as McGhee and Sundin (2012), and Tancos and Cox (2016). Spacers are represented by boxes, uniquely numbered at top of each column. White boxes indicate abscence of a spacer in a given pattern. Spacer numbers that are shaded (teal) are newly discovered in the present study. Spacers are arranged in patterns, depicted in individual rows. Patterns are uniquely numbered at the left hand side of each row. Text color of pattern number indicates who and when it was identified: black, McGhee and Sundin (2012); red, Tancos and Cox (2016); teal, this study. GAP indicates no spacer present between two adjacent repeat segments.

176