Genetic Analysis of Black Raspberry Breeding Germplasm

Thesis

Presented in Partial Fulfillment of the Requirements for the Degree of Master of Science in the Graduate School of The Ohio State University

By

Matthew Willman, B.S.

Graduate Program in Horticulture and Crop Science

The Ohio State University

2019

Thesis Committee:

Jonathan Fresnedo Ramírez, Advisor

Leah McHale

Joseph Scheerens

Copyright by

Matthew Willman

2019

Abstract

U.S. black raspberry (BR) production is currently limited by narrowly adapted, elite germplasm. Improved understanding of genetic control and stability of pomological traits will inform the development of improved BR germplasm and cultivars. To this end, analysis of a multiple-environment trial of two BR mapping populations derived from crosses of commercial cultivars with wild accessions has provided insights into genetic variation, genotype-by-environment interactions (GEI), quantitative trait loci (QTL), and

QTL-by-environment interactions (QEI) of fruit quality traits among diverse field environments. Genetic components and stability of four fruit size traits and six fruit biochemistry were characterized in two mapping populations following their evaluation over three years at four distinct locations representative of current U.S. BR production.

GEI of pomological traits were described using two methods: mixed model analysis and

Bayesian Finlay-Wilkinson regression. Both methods revealed relatively stable genetic control of the four fruit size traits across the tested production environments and less stable genetic control of the six fruit biochemistry traits. Further, Finlay-Wilkinson regression revealed individuals contributing to GEI for each trait. Ten QTL associated with three fruit morphology traits and five QTL associated with two fruit biochemistry traits were identified. Of the fifteen total QTL, eleven exhibited significant QEI. Closely overlapping QTL revealed linkage of several fruit size traits: fruit mass, drupelet count, and seed fraction. Further, alignment of linked markers to the BR genome revealed genomic regions associated with these traits. These and related findings are expected to guide further genetic characterization of BR fruit quality, management of breeding germplasm, and development of improved BR cultivars for U.S. production.

i Dedication

To my parents, who inspired my pursuit of science.

ii Acknowledgements

Thank you to all those who worked on this project before me – those who gathered and continue to develop germplasm, who spent countless hours in the field and lab taking measurements, and who had the foresight to plan and conduct the trials which led to my master’s work. To those I had the pleasure to work with, thank you for your time and direction.

Thank you to The Ohio State University Graduate School and Department of

Horticulture and Crop Science for your fellowship and research associate funding, and to the Department of Horticulture and Crop Science and Hartman Excellence in Fruit

Production Fund for helping me share my research at an international symposium.

Thank you to the OARDC librarians, who were masters at finding any text and forgiving of overdue books.

Thank you to my advisor, Jonathan, for your unwavering support and for keeping me on track, and to my thesis committee, Joe and Leah, for your thoughtful questions, feedback, and expertise.

Thank you to my sister, Anne, for being a constant friend.

Finally, thank you to my wife, Katie, for your patience, enthusiasm, and companionship.

I look forward to many great adventures with you.

iii Vita

April 2013 A.S., Cincinnati State Technical and Community College

December 2015 B.S. Sustainable Plant Systems, The Ohio State University

2017 Graduate Fellow, The Ohio State University

2018 to present Graduate Research Associate, Department of Horticulture

and Crop Science, The Ohio State University

Fields of Study

Major Field: Horticulture and Crop Science

iv Table of Contents

Abstract ...... i

Dedication ...... ii

Acknowledgements ...... iii

Vita ...... iv

List of Figures ...... ix

List of Tables ...... xiii

List of Abbreviations ...... xviii

Introduction ...... 1

Chapter 1. Literature Review ...... 3

Black raspberry biology, cultivation, and domestication ...... 3

Biology ...... 3

Cultivation ...... 4

Commercial production ...... 6

Breeding history and objectives ...... 7

Germplasm ...... 11

Statistical methods of genetic characterization ...... 13

Analysis of variance components ...... 13

Quantitative trait loci mapping ...... 14

Description of genotype-by-environment interactions ...... 15

Genetic characterization of black raspberry and related crops ...... 17

Variance components in raspberry ...... 18

Quantitative trait loci mapping in raspberry ...... 20

v Genotype-by-environment interactions in raspberry and other perennial crops ...... 21

References ...... 23

Figures ...... 32

Tables ...... 34

Chapter 2. Genotype-by-environment interaction analysis of a multiple- environment black raspberry trial ...... 36

Introduction ...... 36

Black raspberry biology and cultivation ...... 36

Genotype-by-environment interaction ...... 37

Fruit size and fruit chemistry ...... 39

Materials and methods ...... 40

Germplasm ...... 40

Phenotypic data ...... 41

DNA marker data & relatedness analysis ...... 42

Mixed model analysis ...... 44

Finlay-Wilkinson regression ...... 45

Results ...... 47

Marker data & relatedness analysis ...... 47

Fruit size traits ...... 48

Fruit chemistry traits ...... 50

Discussion ...... 51

Relatedness analysis ...... 52

Fruit size traits ...... 53

Fruit chemistry traits ...... 58

Conclusions ...... 61

vi References ...... 64

Figures ...... 69

Tables ...... 78

Chapter 3. Quantitative trait loci analysis of a black raspberry multiple environment trial ...... 85

Introduction ...... 85

Materials and Methods ...... 90

Germplasm ...... 90

Phenotypic data ...... 90

Genotyping ...... 91

Linkage map construction ...... 93

QTL mapping ...... 94

Results ...... 97

Genotyping and linkage map construction ...... 97

QTL mapping ...... 98

Discussion ...... 102

Genotyping and linkage map construction ...... 103

QTL mapping ...... 104

Conclusions ...... 115

References ...... 117

Figures ...... 124

Tables ...... 141

Concluding Remarks ...... 148

vii References ...... 152

viii List of Figures

Figure 1.1. Native range of Rubus occidentalis by state and province...... 32

Figure 1.2. U.S. per capita fresh small fruit consumption...... 33

Figure 2.1. Pedigree chart for black raspberry mapping populations ORUS 4304 and

ORUS 4305. Maternal parentage is indicated by a red line. Paternal parentage is

indicated by a blue line...... 69

Figure 2.2. Venn diagram of ORUS 4304 and 4305 individuals evaluated in four trial

locations: Jackson Springs, North Carolina (NC), Geneva, New York (NY), Wooster,

Ohio (OH), and Corvallis, Oregon (OR). 154 individuals were evaluated in every

location. Seven individuals were evaluated in OR only...... 70

Figure 2.3. Venn diagrams of ORUS 4304 and ORUS 4305 individuals sampled in three

production seasons (2013, 2014, and 2015) at four trial locations: Jackson Springs,

North Carolina (NC), Geneva, New York (NY), Wooster, Ohio (OH), and Corvallis,

Oregon (OR)...... 71

Figure 2.4. Relatedness plot for ORUS 4304 parents and progeny with the maternal

parent (ORUS 4158-2) on the y-axis and paternal parent (ORUS 3021-2) on the x-

axis...... 72

Figure 2.5. Pearson correlation coefficient plot of breeding values for fruit size and

chemistry traits. Significant correlations (p<0.01) are shaded in red (negative) or blue

(positive)...... 73

Figure 2.6. Finlay-Wilkinson regression plots for eight black raspberry fruit quality traits

measured in location-by-year environments. Individual performances are regressed

ix over environmental effects. Genotype-by-environment interaction is indicated as

heterogeneity of individual slope effects...... 74

Figure 3.1. Genetic position plotted against physical genomic position for 974 combined

SNP and SSR markers. A reduced slope and gap between groups of markers

suggest reduced recombination and few markers near the centromere. Discontinuous

portions in chromosomes 1, 4, and 6 suggest misalignments within either the linkage

maps or the current genome assembly...... 124

Figure 3.2. Distributions of average fruit mass observed in ten location-by-year

environments...... 125

Figure 3.3. Distributions of average seed mass observed in ten location-by-year

environments...... 126

Figure 3.4. Distributions of average drupelet count observed in ten location-by-year

environments...... 127

Figure 3.5. Distributions of seed fraction observed in ten location-by-year environments.

Seed fraction was calculated as total seed mass / total fruit mass × 100%...... 128

Figure 3.6. Distributions of soluble solid content observed in eleven location-by-year

environments...... 129

Figure 3.7. Distributions of titratable acidity observed in eleven location-by-year

environments...... 130

Figure 3.8. Distributions of anthocyanin content observed in eleven location-by-year

environments. C3G=cyanadin 3-glucoside...... 131

Figure 3.9. Distributions of phenolics content observed in eleven location-by-year

environments. GAE=gallic acid equivalents...... 132

x Figure 3.10. Additive ORUS 3021-2 effects of four fruit mass QTL on linkage groups 1,

2 and 6. Positive or negative values indicate origin of the high value allele from either

NC 84-10-3 or cv. Jewel, in no particular order. Heterogeneous effects indicate QTL-

by-environment interaction...... 133

Figure 3.11. Additive ORUS 3021-2 effects of three drupelet count QTL on linkage

groups 1 and 4. Positive or negative values indicate origin of the high value allele

from either NC 84-10-3 or cv. Jewel, in no particular order. Heterogeneous effects

indicate QTL-by-environment interaction...... 134

Figure 3.12. Additive ORUS 3021-2 effects of three seed fraction QTL on linkage

groups 2, 6, and 7. Positive or negative values indicate origin of the high value allele

from either NC 84-10-3 or cv. Jewel, in no particular order. Heterogeneous effects

indicate QTL-by-environment interaction...... 135

Figure 3.13. Additive ORUS 3021-2 effects of three titratable acidity QTL on linkage

groups 3, 4, and 6. Positive or negative values indicate origin of the high value allele

from either NC 84-10-3 or cv. Jewel, in no particular order. Heterogeneous effects

indicate QTL-by-environment interaction...... 136

Figure 3.14. Additive ORUS 3021-2 effects of two anthocyanin content QTL on linkage

groups 2 and 3. Positive or negative values indicate origin of the high value allele

from either NC 84-10-3 or cv. Jewel, in no particular order. Heterogeneous effects

indicate QTL-by-environment interaction...... 137

Figure 3.15. 95% Bayes credible intervals for linkage group 1 QTL. FrM = fruit mass;

DCt = drupelet count...... 138

xi Figure 3.16. 95% Bayes credible intervals for linkage group 2 QTL. FrM = fruit mass;

SdF = seed fraction; AnC = anthocyanin content...... 138

Figure 3.17. 95% Bayes credible intervals for linkage group 3 QTL. TAc = titratable

acidity; AnC = anthocyanin content...... 139

Figure 3.18. 95% Bayes credible intervals for linkage group 4 QTL. DCt = drupelet

count; TAc = titratable acidity...... 139

Figure 3.19. 95% Bayes credible intervals for linkage group 6 and 7 QTL. FrM = fruit

mass; SdF = seed fraction; TAc = titratable acidity...... 140

xii List of Tables

Table 1.1. Monthly normal maximum and minimum temperatures at research locations

...... 34

Table 1.2. Monthly normal precipitation at research locations testing adaptability of

black raspberry ...... 35

Table 2.1. SNP filtering criteria and counts for ORUS 4304 relationship analysis...... 78

Table 2.2. SNP filtering criteria and counts for ORUS 4305 relationship analysis...... 78

Table 2.3. SNP filtering criteria and counts for additive genomic relationship matrix

estimation...... 78

Table 2.4. Mixed model fit comparison for black raspberry fruit size traits with fixed

environmental effects and random additive genomic-by-environment effects

(GA×E).GA×E variance structures were fit as homogeneous with uniform correlation,

heterogeneous with uniform correlation, or heterogeneous with non-uniform

correlation estimated by factor analysis. Residual variance structures were fit as

independent with homogeneous variance or independent with heterogeneous

variance. Each model was tested against the previous model by a likelihood ratio

test...... 79

Table 2.5. Genome-by-environment correlation estimates (�) and standard errors (SE)

for black raspberry fruit quality traits tested in four locations (NC, NY, OH, OR) and

three years (2013, 2014, 2015)...... 80

Table 2.6. Heritability (ℎ�2) estimates for fruit size traits at 10 location-by-year

environments derived from the best fitting model. Maximum and minimum values are

in bold...... 80

xiii Table 2.7. Variance components for fruit size traits estimated by Bayesian Finlay-

Wilkinson regression. SD=standard deviation; μ= overall mean; �g2=genetic

variance; �b2=slope variance; �h2=environmental variance; �ε2=residual variance 81

Table 2.8. Punctual coefficients of variation (CV %) for fruit size traits derived from

Bayesian Finlay-Wilkinson regression mean estimates. g=genetic; b=slope;

h=environmental; ε=residual ...... 81

Table 2.9. Pearson correlation of individual genetic (g) and slope (b) effects estimated

by Bayesian Finlay-Wilkinson regression. r = Pearson correlation coefficient; p = p-

value ...... 81

Table 2.10. Pearson correlation of breeding values estimated by mixed model and

Bayesian Finlay-Wilkinson regression...... 81

Table 2.11. Mixed model comparison for black raspberry fruit size traits with fixed

environmental effects and random additive genomic-by-environment effects (GA×E).

GA×E variance structures were fit as homogeneous with uniform correlation,

heterogeneous with uniform correlation, or heterogeneous with non-uniform

correlation estimated by factor analysis. Residual variance structures were fit as

independent with homogeneous variance or independent with heterogeneous

variance. Each model was tested against the previous model by a likelihood ratio

test...... 82

Table 2.12. Heritability (ℎ�2) estimates for fruit chemistry traits at 11 location-by-year

environments derived from the best fitting model. Maximum and minimum values are

in bold. TA = titratable acidity, SSC = soluble solid content...... 83

xiv Table 2.13. Variance components for fruit chemistry traits estimated by Markov chain

Monte Carlo simulation. SD=standard deviation; μ=overall mean; �g2=genetic

variance; �b2=slope variance; �h2=environmental variance; �ε2=residual variance 83

Table 2.14. Punctual coefficients of variation (CV %) for fruit chemistry traits derived

from Markov chain Monte Carlo simulation mean estimates. TA=titratable acidity (%

citric acid); SSC=soluble solid content (°Brix); g=genetic; b=slope; h=environmental;

ε=residual ...... 83

Table 2.15. Monthly normal maximum and minimum temperatures at research locations

...... 84

Table 3.1. SNP filtering criteria and counts for ORUS 4304 and ORUS 4305 linkage

maps...... 141

Table 3.2. ORUS 4305 linkage map marker counts and genetic lengths of seven linkage

groups representing the seven chromosomes of the black raspberry genome.

Linkage maps were constructed using JoinMap® 4.1...... 141

Table 3.3. Number of ORUS 4305 individuals with phenotypic data for fruit size (FrM,

SdM, DCt, SdF) and fruit chemistry (TAc, AnC, PhC) within trial environments. In

NC_13, fruit size data were available for 57 individuals and fruit chemistry data were

available for 51 of 103 individuals used to construct linkage maps. Among all

environments, fruit size data were available for 84 individuals, and fruit chemistry

data were available for 80 individuals of 103 individuals used to construct linkage

maps...... 141

Table 3.4. Summary of QTL identified by single trait, multi-environment analysis. FrM =

fruit mass; SdM = seed mass; DCt = drupelet count; SdF = seed fraction; SSC =

xv soluble solid content; TAc = titratable acidity; AnC = anthocyanin content; PhC =

phenolics content...... 142

Table 3.5. Summary of QTL additive and dominance effects by trial environment. Table

entries are ordered by trait, position, then environment. R2 is given as the change in

% variance when that QTL is dropped from the full model. Env = location-by-year

environment; LG = linkage group; LOD = log of odds; FrM = fruit mass; DCt =

drupelet count; SdF= seed fraction; TAc = titratable acidity; AnC = anthocyanin

content...... 143

Table 3.6. Summary of QTL intervals by trial environment. Multiple environment values

were estimated in Genstat® by using a mixed model approach with genetic

correlation modeled by a heterogeneous compound symmetry (CShet) variance-

covariance matrix. Single environment values were estimated in R/qtl by using the

refineqtl() function to optimize the position of the (multiple environment) QTL.

Flanking marker positions were not defined for linkage group 6 due to discontinuity

between linkage map and genome assembly (Figure 3-2). Table entries are ordered

by trait, position, then environment. Env = location-by-year environment; LG = linkage

group; LOD = log of odds; BCI = 95% Bayes credible interval; FrM = fruit mass; DCt

= drupelet count; SdF= seed fraction; TAc = titratable acidity; AnC = anthocyanin

content...... 145

Table 3.7. Black raspberry genomic regions with high synteny to known genes

influencing organ size, fw2.2 and CNR01. Protein sequences were retrieved from

NCBI (https://www.ncbi.nlm.nih.gov). BLAST was performed on the Genome

xvi Database for Rosaceae website (https://www.rosaceae.org/) using the Rubus occidentalis v3.0 genome...... 147

xvii List of Abbreviations

AnC total anthocyanin content ANOVA analysis of variance BCI Bayes Credible Interval BR black raspberry BV breeding value CIM composite interval mapping CP cross pollinator (population type)

CShet heterogeneous compound symmetry DCt Drupelet count FrM fruit (infructescence) mass GEI genotype-by-environment interaction GBS genotyping by sequencing LG linkage group MET multiple environment trail NC North Carolina NY New York OH Ohio OR Oregon PhC total phenolic content QEI QTL-by-environment interaction QTL quantitative trait locus/loci SdF seed fraction SdM seed mass SIM simple interval mapping SNP single nucleotide polymorphism SSC soluble solid content SSR simple sequence repeat TAc titratable acidity

xviii Introduction

Black raspberry (Rubus occidentalis L., BR) is a North American native plant cultivated for its fruit which is consumed by humans in both fresh and processed forms.

Interest in expanding U.S. BR production has been stifled by limited genetic improvement resulting from low genetic diversity in breeding populations. Surveys of wild germplasm have revealed that novel, favorable alleles may be available. Evaluation of wild germplasm through the development of mapping populations and multiple- environment trials will define breeding strategies for BR, such as definition of selection indices and identification of genetic markers for marker-assisted selection. Further, such study may elucidate genetic control of pomological traits such as fruit size and human health related traits such as fruit anthocyanin content.

To this end, a National Institute of Food and Agriculture (NIFA) Specialty Crop

Research Initiative (SCRI)-funded project developed two BR mapping populations of a common parent and trialed cloned progeny over three years (2013-15) at four U.S. research locations: Oregon State University, Corvallis, OR; The Ohio State University,

Wooster, OH; Cornell University, Geneva, NY; and North Carolina State University,

Jackson Springs, NC. Objectives of the project included field assessment of important pomological traits, evaluation of genotype-by-environment interactions, and identification of genetic determinants of pomological traits. Though ample phenotypic and genotype-by-sequence data were generated through this project, a formal genetic characterization is still lacking for most measured traits.

The main objective of this thesis work is to elucidate genetic contributors to fruit quality in BR. Such elucidation will improve our understanding of BR biology and

1 genetics and enable the development of strategies for domestication and breeding. The specific objectives are to (i) estimate and describe genotype-by-environment interactions within diverse U.S. locations for two full-sibling populations, each having wild and cultivated ancestry, and (ii) discover quantitative trait loci contributing to horticulturally-relevant traits within these populations across and within the environments described above.

This research thesis is presented in three chapters. Chapter 1 is a review of BR biology and domestication, methods of genetic characterization, and current genetic characterization of black raspberry and related crops. Chapter 2 reports genetic components of black raspberry including genotype-by-environment interactions observed in a multiple-environment trial of two experimental BR mapping populations.

Chapter 3 reports quantitative trait loci and locus-by-environment interactions from this same study while focusing in one mapping population. Finally, concluding remarks are provided with respect to possible future venues of research regarding BR.

2 Chapter 1. Literature Review

Black raspberry biology, cultivation, and domestication

The following is a summary of known information regarding black raspberry biology, cultivation, and domestication.

Biology

Black raspberry (Rubus occidentalis L., BR) is a member of the Rosaceae family and Idaeobatus (Focke) subgenus, a grouping commonly referred to as raspberries.

BR’s natural distribution spans much of the eastern half of North America north of

Florida and Texas and east of the Rocky Mountains (Figure 1.1) (Jennings 1988;

Dossett et al. 2012a). BR is diploid (2n=2x=14) with an estimated genome size of

250Mbp (VanBuren et al. 2016, 2018; Jibran et al. 2018). Plants are perennial shrubs producing biennial canes. First-year canes (primocanes) are produced in the summer and flower as second-year canes (floricanes) the following spring. Flowers of cultivated and wild BR varieties are hermaphroditic and self-compatible, relying on insects for cross-pollination.

Following pollination, flowers produce an infructescence (hereto referred to as a fruit) composed of approximately 30 to 150 drupelets which are arranged spirally around a receptacle and adhere to each other when ripe. Ripe BR fruits separate easily from the receptacle (i.e. torus) which remains attached to the stem. This is in contrast with blackberry (Rubus subg. Rubus spp.), in which the torus remains attached to the fruit after harvest. After producing fruit, BR second-year canes senesce and die.

Occasionally, BR plants produce fruit on first-year primocanes in addition to second- year floricanes, though factors influencing primocane fruiting in BR are not well 3 understood (Dossett and Finn 2011). However, primocane fruit production has had a large influence in recent crop development of red raspberry (Rubus idaeus L.) (Pritts

2008).

Propagation of BR occurs sexually by seed as well as clonally by shoot tip layering. In shoot tip layering, the apex of sprawling primocanes reach the ground, and roots are initiated at the point of soil contact. Unlike red raspberry, BR plants do not produce underground suckers or root buds. Rather, shoot buds are initiated in a perennial crown.

Cultivation

Earliest records of BR cultivation appear in the 19th century (Jennings 1988).

Though early European explorers described BR fruit, intensive cultivation was likely slowed by an abundance of wild crop and preference for red raspberry cultivars brought from Europe (Jennings 1988). U.S. BR cultivation and breeding intensified in the late

19th century and has since declined. Hedrick (1925) described 189 named varieties in a

New York State Agricultural Experiment Station report. His list includes ‘Munger’, an

1897 Ohio introduction which, due to its suitability for machine harvest, is the most propagated U.S. BR cultivar and primary cultivar grown for the U.S. Northwest Coast

BR processing industry.

U.S. BR production peaked in the early twentieth century, when thousands of acres of BR were cultivated in New York alone (Jennings 1988). Commercial production is now largely concentrated in the Pacific Northwest, where fresh and processed production covered a combined area of 1660 acres and generated a utilized production value of $28 million in 2016 (U.S. Dept. of Agr. 2017a). California fresh market and

4 Oregon processing market production account for most recorded U.S. production; in

2016, California fresh market production spanned 600 acres and generated $22 million utilized value, while Oregon processing market production spanned 950 acres and generated $5 million (U.S. Dept. of Agr. 2017a).

BR is grown to a lesser extent throughout the U.S. Midwest in small, direct-to- consumer operations. Ohio commercial production is estimated to total 50-100 acres

(Gary Gao, Ohio Extension Specialist, personal communication). Additionally, BR is popularly cultivated in Ohio home gardens (Gao et al. 2017). Ohio demand for BR fruit is largely regional, though a formal survey of Ohio production and consumption is lacking (Gary Gao, personal communication). Fruit are hand-harvested and large fruited cultivars, such as ‘Jewel’, ‘MacBlack’, and ‘Bristol’ are favored for Ohio production (Gao et al. 2017).

Renewed interest in BR fruit has followed the discovery of its positive effects on human health (Fang 2015; Amiot et al. 2016; Mazzoni et al. 2016; Kresty et al. 2016).

Additionally, increasing fresh market consumption of other small fruit crops, including red raspberry (R. idaeus), strawberry (Fragaria × ananassa Duchesne), and blueberry

(Vaccinium L.), suggest high market demand for nutrient-dense small fruit like BR

(Figure 1.2) (U.S. Dept. of Agr. 2017b). Regional popularity of BR and agricultural tourism reinforce potential for increased Ohio production and sales. Further, disruptive technologies such as primocane-fruiting BR cultivars and covered production have the potential to expand Ohio BR production (Pritts et al. 2017).

5 Commercial production

BR is produced as a high-value specialty crop in the U.S. Midwest, Northeast, and Northwest regions. There is also active interest in expanding BR production into the

U.S. South (Bradish et al. 2016b). These locations represent diverse environments in which BR is produced (Table 1.1 and Table 1.2).

BR cultivars are propagated clonally in the field by tip layering or in tissue culture.

Crop plants are established and managed in rows. Unlike red raspberry, BR do not produce underground suckers, and new canes emanate from a central crown. Trellis systems may be used to hold canes upright and are necessary for mechanized harvesting. Plantings may be rain-fed and are often irrigated, as drought conditions can negatively affect current-year and future crop performance (Funt and Ross 2013).

Well-established plants producing vigorous canes in the second year may yield a small crop in the third year from the second-year canes. In early summer, primocanes are pruned to promote lateral growth. Lateral branches are then pruned to 20-35 cm in length in early autumn to early spring depending on climate. Floricane fruit is harvested mid-summer. After harvest, these floricanes senesce and are removed. Labor is reduced by pruning primocane lateral branches before removing dead floricanes. BR plantings generally remain in production until accumulation of disease and/or abiotic stress render them unprofitable. Thus, the profitability of a planting is largely dependent on disease and stress avoidance and mitigation. Kuhlman and Mumford (1949) reported the lifetime of an Oregon BR stand as 4 to 10+ years. Likewise, current production can range from three or four seasons to several decades depending on management and disease pressure (Halgren et al. 2007).

6 Black raspberry cultivation consists almost exclusively of summer-fruiting cultivars; however, fall-producing primocane-fruiting cultivars and accessions have been described (Dossett and Finn 2011). A recent expansion of primocane-fruiting trait in red raspberry cultivars has transformed the fresh raspberry industry. Development of primocane-fruiting blackberry is expected to similarly affect the fresh blackberry industry. Likewise, expansion of primocane-fruiting trait in BR may increase production opportunities for this crop. Records of primocane-fruiting trait in BR date back to 1832 with the naming of ‘Ohio Everbearer’; at least 19 primocane fruiting BR cultivars have since been described, though most of these cultivars produced poor quality fruit

(Dossett and Finn 2011). Further, multiple-environment trials have shown low stability of the primocane-fruiting trait in BR cultivars and experimental germplasm (Dossett and

Finn 2011). Currently, U.S. BR primocane production is dominated by a single cultivar,

‘Niwot’ (Tallman 2016). Since its release in 2014, ‘Niwot’ has quickly become popular among Ohio commercial growers (Gary Gao, personal communication).

Breeding history and objectives

Early BR breeding efforts began in 1893 at Geneva, New York (Jennings 1988).

These efforts gave rise to ‘Bristol’ and ‘Dundee’, ancestors of the still-popular ‘Jewel’

(Ourecky and Slate 1973; Jennings 1988). Breeding efforts in this time in combination with naming of wild accessions gave rise to as many as 189 BR varieties, reflecting a large interest in BR production (Hedrick 1925). However, few of these historic varieties have survived to the present day. Genetic marker analysis of extant cultivars suggest that early breeding efforts were hampered by limited genetic diversity within breeding germplasm (Weber 2003; Dossett et al. 2012a). These findings support a previous

7 report of difficulty distinguishing between cultivars by phenotypic analysis (Ourecky

1975).

Challenges in BR breeding brought on by lack of elite genetic diversity have likely dissuaded breeders from investing resources in BR improvement. According to a 2002 report on Rubus breeding, of 30 Rubus breeding programs spanning 19 countries, only

7 of these were actively breeding BR, and none were pursuing BR as their main crop

(Finn and Knight 2002). Only six BR cultivars have been released since 1980:

‘Earlysweet’, ‘Haut’, and ‘Glencoe’, ‘Ohio’s Treasure’ (U.S. Plant Pat. No. 27,871),

‘Explorer’ (U.S. Plant Pat. No. 17,727), and ‘Niwot’ (U.S. Plant Pat. No. 27,131). In the same timeframe, over 100 red raspberry cultivars were released (Finn and Knight

2002).

Breeding objectives for BR include optimization of many complex horticultural and fruit quality traits (Jennings 1988; Graham and Jennings 2008). Ideal accessions produce spine-free and upright canes with strongly attached fruiting laterals, and they express consistent bud break, pest and disease resistances, and high yields of high quality, darkly colored fruit. Processing cultivars require adaptation to machine harvesting. Fresh-market cultivars require large, glossy fruit with inconspicuous seeds and long shelf-life.

BR yields lag behind those for similar crops. In 2016, U.S. commercial black raspberry plantings yielded 6,900 lbs. of fruit per acre harvested; red raspberries produced 14,600 lbs. of fruit per acre in the same year (U.S. Dept. of Agr. 2017a).

Cultivars suffer low yield due in part to high disease susceptibility, which also cuts stand life-time short (Kuhlman and Mumford 1949; Halgren et al. 2007). Current BR breeding

8 efforts are focused on Verticillium wilt resistance, aphid resistance, and tolerance to biotic stress (Bushakra et al. 2016a, 2018).

BR fruit tend to be small and seedy, whereas fresh fruit consumers prefer large fruit with little seeds (Chad Finn, USDA Rubus breeder, personal communication). In red raspberry, selection for increased drupelet count and drupelet size have resulted in larger fruit and increased yields (Daubeny 1996). However, breeding attempts to increase BR fruit size have been met with limited success (Slate and Klein 1952; Drain

1953, 1956).

Pre- and post-budbreak cold tolerance are important traits in regions that experience cold winters and late spring frosts, a common occurrence in the U.S.

Midwest and Southern regions (Gary Gao, personal communication). Heat tolerance is important in southern states, where rapid stand decline has been attributed to high summer temperatures (Bradish et al. 2016b). Fruit chemical components such as soluble solid content, acidity, and anthocyanin content contribute to the appearance, flavor, and potential health benefits of BR fruit, and these traits appeal to both consumers and human health researchers.

Interspecific hybridization has greatly contributed to red raspberry breeding efforts (Williams 1950; Keep 1989), and it has been posited that improvement of BR germplasm will require introgression of related Rubus species (Ourecky 1975).

However, attempts to widen the genetic base of BR through interspecific hybridization have found limited success due to loss of fertility and failure to consolidate important fruit quality traits (Slate and Klein 1952; Drain 1956; Ourecky and Slate 1966; Jennings

1988). For example, one attempt to increase fruit size in BR by crossing with R. idaeus

9 and backcrossing to R. occidentalis was unsuccessful at producing progeny with combined BR fruit qualities and increased fruit size (Slate and Klein 1952). A similar attempt to improve BR adaptation to U.S. Southern climates through interspecific crosses with R. albescens and R. crataegifolius resulted in improved vigor and abiotic stress tolerance at the cost of fruit quality, and none of the resulting accessions were commercialized (Drain 1956).

The only known cultivar bred from an interspecific cross is ‘Earlysweet’, which was selected from progeny of a cross between R. occidentalis cv. Haut and R. leucodermis Douglas ex. Torr. & Gray (Galleta et al. 1998). R. leucodermis is native to

Northwestern U.S. This species appears to contain resistance to bushy dwarf virus, an important disease of BR, and may provide a valuable source of genetics for BR improvement (Finn et al. 2003). R. leucodermis accessions have similar soluble sugar content, higher pH, and lower titratable acidity than BR (Finn et al. 2003). Further, R. leucodermis fruit have been described as having a flat, unappealing flavor, reportedly due to its low acidity (Finn et al. 2003).

R. occidentalis has been used as a donor to R. idaeus for multiple traits including aphid resistance (Keep and Knight 1967; Keep 1989) and increased fruit firmness

(Keep 1989; Daubeny 1996). Following interspecific hybridization of R. occidentalis and

R. idaeus, reclamation of fruit quality traits requires multiple generations of back- crossing to either species. Reclamation of dark-colored and glossy fruit requires many cycles of backcrossing to BR, during which desirable red raspberry traits such as fruit size are easily lost (Slate and Klein 1952). Comparably few generations of back-

10 crossing to R. idaeus are required in red raspberry following hybridization with BR, during which desired BR traits have been retained (Keep 1984, 1989).

Germplasm

Rubus is a widely distributed genus inhabiting all continents except Antarctica. It is estimated to contain 250-700 species (Lawrence et al. 2014). This wide range in species number estimates follows an especially difficult species-level taxonomy due to polyploidy, apomixis, and hybridization. Flora of North America lists 37 Rubus species,

18 of which may be classified as raspberries – having uncovered fruit which detach from the torus upon ripening; of these, 9 inhabit ranges overlapping that of R. occidentalis: R. chamaemorus L., R. deliciosus Torr., R. idaeus, R. illecebrosus Focke, R. neomexicanus A. Gray, R. odoratus L., R. parviflorus Nutt., R. parvifolius L., and R. phoenicolasius Maxim. (Lawrence et al. 2014).

The U.S. National Plant Germplasm System (NPGS) lists germplasm representing 157 Rubus species accessible through the Germplasm Resources

Information Network (GRIN). These include 230 R. occidentalis accessions – 27 cultivars and 203 wild accessions. Cultivars available through GRIN include ‘Munger’,

‘Jewel’, and ‘MacBlack’ as well as open-pollinated seeds of primocane fruiting

‘Explorer’.

Studies have revealed low genetic diversity within elite BR germplasm using both phenotypic and DNA marker analysis (Ourecky 1975; Weber 2003; Dossett et al.

2012a). Comparison between 16 BR cultivars sampled from various origins with 30

Random Amplified Polymorphic DNA (RAPD) markers revealed high levels of genetic similarity between cultivars (Weber 2003). Likewise, comparisons drawn between 21

11 BR cultivars and 137 wild BR accessions by 21 Simple Sequence Repeat (SSR) markers showed high levels of similarity between all sampled cultivars except ‘Explorer’, which was bred for a unique primocane flowering habit (Dossett et al. 2012a). In this same study, SSR marker evaluation of wild germplasm showed high levels of allelic diversity among wild accessions, suggesting wild germplasm contains genetic variation that is currently unexploited in elite material (Dossett et al. 2012a).

Breeding relevance of wild BR germplasm is supported by several experimental studies. In a field study comparing progeny of cultivated and wild accession crosses, high vigor and disease tolerance were described in progeny of a wild accession collected in North Carolina, NC-84-10-3 (Dossett et al. 2008). A separate study identified resistance to the large raspberry aphid (Amphorophora agathonica Hottes) in three wild BR accessions collected in Maine and Ontario, Canada (Bushakra et al.

2018). A. agathonica is an important vector of viruses associated with rapid BR stand decline. The overlapping locations of aphid-resistance conferring genes from these accessions have been mapped, providing a basis for the development of markers which may be used in marker-assisted selection (Bushakra et al. 2015, 2018).

To be adopted by growers, new BR cultivars will need to perform as well or better than current cultivars. Strategies for improving elite germplasm include integration of wild and elite germplasm followed by forward selection or backcrossing to elite germplasm. Important traits exhibited by current elite germplasm in comparison to wild accessions include larger fruit (as exhibited by ‘Jewel’ and ‘MacBlack)’ as well as suitability for machine harvest (as exhibited by ‘Munger’).

12 Statistical methods of genetic characterization

Trait characterization of experimental and cultivated germplasm can inform and increase efficiencies of crop production and breeding efforts. The following provides summary of three general methods used to characterize quantitative genetic components of diverse germplasm: analysis of variance components, mapping of quantitative trait loci, and description of genotype-by-environment interactions.

Analysis of variance components

Analysis of variance (ANOVA) originated with independent formulations of least squares described by Legendre (1806) and Gauss (1809) for astronomy research

(Searle 1989). Variance component models were later formulated by Fisher (1918,

1925) for genetics research (Searle 1989). Statistical analysis of quantitative genetic determinants currently center around partitioning of phenotypic variance into components attributable to genetic and non-genetic (i.e. environmental) factors

(Falconer and Mackay 1996). Genetic components may be partitioned further into additive variance (i.e. breeding value) and dominance deviation (Falconer and Mackay

1996).

Following ANOVA, heritability coefficients may be calculated to measure genetic determination of a quantitative trait. Heritability is expressed as a proportion of total variance that is attributable to breeding value, and it is calculated as the ratio of additive genetic variance over phenotypic variance (Falconer and Mackay 1996). Heritability is central to breeding, as a high heritability coefficient indicates a strong relationship between observed phenotype and breeding value.

13 Quantitative trait loci mapping

Quantitative trait loci (QTL) mapping enables characterization of quantitative traits through the identification and quantification of trait-associated genetic regions.

QTL analysis involves the statistical association of continuous trait data with genetic regions defined by polymorphic genetic markers. When population structure is well defined, QTL analysis can be informed by the construction of genetic linkage maps (Van

Ooijen and Jansen 2013). Selection of trait-linked markers can then be used in addition to or in place of traditional phenotypic selection to advance plant breeding efforts

(Collard et al. 2005; Collard and Mackill 2008).

QTL mapping is most straightforward for populations derived from two diploid parents that are homozygous at all genetic loci, as achieved by repetitive inbreeding or doubled-haploidy (Collard et al. 2005). In this case, all polymorphic markers are biallelic, and parental linkage phases are well defined. In the case where one or both parents are heterozygous, triallelic and tetra-allelic markers are possible, and linkage phases are unknown and must be inferred (Ritter and Salamini 1996; Maliepaard et al. 1997; Wu et al. 2002).

Power of a QTL analysis is dependent on many factors including the quality and quantity of genetic markers used, amount of meiotic recombination occurring in the mapping population, mapping population structure, number of individuals surveyed, and quality of the phenotypic data (Collard et al. 2005).

Linkage association studies generally utilize an F2 or backcross population with parents homozygous (or assumed to be homozygous) at all genetic loci. Such populations have simple structure with known linkage phases, due to homozygous

14 parents. However, homozygous parents are sometimes impossible to obtain due to practical or biological constraints including time, self-incompatibility (Silva and Goring

2001), and severe inbreeding depression (Voillemot and Pannell 2017). In such populations where both parents are heterozygous and without homozygous founders, up to four marker alleles may be present at any single locus, and marker linkage phases are unknown a priori. Segregating markers in these populations may be mapped using a two-way pseudo-testcross strategy, where recombination frequencies are estimated separately for each parent (Maliepaard et al. 1997). Such populations are commonly designated as cross pollinator (CP) populations and can be described by five co- dominant marker segregation types, lm×ll, nn×np, hk×hk ef×eg, and ab×cd (Van Ooijen and Jansen 2013). Segregation types lm×ll and nn×np describe bi-allelic markers that are segregating in one parent and fixed in the other. Segregation type hk×hk describes a bi-allelic marker that segregating in both parents. Segregation types ef×eg and ab×cd describe tri-allelic and tetra-allelic markers, respectively, that are segregating in both parents. Expected segregation patterns are 1:1 for lm×ll and nn×np markers, 1:2:1 for hk×hk markers, and 1:1:1:1 for ef×eg and ab×cd markers.

Description of genotype-by-environment interactions

Genotype-by-environment interaction (GEI) is the differential performance of a set of genotypes over heterogeneous environmental conditions. A multiple-environment trial (MET) is a common experiment used to evaluate and predict genotype performances over multiple environmental variables such as location and/or year. A wide array of statistical approaches are used to explore GEI in crops through METs

(van Eeuwijk 1995). These approaches are generally derived from a classical ANOVA

15 and vary by (i) their representation of GEI by additive or multiplicative parameters

(Kempton 1984) and (ii) their definition of genetic and environmental effects as fixed or random (Smith et al. 2005; Lian and de los Campos 2016).

In a simple analysis of MET data, ANOVA can be used to partition variance into additive genetic, environmental, and GEI components. Though this approach enables quantification of GEI, it does not describe the nature of GEI where it is found to be significant (Kempton 1984). Patterns of GEI may be further described as multiplicative terms in an Additive Main Effect Multiplicative Interaction (AMMI) model (Gollob 1968;

Gabriel 1971, 1978; Johnson and Graybill 1972; Gauch 1988) or in a factorial regression model such as the reaction norm model (Mooers 1921; Yates and Cochran

1938; Finlay and Wilkinson 1963).

Traditional ANOVA, AMMI, and factorial regression analyses consider genetic and environmental effects as fixed. This can be useful when genotype-by-environment data are balanced and direct comparison between levels of genotypes and/or environments are desired. However, it is restrictive in the assessment of unbalanced data, a frequent situation in METs. Further, in breeding and genetics trials, it is often desirable to observe many different genotypes. Thus, treatment of the genotype as a random effect is common in breeding and genetics studies (Patterson et al. 1977; Smith et al. 2005, 2007; Cullis et al. 2014; Hardner 2017; Smith and Cullis 2018; Hardner et al.

2019). A random effects reaction norm model using a Bayesian framework has also been developed (Su et al. 2006; Lian and de los Campos 2016), and this approach enables efficient estimation of genetic, environmental, and GEI parameters using unbalanced data.

16 In addition to their improved use of unbalanced data, random effects and mixed effects models allow the specification of (co)variance matrix structures for random effects (Lian and de los Campos 2016; Hardner et al. 2019) which may account for genetic correlations. Such a genetic covariance matrix can be designed using pedigree and/or genetic marker data representing the genetic resemblances between genotypes.

Several methods for developing genetic covariance matrices from genetic marker data have been developed (VanRaden 2008; Yang et al. 2010; Pérez-Elizalde et al. 2015).

Further, in a mixed model, covariance between fixed environmental effect levels can be estimated through the specification of residual (co)variance matrix structures (Smith et al. 2005). This allows the fitting of homogeneous or heterogeneous residual variance structures, depending on which makes the most parsimonious use of available data.

Study of GEI has been expanded to include QTL as genetic effects in order to define QTL-by-environment interactions (QEI), particularly in field crops and model species (Des Marais et al. 2013; Malosetti et al. 2013; Li et al. 2018). Approaches include overlaying QTL analyses conducted within and among multiple environments

(Malosetti et al. 2013; Li et al. 2018) and mapping estimated GEI effects from a reaction norm model (Li et al. 2018). A mixed model QTL analysis approach has also been developed which accounts for genetic covariance between MET environments, thus enabling investigation of QEI while protecting against Type I error.

Genetic characterization of black raspberry and related crops

Current genetic characterization of BR is limited, following limited breeding progress and economic importance in comparison to similar crop species. The following is a summary of genetic characterization efforts in BR and related, perennial crops.

17 Variance components in raspberry

Genetic improvement in red raspberry has been well demonstrated, and high additive genetic variance has been established for many red raspberry traits, including cropping nodes, multiple fruiting laterals per node, fruit weight, disease resistances (e.g. spur blight, fruit rot, can blight, powdery mildew, leaf spot, Phytophthora root rot), fruit firmness, and primocane fruiting (Daubeny 1996). Characterization of BR genetics is limited, partially due to deficient diversity in R. occidentalis germplasm (Jennings 1988).

Historic studies of BR heritability feature several failed attempts at heritable improvement, though these studies do not report estimated variance components (Slate and Klein 1952; Drain 1953, 1956).

Additional BR trait characterization can be found in studies regarding BR as a source of introgression material for red raspberry improvement. For example, resistance to Rubus aphid (Amphorophora rubi) was described in BR (Keep and Knight 1967) and successfully transferred into red raspberry through introgressive hybridization (Keep

1984). R. occidentalis has also been used as a source of fruit firmness in red raspberry breeding (Keep 1989; Daubeny 1996).

Resurging interest in BR as a high-value crop has led to increased interest in BR genetics. A recent study of diverse BR breeding germplasm found high genetic variance for fruit chemistry properties (i.e. pH, titratable acidity, percent soluble solids, anthocyanin profiles, and total anthocyanin) cane vigor, flowering time, and ripening time (Dossett et al. 2008). In addition, linkage mapping has been used to identify genetic positions of multiple genetic loci conferring resistance to A. agathonica

(Bushakra et al. 2015, 2018; Lightle et al. 2015).

18 Recent study of BR anthocyanin content has followed interest in dietary anthocyanins for their health-promoting characteristics (Fang 2015). Further, anthocyanins are responsible for the dark coloration of BR fruit. Genetic variation for anthocyanin content in wild BR accessions has been reported (Dossett et al. 2010,

2011, 2012b). Total anthocyanin content and total phenolic content were found to vary significantly between progeny groups of controlled BR crosses (Dossett et al. 2010). R. occidentalis accessions lacking anthocyanidin-3-rutinoside, a major component of the

BR anthocyanin modification pathway, have also been described (Dossett et al. 2011).

These mutants were found to have elevated cyanidin-3-sambubioside proportions and low total anthocyanin content (Dossett et al. 2011).

Production environment has also been shown to influence BR biochemistry. A comparison of BR cultivars grown in multiple Ohio production sites found significant differences in antioxidant capacity, anthocyanin content, and titratable acidity between production sites (Ozgen et al. 2008). In addition, this study found no significant differences in antioxidant capacity, anthocyanin content, or titratable acidity between the three tested cultivars (Ozgen et al. 2008), reinforcing a noted lack of genetic diversity in elite BR germplasm. A national MET of R. occidentalis germplasm grown at research stations in Oregon, Ohio, North Carolina, and New York found statistically significant differences in soluble solid content (SSC), pH, titratable acidity (TA), and SSC:TA between research locations as well as between genotypes (Perkins-Veazie et al. 2016).

GEI has been described in several red raspberry cultivar traits. A study of five red raspberry cultivars grown in three Pacific Northwest locations found significant GEI for aborted drupelet count, pH, soluble solid content, and titratable acidity (Burrows and

19 Moore 2002). Another study of three cultivars grown in six Eastern Canada locations identified two mega-environments based on GEI observed for yield (Sullivan et al.

2002).

Quantitative trait loci mapping in raspberry

Genetic mapping studies are more numerous for red raspberry than for BR. In an ongoing study at the James Hutton Institute in Scotland, an F1 population derived from a cross between 1930s cv. Latham and modern cv. Glen Moy has been used to elucidate control of many traits in red raspberry including cane spininess and root sucker density and diameter (Graham et al. 2004), root vigor and resistance to Phytophthora root rot

(Graham et al. 2011), cane pubescence (Graham et al. 2006), anthocyanin content and fruit color (Kassim et al. 2009; McCallum et al. 2010), phenolic content and antioxidant capacity (Dobson et al. 2012), and bud break and fruit ripening (Graham et al. 2009;

Simpson et al. 2017). In addition to identification of QTL for fruit ripening, Simpson et al.

(2017) identified overlap between mapped QTL and expression of candidate genes involved in fruit softening. Two anthocyanin QTL-linked markers identified by Kassim et al. (2009) have also been associated with anthocyanin content in progeny of an R. occidentalis × R. idaeus cross (Bushakra et al. 2013). However, the majority of ‘Latham’

× ‘Glen Moy’ QTL have not been validated outside the initial test population.

Twenty-seven anthocyanin-associated QTL were described and mapped in F1 progeny of an R. occidentalis × R. idaeus cross (Bushakra et al. 2012). Seven of these

QTL mapped to the BR parent, two of which were identifiable over all three years of testing (Bushakra et al. 2012). Instability over years observed in most anthocyanin-

20 associated QTL from this study indicate possible QEI and suggest a strong environmental influence on BR anthocyanin content (Bushakra et al. 2012).

Trait-mapping in R. occidentalis germplasm is currently limited to single loci associated with aphid resistance (Bushakra et al. 2015, 2018). Resistance-associated alleles in this germplasm follow Mendelian segregation patterns for single-gene dominant resistance and originated in R. occidentalis accessions collected in Maine,

U.S. and Ontario, Canada (Dossett and Finn 2010; Bushakra et al. 2018).

Genotype-by-environment interactions in raspberry and other perennial crops

GEI is a particular concern in perennial fruit crops, where breeding efforts are often restricted by long juvenile periods, self-incompatibility, susceptibility to inbreeding- depression, and selection of many complex traits (Byrne 2012). Cultivar development of a perennial fruit crop is expected to require ten years at minimum following the final cross (Byrne 2012). Thus, it is important that final selections have high genetic merit and stable trait expression across production environments. Estimation of GEI through mixed model analysis of METs is becoming increasingly common in perennial crops, including: sugarcane (Smith et al. 2007), pine (Smith and Cullis 2018), macadamia

(Hardner et al. 2001; Hardner 2017), and sweet cherry (Hardner et al. 2019). These studies enable the identification of high-performing and stable individuals as well as breeding value predictions for unbalanced METs. Several statistical tools enable mixed model analysis of MET data, including ASReml (Gilmour et al. 2015) and packages in the R Statistical Language and Framework (R Core Team, 2018), such as lmer (Bates et al., 2015).

21 GEI characterization of BR is currently limited to simple descriptive analyses of few traits. Differential performance of fruit firmness and vigor were reported for some experimental BR genotypes trialed in Oregon, Ohio, New York, and North Carolina, though no GEI statistics were reported (Bushakra et al. 2016b). A summary of GEI is available for the primocane-fruiting trait, though factors influencing GEI of this trait remain poorly understood (Dossett and Finn 2011).

Currently, there is no published report of QTL-by-environment interactions in BR.

22 References

Amiot MJ, Riva C, Vinet A (2016) Effects of dietary polyphenols on metabolic syndrome features in humans: a systematic review. Obes Rev 17:573–586. https://www.doi.org/10.1111/obr.12409

Bates D, Mächler M, Bolker B, Walker S (2015) Fitting linear mixed-effects models using lme4. J Stat Softw 67:1. https://www.doi.org/10.18637/jss.v067.i01

Bradish CM, Fernandez GE, Bushakra JM, et al (2016) Evaluation of vigor and winter hardiness of black raspberry breeding populations (Rubus) grown in the southeastern US. Acta Hortic 1133:129–134 .https://www.doi.org/10.17660/ActaHortic.2016.1133.19

Burrows C, Moore PP (2002) Genotype × environment effects on raspberry fruit quality. Acta Hortic 585, 467–473. https://www.doi.org/10.17660/ActaHortic.2002.585.75

Bushakra JM, Bassil NV, Weiland JE, et al (2016a) Comparative RNA-seq for the investigation of tolerance to Verticillium wilt in black raspberry. Acta Hortic 1133:103–114. https://www.doi.org/10.17660/ActaHortic.2016.1133.16

Bushakra JM, Bradish CM, Weber CA, et al (2016b) Toward understanding genotype × environment interactions in black raspberry (Rubus occidentalis L.). Acta Hortic 1117:25–30. https://www.doi.org/10.17660/ActaHortic.2016.1117.5

Bushakra JM, Bryant DW, Dossett M, et al (2015) A genetic linkage map of black raspberry (Rubus occidentalis) and the mapping of Ag 4 conferring resistance to the aphid Amphorophora agathonica. Theor Appl Genet 128:1631–1646. https://www.doi.org/10.1007/s00122-015-2541-x

Bushakra JM, Dossett M, Carter KA, et al (2018) Characterization of aphid resistance loci in black raspberry (Rubus occidentalis L.). Mol Breed 38:83. https://www.doi.org/10.1007/s11032-018-0839-5

Bushakra JM, Krieger C, Deng D, et al (2013) QTL involved in the modification of cyanidin compounds in black and red raspberry fruit. Theor Appl Genet 126:847– 865. https://www.doi.org/10.1007/s00122-012-2022-4

Bushakra JM, Stephens MJ, Atmadjaja AN, et al (2012) Construction of black (Rubus occidentalis) and red (R. idaeus) raspberry linkage maps and their comparison to the genomes of strawberry, apple, and peach. Theor Appl Genet 125:311–327. https://www.doi.org/10.1007/s00122-012-1835-5

Byrne DH (2012) Trends in fruit breeding. In: Fruit breeding. Springer, New York; London, pp 3–36

Collard BCY, Jahufer MZZ, Brouwer JB, Pang ECK (2005) An introduction to markers, quantitative trait loci (QTL) mapping and marker-assisted selection for crop 23 improvement: The basic concepts. Euphytica 142:169–196. https://www.doi.org/10.1007/s10681-005-1681-5

Collard BCY, Mackill DJ (2008) Marker-assisted selection: an approach for precision plant breeding in the twenty-first century. Philos Trans Royal Soc B 363:557–572

Cullis BR, Jefferson P, Thompson R, Smith AB (2014) Factor analytic and reduced animal models for the investigation of additive genotype-by-environment interaction in outcrossing plant species with application to a Pinus radiata breeding programme. Theor Appl Genet 127:2193–2210

Daubeny HA (1996) Brambles. In: Janick J, Moore JN (eds) Fruit Breeding, Vine and Small Fruits. Wiley, New York, pp 109–190

Des Marais DL, Hernandez KM, Juenger TE (2013) Genotype-by-environment interaction and plasticity: exploring genomic responses of plants to the abiotic environment. Annu Rev Ecol Evol S 44:5–29. https://www.doi.org/10.1146/annurev-ecolsys-110512-135806

Dobson P, Graham J, Stewart D, et al (2012) Over-seasons analysis of quantitative trait loci affecting phenolic content and antioxidant capacity in raspberry. J Agric Food Chem 60:5360–5366. https://www.doi.org/10.1021/jf3005178

Dossett M, Bassil NV, Lewers KS, Finn CE (2012a) Genetic diversity in wild and cultivated black raspberry (Rubus occidentalis L.) evaluated by simple sequence repeat markers. Genet Resour Crop Evol 59:1849–1865. https://www.doi.org/10.1007/s10722-012-9808-8

Dossett M, Finn CE (2011) Primocane-fruiting in black raspberry (Rubus occidentalis L.). J Am Pomol Soc 65:56–61

Dossett M, Finn CE (2010) Identification of resistance to the large raspberry aphid in black raspberry. J Am Soc Hortic Sci 135:438–444

Dossett M, Lee J, Finn CE (2012b) Anthocyanin content of wild black raspberry germplasm. Acta Hortic 946:43–48. https://www.doi.org/10.17660/ActaHortic.2012.946.3

Dossett M, Lee J, Finn CE (2008) Inheritance of phenological, vegetative, and fruit chemistry traits in black raspberry. J Am Soc Hortic Sci 133:408–417

Dossett M, Lee J, Finn CE (2010) Variation in anthocyanins and total phenolics of black raspberry populations. J Funct Foods 2:292–297. https://www.doi.org/10.1016/j.jff.2010.10.004

Dossett M, Lee J, Finn CE (2011) Characterization of a novel anthocyanin profile in wild black raspberry mutants: An opportunity for studying the genetic control of

24 pigment and color. J Funct Foods 3:207–214. https://www.doi.org/10.1016/j.jff.2011.04.003

Drain BD (1953) Some inheritance data with black raspberries. Proc Amer Soc Hort Sci 61:231–234

Drain BD (1956) Inheritance in black raspberry species. Proc Amer Soc Hort Sci 68:169–170

Falconer DS, Mackay TFC (1996) Introduction to quantitative genetics, 4th ed. Burnt Mill, England: Longman

Fang J (2015) Classification of fruits based on anthocyanin types and relevance to their health effects. Nutrition 31:1301–1306. https://www.doi.org/10.1016/j.nut.2015.04.015

Finlay KW, Wilkinson GN (1963) The analysis of adaptation in a plant-breeding programme. Aust J Agric Res 14:742–754

Finn C, Knight VH (2002) What’s going on in the world of Rubus breeding? Acta Hortic 585:31–38. https://www.doi.org/10.17660/ActaHortic.2002.585.1

Finn C, Wennstrom K, Link J, Ridout J (2003) Evaluation of Rubus leucodermis populations from the Pacific Northwest. HortScience 38:1169–1172

Fisher RA (1918) The correlation between relatives on the supposition of Mendelian inheritance. Trans R Soc, Edinburgh 52:399–433

Fisher RA (1925) Statistical methods for research workers, 1st edn. Oliver and Boyd, Edinburgh and London

Funt RC, Ross DS (2013) Soil and water management. In: Raspberries. CABI, pp 103– 120

Gabriel KR (1971) The biplot graphic display of matrices with application to principal component analysis. Biometrika 58:453–467

Gabriel KR (1978) Least squares approximation of matrices by additive and multiplicative models. J R Stat Soc B 40:186–196

Galleta GJ, Maas JL, Enns JM (1998) “Earlysweet” black raspberry. Fruit Var J 52:123– 124

Gao G, Draper E, Lewandowski R (2017) Raspberries for the home fruit planting. https://ohioline.osu.edu/factsheet/hyg-1421

Gauch HG (1988) Model selection and validation for yield trials with interaction. Biometrics 44:705–715. https://www.doi.org/10.2307/2531585

25 Gauss KF (1809) Theoria motus corporum celestrium in sectionibus conics solem ambientium. Perthes and Besser: Hamburg

Gilmour AR, Gogel B, Cullis BR, et al (2015) ASReml User Guide Release 4.1. VSN International Ltd, Hemel Hempstead

Gollob HF (1968) A statistical model which combines features of factor analytic and analysis of variance techniques. Psychometrika 33:73–115. https://www.doi.org/10.1007/BF02289676

Graham J, Hackett CA, Smith K, et al (2011) Towards an understanding of the nature of resistance to Phytophthora root rot in red raspberry. Theor Appl Genet 123:585– 601. https://www.doi.org/10.1007/s00122-011-1609-5

Graham J, Hackett CA, Smith K, et al (2009) Mapping QTLs for developmental traits in raspberry from bud break to ripe fruit. Theor Appl Genet 118:1143–1155. https://www.doi.org/10.1007/s00122-009-0969-6

Graham J, Jennings N (2008) Raspberry breeding. In: Jain S, Priyadarshan M (eds) Breeding tree crops. IBH & Science Publication Inc, Oxford

Graham J, Smith K, MacKenzie K, et al (2004) The construction of a genetic linkage map of red raspberry (Rubus idaeus subsp. idaeus) based on AFLPs, genomic- SSR and EST-SSR markers. Theor Appl Genet 109:740–749

Graham J, Smith K, Tierney I, et al (2006) Mapping gene H controlling cane pubescence in raspberry and its association with resistance to cane botrytis and spur blight, rust and cane spot. Theor Appl Genet 112:818–831

Halgren A, Tzanetakis IE, Martin RR (2007) Identification, characterization, and detection of black raspberry necrosis virus. Phytopathology 97:44–50. https://www.doi.org/10.1094/PHYTO-97-0044

Hardner C (2017) Exploring opportunities for reducing complexity of genotype-by- environment interaction models. Euphytica 213:248. https://www.doi.org/10.1007/s10681-017-2023-0

Hardner C, Winks C, Stephenson R, Gallagher E (2001) Genetic parameters for nut and kernel traits in macadamia. Euphytica 117:151–161

Hardner CM, Hayes BJ, Kumar S, et al (2019) Prediction of genetic value for sweet cherry fruit maturity among environments using a 6K SNP array. Hortic Res 6:6. https://www.doi.org/10.1038/s41438-018-0081-7

Hedrick UP (1925) The small fruits of New York. J.B. Lyon Company, Albany

Jennings DL (1988) Raspberries and blackberries: their breeding, diseases and growth. Academic Press, San Diego, CA

26 Jibran R, Dzierzon H, Bassil N, et al (2018) Chromosome-scale scaffolding of the black raspberry (Rubus occidentalis L.) genome based on chromatin interaction data. Hortic Res 5:8. https://www.doi.org/10.1038/s41438-017-0013-y

Johnson DE, Graybill FA (1972) An analysis of a two-way model with interaction and no replication. J Am Stat Assoc 67:862. https://www.doi.org/10.2307/2284651

Kassim A, Poette J, Paterson A, et al (2009) Environmental and seasonal influences on red raspberry anthocyanin antioxidant contents and identification of quantitative traits loci (QTL). Mol Nutr Food Res 53:625–634. https://www.doi.org/10.1002/mnfr.200800174

Keep E (1989) Breeding red raspberry for resistance to diseases and pest. In: Janick J (ed) Plant Breeding Reviews. Timber Press, Portland, Oregon, pp 245–322

Keep E (1984) Breeding Rubus and Ribes crops at East Malling. Sci Hortic 35:54–71

Keep E, Knight RL (1967) A new gene from Rubus occidentalis L. for resistance to strains 1, 2, and 3 of the rubus aphid, Amphorophora rubi Kalt. Euphytica 16:209–214. https://www.doi.org/10.1007/BF00043456

Kempton RA (1984) The use of biplots in interpreting variety by environment interactions. J Agr Sci 103:123. https://www.doi.org/10.1017/S0021859600043392

Kresty LA, Mallery SR, Stoner GD (2016) Black raspberries in cancer clinical trials: past, present and future. J Berry Res 6:251–261. https://www.doi.org/10.3233/JBR- 160125

Kuhlman GW, Mumford DC (1949) Cost of producing black raspberries for processing in the Willamette Valley, Oregon. Agricultural Experiment Station, Oregon State College, Corvallis

Lawrence AA, Goldman DH, Macklin JA, Moore G (2014) Rubus. In: Flora of North America Editorial Committee, eds. 1993+. Flora of North America North of Mexico. 20+ vols. New York and Oxford. Vol. 9. On-line version: http://www.efloras.org/florataxon.aspx?flora_id=1&taxon_id=128837

Legendre AM (1806) Nouvelles methodes pour la determination des orbites des cometes; avec un supplement contenant divers perfectionnements de ces methodes et leur application aux deux cometes de 1805. Courcier, Paris

Li X, Guo T, Mu Q, et al (2018) Genomic and environmental determinants and their interplay underlying phenotypic plasticity. Proc Natl Acad Sci U S A 115:6679– 6684. https://www.doi.org/10.1073/pnas.1718326115

27 Lian L, de los Campos G (2016) FW: an R package for Finlay–Wilkinson regression that incorporates genomic/pedigree information and covariance structures between environments. G3 6:589–597. https://www.doi.org/10.1534/g3.115.026328

Lightle D, Dossett M, Ebert T, et al (2015) Effects of three novel resistant black raspberry selections on Amphorophora agathonica feeding behavior and performance. Arthropod Plant Interact 9:487–496. https://www.doi.org/10.1007/s11829-015-9390-z

Maliepaard C, Jansen J, Van Ooijen JW (1997) Linkage analysis in a full-sib family of an outbreeding plant species: overview and consequences for applications. Genet Res 70:237–250. https://www.doi.org/10.1017/S0016672397003005

Malosetti M, Ribaut J-M, van Eeuwijk FA (2013) The statistical analysis of multi- environment data: modeling genotype-by-environment interaction and its genetic basis. Front Physiol 4:44. https://www.doi.org/10.3389/fphys.2013.00044

Mazzoni L, Perez-Lopez P, Giampieri F, et al (2016) The genetic aspects of berries: from field to health: The genetic aspects of berries. J Sci Food Agric 96:365–371. https://www.doi.org/10.1002/jsfa.7216

McCallum S, Woodhead M, Hackett CA, et al (2010) Genetic and environmental effects influencing fruit colour and QTL analysis in raspberry. Theor Appl Genet 121:611–627. https://www.doi.org/10.1007/s00122-010-1334-5

Mooers CA (1921) The agronomic placement of varieties. J Am Soc Hortic Sci 13:337– 352

Ourecky DK (1975) Brambles. In: Janick J, Moore JN (eds) Advances in fruit breeding. Purdue University Press, West Lafayette, IN, pp 98–129

Ourecky DK, Slate GL (1973) Jewel black raspberry. New York’s Food and Life Sci Bul 35

Ourecky DK, Slate GL (1966) Hybrid vigor in Rubus occidentalis x Rubus leucodermis seedlings. In: Proceedings of the XVII International Horticultural Congress. Abstract 277

Ozgen M, Wyzgoski FJ, Tulio AZ, et al (2008) Antioxidant capacity and phenolic antioxidants of Midwestern black raspberries grown for direct markets are influenced by production site. HortScience 43:2039–2047

Patterson HD, Silvey V, Talbot M, Weatherup STC (1977) Variability of yields of cereal varieties in U. K. trials. J Agr Sci 89:239. https://www.doi.org/10.1017/S002185960002743X

Pérez-Elizalde S, Cuevas J, Pérez-Rodríguez P, Crossa J (2015) Selection of the bandwidth parameter in a Bayesian kernel regression model for genomic-

28 enabled prediction. J Agric Biol Environ Stat 20:512–532. https://www.doi.org/10.1007/s13253-015-0229-y

Perkins-Veazie P, Ma G, Fernandez GE, et al (2016) Black raspberry fruit composition over two years from seedling populations grown at four US geographic locations. Acta Hortic 1133:335–338. https://www.doi.org/10.17660/ActaHortic.2016.1133.52

Pritts M (2008) Primocane-fruiting raspberry production. HortScience 43:1640–1641. https://www.doi.org/10.21273/HORTSCI.43.6.1640

Pritts M, McDermott L, Demchak K, et al (2017) High tunnel raspberries and blackberries. Dept. of Horticulture, Cornell University

R Core Team (2018) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria

Ritter E, Salamini F (1996) The calculation of recombination frequencies in crosses of allogamous plant species with applications to linkage mapping. Genet Res 67:55. https://www.doi.org/10.1017/S0016672300033474

Searle SR (1989) Variance components - some history and a summary account of estimation methods. J Anim Breed Genet 106:1–29. https://www.doi.org/10.1111/j.1439-0388.1989.tb00209.x

Silva NF, Goring DR (2001) Mechanisms of self-incompatibility in flowering plants. Cell Mol Life Sci 58:1988–2007

Simpson CG, Cullen DW, Hackett CA, et al (2017) Mapping and expression of genes associated with raspberry fruit ripening and softening. Theor Appl Genet 130:557–572. https://www.doi.org/10.1007/s00122-016-2835-7

Slate GL, Klein LG (1952) Black raspberry breeding. Proc Amer Soc Hort Sci 59:266– 268

Smith AB, Cullis BR (2018) Plant breeding selection tools built on factor analytic mixed models for multi-environment trial data. Euphytica 214:143. https://www.doi.org/10.1007/s10681-018-2220-5

Smith AB, Cullis BR, Thompson R (2005) The analysis of crop cultivar breeding and evaluation trials: an overview of current mixed model approaches. J Agr Sci 143:449–462. https://www.doi.org/10.1017/S0021859605005587

Smith AB, Stringer JK, Wei X, Cullis BR (2007) Varietal selection for perennial crops where data relate to multiple harvests from a series of field trials. Euphytica 157:253–266. https://www.doi.org/10.1007/s10681-007-9418-2

29 Su G, Madsen P, Lund MS, et al (2006) Bayesian analysis of the linear reaction norm model with unknown covariates. J Anim Sci 84:1651–1657. https://www.doi.org/10.2527/jas.2005-517

Sullivan JA, Yan W, Privé JP (2002) Genotype/genotype × environment biplot analysis for cultivar evaluation and mega-environment investigation in primocane-fruiting red raspberry. J Am Soc Hortic Sci 127:776–780. https://www.doi.org/10.21273/JASHS.127.5.776

Tallman PH (2016) Black raspberry plant named ‘Niwot’. US PP27,131. http://patft.uspto.gov/netacgi/nph- Parser?Sect1=PTO2&Sect2=HITOFF&p=1&u=%2Fnetahtml%2FPTO%2Fsearch - bool.html&r=1&f=G&l=50&co1=AND&d=PTXT&s1=%22black+raspberry%22&s2 =Niwot&OS=

U.S. Dept. of Agr. (2017a) National statistics for raspberries. https://www.ers.usda.gov/data-products/fruit-and-tree-nut-data/data-by- commodity/

U.S. Dept. of Agr. (2017b) Fruit and tree nut yearbook tables. https://www.ers.usda.gov/data-products/fruit-and-tree-nut-data/fruit-and-tree-nut- yearbook-tables/ van Eeuwijk FA (1995) Linear and bilinear models for the analysis of multi-environment trials: I. An inventory of models. Euphytica 84:1–7. https://www.doi.org/10.1007/BF01677551

Van Ooijen JW, Jansen J (2013) Genetic mapping in experimental populations. Cambridge University Press, New York

VanBuren R, Bryant D, Bushakra JM, et al (2016) The genome of black raspberry (Rubus occidentalis). Plant J 87:535–547. https://www.doi.org/10.1111/tpj.13215

VanBuren R, Wai CM, Colle M, et al (2018) A near complete, chromosome-scale assembly of the black raspberry (Rubus occidentalis) genome. GigaScience 7:1– 9. https://www.doi.org/10.1093/gigascience/giy094

VanRaden PM (2008) Efficient methods to compute genomic predictions. J Dairy Sci 91:4414–4423. https://www.doi.org/10.3168/jds.2007-0980

Voillemot M, Pannell JR (2017) Inbreeding depression is high in a self-incompatible perennial herb population but absent in a self-compatible population showing mixed mating. Ecol Evol 7:8535–8544. https://www.doi.org/10.1002/ece3.3354

Weber CA (2003) Genetic diversity in black raspberry detected by RAPD markers. HortScience 38:4

30 Williams CF (1950) Influence of parentage in species hybridization of raspberries. Proc Amer Soc Hort Sci 56:149–156

Wu R, Ma C-X, Painter I, Zeng Z-B (2002) Simultaneous maximum likelihood estimation of linkage and linkage phases in outcrossing species. Theor Popul Biol 61:349– 363. https://www.doi.org/10.1006/tpbi.2002.1577

Yang J, Benyamin B, McEvoy BP, et al (2010) Common SNPs explain a large proportion of the heritability for human height. Nat Genet 42:565–569. https://www.doi.org/10.1038/ng.608

Yates E, Cochran WG (1938) The analysis of groups of experiments. J Agric Sci Camb 28:556–580

31 Figures

Figure 1.1. Native range of Rubus occidentalis by state and province. https://plants.usda.gov/core/profile?symbol=ruoc. Accessed May 4, 2018.

32 US Per Capita Small Fruit Consumption, Fresh 2.00 8.00 7.00 1.50 6.00 5.00 4.00 1.00 3.00 2.00 0.50

Strawberries (pounds) 1.00

0.00 0.00 Blueberries & Raspberries (pounds) 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 Strawberries Blueberries Raspberries

Figure 1.2. U.S. per capita fresh small fruit consumption.

33 Tables

Table 1.1. Monthly normal maximum and minimum temperatures at research locations [Ohio State University (Ohio Agricultural Research and Development Center, Wooster, Wayne County); Oregon State University (Lewis-Brown Farm, Corvallis, Benton County); Cornell University (New York State Agricultural Experiment Station, Geneva, Seneca County); and North Carolina State University (Sandhills Research Station, Jackson Springs, Montgomery County)] testing adaptability of black raspberry (1981- 2010, NOAA). Daily Temperature (°F) Corvallis, OR Wooster, OH Jackson Springs, NC Geneva, NY High Low High Low High Low High Low Jan 47.1 34.2 34.5 18.2 50.2 30.3 30.9 16 Feb 50.9 34.7 38.1 19.9 54.3 32.9 33.2 17.5 Mar 56.1 37.3 48 27.3 62.3 38.2 41.3 24.8 Apr 60.8 39.6 61 37.5 71.7 46.3 54.9 36.2 May 67.2 44.1 70.9 47.3 78.9 55.2 66.9 46.5 Jun 73.2 48.5 79.5 56.5 86.2 64.3 75.9 56.4 Jul 81.7 51.8 83.1 60.2 89 68 80 61 Aug 82.6 51.1 81.7 58.4 87.4 67 78.4 59.3 Sep 77 47.7 74.6 50.8 81.6 60.4 71.1 51.7 Oct 64.7 41.7 62.7 40.1 71.8 49 59 41 Nov 52.6 37.9 50.7 32.1 62.8 41.5 47.4 32.5 Dec 45.8 33.4 38.4 22.8 53 33.3 35.9 22.5 Summary 63.3 41.8 60.3 39.3 70.8 48.9 56.2 38.8

34 Table 1.2. Monthly normal precipitation at research locations testing adaptability of black raspberry (1981-2010, NOAA). Precipitation (in.) OR OH NC NY Jan 6.4 2.29 3.88 1.71 Feb 5.11 1.97 3.54 1.49 Mar 4.44 2.79 4.04 2.35 Apr 2.91 3.65 3.05 2.85 May 2.31 4.45 3.09 3.06 Jun 1.52 4.45 4.05 3.67 Jul 0.49 4.28 5.03 3.47 Aug 0.53 3.88 5.12 3.05 Sep 1.25 3.43 4.22 3.48 Oct 3.1 3.07 3.72 3.33 Nov 6.94 3.11 3.32 2.86 Dec 7.71 2.71 3.27 2.21 Summary 42.71 40.08 46.33 33.53

35 Chapter 2. Genotype-by-environment interaction analysis of a multiple-environment black raspberry trial

Introduction

Black raspberry biology and cultivation

Black raspberry (BR, Rubus occidentalis L.) is a minor U.S. crop and a source of germplasm for red raspberry breeding. Black raspberry is endemic to eastern North

America where it is enjoyed both as a wild forage and commercially cultivated crop.

Recent interest in BR has followed discoveries of potential health benefits associated with its fruit (Amiot et al. 2016; Kresty et al. 2016) and general increase in U.S. demand for fresh small fruit (U.S. Dept. of Agr. 2017).

U.S. BR production is divided into fresh and processing markets, and growing regions vary in geography and climate. Processing market production is concentrated in the U.S. Northwest region and is primarily dependent on ‘Munger’, a cultivar suitable for machine harvesting. Fresh market production is spread among the U.S. Northwest,

Northeast, and Midwest regions and is largely dependent on ‘Jewel’ and ‘MacBlack’

(Ourecky and Slate 1973; Weber 2015), two cultivars selected for consistent performance and large fruits. There is also interest in expanding BR production into the

U.S. Southern region, where it is currently limited by high summer temperatures

(Bradish et al. 2016a).

Black raspberry domestication has a short history attributable to availability of wild fruit and a general preference for red raspberry cultivars imported from Europe

(Jennings 1988). The U.S. production is currently limited due to scant cultivar selection for growers. Further, breeding efforts have likely been limited by narrow elite germplasm

36 (Weber 2003; Dossett et al. 2012a). Introgression of wild accessions into current elite germplasm may provide breeders with increased allelic diversity needed to make genetic gains in relevant BR traits (Dossett et al. 2012a).

Black raspberry production may be increased by breeding improved cultivars.

Ideal cultivars for fresh and processing markets are high yielding and flavorful with good resistance to pests and diseases and a wide production range. Target traits for fresh market production also include large, dark fruit with a small seed fraction.

Genotype-by-environment interaction

BR is cultivated in diverse U.S. geographic regions. Given limited resources for

BR breeding, efficient assessment of breeding germplasm among diverse production environments is paramount to successful genetic improvement of this crop. Genotype- by-environment interactions (GEI) are observed as differential genetic response of accessions grown in heterogeneous environments via a multiple-environment trial

(MET).

GEI characterization of BR is currently limited to simple descriptive analyses of few traits. Differential performance of fruit firmness and vigor were reported for some experimental BR genotypes trialed in Oregon, Ohio, New York, and North Carolina, though no GEI statistics were reported (Bushakra et al., 2016b). A summary of GEI is available for the primocane-fruiting trait, and factors influencing GEI of this trait are not well understood (Dossett and Finn, 2011).

Multiple methods exist for analysis of GEI (van Eeuwijk 1995). Due to high costs associated with establishment and maintenance, METs of perennial crops are often unbalanced and often lack within-environment replication. Mixed effects models

37 optimize use of unbalanced data and allow estimation of complex variance-covariance structures for random effects (Smith et al. 2005). This approach is becoming increasingly common in perennial crops, including: sugarcane (Smith et al. 2007), pine

(Smith and Cullis 2018), macadamia (Hardner et al. 2001; Hardner 2017), and sweet cherry (Hardner et al. 2019). Further, fitting of random genetic effects allows definition of individual genetic covariance with an additive genomic relationship matrix (Hardner et al. 2019) which can be generated through a known pedigree, by use of genome-wide molecular markers, or by a combination of the two (Muñoz et al. 2014). GEI is quantified for each environment as heterogeneous genetic effects and heritability coefficients.

The Finlay-Wilkinson regression model is a factorial regression model where environmental effect is used as a concomitant variable on the environmental factor

(Finlay and Wilkinson 1963; van Eeuwijk 1995). A Bayesian framework may be used to fit random genetic effects in a Finlay-Wilkinson regression model (Su et al. 2006; Lian and de los Campos 2016). Advantages of the Bayesian framework include efficient treatment of unbalanced data and incorporation of genetic relationships through pedigree or marker data. GEI is quantified for each individual as a linear regression of phenotype over environmental effects.

Applied to single-trait MET data, these methods allow elucidation of patterns in

GEI through several statistics, including pairwise genetic correlation between environments, proportion of variance attributable to genetics within each environment, and phenotypic stability estimates for each individual. These statistics may then be used to identify environments and genotypes driving GEI in BR complex traits.

38 Fruit size and fruit chemistry

Fruit size and its components are important traits for fresh market raspberry production. Historical trends in fresh red raspberry (Rubus idaeus) have shown heritable increases in infructescence (fruit) size through selection of increased drupelet count and drupelet size (Daubeny 1996). A similar trend may be expected for the case of BR, but formal genetic analyses are lacking. Current limitations to BR fresh market production include small fruit which are perceived by consumers as very seedy (Chad Finn, USDA

BR breeder, personal communication). Increased fruit size and reduced seediness are thus important goals to the fresh BR market. Breeding progress for increased fruit size and reduced seediness will rely on the identification and selection of genetic components contributing to these traits.

There is considerable interest in breeding for improved fruit flavor in horticultural crops (Klee 2010). SSC and TA are important components of flavor. In consumer testing studies, low SSC and high TA have been associated with high sourness and disliking of red raspberry fruit, whereas high soluble solid content has been associated with sweetness and liking of red raspberry fruit (Shamaila et al. 1993; Villamor et al.

2013). Both SSC and TA are complex traits that vary between genetic accessions and production locations (Mazur et al. 2014b), as well as during fruit development (Graham and Brennan 2018) of red raspberry.

There is also interest in exploring BR phenolic content and composition. BR fruit are rich in anthocyanins and other phenolic compounds, the dietary benefits of which is an active area of research (Fang 2015; Amiot et al. 2016; Kresty et al. 2016).

Anthocyanin content also contributes to dark coloration of BR fruit. Appearance is an

39 important component of flavor (Klee 2010), and darkly colored BR fruit are preferred over lightly colored fruit. Anthocyanins also function as a dye in processed BR products

(Dossett et al. 2010). High acidity and low pH promote anthocyanin stability, making them important traits for the raspberry processing market (Daubeny 1996).

To study breeding potential of fruit size and fruit chemistry traits in BR, two mapping populations having wild and cultivated ancestry were developed and tested at four locations representative of U.S. production regions (Dossett 2011). Fruit size and fruit chemistry data were collected in three growing seasons at each location. Measured traits include: infructescence (fruit) mass, seed mass, drupelet count, seed fraction, titratable acidity, pH, soluble solid content, SSC:TA, total anthocyanin content, and total phenolics content. The objective of the present study is to quantify and characterize GEI via a multiple-environment trial for these traits among diverse BR production environments. Two statistical approaches are used, linear mixed model analysis and

Bayesian Finlay-Wilkinson regression.

Materials and methods

Germplasm

Two mapping populations, ORUS 4304 and ORUS 4305, were constructed through controlled pollination in Corvallis, Oregon. ORUS 4304 and ORUS 4305 are composed of 192 and 115 individuals, respectively. Founders for both populations included the fresh market cultivar, ‘Jewel’ (Ourecky and Slate 1973), and the wild accession, NC 84-10-3. Wild accessions ORUS 3817-2 and ORUS 3778-1 are founders for ORUS 4304 and ORUS 4305, respectively (Figure 2.1).

40 Population grandparents were selected for their exhibition of desirable traits including desirable fruit quality, high field vigor, and aphid resistance. ‘Jewel’ is a popular fresh market BR cultivar selected for producing large, firm, good quality fruit

(Ourecky and Slate 1973). NC 84-10-3 was collected from North Carolina where it was selected for tolerance to biotic and abiotic stress. ORUS 3817-2 and ORUS 3778-1 were collected in Maine and Ontario, Canada, respectively, and have each displayed resistance to the large raspberry aphid (Amphorophora agathonica), an important vector of viruses associated with rapid BR stand decline (Dosset and Finn 2010; Bushakra et al. 2018).

Phenotypic data

ORUS 4304 and ORUS 4304 progeny, parents, and common ancestor, ‘Jewel’, were planted at four U.S. locations [Corvallis, Oregon (OR); Wooster, Ohio (OH);

Jackson Springs, North Carolina (NC); and Geneva, New York (NY)] in the spring of

2012. Fertilizer, pesticides, and irrigation were applied according to local recommendations. In all locations, first-year canes were tipped (i.e. cut to ~1m in length) each summer to promote lateral branch growth.

Ripe fruit were hand-collected at trial locations in the summers of 2013, 2014, and 2015 in a total of eleven location-by-year environments. Fruit were determined to be ripe when once they were easily removed from the torus. Collected fruit were frozen and sent to Kannapolis, North Carolina, where they were analyzed for chemical and physical fruit quality traits: titratable acidity (TA, % citric acid), pH, soluble solid content

(SSC, °Brix), total anthocyanin content (mg/kg cyanidin 3-glucoside equivalents), total phenolics content (mg/kg gallic acid equivalents), average infructescence (fruit) mass,

41 and average seed mass (Perkins-Veazie et al. 2016). SSC and TA were used to calculate the ratio, SSC:TA. Drupelet count was calculated as number of seeds per fruit.

Seed fraction was calculated as seed mass per fruit mass.

Most individuals (parents and progeny) were replicated across locations; however, not all individuals were replicated across trial locations (Figure 2.2). Further, phenotypic data collected within locations were incomplete across trial years (Figure

2.3). Progeny individuals were not replicated within locations. Fruit and seed size traits were not recorded in Oregon in 2013. Due to poor plant performance, no fruit were collected in North Carolina in 2015.

DNA marker data & relatedness analysis

DNA was extracted, libraries were prepared, and genotyping-by-sequencing

(GBS) data were collected in three separate batches composed of a total of seven sequencing runs. Due to logistic situations, GBS runs were completed in Oregon, North

Carolina, and Ohio. In all, four GBS runs were completed in Oregon, two in North

Carolina, and one in Ohio. Initial GBS data were collected in Oregon and North Carolina as described by Bushakra et al. (2015) and Cash (2016), respectively. Preliminary analysis of these data revealed low sequence depth for 145 individuals, and additional sequence data were collected in Ohio via a single GBS run with 72 available individuals of the 145 that had low depth. In Ohio, leaf samples were collected, lyophilized, and held in dry storage before DNA extraction. Leaf tissue was ground in liquid nitrogen by mortar and pestle, and DNA was extracted using a modified CTAB protocol (Lodhi et al.

1994). Following RNase treatment, DNA was further purified using the Genomic DNA

Clean & Concentrator-10 kit (Zymo Research, Irvine, CA). Genomic DNA concentration

42 and quality were initially assessed by a NanoDrop ND-1000 spectrophotometer

(NanoDrop Technologies, Wilmington, DE) and subsequently assessed using a Qubit 4 fluorometer (Thermo Fisher, Waltham, MA) and by gel electrophoresis.

In Ohio, genomic DNA samples were submitted to The Ohio State University

Molecular and Cellular Imaging Center (Wooster, OH, USA) for GBS library construction and sequencing. GBS libraries were constructed following Elshire et al. (2011). Briefly, genomic DNA samples were digested with ApeKI and ligated to adapters compatible with ApeKI sticky ends and containing unique barcodes with partial Illumina® adaptor sequences. Following PCR amplification using Illumina® Nextera adapters (Illumina,

Inc.), libraries were pooled and quantified using a Qubit® fluorometer (Thermo Fisher,

Waltham, MA, USA), and size distribution and concentration was confirmed using the

4200 TapeStation (Agilent Technologies, Santa Clara, CA, USA). The pooled library was analyzed by single-read sequencing of 151 cycles on the NextSeq (Illumina, Inc.) instrument at the HudsonAlpha Institute for Biotechnology (Huntsville, AL, USA).

All GBS data, representing 305 individuals, were analyzed simultaneously. Data were initially subjected to quality control and adapter removal using FastQC v. 0.11.7

(Andrews 2010) and Trimmomatic v. 0.36 (Bolger et al. 2014). Single nucleotide polymorphisms (SNPs) were identified using the TASSEL 5 GBS v2 pipeline (Glaubitz et al. 2014) and the BR reference genome sequence (VanBuren et al. 2018). SNP- guided relationship analyses were conducted separately for populations ORUS 4304 and ORUS 4305 using VCFtools v. 1.16 (Danecek et al. 2011) and the method described by Yang et al. (2010). Estimated relationship statistics were then checked against pedigree population structure to confirm relatedness of population progeny.

43 To describe additive genetic covariance between individuals, an additive genomic relationship matrix (GA) was generated by Euclidean distance estimation (Nei

1987) of SNP markers using R package synbreed v 0.12.9 (Wimmer et al. 2012).

Mixed model analysis

Phenotypic data were checked to approximately meet assumptions of normal distribution for each trait-by-environment combination by visual examination of histograms. Phenotypic data were checked for homoscedasticity for each trait among environments by evaluation of residuals in a multivariate fixed effect ANOVA.

To estimate within-environment variance parameters, multiple genomic and residual variance structures were fit for each phenotypic trait using ASReml v. 4.1

(Gilmour et al. 2015) and statistical methods described by Hardner et al. (2019). Briefly, a general mixed model was fit with fixed location-by-year environment effects and random additive genomic effects captured in a relationship matrix. Variation of observations were assumed to follow a multivariate normal distribution. A GA matrix was generated as described above and incorporated into the model.

Additive genomic-by-environment variance-covariance matrix (GA×E) structures and residual variance-covariance matrix (R) structures were modelled as described by

Isik et al. (2017). Four models were fit for each trait. Briefly, GA×E structures were fit as homogeneous with uniform correlation (Model 1), heterogeneous with uniform correlation (Models 2 & 3), and heterogeneous with non-uniform correlation estimated by factor analysis (Model 4). Residual variance structures were fit as independent with homogeneous variance (Models 1 & 2) and independent with heterogeneous variance

44 (Models 3 & 4). The most parsimonious model (i.e. the simplest model with the greatest explanatory power) was selected for all traits by likelihood ratio tests (Isik et al. 2017).

Uniform genomic correlation among the location-by-year environments was estimated as:

� = , Equation 1 ×

where � was the pairwise genomic correlation among environments, � was the

additive genomic variance, and �× was the genome-by-environment variance.

Additive genomic variance was thus estimated as:

� = (� + �×) ∗ �, Equation 2

where (� + �×) was the diagonal element of the GA×E structure and � was the off-diagonal. Heritability was estimated within each environment as:

ℎ = , Equation 3

where and were the estimated additive genomic and residual variances, � � respectively, for the jth location-by-year environment.

To explore correlation between traits, individual breeding values (BVs) were predicted across environments by best linear unbiased prediction of individual genetic effects. Pairwise Pearson correlation coefficients were then estimated between BVs for all traits.

Finlay-Wilkinson regression

To explore reaction norm patterns of individuals across location-by-year environments (i.e. individual phenotypic expression across a range of environments), a

Bayesian Finlay-Wilkinson regression model was fit for each trait using R package FW

45 (Lian and de los Campos 2016). In a reaction norm model, the phenotypic record of the ith individual observed in the jth environment is modeled as:

� = � + � + ℎ + �ℎ + �, Equation 4

th th where gi is the main effect of the i individual, hj is the main effect of the j environment, bi + 1 is the change of expected individual performance per unit change of hj, and εij is an error term. In Bayesian inference, this model is expressed in the following likelihood:

�(�|�) = ∏ N( � + � + ℎ + �ℎ, � ), Equation 5

where �(�|�) is the conditional probability of the observed phenotypic

observations (y) given a matrix of estimated parameters (θ = {μ, g, b, h, � , � , � , � }).

This conditional distribution is equal to the product of probabilistic distributions

(assumed to be normal) with an estimated mean � (from Equation 4) and variance � , for each individual sampled by Markov chain Monte Carlo simulation. Thus, each trait was analyzed considering probabilistic properties of genetic effects, environmental effects, and GEI.

A GA matrix was fit as a covariance structure describing covariance between individual genetic effects as well as between individual slope effects. This GA matrix was identical to that used in the mixed model analysis described above.

Variance components (� , � , � , and � ) were estimated based on the posterior distributions from three Markov chain Monte Carlo simulations with 60,000 iterations, each followed by removal of the first 10,000 iterations and a thinning interval of 5

(N=10,000). Convergence of the chains was determined by visual observation of the

46 corresponding trace plots. Coefficients of variation were computed using sample distribution means to assess dispersion of variance components around trait means.

th Instability of the i individual was interpreted as deviation of individual slope (bi)

th from zero. BV of the i individual was interpreted as individual genetic effect (gi).

Significance of each effect was determined by highest posterior density interval estimation using the R package coda v. 0.19.2 (Plummer et al. 2006).

To explore relationships between BV and phenotypic stability, Pearson correlation coefficients were computed between individual slopes and genetic effects for each trait. To explore relatedness between mixed model analysis and Finlay-Wilkinson regression results, Pearson correlation coefficients were computed between BVs predicted by each method.

Results

Marker data & relatedness analysis

A total of 242,391 SNPs were identified among 305 individuals using the

TASSEL 5 GBS v2 pipeline. Following SNP-calling, 14 individuals were excluded from further analysis following inability to call reliable marker genotypes due to low marker depth. Relatedness analysis of ORUS 4304 was conducted using 992 filtered SNPs

(Table 2.1). Relatedness analysis of ORUS 4305 was conducted using 1763 filtered

SNPs (Table 2.2). Relatedness analysis of ORUS 4304 indicated that 38 progeny individuals resulted from self-pollination of the maternal parent and one additional off- type resulted from an unknown cross (Figure 2.4). Relatedness analysis of ORUS 4305 indicated five off-types of unknown origin: 4305-38, 4305-39, 4305-41, 4305-59, 4305-

47 65. In total, thirteen ORUS 4304 individual and seven ORUS 4305 individuals were removed from the study due to low marker depth or genetic off-types.

A GA matrix was estimated for the remaining 285 individuals [174 ORUS 4304 progeny, 107 ORUS 4305 progeny, parents (3021-2, 4153-1, 4158-2), and a common grandparent (‘Jewel’)] using 427 SNPs (Table 2.3). This GA matrix was used to define variance-covariance of additive genomic effects in mixed model analysis and Finlay-

Wilkinson regression analysis of all phenotypic traits.

Fruit size traits

Mixed model analysis

Data sets for each fruit size trait (fruit mass, seed mass, drupelet count, and seed fraction) were determined to be homoscedastic and normally distributed. A model fitting heterogeneous GA×E variance with uniform correlation between environments and independent heterogeneous residual variance (Model 3) was found to be the most parsimonious model (i.e. the simplest model with the greatest explanatory power) for most traits (p<0.05, Table 2.4). For fruit mass, an extended factor analytic model which estimates non-uniform GA×E correlation structure (Model 4) was found to be the most parsimonious model (p<0.05, Table 2.4). However, near-zero estimates of residual variance limit interpretation of extended factor analysis solutions for this trait. Thus, a single model fitting heterogeneous GA×E variance with uniform correlation between environments and independent heterogeneous residual variance (Model 3) was selected, as it best fits all fruit size traits.

Genome-by-environment correlations were estimated for each trait (Table 2.5).

Correlations were highest for drupelet count (0.81, s.e. = 0.049) and lowest for seed

48 mass (0.56, s.e. = 0.057). Heritability estimates for each trait-by-environment combination are summarized in Table 2.6. Heritability ranged 0.01 to 0.70. Fruit size heritability estimates were highest in OH and lowest in NC and NY.

Correlations between trait BVs are summarized in Figure 2.5. There were significant (p<0.01) correlations between drupelet count and fruit mass (r=0.74), drupelet count and seed fraction (r=0.40), and drupelet count and seed mass (r=-0.32).

Finlay-Wilkinson Regression

Finlay-Wilkinson regression was used to model and plot reaction norms over the ten trial environments (Figure 2.6). Markov chain convergence was observed for all fruit size traits: fruit mass, seed mass, drupelet count, and seed fraction. Variance components estimated from posterior distributions MCMC chain simulations are summarized in Table 2.7, and coefficients of variation are summarized in Table 2.8. Low slope variance can be interpreted as high genetic covariance between test environments. Genetic coefficients of variation range from 7.6 to 19.1, and slope coefficients of variation range from 3.8 to 617.8 (Table 2.8). There was a significant positive correlation between genetic effect and slope effect for each fruit size trait

(r=0.18 to 0.57, p£0.0001 to 0.006, Table 2.9). In other words, individuals with high breeding values tended to perform much better in high-performing environments than in low-performing environments. There was also a positive correlation between BVs estimated by mixed model analysis and Finlay-Wilkinson regression for each fruit size trait (r=0.93 to 0.98, Table 2.10).

49 Fruit chemistry traits

Mixed model analysis

Data for each fruit chemistry trait (TA, pH, SSC, SSC:TA, anthocyanin content, and phenolic content) were determined to be homoscedastic and normally distributed. A model fitting heterogeneous GA×E variance with uniform correlation between environments and independent heterogeneous residual variance (Model 3) was found to be the most parsimonious model for most traits (p<0.05, Table 2.11). For phenolic content, an extended factor analytic model which estimates non-uniform GA×E correlation structure (Model 4) was found to be the most parsimonious model (p<0.05,

Table 2.11). However, near-zero estimates of residual variance limit interpretation of extended factor analysis solutions for this trait. Thus, a single model fitting heterogeneous GA×E variance with uniform correlation between environments and independent heterogeneous residual variance (Model 3) was selected, as it best fits all fruit chemistry traits.

Genome-by-environment correlations were estimated for each trait (Table 2.5).

Correlations were highest for SSC:TA (0.50, s.e. = 0.63) and lowest for phenolic content

(0.05, s.e. = 0.31). Narrow-sense heritability estimates for each trait-by-environment combination are summarized in Table 2.12. Heritability estimates for total anthocyanin content, TA, pH, and SSC:TA were all highest in OR. Heritability estimates for SSC were highest in OH. Heritability estimates for total phenolics content were low (≤0.05) in all environments.

Correlations between trait BVs are summarized in Figure 2.5. There was a high significant (p<0.01) negative correlation between TA and pH (r=-0.85) and between TA

50 and SSC:TA (r=-0.86); and there was a high positive correlation between pH and

SSC:TA (r=0.85).

Finlay-Wilkinson Regression

Finlay-Wilkinson regression was used to model and plot reaction norms for TA, pH, SSC, and SSC:TA (Figure 2.6). Markov chain convergence was observed for all traits except anthocyanin content and phenolics content; therefore, no reliable results may be reported for these two traits using this approach. Variance components estimated from posterior distributions MCMC chain simulations are summarized in

Table 2.13, and coefficients of variation are summarized in Table 2.14. There was a significant positive correlation between genetic effect and slope effect for each fruit chemistry trait except TA (r=0.15 to 0.56, p£0.0001 to 0.018, Table 2.9). There was also a positive correlation between BVs estimated by mixed model analysis and Finlay-

Wilkinson regression for each fruit size trait (r=0.82 to 0.97, p£0.0001, Table 2.10).

Discussion

In this study, variance components influencing ten fruit quality traits were estimated for eleven location-by-year trial environments. Two methods, linear mixed model analysis, and a Bayesian Finlay-Wilkinson regression were applied. Both methods followed confirmation of expected pedigree structure through SNP-driven relatedness analysis and incorporated a SNP-derived GA matrix to define additive genomic covariance between individuals.

In the following, fruit size and fruit chemistry traits are discussed in light of mixed model analysis and Finlay-Wilkinson regression results. Genotype-by-environment interaction is presented for each trait as (i) genome-by-environment correlations

51 estimated across location-by-year environments, (ii) heterogeneous variance components and narrow-sense heritability coefficients estimated by environment, and

(iii) heterogenous phenotypic stability of individuals across environments. Comparisons are drawn between current study findings and trait analyses of related crop species.

Relatedness analysis

Statistical analysis of ~1000 evenly-distributed genome-wide SNPs revealed 39 potential off-type individuals in ORUS 4304 (Figure 2.4). Skewed segregation of single- gene aphid resistance trait has been observed in ORUS 4304 in multiple studies

(Dossett and Finn 2010; Bushakra et al. 2018). A possible explanation for skewed segregation of this single-gene trait is the presence of off-type individuals in ORUS

4304.

Close relatedness to the maternal parent of the 38 ORUS 4304 off-types suggests that these individuals resulted from self-fertilization. This inference is supported by Mendelian segregation of a single-gene aphid resistance trait (Bushakra et al. 2018) within the putative self-fertilized progeny by a chi-squared test, c2 (2,

N=38)=0.04, p=0.85. Unlike some fruit crop species, black raspberry has been reported to produce self-compatible flowers (Jennings 1988), and no self-incompatibility mechanism in black raspberry has been described, further supporting the conclusion that progeny individuals with close resemblance to the maternal parent originated from self-pollination.

The five ORUS 4305 identified in this study as off-types match those reported by

Bushakra et al. (2015) as having incongruous SNP data, and they were removed from linkage map construction in the cited study.

52 Because inclusion of off-types may invalidate assumptions relating to identity by descent, all off-type individuals of undetermined origin were removed from GA matrix construction in the present study and hence from further analysis. However, the ORUS

4304 progeny individuals resulting from putative self-pollination were retained, as they share all possible alleles in common with the maternal parent.

A GA matrix was constructed using a 427 count subset of 242,391 possible

SNPs. This subset was defined by a series of filters selecting for markers that are biallelic, evenly distributed through the genome, and sequenced with high depth and low missing data among all individuals. Filtering criteria are outlined in Table 2.3 and were selected to isolate markers with high confidence and few missing data. Number of markers used is less than typical for genetic analysis. For example, 1,617 markers selected from a 6k SNP array were used to estimate variance components of sweet cherry tested in eight environments (Hardner et al. 2019). However, individuals examined in the current study include only full-siblings, half-siblings, and three parents; thus, few recombination events (recent or ancestral) are expected between individuals, and kinship may be described by fewer, high-quality molecular markers than for studies with complex population structures.

Fruit size traits

Fruit size and its components, including drupelet count and seed size, are important traits for fresh market raspberry production. Current limitations to BR fresh market production include small fruit which are perceived by consumers as very seedy

(Chad Finn, personal communication). Increased fruit size and reduced seediness are thus important goals to the fresh BR industry. Breeding progress for increased fruit size

53 and reduced seediness will rely on the identification and consolidation of genetic components contributing to these traits.

Given interest in improving BR fresh fruit production within several U.S. geographical regions and limited resources for BR breeding, efficient assessment of breeding germplasm among diverse production environments is paramount to successful genetic improvement of this crop.

Mixed model analysis

GEI for BR fruit traits are described here as genome-by-environment correlations and heritability coefficients arising from the estimation of heterogeneous genetic and residual variances. These estimates are informed by a SNP-derived GA matrix, which allowed fitting of heterogeneous variance structures although individuals were not replicated within trial locations (Hardner et al. 2019).

High genome-by-environment correlation values (0.56 to 0.81) were estimated for average fruit mass, average seed mass, drupelet count, and seed fraction, indicating limited genotype-by-environment interaction for these traits (Table 2.6). Drupelet count genome-by-environment effects were strongly correlated across environments (0.81), suggesting this trait is fairly stable across the trial environments and strongly influenced by genetic factors. Drupelet count also presented the highest narrow-sense heritability coefficients (0.17 to 0.70) of all ten studied traits (Table 2.7). In addition, drupelet count presented high narrow-sense heritability coefficients (>0.50) in all OH and OR trial years, suggesting strong influence by additive genetic factors within these locations.

Seed fraction heritability estimates were also highest in OR and OH (0.29 to 0.44), suggesting moderate control of additive genetic factors in these environments.

54 Heritability estimates for fruit mass, seed mass, and drupelet count were all lowest in NC_2014, suggesting one or more properties of this environment may dampen the influence of genetic factors identified in other trial environments. North Carolina raspberry production has been identified as particularly challenging due to high summer temperatures (Bradish et al. 2016a). During the observed initial bloom time in NC_2014

(May 1 through May 10), local daily high temperatures were particularly high, ranging from 19 to 32 °C (U.S. Dept. of Commerce 2019). Though a study measuring influences of high temperature on BR fruit quality is currently lacking, high temperature during flowering in has been associated with reduced stigma receptivity, pistil density, and drupelet set in cultivated blackberry, Rubus L. subgenus Rubus Watson (Stanton et al.

2007). Similarly, reduction in fruit size and yield has been observed in response to heat stress in cultivated strawberry, Fragaria × ananassa Duch. (Kadir et al. 2006).

Observed genotype-by-environment interaction for drupelet count and fruit size among all environment may have been exacerbated by heterogeneous field conditions, non-uniform susceptibility of heat stress among individuals, and non-uniform individual bloom time among individuals.

Interestingly, NY heritability estimates for seed fraction were extremely low (0.01-

0.03), suggesting one or more properties of this environment may dampen the influence of additive genetic factors identified in other trial environments. Presently, an environmental variable which may explain this low heritability is unknown.

Significant correlation of BVs for drupelet count with fruit mass, seed mass, and seed fraction suggests that these fruit size traits may share genetic components. High positive correlation between drupelet count and fruit mass (r=0.74) suggests that

55 drupelet count is an important component of fruit mass, which is a product of drupelet count and average drupelet mass. Significant positive correlation between drupelet count and fruit size has also been observed in red raspberry (Woznicki et al. 2016).

Fruit mass and drupelet count are each components of total yield and important qualities for fresh market production. High drupelet count has been associated with high yield in red raspberry cultivar development (Daubeny 1996), and selection for high drupelet count may lead to larger BR fruit. Large fruit has also been associated with high yield in red raspberry and is an important trait for fresh market breeders and producers (Sanford et al. 1985; Daubeny 1996).

Positive correlation between drupelet count and seed fraction suggests that increased drupelet count bears a cost of increased relative seed mass. This suggests a tradeoff to fresh market breeders and producers, as the fresh fruit market prefers large fruit with a small seed portion. Negative correlation between drupelet count and average seed mass suggests that increased drupelet production bears a cost of reduced seed size. This suggests a second tradeoff, more desirable than the first, as small seeds in high-drupelet fruit may be less recognizable to the consumer than large seeds.

Reduced seediness is an important goal in many fruit breeding programs, as evidenced by fresh market popularity of seedlessness in grapes (Karaagac et al. 2012;

Royo et al. 2018), citrus (Vardi et al. 2008), and melon. Seediness is a particular concern in BR, as available cultivars are often perceived as very seedy, particularly in comparison to red raspberry. Selection for reduced seed mass and reduced seed fraction may generate BR accessions with reduced seediness. However, studies of perceived seediness in Rubus are sparse.

56 Consumer panel testing in blackberry has shown significant correlation between perceived seediness and seed fraction (Sebesta et al. 2013), and similar trends may be observed for BR. Contrasting this, there is limited evidence that seed mass may be a poor predictor of perceived seediness. For example, if seed mass is a good predictor of perceived seediness, one may expect cultivated red raspberry fruits to produce smaller seeds than those of BR. The average seed size in cultivated red raspberry (Rubus idaeus) grown in Corvallis, OR has been reported as 1.7 mg (Hummer and Peacock

1994). The average black raspberry seed mass of OR fruit in the present study is 1.8 mg, very similar to that reported for red raspberry. Though direct statistical comparison of these two studies cannot be conclusive, these observations suggest that perceived seediness may be influenced by factors aside from seed mass.

Seediness in Rubus may be attributed to additional factors such as seed composition, and further study of seed physical traits (e.g. shape, density, adherence to flesh) and chemical composition (e.g. lignin content) may elucidate these trends.

Finlay-Wilkinson regression

Bayesian Finlay-Wilkinson estimation of variance components are summarized in

Table 2.7 and Table 2.8. Low slope variance was observed for drupelet count, indicating low genotype-by-environment interaction for this trait. This finding agrees with high genome-by-environment correlation estimated for drupelet count by the mixed model approach. High slope variance was observed for seed fraction, indicating high genotype-by-environment interaction for seed fraction. This finding agrees somewhat with lower genome-by-environment correlation estimated for seed fraction by the mixed model approach.

57 Finlay-Wilkinson regression plots reveal little about GEI trends in the four fruit size traits, though phenotypic variance of fruit size, seed size, and seed fraction appear to be slightly higher in high-performing environments than in low-performing environments. Genetic and slope effects correlate significantly (p<0.01) for all four fruit size traits, suggesting a tradeoff between high breeding values and stability across environments (Simmonds 1991).

In BR, likely breeding targets include selection for high fruit mass, low seed mass, and low seed fraction. Thus, selection of genotypes with high mass and high drupelet count may coincide with selection of unstable genotypes which perform especially well in high fruit mass and high drupelet count environments. On the other hand, selection for genotypes with low seed mass and low seed fraction may coincide with selection of stable genotypes which perform relatively similar in all environments.

Correlation between genetic effect and slope effect is a common observation in biological studies using Finlay-Wilkinson regression, and it has been posited that such correlations suggest an environment which lacks genotypic differences (Hardwick

1981).

Correlation was observed between BVs estimated by mixed model and Finlay-

Wilkinson approaches (r=0.93 to 0.98), suggesting these methods are complementary for the estimation of genetic effects of these complex fruit size traits.

Fruit chemistry traits

There is considerable interest in breeding for improved fruit flavor in horticultural crops (Klee 2010). SSC and TA are important components of flavor. In consumer testing studies, low SSC and high TA have been associated with high sourness and

58 disliking of red raspberry fruit, whereas high soluble solid content has been associated with sweetness and liking of red raspberry fruit (Shamaila et al. 1993; Villamor et al.

2013). Both SSC and TA are complex traits that vary between genetic accessions and production locations (Mazur et al. 2014b), as well as during fruit development (Graham and Brennan 2018) of red raspberry.

There is also interest in exploring BR phenolic content and composition. BR fruit are rich in anthocyanins and other phenolic compounds, the dietary benefits of which is an active area of research (Fang 2015; Amiot et al. 2016; Kresty et al. 2016).

Anthocyanin content also contributes to dark coloration of BR fruit. Appearance is an important component of flavor (Klee 2010), and darkly colored BR fruit are preferred over lightly colored fruit. Anthocyanins also function as a dye in processed BR products

(Dossett et al. 2010). High acidity and low pH promote anthocyanin stability, making them important traits for the raspberry processing market (Daubeny 1996).

Mixed model analysis

Heritability of anthocyanin content, TA, and pH were all highest in OR, suggesting OR is an ideal location to investigate genetic control these traits and make breeding selections. This is advantageous, as a majority of the BR processing industry is currently located in Oregon, and breeding efforts for improved processing BR cultivars are currently underway. Interestingly, anthocyanin content breeding values were found to correlate significantly (p<0.01) with those for TA and pH (Figure 2.5), suggesting high TA and low pH may promote anthocyanin accumulation within BR fruit.

It is also possible that high TA and low pH promote anthocyanin stability within fruit and/or during sample preparation, a notion which agrees with previous assessment that

59 high TA contributes to anthocyanin stability (Daubeny 1996). Further, significant correlation between TA and total anthocyanin content suggest that selection for high TA will coincide with selection for high anthocyanin content, if that is the desired objective.

Low genome-by-environment correlation was estimated for total phenolics content (0.05), suggesting this trait is unstable and highly influenced by GEI among the trial environments (Table 2.5). Low heritability values were estimated for phenolics content in all environments (0.01 to 0.05), suggesting limitations for rapid genetic improvement of phenolic content. In red raspberry, accession phenolic content has also been shown to interact significantly with harvest year (Mazur et al. 2014b). In contrast, red raspberry fruit phenolic content has also been observed as stable between harvest years (Dobson et al. 2012). Red raspberry phenolic content has also been shown to decrease in response to decreased temperature during fruit maturation (Remberg et al.

2010).

High GEI for phenolic content suggests that genetic and environmental components influencing this trait are extremely complex. Raspberry fruit contains a large diversity of phenolic compounds including anthocyanins, phenolic acids, flavanols, and tannins (Graham and Brennan 2018). In addition, the method used to measure phenolic content in the current study measures abundance of all antioxidant compounds, including but not limited to phenolics (Singleton et al. 1999).

Further study of phenolic groups may elucidate additive components of BR phenolic content. This is supported by a relatively high genome-by-environment correlation observed for anthocyanin content (0.40, Table 2.5) and higher heritability

60 observed in all locations (Table 2.12). In addition, correlation was observed between anthocyanin content BVs and phenolic content BVs (r=0.45, Figure 5).

Finlay-Wilkinson regression

Bayesian Finlay-Wilkinson estimation of variance components of TA, pH, SSC, and SSC:TA are summarized in Table 2.13 and Table 2.14. Just as genome-by- environment correlation was highest for SSC:TA (Table 2.5), slope variance was lowest for SSC:TA (Table 2.14), indicating SSC:TA is more stable across environments than

TA, pH, or SSC. Slope variance for TA and pH were very high (CVb=2168 to 3122), indicating low stability of these traits across the trial environments.

Finlay-Wilkinson regression of SSC:TA reveals large heterogeneity of individual slopes (Figure 2.6). Genetic and slope effects correlate significantly (p<0.05) for pH, and very significantly (p<0.0001) for SSC and SSC:TA, suggesting a tradeoff between high breeding values and stability across environments for these traits. Correlation between genetic effect and slope effect is a common observation in biological studies using Finlay-Wilkinson regression, and it has been posited that such correlations suggest an environment which lacks genotypic differences (Hardwick 1981).

Correlation was observed between BVs estimated by mixed model and Finlay-

Wilkinson approaches (r=0.82-0.97), suggesting these methods are complementary for the estimation of genetic effects of these complex fruit chemistry traits.

Conclusions

The objective this study was to quantify and characterize GEI for fruit size and fruit chemistry traits among diverse BR production environments. Approaches used include the fitting of mixed models with complex variance-covariance structures and a

61 Bayesian Finlay-Wilkinson regression model. Genotype-by-environment interactions were presented for each trait as (i) genome-by-environment correlations estimated across location-by-year environments, (ii) heterogeneous variance components and narrow-sense heritability coefficients estimated by environment, and (iii) heterogenous phenotypic stability of individuals across environments.

Notable findings include low GEI for drupelet count supported by high genome- by-environment correlation, high narrow-sense heritability, and high stability among trial location-by-year environments. This suggests that rapid improvement may be made for this drupelet count in BR. However, low heritability in one location-by-year environment

(NC_2014) suggests one or more characteristics of this environment may dampen genetic effects on drupelet count. Additional findings include relatively high narrow- sense heritability of fruit chemistry traits within OR and OH as well as high narrow- sense heritability of fruit size traits within OH. A significant correlation was observed between individual breeding values and stability effects for fruit mass, seed mass, drupelet count, seed fraction, pH, SSC, and SSC:TA.

High heritability estimated for TA, SSC:TA, and anthocyanin content in OR suggest that rapid improvements may be made for these traits through selective breeding. However, low heritability for phenolic content in all environments suggests limited potential for selective breeding of phenolic content in BR.

Though the presented work identified patterns in GEI, identification and characterization of causal factors will require further experimental study. Narrow-sense heritability estimates reported in this work may be used to group environments and identify covariates, such as temperature or water availability. Further, identification of

62 quantitative trait loci may elucidate trends in genetic contributions to phenotypic variability of each trait.

63 References

Amiot MJ, Riva C, Vinet A (2016) Effects of dietary polyphenols on metabolic syndrome features in humans: a systematic review. Obes Rev 17:573–586. https://www.doi.org/10.1111/obr.12409

Andrews S (2010) FastQC: a quality control tool for high throughput sequence data. Available online at: http://www.bioinformatics.babraham.ac.uk/projects/fastqc

Bolger AM, Lohse M, Usadel B (2014) Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30:2114–2120. https://www.doi.org/10.1093/bioinformatics/btu170

Bradish CM, Bushakra JM, Dossett M, et al (2016) Genotyping and phenotyping heat tolerance in black raspberry (Rubus occidentalis L.). Acta Hortic 1127:321–324. https://www.doi.org/10.17660/ActaHortic.2016.1127.50

Bushakra JM, Bryant DW, Dossett M, et al (2015) A genetic linkage map of black raspberry (Rubus occidentalis) and the mapping of Ag 4 conferring resistance to the aphid Amphorophora agathonica. Theor Appl Genet 128:1631–1646. https://www.doi.org/10.1007/s00122-015-2541-x

Bushakra JM, Dossett M, Carter KA, et al (2018) Characterization of aphid resistance loci in black raspberry (Rubus occidentalis L.). Mol Breed 38:83. https://www.doi.org/10.1007/s11032-018-0839-5

Cash CB (2016) Studies of linkage mapping, trait heritability, and pedigrees for breeding improvement of southeastern black and red raspberry. North Carolina State University

Danecek P, Auton A, Abecasis G, et al (2011) The variant call format and VCFtools. Bioinformatics 27:2156–2158. https://www.doi.org/10.1093/bioinformatics/btr330

Daubeny HA (1996) Brambles. In: Janick J, Moore JN (eds) Fruit Breeding, Vine and Small Fruits. Wiley, New York, pp 109–190

Dobson P, Graham J, Stewart D, et al (2012) Over-seasons analysis of quantitative trait loci affecting phenolic content and antioxidant capacity in raspberry. J Agric Food Chem 60:5360–5366. https://www.doi.org/10.1021/jf3005178

Dossett M, Bassil NV, Lewers KS, Finn CE (2012) Genetic diversity in wild and cultivated black raspberry (Rubus occidentalis L.) evaluated by simple sequence repeat markers. Genet Resour Crop Evol 59:1849–1865. https://www.doi.org/10.1007/s10722-012-9808-8

Dossett M, Finn CE (2010) Identification of resistance to the large raspberry aphid in black raspberry. J Am Soc Hortic Sci 135:438–444

64 Dossett M, Lee J, Finn CE (2010) Variation in anthocyanins and total phenolics of black raspberry populations. J Funct Foods 2:292–297. https://www.doi.org/10.1016/j.jff.2010.10.004

Elshire RJ, Glaubitz JC, Sun Q, et al (2011) A robust, simple genotyping-by-sequencing (GBS) approach for high diversity species. PLoS ONE 6:e19379. https://www.doi.org/10.1371/journal.pone.0019379

Fang J (2015) Classification of fruits based on anthocyanin types and relevance to their health effects. Nutrition 31:1301–1306. https://www.doi.org/10.1016/j.nut.2015.04.015

Finlay KW, Wilkinson GN (1963) The analysis of adaptation in a plant-breeding programme. Aust J Agric Res 14:742–754

Gilmour AR, Gogel B, Cullis BR, et al (2015) ASReml User Guide Release 4.1. VSN International Ltd, Hemel Hempstead

Glaubitz JC, Casstevens TM, Lu F, et al (2014) TASSEL-GBS: a high capacity genotyping by sequencing analysis pipeline. PLoS ONE 9:2. https://www.doi.org/10.1371/journal.pone.0090346

Graham J, Brennan R (eds) (2018) Raspberries: breeding, challenges, and advances. Springer Science+Business Media, New York, NY

Hardner C (2017) Exploring opportunities for reducing complexity of genotype-by- environment interaction models. Euphytica 213:248. https://www.doi.org/10.1007/s10681-017-2023-0

Hardner C, Winks C, Stephenson R, Gallagher E (2001) Genetic parameters for nut and kernel traits in macadamia. Euphytica 117:151–161

Hardner CM, Hayes BJ, Kumar S, et al (2019) Prediction of genetic value for sweet cherry fruit maturity among environments using a 6K SNP array. Hortic Res 6:6. https://www.doi.org/10.1038/s41438-018-0081-7

Hardwick RC (1981) The analysis of genotype × environment interactions: what does it mean if varietal stability is linearly related to varietal performance? Euphytica 30:217–221

Hummer KE, Peacock DN (1994) Seed dimension and weight of selected Rubus species. HortScience 29:1034–1036. https://www.doi.org/10.21273/HORTSCI.29.9.1034

Isik F, Holland J, Maltecca C (2017) Multienvironmental trials. In: Genetic data analysis for plant and animal breeding. Springer International Publishing

65 Jennings DL (1988) Raspberries and blackberries: their breeding, diseases and growth. Academic Press, San Diego, CA

Kadir S, Sidhu G, Al-Khatib K (2006) Strawberry (Fragaria × ananassa Duch.) growth and productivity as affected by temperature. HortScience 41:1423–1430. https://www.doi.org/10.21273/HORTSCI.41.6.1423

Karaagac E, Vargas A, Andrés M, et al (2012) Marker assisted selection for seedlessness in table grape breeding. Tree Genet Genomes 8:1003–1015. https://www.doi.org/10.1007/s11295-012-0480-0

Klee HJ (2010) Improving the flavor of fresh fruits: genomics, biochemistry, and biotechnology: Tansley review. New Phytol 187:44–56. https://www.doi.org/10.1111/j.1469-8137.2010.03281.x

Kresty LA, Mallery SR, Stoner GD (2016) Black raspberries in cancer clinical trials: past, present and future. J Berry Res 6:251–261. https://www.doi.org/10.3233/JBR- 160125

Lian L, de los Campos G (2016) FW: an R package for Finlay–Wilkinson regression that incorporates genomic/pedigree information and covariance structures between environments. G3 6:589–597. https://www.doi.org/10.1534/g3.115.026328

Lodhi MA, Ye G-N, Weeden NF, Reisch BI (1994) A simple and efficient method for DNA extraction from grapevine cultivars and Vitis species. Plant Mol Biol Report 12:6–13. https://www.doi.org/10.1007/BF02668658

Mazur SP, Nes A, Wold A-B, et al (2014) Quality and chemical composition of ten red raspberry (Rubus idaeus L.) genotypes during three harvest seasons. Food Chem 160:233–240. https://www.doi.org/10.1016/j.foodchem.2014.02.174

Muñoz PR, Resende MFR, Gezan SA, et al (2014) Unraveling additive from nonadditive effects using genomic relationship matrices. Genetics 198:1759–1768. https://www.doi.org/10.1534/genetics.114.171322

Nei M (1987) Genetic distance between populations. In: Molecular evolutionary genetics. Columbia University Press, New York, pp 208–253

Ourecky DK, Slate GL (1973) Jewel black raspberry. New York’s Food and Life Sci Bul 35

Perkins-Veazie P, Ma G, Fernandez GE, et al (2016) Black raspberry fruit composition over two years from seedling populations grown at four US geographic locations. Acta Hortic 1133:335–338. https://www.doi.org/10.17660/ActaHortic.2016.1133.52

Plummer M, Best N, Cowles K, Vines K (2006) CODA: convergence diagnosis and output analysis for MCMC. R News 6:7–11

66 Remberg SF, Sønsteby A, Aaby K, Heide OM (2010) Influence of postflowering temperature on fruit size and chemical composition of Glen Ample raspberry (Rubus idaeus L.). J Agric Food Chem 58:9120–9128. https://www.doi.org/10.1021/jf101736q

Royo C, Torres-Pérez R, Mauri N, et al (2018) The major origin of seedless grapes is associated with a missense mutation in the MADS-box gene VviAGL11. Plant Physiol 177:1234–1253. https://www.doi.org/10.1104/pp.18.00259

Sanford JC, Ourecky DK, Reich JE (1985) ‘Titan’ red raspberry. HortScience 20:1133– 1134

Sebesta B, Clark JR, Threlfall RT, Howard LR (2013) Characterization of seediness attributes of blackberry genotypes. Discovery, the student journal of Dale Bumpers College of Agricultural, Food and Life Sciences 14:72–79

Shamaila M, Skura B, Daubeny H, Anderson A (1993) Sensory, chemical and gas chromatographic evaluation of five raspberry cultivars. Food Res Int 26:443–449. https://www.doi.org/10.1016/0963-9969(93)90090-6

Simmonds NW (1991) Selection for local adaptation in a plant breeding programme. Theor Appl Genet 82:363–367. https://www.doi.org/10.1007/BF02190624

Singleton VL, Orthofer R, Lamuela-Raventos RM (1999) Analysis of total phenolics and other oxidation substrates and antioxidants by means of Folin-Ciocalteu reagent. Methods Enzymol 299:152–177

Smith AB, Cullis BR (2018) Plant breeding selection tools built on factor analytic mixed models for multi-environment trial data. Euphytica 214:143. https://www.doi.org/10.1007/s10681-018-2220-5

Smith AB, Cullis BR, Thompson R (2005) The analysis of crop cultivar breeding and evaluation trials: an overview of current mixed model approaches. J Agr Sci 143:449–462. https://www.doi.org/10.1017/S0021859605005587

Smith AB, Stringer JK, Wei X, Cullis BR (2007) Varietal selection for perennial crops where data relate to multiple harvests from a series of field trials. Euphytica 157:253–266. https://www.doi.org/10.1007/s10681-007-9418-2

Stanton MA, Scheerens JC, Funt RC, Clark JR (2007) Floral competence of primocane- fruiting blackberries Prime-Jan and Prime-Jim grown at three temperature regimens. HortScience 42:508–513. https://www.doi.org/10.21273/HORTSCI.42.3.508

Su G, Madsen P, Lund MS, et al (2006) Bayesian analysis of the linear reaction norm model with unknown covariates. J Anim Sci 84:1651–1657. https://www.doi.org/10.2527/jas.2005-517

67 U.S. Dept. of Agr. (2017) Fruit and tree nut yearbook tables. https://www.ers.usda.gov/data-products/fruit-and-tree-nut-data/fruit-and-tree-nut- yearbook-tables/

U.S. Dept. of Commerce (2019) Local climatological data dataset. https://www.ncdc.noaa.gov/cdo- web/datasets/GHCND/stations/GHCND:USC00314464/detail van Eeuwijk FA (1995) Linear and bilinear models for the analysis of multi-environment trials: I. An inventory of models. Euphytica 84:1–7. https://www.doi.org/10.1007/BF01677551

VanBuren R, Wai CM, Colle M, et al (2018) A near complete, chromosome-scale assembly of the black raspberry (Rubus occidentalis) genome. GigaScience 7:1– 9. https://www.doi.org/10.1093/gigascience/giy094

Vardi A, Levin I, Carmi N (2008) Induction of seedlessness in citrus: from classical techniques to emerging biotechnological approaches. J Am Soc Hortic Sci 133:117–126. https://www.doi.org/10.21273/JASHS.133.1.117

Villamor RR, Daniels CH, Moore PP, Ross CF (2013) Preference mapping of frozen and fresh raspberries. J Food Sci 78:S911–S919. https://www.doi.org/10.1111/1750- 3841.12125

Weber C (2015) Performance of new berry varieties at the NYSAES in Geneva, NY. New York Fruit Quarterly 23:5–9

Weber CA (2003) Genetic diversity in black raspberry detected by RAPD markers. HortScience 38:4

Wimmer V, Albrecht T, Auinger H-J, Schön C-C (2012) synbreed: a framework for the analysis of genomic prediction data using R. Bioinformatics 28:2086–2087. https://www.doi.org/10.1093/bioinformatics/bts335

Woznicki TL, Heide OM, Remberg SF, Sønsteby A (2016) Effects of controlled nutrient feeding and different temperatures during floral initiation on yield, berry size and drupelet numbers in red raspberry (Rubus idaeus L.). Sci Hortic 212:148–154. https://www.doi.org/10.1016/j.scienta.2016.09.047

Yang J, Benyamin B, McEvoy BP, et al (2010) Common SNPs explain a large proportion of the heritability for human height. Nat Genet 42:565–569. https://www.doi.org/10.1038/ng.608

68 Figures

Figure 2.1. Pedigree chart for black raspberry mapping populations ORUS 4304 and ORUS 4305. Maternal parentage is indicated by a red line. Paternal parentage is indicated by a blue line.

69 NY OH NC 5 0 OR 5 2 0

5 7 8 29

154

1 6

25 14

3

Figure 2.2. Venn diagram of ORUS 4304 and 4305 individuals evaluated in four trial locations: Jackson Springs, North Carolina (NC), Geneva, New York (NY), Wooster, Ohio (OH), and Corvallis, Oregon (OR). 154 individuals were evaluated in every location. Seven individuals were evaluated in OR only.

70 North Carolina New York NC_2014 NY_2015

15 7

6 41 185 141 16 4 NY_2013 8 NY_2014 12

NC_2013

Ohio Oregon OH_2015 OR_2015

0 18

2 9 19 75 209 110 0 0 3 6 OH_2013 2 OH_2014 OR_2013 7 OR_2014

Figure 2.3. Venn diagrams of ORUS 4304 and ORUS 4305 individuals sampled in three production seasons (2013, 2014, and 2015) at four trial locations: Jackson Springs, North Carolina (NC), Geneva, New York (NY), Wooster, Ohio (OH), and Corvallis, Oregon (OR).

71 4158_2 0.6 putative selfs true−type individuals additional off−type 0.4 0.2 4158 − 2 0.0 − 0.2 3021_2

−0.2 0.0 0.2 0.4 0.6

3021−2

Figure 2.4. Relatedness plot for ORUS 4304 parents and progeny with the maternal parent (ORUS 4158-2) on the y-axis and paternal parent (ORUS 3021-2) on the x-axis.

72 Seed_massDrupelet_countSeed_fraction pH SSC_per_TAAnthocyanin_contentPhenolics_content TA SSC 1 Fruit_mass 0.15 0.74 −0.01 0.03 −0.18−0.29 −0.2 −0.24−0.09 0.8 Seed_mass −0.32 0.07 −0.04−0.03 0.25 0.13 0.1 −0.29 0.6 Drupelet_count 0.4 0.09 −0.13−0.18−0.21−0.35−0.04 0.4

Seed_fraction 0.16 −0.05 0.34 −0.02−0.22−0.22 0.2

TA −0.85 0.1 −0.86 0.23 0.13 0

pH 0.1 0.85 −0.18−0.16 −0.2

SSC 0.26 0.13 −0.14 −0.4 −0.6 SSC_per_TA −0.15−0.21 −0.8 Anthocyanin_content 0.45 −1

Figure 2.5. Pearson correlation coefficient plot of breeding values for fruit size and chemistry traits. Significant correlations (p<0.01) are shaded in red (negative) or blue (positive).

73 Figure 2.6. Finlay-Wilkinson regression plots for eight black raspberry fruit quality traits measured in location-by-year environments. Individual performances are regressed over environmental effects. Genotype-by-environment interaction is indicated as heterogeneity of individual slope effects. a: average infructescence mass b: average seed mass c: average drupelet count per infructescence d: seed fraction e: titratable acidity f: pH g: soluble solid content h: soluble solid content : titratable acidity

74 a.

NC14 NC13 OH15 OR14 NY13

● ● ● ● 3021_2 ● ● 3 ● ● ● ● ● ● ● ● 4153_1 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 4158_2 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 2 ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● 4304_10 ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● 4304_100 ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● 1 ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● 4304_101 ●● ● ● ●● ● ● ● 4304_102

Infructescence Mass ● ● ● 4304_103 4304_104 0.0 0.2 0.4 0.6 4304_107 4304_109 NY14 OR15 OH14 NY15 OH13 4304_11 Environment effect 4304_110 b.

OR15 NC13 NC14 NY13 NY14

● 3021_2 ● ●

3 ● ● 4153_1 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 4158_2 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 2 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 4304_10 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 4304_100 ● ● ● ● ● Seed Mass ● 1 ● 4304_101 4304_102 ● 4304_103 4304_104 0.0 0.1 0.2 0.3 0.4 4304_107 4304_109 OH15 OH13 OH14 NY15 OR14 4304_11 Environment effect 4304_110 c.

NC13 OH15 OH14 NY13 NY15

● ● ● ● ● ● ● ●● ● ● ● ● ● ●● ● ● ● ● ●● ● ●● ● ● ● ● ●● ● ●● ● ● ● ●● ● ●● ● ount ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ● ●● ● c ● ● ● ● ●● ● ●● ● ● ● ● ● ●●●● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● Drupelet ● ● ● ● 20 60 100 140 ● ● 4304_104 −2 0 2 4304_107 4304_109 NC14 NY14 OH13 NY13 OR15 4304_11 Environment effect 4304_110

75 d.

NC13 OH14 OR15 NY14

● 3021_2

(%) ● ● 4153_1 4158_2 ● ● ● ● 4304_10 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 4304_100 ●● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● ● ● ●● ● ●● ● ● ● 4304_101 ● ● ●● ● ●● ● ● ● ● ● ●● ●

Seed Fraction ● ●● ● 4304_102 ● 0 10 20 30 4304_103 4304_104 −0.5 0.0 0.5 1.0 1.5 2.0 2.5 4304_107 4304_109 OH13 NY13 NC14 OR14 NY15 4304_11 Environment effect 4304_110 e.

NY15 OR15 NY14 OH14 NC13

● ● ● ● ● ● ● ● ● ● ● ● 3021_2 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 4153_1 ● ● ● ● ● ● ● ● 1.5 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 4158_2 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 4304_10 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 1.0 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 4304_100 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 4304_101 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● Titratable Acidity Titratable ● ● ● ● ● ● ● ● ● ● ● ● 4304_102 0.5 ● ● ● ● 4304_103 4304_104 −0.2 −0.1 0.0 0.1 4304_107 4304_109 OR13 OR14 NY13 OH15 NC14 OH13 4304_11 Environment effect 4304_110 f.

OH14 NY15 NY13 NC14 OR14

● ● ● ● ● ● 3021_2 ● ● ● ● ● ● ● ● ● ● ● ●

4.5 ● ● ● ● ● ● ● ● ● 4153_1 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● 4158_2 ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● 4.0 ●● ● ● ● ● ● ● ● ● ● 4304_10 pH ●● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● 4304_100 ●● ● ● ● ● ● ● ●● ● ● ● ● ● ● 3.5 ●● ● ● ● ● ●● ● ● ● ● 4304_101 ● ● ● ● ● ● ● ● ● 4304_102 ● 3.0 4304_103 4304_104 −0.1 0.0 0.1 0.2 0.3 4304_107 4304_109 OH13 NC13 OH15 OR15 NY14 OR13 4304_11 Environment effect 4304_110

76 g.

OR13 NY15 OH15 OR14 NY14

●

● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ●●●●●● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ●●●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ● ● Soluble solid content Soluble ● ● 5 10 15 20 4304_104 −1.0− 0.5 0.0 0.5 1.0 1.5 2.0 4304_107 4304_109 NC13 NY13 OH13 OH14 NC14 OR15 4304_11 Environment effect 4304_110 h.

OH13 OH15 OR13 NC14 NY14

●

● ● ●● ● ●● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ● ●● ● ● ● ●● ●● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ●● ●● ● 10 20 30 40● 50 ● ● ● ● ● ●● ●● ● ● ● ● ● ●● ●● ● ● ● ● ● 4304_104 0 1 2 3 4304_107

Soluble solid content : titratable acidity solid content : titratable Soluble 4304_109 NC13 NY13OH14 NY15 OR14 OR15 4304_11 Environment effect 4304_110

77 Tables

Table 2.1. SNP filtering criteria and counts for ORUS 4304 relationship analysis. Minimum minor allele frequency 0.01 Minimum sequencing depth (per individual) 6 Maximum missing allele frequency 0.80 Thinning interval (bp) 100,000 Number of individuals (progeny + parents) 178 Number of sites post-filtering 992

Table 2.2. SNP filtering criteria and counts for ORUS 4305 relationship analysis. Minimum minor allele frequency 0.01 Minimum sequencing depth (per individual) 6 Maximum missing allele frequency 0.80 Thinning interval (bp) 100,000 Number of individuals (progeny + parents) 115 Number of sites post-filtering 1763

Table 2.3. SNP filtering criteria and counts for additive genomic relationship matrix estimation. Minimum minor allele frequency 0.05 Minimum sequencing depth (per individual) 7 Maximum missing allele frequency 0.90 Thinning interval (bp) 100,000 Number of individuals (progeny + parents) 285 Number of sites post filtering 427

78 Table 2.4. Mixed model fit comparison for black raspberry fruit size traits with fixed environmental effects and random additive genomic-by-environment effects (GA×E).GA×E variance structures were fit as homogeneous with uniform correlation, heterogeneous with uniform correlation, or heterogeneous with non-uniform correlation estimated by factor analysis. Residual variance structures were fit as independent with homogeneous variance or independent with heterogeneous variance. Each model was tested against the previous model by a likelihood ratio test. # of Log Trait Modela parameters p-value Likelihood fit 1 3 -5608.58 Drupelet 2 12 -5543.82 7.44E-24 *** count 3 21 -5526.54 3.56E-05 *** 4 30 -5519.69 0.067 1 3 1079.19 2 12 1145.7 1.42E-24 *** Fruit mass 3 21 1172.41 1.22E-08 *** 4 30 1179.93 0.045 * 1 3 1855.79 2 12 1964.25 4.65E-42 *** Seed mass 3 21 1988.56 9.78E-08 *** 4 30 1995.14 0.078 1 3 6653.61 Seed 2 12 7040.65 4.03E-161 *** fraction 3 21 7054.96 3.75E-04 *** 4 30 7049.82 na ***p<0.001; **p<0.01; *p<0.05 a Model 1: homogeneous var(GA×E), homogeneous var(R) Model 2: homogeneous var(GA×E), heterogeneous var(R) Model 3: heterogeneous var(GA×E), heterogeneous var(R) Model 4: factor analytic var-cov(GA×E), heterogeneous var(R)

79 Table 2.5. Genome-by-environment correlation estimates (�) and standard errors (SE) for black raspberry fruit quality traits tested in four locations (NC, NY, OH, OR) and three years (2013, 2014, 2015). Trait � SE � / SE p-value Fruit mass 0.61 0.069 8.81 1.7×10-17 Seed mass 0.56 0.057 9.82 3.3×10-21 Drupelet count 0.81 0.049 16.43 1.3×10-54 Seed fraction 0.59 0.093 6.33 6.2×10-10 Titratable acidity (TA) 0.47 0.048 9.83 3.0×10-21 pH 0.37 0.057 6.48 2.5×10-10 Soluble solid content (SSC) 0.27 0.066 4.09 5.1×10-5 SSC:TA 0.50 0.063 7.97 1.1×10-14 Total anthocyanin content 0.40 0.070 5.68 2.5×10-8 Total phenolics content 0.05 0.031 1.16 0.011

Table 2.6. Heritability (ℎ ) estimates for fruit size traits at 10 location-by-year environments derived from the best fitting model. Maximum and minimum values are in bold. Heritability Fruit Seed Drupelet Seed mass mass count fraction NC 2013 0.38 0.40 0.49 0.17 NC 2014 0.17 0.25 0.17 0.29 NY 2013 0.39 0.37 0.66 0.03 NY 2014 0.28 0.40 0.40 0.01 NY 2015 0.31 0.33 0.54 0.05 OH 2013 0.51 0.50 0.56 0.31 OH 2014 0.53 0.38 0.70 0.44 OH 2015 0.41 0.42 0.64 0.27 OR 2014 0.38 0.31 0.59 0.29 OR 2015 0.31 0.45 0.58 0.43

80 Table 2.7. Variance components for fruit size traits estimated by Bayesian Finlay- Wilkinson regression. SD=standard deviation; μ=overall mean; � =genetic variance; � =slope variance; � =environmental variance; � =residual variance

Seed mass Fruit mass (g) (mg) Drupelet count Seed fraction (%) Mean (SD) Mean (SD) Mean (SD) Mean (SD) μ 1.29 (0.11) 1.79 (0.07) 63.12 (4.23) 7.44 (0.22) � �� 0.04 (0.01) 0.02 (0.00) 145.94 (15.90) 0.31 (0.05) � �� 0.17 (0.05) 0.14 (0.04) 5.71 (2.47) 0.83 (0.33) � �� 0.25 (0.12) 0.12 (0.05) 8.90 (4.19) 1.33 (0.64) � �� 0.08 (0.00) 0.04 (0.00) 117.68 (4.46) 1.58 (0.06)

Table 2.8. Punctual coefficients of variation (CV %) for fruit size traits derived from Bayesian Finlay-Wilkinson regression mean estimates. g=genetic; b=slope; h=environmental; ε=residual

Fruit mass Seed mass Seed fraction (g) (mg) Drupelet count (%)

CVg 15.5 7.9 19.1 7.5

CVb 32.0 20.9 3.8 12.3

CVh 38.8 19.4 4.7 15.5

CVε 21.9 11.2 17.2 16.9

Table 2.9. Pearson correlation of individual genetic (g) and slope (b) effects estimated by Bayesian Finlay-Wilkinson regression. r = Pearson correlation coefficient; p = p-value

Seed Drupelet Seed Fruit mass TA pH SSC SSC:TA mass count fraction r 0.57 0.38 0.18 0.19 -0.04 0.15 0.32 0.56

p <0.001 <0.001 0.006 0.003 0.57 0.018 <0.001 <0.001

Table 2.10. Pearson correlation of breeding values estimated by mixed model and Bayesian Finlay-Wilkinson regression. Fruit Seed Drupelet Seed TA pH SSC SSC:TA mass mass count fraction 0.93 0.96 0.98 0.95 0.97 0.97 0.97 0.82

81 Table 2.11. Mixed model comparison for black raspberry fruit size traits with fixed environmental effects and random additive genomic-by-environment effects (GA×E). GA×E variance structures were fit as homogeneous with uniform correlation, heterogeneous with uniform correlation, or heterogeneous with non-uniform correlation estimated by factor analysis. Residual variance structures were fit as independent with homogeneous variance or independent with heterogeneous variance. Each model was tested against the previous model by a likelihood ratio test. # of Log Trait Modela parameters p-value Likelihood fit

1 3 2138.65 4.84E-24 Titratable 2 13 2205.29 5.00E-24 *** acidity (TA) 3 23 2224.66 1.41E-05 *** 4 33 2228.85 2.96E-01 1 3 2039.77 4.01E-29 2 13 2074.51 2.79E-11 *** pH 3 23 2083.24 3.24E-02 * 4 33 2088.9 0.167 1 3 -1773.64 4.87E-14 Soluble solid 2 13 -1537.28 1.48E-95 *** content (SSC) 3 23 -1506.57 9.75E-10 *** 4 33 -1501.43 2.08E-01 1 3 -3320.73 2.29E-22 2 13 -3087 1.97E-94 *** SSC/TA 3 23 -3045.35 5.64E-14 *** 4 33 -3039.68 0.166 1 3 -3508.57 1.58E-13 Anthocyanin 2 13 -3412.69 4.21E-36 *** content 3 23 -3395.4 7.36E-05 *** 4 33 -3387.57 0.055 1 3 -2743.7 3.93E-41 Phenolics 2 13 -2486.44 1.74E-104 *** content 3 23 -2462.8 4.20E-07 *** 4 33 -2430.16 1.79E-10 *** ***p<0.001; **p<0.01; *p<0.05 a Model 1: homogeneous var(GA×E), homogeneous var(R) Model 2: homogeneous var(GA×E), heterogeneous var(R) Model 3: heterogeneous var(GA×E), heterogeneous var(R) Model 4: factor analytic var-cov(GA×E), heterogeneous var(R)

82 Table 2.12. Heritability (ℎ ) estimates for fruit chemistry traits at 11 location-by-year environments derived from the best fitting model. Maximum and minimum values are in bold. TA = titratable acidity, SSC = soluble solid content. Heritability Anthocyanin Phenolics TA pH SSC SSC:TA Content Content NC 2013 0.15 0.26 0.02 0.12 0.22 0.02 NC 2014 0.14 0.17 0.09 0.16 0.18 0.03 NY 2013 0.40 0.23 0.16 0.25 0.11 0.01 NY 2014 0.33 0.27 0.10 0.41 0.16 0.05 NY 2015 0.20 0.19 0.14 0.06 0.05 0.02 OH 2013 0.38 0.23 0.22 0.40 0.30 0.02 OH 2014 0.34 0.19 0.17 0.32 0.15 0.04 OH 2015 0.22 0.13 0.15 0.27 0.23 0.01 OR 2013 0.33 0.22 0.02 0.39 0.09 0.01 OR 2014 0.44 0.30 0.10 0.42 0.33 0.02 OR 2015 0.47 0.19 0.05 0.37 0.15 0.03

Table 2.13. Variance components for fruit chemistry traits estimated by Markov chain Monte Carlo simulation. SD=standard deviation; μ=overall mean; � =genetic variance; � =slope variance; � =environmental variance; � =residual variance TA pH SSC SSC:TA Mean (SD) Mean (SD) Mean (SD) Mean (SD) μ 1.031 (0.067) 3.832 (0.057) 11.611 (0.302) 10.253 (0.597) � �� 0.014 (0.002) 0.010 (0.002) 0.355 (0.054) 1.782 (0.347) � �� 0.094 (0.031) 0.105 (0.031) 0.357 (0.107) 1.406 (0.556) � �� 0.076 (0.031) 0.085 (0.035) 1.491 (0.654) 3.856 (1.927) � �� 0.032 (0.001) 0.037 (0.001) 2.070 (0.074) 9.871 (0.357)

Table 2.14. Punctual coefficients of variation (CV %) for fruit chemistry traits derived from Markov chain Monte Carlo simulation mean estimates. TA=titratable acidity (% citric acid); SSC=soluble solid content (°Brix); g=genetic; b=slope; h=environmental; ε=residual TA pH SSC SSC:TA

CVg 11.5 2.6 5.1 13.0

CVb 29.7 8.5 5.1 11.6

CVh 26.7 7.6 10.5 19.2

CVε 17.4 5.0 12.4 30.6

83 Table 2.15. Monthly normal maximum and minimum temperatures at research locations [Ohio State University (Ohio Agricultural Research and Development Center, Wooster, Wayne County); Oregon State University (Lewis-Brown Farm, Corvallis, Benton County); Cornell University (New York State Agricultural Experiment Station, Geneva, Seneca County); and North Carolina State University (Sandhills Research Station, Jackson Springs, Montgomery County)] testing adaptability of black raspberry (1981- 2010, NOAA). Daily Temperature (°F) Corvallis, OR Wooster, OH Jackson Springs, NC Geneva, NY High Low High Low High Low High Low Jan 47.1 34.2 34.5 18.2 50.2 30.3 30.9 16 Feb 50.9 34.7 38.1 19.9 54.3 32.9 33.2 17.5 Mar 56.1 37.3 48 27.3 62.3 38.2 41.3 24.8 Apr 60.8 39.6 61 37.5 71.7 46.3 54.9 36.2 May 67.2 44.1 70.9 47.3 78.9 55.2 66.9 46.5 Jun 73.2 48.5 79.5 56.5 86.2 64.3 75.9 56.4 Jul 81.7 51.8 83.1 60.2 89 68 80 61 Aug 82.6 51.1 81.7 58.4 87.4 67 78.4 59.3 Sep 77 47.7 74.6 50.8 81.6 60.4 71.1 51.7 Oct 64.7 41.7 62.7 40.1 71.8 49 59 41 Nov 52.6 37.9 50.7 32.1 62.8 41.5 47.4 32.5 Dec 45.8 33.4 38.4 22.8 53 33.3 35.9 22.5 Summary 63.3 41.8 60.3 39.3 70.8 48.9 56.2 38.8

84 Chapter 3. Quantitative trait loci analysis of a black raspberry multiple environment trial

Introduction

Black raspberry (Rubus occidentalis L., BR) is a North American native member of the Rosaceae family. In United States agriculture, BR is a minor, high-value crop and a source of germplasm for red raspberry breeding. Interest in developing improved BR varieties has followed discoveries of potential health benefits associated with its fruit

(Amiot et al. 2016; Kresty et al. 2016). However, BR germplasm development has lagged behind that of other Rubus crops such as red raspberry (R. idaeus L.) and blackberry (Rubus spp.). Despite perceived market demand for BR fruit, production is severely limited by insufficient elite germplasm.

Production of BR is limited by failure to meet producer and consumer demands, which include large fruit with low perceptible seediness in fresh fruit markets and dark- colored fruit in both processing and fresh fruit markets (Chad Finn, USDA BR breeder, personal communication). Production of BR is also limited directly by various diseases including Verticillium wilt (Bushakra et al. 2016a) and aphid-transmitted viruses (Halgren et al. 2007; Martin et al. 2013). Successful domestication and breeding of BR will require discovery and consolidation of genetic components controlling multiple traits including fruit quality traits, disease resistance/tolerance, and abiotic stress tolerance.

Target fruit quality traits for fresh market BR production include increased fruit size, reduced seed size, and reduced seed fraction. Targets for both fresh market and processing market BR include fruit with high sugar content. Targets for processing market production include high acidity, which is associated with anthocyanin stability.

Additionally, anthocyanin and phenolic contents are of interest in BR for their implicated 85 effects on human health. Genetic interrogation of these traits will enable domestication and breeding of BR via marker assisted selection as well as candidate gene identification. Candidate gene identification may be further assisted due to synteny observed between BR and other Rosaceous crop species (Bushakra et al. 2012).

Genetic analysis of BR may also leverage recently available tools including a high- quality chromosome-scale BR genome assembly (VanBuren et al. 2018).

To pursue the domestication and breeding of BR, information on quantitative trait loci (QTL) is needed. QTL mapping involves the statistical association of complex traits measured in a population with finite genetic regions defined by molecular markers such as single nucleotide polymorphisms (SNPs) or simple sequence repeats (SSRs). These associations may then be used to characterize genetic trait components, identify candidate causal genes, and enable marker assisted selection (Collard et al. 2005).

Though associations may be measured without a linkage map, use of a genetic linkage map provides improved resolution as well as increased power to identify and differentiate between QTL. Linkage mapping also enables identification of linked markers which may be used in marker assisted selection. Construction of a linkage map requires that the measured individuals have segregating markers and a known population structure. The first reported BR linkage map was constructed using high- resolution melting (HRM) markers observed in progeny of an F1 R. occidentalis × R. idaeus cross (Bushakra et al. 2012). Bushakra et al. (2012) used this map to define synteny between Rubus and other Rosaceous crop genera: Fragaria (strawberry),

Malus (apple), and Prunus (almond, cherry, peach, plum, et al.). Multiple linkage maps have been constructed using HRM, simple sequence repeat (SSR), and single

86 nucleotide polymorphism (SNP) markers observed in F1 populations of BR (R. occidentalis × R. occidentalis) in order to characterize a locus associated with aphid resistance (Bushakra et al. 2015, 2018).

Linkage association studies generally utilize an F2 or backcross population with parents homozygous (or assumed to be homozygous) at all genetic loci. Such populations have simple structure with known linkage phases, due to homozygous parents. However, homozygous parents are sometimes impossible to obtain due to practical or biological constraints including time, self-incompatibility (Silva and Goring

2001), and severe inbreeding depression (Voillemot and Pannell 2017). In such populations where both parents are heterozygous and without homozygous founders, up to four marker alleles may be present at any single locus, and marker linkage phases are unknown a priori. Segregating markers in these populations may be mapped using a two-way pseudo-testcross strategy, where recombination frequencies are estimated separately for each parent (Maliepaard et al. 1997). Such populations are commonly designated as cross pollinator (CP) populations and can be described by five co- dominant marker segregation types, lm×ll, nn×np, hk×hk ef×eg, and ab×cd (Van Ooijen and Jansen 2013). Segregation types lm×ll and nn×np describe bi-allelic markers that are segregating in one parent and fixed in the other. Segregation type hk×hk describes a bi-allelic marker that segregating in both parents. Segregation types ef×eg and ab×cd describe tri-allelic and tetra-allelic markers, respectively, that are segregating in both parents. Expected segregation patterns are 1:1 for lm×ll and nn×np markers, 1:2:1 for hk×hk markers, and 1:1:1:1 for ef×eg and ab×cd markers.

87 Trait-linkage association is simplest for observations made in a single environment and may use a standard analysis of variance (ANOVA); however, verification and selection for a trait-linked marker require observations made in more than one location and/or production season.

A multiple-environment trial (MET) is a common experiment used to evaluate and predict genotype performances over multiple locations, years, and/or other environmental variables. METs enable evaluation of genetic effects, environmental effects, and genotype-by-environment interactions (GEI). METs also enable the identification of QTL-by-environment interactions (QEI), which may be used to define stability of QTL in heterogeneous environments and, in some cases, explore environmental covariates which influence QTL expression. Loci that exhibit high QEI are considered unstable in their contribution to the associated trait. Likewise, loci that exhibit low QEI may be considered stable within the trial conditions. Breeding programs can benefit from characterization of QEI by prioritizing loci that are stable across target environments. Special attention should be given to cross-over QEI, in which a given allele has contrasting effects within separate environments. If the contrasting effects are large, it may be appropriate to group environments based on QEI. A breeder may then prioritize individual environmental groups and/or make separate selections for separate groups.

Increased individual observations recorded in a MET, due to replication of genotypes across environments, increases statistical power to identify significant QTL as compared to a single environment trial. However, when analyzing observations made of identical individuals across MET environments, a standard two-way ANOVA ignores

88 genetic correlation and may be too liberal when the assumption of sphericity is not met

(Piepho 2005). In this case, sphericity is the assumption of equal variances of differences between all pairs of observations for a common genotype. Repeated- measures multivariate ANOVA can accommodate observations that violate the assumption of sphericity; however, this approach requires complete data for each genotype-by-environment combination (Piepho 2005). On the other hand, mixed model analysis with appropriately specified variance-covariance structures enables modeling of genetic correlation in a framework which is flexible to missing data (Piepho 2005).

Mixed model QEI studies have been reported for several agronomic crops, including maize (Boer et al. 2007; Malosetti et al. 2008), wheat (Mathews et al. 2008), and sugarcane (Pastina et al. 2012; Margarido et al. 2015); however, QEI analyses of

BR and fruit crops in general are lacking. Such analyses will enable a more robust dissection of complex traits than standard, single-environmental analyses alone.

Further, QEI analysis of fruit quality traits in BR will inform breeding decisions through the discovery of QTL which are stable across heterogeneous production environments and the identification of trait-linked markers which may contribute to marker assisted selection.

To this end, two BR mapping populations were previously constructed using breeding germplasm, and phenotypic data for these populations were collected over three years in four research locations representative of U.S. BR production. The objective of the current study is to identify genetic loci associated with fruit size traits

[fruit (infructescence) mass (FrM), seed mass (SdM), drupelet count (DCt), seed fraction

(SdF)] and fruit chemistry traits [titratable acidity (TAc), soluble solid content (SSC), total

89 anthocyanin content (AnC), and total phenolics content (PhC)] observed in an MET spanning four locations [North Carolina (NC), New York (NY), Ohio (OH), and Oregon

(OR)] and three years (2013, 2014, and 2015) through QTL analysis.

Materials and Methods

Germplasm

Two mapping populations, ORUS 4304 and ORUS 4305, were constructed through controlled pollination in Corvallis, Oregon. ORUS 4304 and ORUS 4305 are composed of 192 and 115 individuals, respectively. Founders for both populations included the fresh market cultivar, ‘Jewel’ (Ourecky and Slate 1973), and the wild accession, NC 84-10-3. Wild accessions ORUS 3817-2 and ORUS 3778-1 are founders for ORUS 4304 and ORUS 4305, respectively (Figure 2.1).

Phenotypic data

ORUS 4304 and ORUS 4304 progeny, parents, and common ancestor, ‘Jewel’, were planted at four U.S. locations [Corvallis, Oregon (OR); Wooster, Ohio (OH);

Jackson Springs, North Carolina (NC); and Geneva, New York (NY)] in the spring of

2012. Fertilizer, pesticides, and irrigation were applied according to local recommendations. In all locations, first-year canes were tipped (i.e. cut to ~1m in length) each summer to promote lateral branch growth.

Ripe fruit were collected at trial locations in the summers of 2013, 2014, and

2015 in a total of 11 location-by-year environments. Collected fruit were frozen and sent to Kannapolis, North Carolina, where they were analyzed for chemical and physical fruit quality traits: titratable acidity (% citric acid), soluble solid content (°Brix), total anthocyanin content (mg/kg cyanidin 3-glucoside equivalents), total phenolics content

90 (mg/kg gallic acid equivalents), average fruit (infructescence) mass, and average seed mass (Perkins-Veazie et al. 2016). Drupelet count was calculated as an average number of seeds per fruit. Seed fraction was calculated as an average percent seed mass per fruit mass. Most individuals were replicated across locations; however, progeny individuals were not replicated across trial locations. Further, phenotypic data collected within locations were incomplete across trial years. Fruit and seed size traits were not recorded in OR_2013. Due to poor plant performance, no fruit were collected in NC_2015.

Phenotypic data were checked to meet assumptions of normal distribution for each trait-by-environment combination by visual examination of histograms. Phenotypic data were checked for homoscedasticity for each trait among environments by evaluation of residuals in a multivariate fixed effect ANOVA.

Genotyping

DNA extraction, preparation, and genotyping-by-sequencing (GBS) data were collected following Elshire et al. (2011) in three separate batches and seven sequencing runs. Four GBS runs were completed in Oregon, two in North Carolina, and one in Ohio.

Initial GBS data were collected in Oregon and North Carolina as described by Bushakra et al. (2015) and Cash (2016), respectively. Preliminary analysis of these data revealed low sequence depth for several individuals, and additional sequence data were collected in Ohio via a single GBS run with available individuals. In Ohio, leaf samples were collected, lyophilized, and held in dry storage before DNA extraction. Leaf tissue was ground in liquid nitrogen by mortar and pestle, and DNA was extracted using a modified CTAB protocol (Lodhi et al. 1994). Following RNase treatment, DNA was

91 further purified using the Genomic DNA Clean & Concentrator-10 kit (Zymo Research,

Irvine, CA). Genomic DNA concentration and quality were initially assessed by a

NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE) and subsequently assessed using a Qubit 4 fluorometer (Thermo Fisher, Waltham, MA) and by gel electrophoresis.

In Ohio, genomic DNA samples were submitted to The Ohio State University

Molecular and Cellular Imaging Center (Wooster, OH, USA) for GBS library construction and sequencing. GBS libraries were constructed following Elshire et al. (2011). Briefly, genomic DNA samples were digested with ApeKI and ligated to adapters compatible with ApeKI sticky ends and containing unique barcodes with partial Illumina® adaptor sequences. Following PCR amplification using Illumina® Nextera adapters (Illumina,

Inc.), libraries were pooled and quantified using a Qubit® fluorometer (Thermo Fisher,

Waltham, MA, USA), and size distribution and concentration was confirmed using the

4200 TapeStation (Agilent Technologies, Santa Clara, CA, USA). The pooled library was analyzed by single-read sequencing of 151 cycles on the NextSeq (Illumina, Inc.) instrument at the HudsonAlpha Institute for Biotechnology (Huntsville, AL, USA).

The seven total GBS runs representing 305 individuals were analyzed simultaneously. Data were initially subjected to quality control and adapter removal using FastQC v. 0.11.7 (Andrews 2010) and Trimmomatic v. 0.36 (Bolger et al. 2014).

Single nucleotide polymorphisms (SNPs) were identified using the TASSEL 5 GBS pipeline (Glaubitz et al. 2014) version 2 (https://www.maizegenetics.net/tassel), and the

BR reference genome sequence (VanBuren et al. 2018). Alignment of sequence data to the reference genome was performed using BWA v. 0.7.17 (Li and Durbin 2009).

92 SNP-guided relationship analyses were conducted separately for populations

ORUS 4304 and ORUS 4305 using VCFtools v. 1.16 (Danecek et al. 2011) and the method described by Yang et al. (2010). Estimated relationship statistics were then checked against pedigree population structure to confirm relatedness of population progeny.

For linkage map construction and QTL analysis, SNPs were filtered separately for ORUS 4304 and ORUS 4305 using VCFtools v. 1.16 (Danecek et al. 2011). Prior to

SNP filtering, 14 individuals (2 from ORUS 4305 and 12 from ORUS 4304) were removed due to low marker depth. Simple sequence repeat (SSR) data used for linkage map construction and QTL analysis were those collected by Bushakra et al. (2015) and

Cash (2016). Briefly, SSR primer pairs were collected from multiple sources, DNA amplification was performed using fluorescent labeling polymerase chain reaction

(PCR), and fragment lengths were determined using a capillary genetic analyzer

(Bushakra et al. 2015; Cash 2016).

Linkage map construction

All markers were recoded to match outbreeding segregation types (lm×ll, nn×np, hk×hk, ef×eg, and ab×cd) as described by Van Ooijen and Jansen (2013). Expected segregation patterns are 1:1 for lm×ll and nn×np markers, 1:2:1 for hk×hk markers, and

1:1:1:1 for ef×eg and ab×cd markers. A Chi-square test was applied to each marker to test for segregation distortion, and markers with significant deviations from expected segregation patterns (p≤0.05) were discarded.

Linkage maps were constructed using JoinMap v. 4.1 (Van Ooijen 2006). Missing genotypic data can result in erroneous marker placement; thus, individuals with ≥30%

93 missing marker data were excluded from map construction and subsequent analysis.

Due to a low number of conserved markers and low density of markers per linkage group in ORUS 4304, maps were only constructed for ORUS 4305. Mapped individuals were selected based on availability of phenotypic data for several traits, including but not limited to those described in the present study.

An Independence Likelihood of Odds (LOD) threshold of 10 was used to establish linkage groups (LG). Loci within each LG were then subjected to the maximum likelihood (ML) mapping algorithm. A nearest neighbor stress parameter (N.N. stress) was used to identify misplaced loci, and an N.N. stress threshold of 5 cM was used to remove unreliable markers. Remaining markers were again subjected to the ML algorithm, and this cycle was repeated until all markers fell below the N.N. stress threshold immediately following ML mapping. Seven maps were constructed representing the seven chromosomes of the BR genome.

QTL mapping

Multiple-environment linkage analysis was performed for each trait using GenStat v. 19.1 routines: QMQTLSCAN, QMESTIMATE, and QMBACKSELECT (VSN

International 2018). The ML algorithm used in linkage map construction assumes independent recombination frequency for each marker; thus the Haldane mapping function was selected during genetic predictor calculation in Genstat®. A heterogeneous compound symmetry variance-covariance structure (CShet) was used to model genetic (co)variance. The CShet structure specifies heterogeneous variance and uniform covariance between environments, as was determined most appropriate by

® previous analysis of GEI. Selection of the CShet structure was confirmed in Genstat by

94 using the Schwarz Information Criterion to select the most parsimonious model.

Candidate QTL were first identified by simple interval mapping (SIM) by fitting the model:

� = � + � + � � + � � + � � + � + � Equation 6 where

� is the trait mean for genotype i in environment j

� is the overall mean

� is the environment j main effect

� is the additive QTL effect for environment j at the position being tested

� is the additive genetic predictor for genotype i at the position being tested

� is the additive QTL effect of the second parent for environment j at the position

being tested

� is the additive genetic predictor of the second parent for genotype i at the

position being tested

� is the dominance QTL effect for environment j at the position being tested

� is the dominance genetic predictor for genotype i at the position being tested

� is the genetic residual for genotype i in environment j

� is the unit error for genotype i in environment j.

Significance thresholds were determined by the Genstat® default method of Li and Ji (2005). Candidate QTL identified by SIM were fit as cofactors by composite interval mapping (CIM). Cofactors falling below the significance threshold were removed, new candidate QTL were selected, and CIM was repeated until candidate

QTL selection did not change. Candidate QTL identified by CIM were then fit to a final

95 multiple-environment model by backward selection to estimate additive and dominance effects.

As indicated in Equation 6, additive effects were estimated for each parent separately. In a CP population, each locus comprises a maximum of four possible alleles, each contributed by one of four grandparents. Though linkage phase of a CP population cannot be defined a priori (as it is in a population derived from inbred individuals), it must be estimated during linkage map construction. Thus, additive grandparental genetic contributions can be estimated within each CP parent. This is particularly relevant in cases where two or more QTL identified on a single LG, as it may allow identification of linked genetic effects.

Following multiple-environment model selection in Genstat®, multiple-QTL, single-environment models were then fit for each trait using R/qtl v. 1.42.8 (Broman et al. 2003) to (i) determine significance for each QTL-by-environment combination, (ii) estimate the proportion of variance explained by each QTL within each environment, and (iii) estimate 95% Bayes Credible Intervals (BCI) for each significant QTL-by- environment combination. Significance was determined for each QTL-by-environment combination by estimation of a point-wise p-value for each locus via drop-one analysis of variance (ANOVA) of a strictly additive QTL model. Loci with a p-value ≤ 0.05 were determined to be significant. Before BCI estimation, within-environment QTL positions were optimized using the refineqtl() function in R/qtl. Genomic position of each QTL region was then defined using the physical positions of markers flanking each BCI. QTL intervals were visualized using MapChart v2.32 (Voorrips 2002).

96 QTL were named according to recommendations provided by Genome Database for Rosaceae (GDR). This naming follows a formula outlined in the following example: qRoc-FrM1.2 where q = quantitative trait

Roc = Rubus occidentalis

FrM = trait code, in this case it is fruit mass

1 = linkage group

2 = 2nd chronological QTL reported for this trait on this linkage group

Results

Genotyping and linkage map construction

A total of 242,391 SNPs were identified using the TASSEL 5 GBS v2 pipeline. Of these, 4367 SNPs passed initial filtering criteria for ORUS 4305 (Table 3.1). Of these,

1623 SNPs passed criteria for expected segregation patterns and were input into

JoinMap® along with 72 SSRs (previously reported by Bushakra et al. 2015) for a total of 1695 genetic markers. Prior to map construction, six individuals were discarded due to high marker missingness (≥30%).

Seven linkage maps were constructed spanning a total of 1230.7 cM and comprising 974 markers (931 SNP and 43 SSR) and 103 individuals (Table 3.2).

Linkage group lengths ranged from 128 to 196 cM and contained 96 to 171 markers.

Collinearity of mapped genetic position to genomic position is illustrated in Figure 3.1.

97 QTL mapping

Phenotypic data were determined to be normally distributed for each trait-by- environment combination (Figure 3.2 through Figure 3.9) and homoscedastic for each trait among environments by evaluation of residuals in a multivariate fixed effect

ANOVA. Available phenotypic data for mapped individuals ranged from 47 to 71 observations per trait-by-environment combination (Table 3.3). Of the 103 individuals included in the final linkage map, fruit size and fruit chemistry trait data were collected in one or more environments for 84 and 80 individuals, respectively.

Fourteen QTL were identified and associated with five of the eight traits analyzed

(Table 3.4). Of these fourteen QTL, significant QEI were found for ten QTL, and significant dominance effects were found for seven QTL (Table 3.5).

Marker-flanked genomic regions containing 95% BCI ranged from 0.4 to 40.5

Mbp in length (Table 3.6). Due to discontinuity between genetic and physical positions at LG 6 (Figure 3.1), genomic positions for putative QTL observed on LG 6 are not reported in this study.

Fruit size traits

Single-trait, multiple-environment linkage analysis was used to identify trait- associated regions for FrM, DCt, and SdF. No QTL were identified for SdM; however,

SIM performed separately for each location-by-year environment found 2 SdM QTL on

LGs 4 and 7 within environments OH_15 and OH_13 respectively. One or more fruit size QTL were identified on LGs 1, 2, 4, 6, and 7.

98 Fruit Mass

Four QTL were associated with FrM, named qRoc-FrM1.1, qRoc-FrM1.2, qRoc-

FrM2.1, and qRoc-FrM6.1 (Table 3.4). Two of these QTL, qRoc-FrM1.1 and qRoc-

FrM6.1, showed stable effects across the ten location-by-year environments (Figure

3.10), and the additive effect of each stable QTL accounted for approximately 70 mg fruit mass (Table 3.5). Two QTL, qRoc-FrM1.2 and qRoc-FrM2.1, showed significant

QEI and unstable effects across environments (Figure 3.10). The effects of qRoc-

FrM1.2 and qRoc-FrM2.1 were significant in three of the ten trial environments (Table

3.5). Further, the effects of qRoc-FrM1.2 and qRoc-FrM2.1 were significant in both NC trial years. Additive effect of qRoc-FrM1.2 was greatest in NC_13 where it accounted for approximately 250 mg fruit mass and 49% of total phenotypic variance in this environment (Table 3.5).

Genomic intervals for FrM QTL, defined by 95% BCI, spanned a median of 16.9

Mbp and ranged from 1.6 to 37.8 Mbp (Table 3.6). Credible intervals for qRoc-FrM1.1 and qRoc-FrM1.2 overlap DCt-associated loci (Figure 3.15). Genetic position was shared between qRoc-FrM1.1 and a DCt-associated locus, qRoc-DCt1.1 (Table 3.4).

Intervals for qRoc-FrM2.1 overlap loci associated with SdF and AnC. Intervals for qRoc-

FrM6.1 overlap loci associated with SdF and TAc (Figure 3.19).

Drupelet Count

Three QTL were associated with DCt, named qRoc-DCt1.1, qRoc-DCt1.2, and qRoc-DCt4.1 (Table 3.4). One of these QTL, qRoc-DCt1.1, showed stable effects across the ten location-by-year environments (Figure 3.10), which accounted for approximately seven drupelets per fruit (Table 3.5). Two QTL, qRoc-DCt1.2 and qRoc-

99 DCt4.1, showed significant QEI and unstable effects across environments (Figure 3.10).

Additive effect of qRoc-DCt1.2 was greatest in NC_13, where it accounted for approximately 12 drupelets per fruit and 56% of total phenotypic variance in this environment (Table 3.5). Effects of qRoc-DCt1.1 and qRoc-DCt4.1 were significant in one or more years at each trial location, suggesting they are fairly stable. Effects of qRoc-DCt1.2 were significant in three of the ten location-by-year environments, including both NC trial years.

Genomic intervals for DCt QTL, defined by 95% BCI, spanned a median of 1.6

Mbp and ranged from 0.3 to 23.9 Mbp (Table 3.6). Genetic position was shared between qRoc-DCt1.1 and a FrM-associated locus, qRoc-FrM1.1 (Figure 3.15).

Credible intervals for qRoc-DCt1.1 and qRoc-DCt1.2 overlap loci associated with FrM

(Figure 3.15). Intervals for qRoc-DCt4.1 overlap a locus associated with TAc (Figure

3.18).

Seed Fraction

Three QTL were associated with SdF, named qRoc-SdF2.1, qRoc-SdF6.1, and qRoc-SdF7.1 (Table 3.4). One QTL, qRoc-SdF6.1, showed stable effects across the ten location-by-year environments (Figure 3.12). Additive effect of qRoc-SdF6.1 equaled approximately 0.23% seed mass per fruit mass (Table 3.5). Interestingly, additive

ORUS 3021-2 effect of qRoc-SdF7.1 is positive in NC_14 and negative in OH_14 and

OH_15, indicating that for these environments, the high value allele was inherited from separate grandparents (see Figure 3.12, Table 3.5). However, this cross-over QEI is only observed in the maternal parent, ORUS 3021-2, and not in the paternal parent,

ORUS 4153-1 (Table 3.5).

100 Genomic intervals for SdF QTL, defined by 95% BCI, spanned a median of 9.7

Mbp and ranged from 2.1 to 35.1 Mbp (Table 3.6). Credible intervals for qRoc-SdF2.1 overlap loci associated with FrM and AnC (Figure 3.16). Intervals for qRoc-SdF6.1 overlap loci associated with FrM and TAc (Figure 3.19).

Fruit chemistry traits

Single-trait, multiple-environment linkage analysis was used to identify trait- associated regions for TAc and AnC. No significant QTL were identified for SSC or PhC, and this finding was confirmed by SIM performed separately for each location-by-year environment. One or more fruit chemistry QTL were identified on LGs 2, 3, 4, and 6.

Anthocyanin Content

Two QTL were associated with AnC, named qRoc-AnC2.1 and qRoc-AnC3.1

(Table 3.4). Significant QEI and unstable effects across environments were observed for both AnC QTL (Figure 3.14). Effects of qRo_AnC2.1 and qRoc-AnC3.1 were significant in three and four environments, respectively (Table 3.5). Further, effects of both QTL were significant in OH_14 and OH_15. Interestingly, additive ORUS 3021-2 effect of qRoc-AnC2.1 is positive at OH_14 and negative at OH_15 and OR_13, indicating that for these environments, the high value allele was inherited from separate grandparents (see Figure 3.14, Table 3.5). However, this cross-over QEI is only observed in the maternal parent, ORUS 3021-2, and not in the paternal parent, ORUS

4153-1 (Table 3.5).

Genomic intervals for AnC QTL, defined by 95% BCI, spanned a median of 18.7

Mbp and ranged from 2.8 to 38.5 Mbp (Table 3.6). Credible intervals for qRoc-AnC2.1

101 overlap loci associated with FrM and SdF (Figure 3.16). Intervals for qRoc-AnC3.1 overlap a locus associated with TAc (Figure 3.17).

Titratable Acidity

Three QTL were associated with TAc, named qRoc-TAc3.1, qRoc-TAc4.1, and qRoc-TAc6.1 (Table 3.4). Significant QEI and unstable effects across environments were observed for all TAc QTL (Figure 3.13). Effects of qRoc-TAc3.1 were significant in one or more years at all trial locations, suggesting its effect is fairly stable. Interestingly, additive ORUS 3021-2 effect of qRoc-TAc4.1 is positive at NY_13 and negative at

OH_15, indicating that for these environments, the high value allele was inherited from separate grandparents (see Figure 3.13, Table 3.5). However, this cross-over QEI is only observed in the maternal parent, ORUS 3021-2, and not in the paternal parent,

ORUS 4153-1 (Table 3.5).

Genomic intervals for TAc QTL, defined by 95% BCI, spanned a median of 23

Mbp and ranged from 3.3 to 40.5 Mbp (Table 3.6). Credible intervals for qRoc-TAc3.1 overlap a locus associated with AnC (Figure 3.17). Intervals for qRoc-TAc4.1 overlap with a locus associated with DCt (Figure 3.18). Intervals for qRoc-TAc6.1 overlap loci associated with FrM and SdF (Figure 3.19).

Discussion

Interest in BR for its flavor and health benefits warrants the development of improved production systems and adapted cultivars. However, efforts to breed and domesticate BR have shown limited success (Jennings 1988). Low genetic diversity within cultivated BR varieties relative to wild BR populations suggest limited sampling of wild germplasm within cultivated BR (Weber 2003; Dossett et al. 2012a). Increasing the

102 diversity of elite BR germplasm through controlled crosses with wild accessions will provide novel qualitative traits, such as resistance to virus-transmitting aphids (Dossett and Finn 2010; Bushakra et al. 2018), and may provide novel genetic contributions to quantitative traits such as fruit size and anthocyanin content. Optimal selection for breeding will rely on characterization of breeding germplasm to identify genetic components as well as GEI among production environments. In this study, linkage mapping was used to identify and characterize fruit quality QTL within a BR F1 population featuring cultivated and wild progenitors as well as QEI among four distinct geographic locations where BR are grown.

Genotyping and linkage map construction

Linkage maps presented in this study were constructed using SNP marker data gathered from five separate GBS runs and aligned to the BR genome (VanBuren et al.

2018) as well as SSR data previously collected by Bushakra et al. (2015) and Cash

(2016). Alignment of SNP markers to the BR reference genome allows direct comparison of genetic and genomic positions (Figure 3.1). These comparisons show expected patterns of collinearity between the generated linkage maps and current genome assembly for most LGs. However, discontinuous portions in chromosomes 1, 4, and 6 suggest errors in alignment of GBS data to the genome, linkage map construction, or the current BR genome assembly.

Three drafts of the BR genome assembly have been presented to date. The first featured Illumina paired-end sequence data assembled into 9245 contiguous sequences (contigs) and 2226 scaffolds (VanBuren et al. 2016). The second draft used high-throughput chromatin confirmation capture (Hi-C) and Proximity Guided Assembly

103 (PGA) to assemble draft 1 scaffolds to chromosome scale (Jibran et al. 2018). The third and most recent draft combined this Hi-C approach with Pacific Biosciences (PacBio) long-read sequence data, greatly increasing coverage and average contig length

(VanBuren et al. 2018). Though the latest assembly shows higher quality than either previous draft, the orientation of some contigs was unclear due to lack of guidance by a linkage map. The linkage map here developed for ORUS 4305 may provide guidance to verify and address issues in assembly of chromosomes 1, 4, and 6.

GBS provides reduced representation of full sequence data following digestion of genomic DNA. Selection of an appropriate enzyme is important, as sequence data quality varies by digestion enzyme used (Sonah et al. 2013; Nguyên et al. 2018). In this study, the use of ApeK1 was continued in order to keep congruency with previous GBS data produced in Oregon and North Carolina. Selection of the ApeKI enzyme for GBS digestion predated availability of a high-quality genome assembly. In silico digestions with the current genome assembly may be used to identify additional enzymes for GBS in BR.

QTL mapping

Linkage mapping of quantitative traits provides a basis for the implementation of marker assisted selection and the identification of causal genes (Collard et al. 2005).

Large BCI reflect limited statistical power to discern genes influencing an observed trait, whereas narrow BCI enable identification of markers which are strongly associated with that trait through genetic linkage. For marker assisted selection in a breeding program, ideal QTL have narrow BCI and are stable within target environments. Narrow BCI are

104 also desirable for causal gene identification, as a narrowed region should contain fewer candidate genes.

Mixed model QTL analysis of MET data accounts for genetic correlation across environments through the fitting of environmental variance-covariance structures, thus optimizing statistical power while minimizing Type I error (i.e. false-positive QTL detection) (Piepho 2005). This approach provides a conservative starting point for the identification of significant QTL and estimation of QEI. Fifteen QTL influencing five fruit quality traits in BR are identified herein, and their additive and dominance effects are estimated within each location-by-year environment.

QTL mapping is commonly used to explore genetic bases of crop domestication

(Miller and Gross 2011). Signatures of perennial crop domestication include instability of

QTL across trial years (i.e. QEI) and low prevalence of QTLs that explain a large proportion (>20%) of phenotypic variance (Miller and Gross 2011). In contrast, domestication traits of annual crops are often caused by few QTL of large effect accounting for >20% of phenotypic variation (Miller and Gross 2011). Four of the fifteen

QTL identified in this study lacked QEI and appeared stable across all location-by-year environments. Ten of the fifteen QTL identified in this study account for >20% of phenotypic variance in one or more trial environments, and three of these QTL account for >45% of phenotypic variance in one or more environments. QTL accounting for

>45% of phenotypic variance are all located on LG 1, and they include one QTL associated with fruit mass (qRoc-FrM1.2) and two QTL associated with drupelet count

(qRoDCt1.1 and qRoDCt1.2). The discovery of stable fruit quality QTL of high effect

105 suggests that BR domestication shares patterns of genetic basis typically observed in inbreeding, annual crop species.

Three QTL (qRoc-SdF7.1, qRoc-TAc4.1, and qRoc-AnC2.1) were estimated to have diverging additive effects in separate environments, indicating cross-over QEI at these loci. For example, qRoc-SdF7.1 shows a positive additive effect in NC_14 and negative effects in OH_14 and OH_15. In all three cases, this divergence is only observed in the maternal parent, ORUS 3021-2, and not in the paternal parent, ORUS

4153-1. Selection and future study of these QTL should follow special consideration of the target environment in which the trait will be observed.

It is noted that all QTL identified on a shared LG also have overlapping regions, as defined by 95% BCI estimated within each environment (Figure 3.15 through Figure

3.19). Some coincidental co-localization is to be expected assuming random distribution of QTL positions. Coincidental co-localization in this study is especially likely given that several QTL regions extend most (>50%) of the genetic distance of their respective

LGs. Thus, special attention should be given to co-localization of narrow QTL regions and putative causal variants.

Large QTL BCI reflect limited statistical power to discern genes with significant effect on the observed traits. Phenotypic data were not collected from all mapped individuals. The ORUS 4305 linkage map contained 109 individuals, 84 of which were phenotyped in one or more locations. Further, precision of these measurements is limited by low GBS sequence depth and unreplicated phenotypic observations.

Increasing the available within-individual data through increased sampling of sequence data or phenotypic data may improve accuracy and precision of QTL identification. In

106 future mapping studies, increasing the number of individuals observed may also provide increased statistical power to identify additional QTL of finer resolution.

Efforts were made in the present study to maximize useful data while controlling for erroneous data due to low GBS sequencing depth and inconsistent population structure. Consequentially, some individuals with available phenotypic data were removed from QTL analysis due to insufficient marker data. Additionally, many markers were removed from analysis due to low depth across individuals. Future emphasis on deeper sequencing of individuals with available phenotypic data may better utilize available phenotypic data and alleviate this issue in future studies.

Heterogeneous BCI estimated among environments may be due to heterogeneous QTL effects within each environment (i.e. QEI) and limited, unbalanced data. Li et al. (2010) recommend a population size of over 200 individuals when mapping traits in a doubled-haploid biparental population, noting decreased accuracy in

QTL position and effect estimates with lower population sizes. The present study retained only 84 individuals for QTL mapping, less than half that recommended by Li et al. (2010).

Though the observed population size was less than ideal for QTL analysis, it should be recognized that the MET described in this study required a large phenotyping effort and collaboration between multiple research groups situated across the United

States. Validation and fine-mapping of QTL may be most efficiently performed in single locations that show high trait heritability and significant QTL for the trait of interest.

As a next step, polymerase chain reaction (PCR) markers may be developed through Kompetitive Allele Specific PCR (Semagn et al. 2014) or amplicon sequencing

107 (Fresnedo-Ramírez et al. 2019; Fresnedo-Ramírez et al. 2017; Yang et al. 2016). These markers may be used to track and validate the QTL in separate populations and then used for marker-assisted breeding of BR. Candidate genes may also be identified by comparative genomics with related plant species.

Fruit size traits

Fruit size traits influence yield and marketability of fresh market fruit. Fresh raspberry consumers prefer large fruit with inconspicuous seeds. An important hurdle to expanded production of fresh BR is this crop’s tendency to produce small fruit with large, conspicuous seeds. The current study identifies BR QTL associated with two fruit size parameters, FrM and DCt, and one fruit seediness parameter, SdF. Identification of

DCt and SdF QTL on separate LGs suggest these traits may be selected for independently in BR breeding germplasm. Further narrowing of these QTL regions may reveal candidate genes controlling these traits.

Fruit mass and drupelet count

Fruit mass and drupelet count are both components of fruit size and important traits for BR breeding. Though data related to breeding improvements of BR are lacking, increased fruit size has been met in red raspberry through genetic increases in drupelet count and drupelet size (Jennings 1988).

Three fruit size QTL identified in this study accounted for a very large proportion

(>45%) of observed phenotypic variance: qRoc-FrM1.2, qRoc-DCt1.1, and qRocDCt1.2.

The effects of qRoc-FrM1.2 and qRoc-FrM2.1 were significant in three of the ten trial environments (Table 3.5). Further, the effects of qRoc-FrM1.2 and qRoc-FrM2.1 were significant in both NC trial years, suggesting the genetic effects of these loci are

108 particularly prominent in NC. Effects of qRoc-DC1.1 and qRoc-DCt4.1 were significant in one or more years at each trial location, suggesting they are fairly stable and good candidates for selection. Effects of qRoc-DCt1.2 were significant in three of the ten location-by-year environments, including both NC trial years, suggesting genetic effect of this locus is particularly prominent in NC.

Fruit mass QTL have been previously identified in red raspberry, including fruit mass and fruit firmness QTL on three linkage groups (Simpson et al. 2017) which are equivalent to BR LGs 5, 6, and 7, as defined by Bushakra et al. (2012). Three fruit size

QTL were identified across LGs 6 and 7: qRoc-FrM6.1, qRoc-SdF6.1, and qRoc-

SdF7.1. Synteny between cultivated R. idaeus and R. occidentalis has been loosely established through interspecific hybridization and the development of an R. occidentalis × R. idaeus linkage map (Bushakra et al. 2012). Direct comparison between genome assemblies for BR (VanBuren et al. 2018) and R. idaeus cv. Glen Moy

(Hackett et al. 2018) may enable greater exploration of synteny and identification of overlapping QTL for these two species.

A candidate gene approach may also be appropriate for exploration of fruit size traits BR. Several candidate genes for fruit mass in red raspberry have been reported by Simpson et al. (2017). Homologous genes may be identified using the BR genome assembly and their positions contrasted to those QTL positions identified in the current study.

Several genes have been reported to control organ size in plant species including tomato (Solanum lycopersicum) fw2.2 (Frary 2000) and maize (Zea mays) Cell

Number Regulator (CNR) family genes (Guo et al. 2010). CNRs are putative orthologs

109 of fw2.2, and expression of CNR1 has been shown to reduce plant and organ size by reducing cell number (Guo et al. 2010). This reduction in plant organ size is also strongly correlated with copy number of CNR1 (Guo et al. 2010). CNR homologs have been identified in Prunus overlapping fruit size QTL in sweet cherry (P. avium, see De

Franceschi et al. 2013). Preliminary BLAST alignment (Altschul et al. 1990) of fw2.2 and

CNR1 proteins to the BR genome revealed homology between both proteins and the

BR genome on chromosomes 2, 4, and 5 (Table 3.7). Some BLAST hits on BR chromosome 4 fall within credible intervals for qRoc-DCt4.1 and may warrant further investigation.

Seed mass and seed fraction

Seed mass and seed fraction are both components of seediness and important in

Rubus crop traits, particularly in fresh fruit markets. Additional traits influencing perceptible seediness but not included in this study include seed shape and seed composition (Sebesta et al. 2013). Seed mass tends to be a quantitative and highly heritable trait in Rubus (Jennings 1988). Large seeds (>3.0 mg) are considered objectionable in blackberry breeding programs (Sebesta et al. 2013). Seed size data are not readily available for cultivated red raspberry, though moderate seed mass (1.7 mg) has been described for wild-type R. idaeus (Hummer and Peacock 1994). Seed mass observed in this study for BR (�̅=1.86, s=0.24) were well below those described for cultivated blackberry and near those described for wild R. idaeus. No seed mass QTL were identified in the current study, suggesting genetic variance for this trait is low in

ORUS 4305 or contributable to genes of low effect that were not detected in this study.

110 Though SdM is an important trait for Rubus breeding, SdF, calculated as a percentage of fresh seed mass over fresh fruit mass, may be a more important contributor to perceptible seediness. A 1931 study of Rubus fruit size and seediness traits found similar SdM between cultivated BR and red raspberry; however, SdF values observed for cultivated BR in this study were much greater (approximately twice) those observed for cultivated red raspberry (Darrow and Sherwood 1931). Such observations led Darrow and Sherwood (1931) to suggest that proportion of seed weight to total berry weight is a more important seediness factor than seed size. Typical SdF observed in red raspberry breeding germplasm averages 4% (Chad Finn, unpublished data) whereas mean BR SdF observed in the present study was 8.0%.

SdF is related to FrM, as increased FrM may be achieved through increased seed or non-seed (i.e. flesh) portions of the fruit. Large fruits are desirable for fresh market production; however, increased FrM accompanied by increased seediness is undesirable. Two SdF QTL overlap QTL controlling FrM, one each on LGs 2 and 6, suggesting these traits may share pleiotropic or linked genetic components. High value alleles for both traits originate from the same grandparent at the overlapping QTL on LG

2, indicating positive correlation between FrM and SdF at this locus. Thus, selection for increased FrM at qRoc-FrM2.1 may bear a tradeoff of increased SdF. In contrast, high value alleles originate from opposite grandparents at the overlapping QTL on LG 6, indicating negative correlation between FrM and SdF at this locus (Table 3.5). Thus, selection for increased FrM at qRoc-FrM6.1 may complement selection of reduced SdF.

These findings emphasize that selection in BR for increased FrM should consider related seediness traits, including SdF.

111 Fruit chemistry traits

Fruit chemical composition in raspberry is greatly influenced by fruit maturity, with anthocyanin and sugar content increasing and acidity decreasing as fruit matures

(Hancock et al. 2018). Fruit chemical composition traits in ripe raspberry fruits have been shown to be greatly influenced by environmental factors including location (Ozgen et al. 2008), harvest season (Mazur et al. 2014b), and photoperiod (Mazur et al. 2014a).

The current study presents three AnC QTL and two TAc QTL.

Anthocyanin content

Numerous anthocyanin compounds have been identified and characterized in

Rubus fruit, including BR (Lee et al. 2012). Dietary anthocyanins and other fruit phenolics are associated with improved human health, and important areas of research include determining mechanisms by which these phytochemicals influence human health and developing crop varieties with improved levels of beneficial compounds

(Mazzoni et al. 2016). Anthocyanin biosynthesis enzymology and localization are well characterized in model plant species; however, improved understanding of genetic regulation of anthocyanin biosynthesis in crop species is needed to develop crops with increased anthocyanin levels (Chaves-Silva et al. 2018). A wealth of genes regulating anthocyanin biosynthesis and accumulation of anthocyanins have been discovered and characterized in Rosaceous small fruit, including cultivated strawberry (Fragaria × ananassa, see Scalzo et al. 2005; Allan et al. 2008; Griesser et al. 2008; and Zorrilla-

Fontanesi et al. 2011), wild strawberry (F. vesca, see Shulaev et al. 2011), red raspberry (R. idaeus), and Korean black raspberry (R. coreanus, see Hyun et al. 2014).

112 AnC QTL identified in the present study may inform allelic variation for genetic components regulating anthocyanin biosynthesis and accumulation in BR. Alignment of candidate genes studied in other Rosaceous species to the BR genome assembly may provide candidate genes for chemical composition trait control in BR.

AnC in this study is total anthocyanin content of ripe fruit measured as cyanidin

3-glucoside equivalents. Total anthocyanin is a complex BR trait consisting of multiple anthocyanin compounds (Paudel et al. 2014) and influenced by many biological factors.

Increased focus on distinct anthocyanin compounds, anthocyanin precursors, or anthocyanin biosynthesis pathway enzymes may provide a clearer signal for genetic analysis in a MET as the present, and a more detailed understanding of anthocyanin biosynthesis and storage in BR.

Titratable Acidity

Acidity is an important component of flavor in fruit crops (Klee 2010). Acidity levels are also important to processing fruit markets, as anthocyanin stability and color are both dependent on acidity and pH (Khoo et al. 2017). The production and accumulation of acidity in fruit is under control of many complex factors (Etienne et al.

2013). The current study reports titratable acidity as % citric acid, as is common in raspberry studies (Malowicki et al. 2008; Vool et al. 2009). A balance of sugar content and acidity is considered important for raspberry fruit flavor; however genetic control of these traits tends to be unstable over different years and locations (Graham and

Brennan 2018).

Effects of qRoc-TAc3.1 were significant in one or more years at all trial locations, suggesting its effect is fairly stable, and selection for increased or reduced acidity may

113 be considered through phenotypic or marker assisted selection. This is relevant to the

BR processing market, where increased fruit acidity may improve anthocyanin stability and improved pigmentation of products containing BR extracts.

Candidate genes for titratable acidity have previously been mapped in peach

(Etienne et al. 2002). In red raspberry, acid content has been found to vary between cultivars (Krüger et al. 2011; Mazur et al. 2014b) and growing environments (Mazur et al. 2014b, a). However, reported QTL associated with raspberry TAc are lacking.

Soluble solid content and phenolic content

SSC is commonly used as a measure of sugar content, which is an important component of fruit flavor (Klee 2010). No SSC QTL were identified in the present study, suggesting genetic variation for SSC in ORUS 4305 is either lacking or conferred by

QTL of low effect and undetectable given available data. SSC QTL have been reported in other Rosaceous crop species, including strawberry, (Zorrilla-Fontanesi et al. 2011), peach (Fresnedo-Ramírez et al. 2015), and apple (Guan et al. 2015). Reports of sugar profiles (e.g. sucrose, fructose, and glucose content) are currently lacking in raspberry, and future study of SSC in BR may consider these components, as has been done in other Rosaceous crops such as peach (Etienne et al. 2002).

PhC is an important trait due to its association with human health benefits. No

PhC QTL were identified in the present study. However, PhC QTL have been reported in red raspberry (Dobson et al. 2012). Further, PhC has been shown to be strongly positively correlated with AnC in BR (Ozgen et al. 2008). Thus, it is possible that genetic variation for PhC in the present study was undetected due to limited sample size.

114 Breeders wishing to select for PhC may focus on AnC, for which several QTL were identified in the present study.

Conclusions

Fifteen QTL associated with five fruit quality traits were identified in this study: fruit mass, drupelet count, seed fraction, titratable acidity, and total anthocyanin content.

Multiple QTL identified for each trait confirms the quantitative nature of these traits. High variance explained by several QTL confirms the potential to selectively breed for these traits through phenotypic or marker assisted selection. Effects of some QTL were very high, suggesting genetic gains may be quickly made through phenotypic selection alone. Additionally, genomic positions of QTL were identified, enabling development of primer pairs for further investigation of QTL, including QTL verification, marker-assisted selection, and candidate gene selection. Future steps include verification of QTL in additional BR germplasm and candidate gene exploration. Future QTL analyses in BR will benefit from increased individual data, especially regarding phenotypic data. QTL analysis of ORUS 4304 was unsuccessful due to limited depth of GBS data and discrepancies found during marker-guided relatedness confirmation. However, further analysis of ORUS 4304 may be productive following collection of additional GBS data and assuming no further discrepancies exist. Results of this analysis confirm that wild germplasm provides genetic variation for important BR fruit quality traits. This study also establishes a basis for further exploration of trait-associated genomic regions.

Characterization these traits will improve our understanding of genetic components controlling fruit quality in BR and related crop species. In addition, careful

115 characterization and selection of these traits will the production and marketability of future BR cultivars.

116 References

Allan AC, Hellens RP, Laing WA (2008) MYB transcription factors that colour our fruit. Trends Plant Sci 13:99–102. https://www.doi.org/10.1016/j.tplants.2007.11.012

Altschul SF, Gish W, Miller W, et al (1990) Basic local alignment search tool. J Mol Bio 215:403–410. https://www.doi.org/10.1016/S0022-2836(05)80360-2

Amiot MJ, Riva C, Vinet A (2016) Effects of dietary polyphenols on metabolic syndrome features in humans: a systematic review. Obes Rev 17:573–586. https://www.doi.org/10.1111/obr.12409

Andrews S (2010) FastQC: a quality control tool for high throughput sequence data. Available online at: http://www.bioinformatics.babraham.ac.uk/projects/fastqc

Boer MP, Wright D, Feng L, et al (2007) A mixed-model quantitative trait loci (QTL) analysis for multiple-environment trial data using environmental covariables for QTL-by-environment interactions, with an example in maize. Genetics 177:1801– 1813. https://www.doi.org/10.1534/genetics.107.071068

Bolger AM, Lohse M, Usadel B (2014) Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30:2114–2120. https://www.doi.org/10.1093/bioinformatics/btu170

Broman KW, Wu H, Sen S, Churchill GA (2003) R/qtl: QTL mapping in experimental crosses. Bioinformatics 19:889–890. https://www.doi.org/10.1093/bioinformatics/btg112

Bushakra JM, Bassil NV, Weiland JE, et al (2016) Comparative RNA-seq for the investigation of tolerance to Verticillium wilt in black raspberry. Acta Hortic 1133:103–114. https://www.doi.org/10.17660/ActaHortic.2016.1133.16

Bushakra JM, Bryant DW, Dossett M, et al (2015) A genetic linkage map of black raspberry (Rubus occidentalis) and the mapping of Ag 4 conferring resistance to the aphid Amphorophora agathonica. Theor Appl Genet 128:1631–1646. https://www.doi.org/10.1007/s00122-015-2541-x

Bushakra JM, Dossett M, Carter KA, et al (2018) Characterization of aphid resistance loci in black raspberry (Rubus occidentalis L.). Mol Breed 38:83. https://www.doi.org/10.1007/s11032-018-0839-5

Bushakra JM, Stephens MJ, Atmadjaja AN, et al (2012) Construction of black (Rubus occidentalis) and red (R. idaeus) raspberry linkage maps and their comparison to the genomes of strawberry, apple, and peach. Theor Appl Genet 125:311–327. https://www.doi.org/10.1007/s00122-012-1835-5

117 Cash CB (2016) Studies of linkage mapping, trait heritability, and pedigrees for breeding improvement of southeastern black and red raspberry. North Carolina State University

Chaves-Silva S, Santos AL dos, Chalfun-Júnior A, et al (2018) Understanding the genetic regulation of anthocyanin biosynthesis in plants – Tools for breeding purple varieties of fruits and vegetables. Phytochemistry 153:11–27. https://www.doi.org/10.1016/j.phytochem.2018.05.013

Collard BCY, Jahufer MZZ, Brouwer JB, Pang ECK (2005) An introduction to markers, quantitative trait loci (QTL) mapping and marker-assisted selection for crop improvement: The basic concepts. Euphytica 142:169–196. https://www.doi.org/10.1007/s10681-005-1681-5

Danecek P, Auton A, Abecasis G, et al (2011) The variant call format and VCFtools. Bioinformatics 27:2156–2158. https://www.doi.org/10.1093/bioinformatics/btr330

Darrow GM, Sherwood H (1931) Seed and berry size of cane fruits. Proc Amer Soc Hort Sci 28:194–199

De Franceschi P, Stegmeir T, Cabrera A, et al (2013) Cell number regulator genes in Prunus provide candidate genes for the control of fruit size in sweet and sour cherry. Mol Breed 32:311–326. https://www.doi.org/10.1007/s11032-013-9872-6

Dobson P, Graham J, Stewart D, et al (2012) Over-seasons analysis of quantitative trait loci affecting phenolic content and antioxidant capacity in raspberry. J Agric Food Chem 60:5360–5366. https://www.doi.org/10.1021/jf3005178

Dossett M, Bassil NV, Lewers KS, Finn CE (2012) Genetic diversity in wild and cultivated black raspberry (Rubus occidentalis L.) evaluated by simple sequence repeat markers. Genet Resour Crop Evol 59:1849–1865. https://www.doi.org/10.1007/s10722-012-9808-8

Dossett M, Finn CE (2010) Identification of resistance to the large raspberry aphid in black raspberry. J Am Soc Hortic Sci 135:438–444

Elshire RJ, Glaubitz JC, Sun Q, et al (2011) A robust, simple genotyping-by-sequencing (GBS) approach for high diversity species. PLoS ONE 6:e19379. https://www.doi.org/10.1371/journal.pone.0019379

Etienne A, Génard M, Lobit P, et al (2013) What controls fleshy fruit acidity? A review of malate and citrate accumulation in fruit cells. J Exp Biol 64:1451–1469. https://www.doi.org/10.1093/jxb/ert035

Etienne C, Rothan C, Moing A, et al (2002) Candidate genes and QTLs for sugar and organic acid content in peach [Prunus persica (L.) Batsch]. Theor Appl Genet 105:145–159. https://www.doi.org/10.1007/s00122-001-0841-9

118 Frary A (2000) fw2.2: a quantitative trait locus key to the evolution of tomato fruit size. Science 289:85–88. https://www.doi.org/10.1126/science.289.5476.85

Fresnedo-Ramírez J, Bink MCAM, van de Weg E, et al (2015) QTL mapping of pomological traits in peach and related species breeding germplasm. Mol Breed 35:. https://www.doi.org/10.1007/s11032-015-0357-7

Fresnedo-Ramírez J, Yang S, Sun Q, et al (2017) An integrative AmpSeq platform for highly multiplexed marker-assisted pyramiding of grapevine powdery mildew resistance loci. Mol Breed 37:145 https://doi.org/10.1007/s11032-017-0739-0

Fresnedo-Ramírez J, Yang S, Sun Q, et al (2019) Computational analysis of AmpSeq data for targeted, high-throughput genotyping of amplicons. Front Plant Sci. 10:599. https://www.doi.org/10.3389/fpls.2019.00599

Glaubitz JC, Casstevens TM, Lu F, et al (2014) TASSEL-GBS: a high capacity genotyping by sequencing analysis pipeline. PLoS ONE 9:2. https://www.doi.org/10.1371/journal.pone.0090346

Graham J, Brennan R (eds) (2018) Raspberries: breeding, challenges, and advances. Springer Science+Business Media, New York, NY

Griesser M, Hoffmann T, Bellido ML, et al (2008) Redirection of flavonoid biosynthesis through the down-regulation of an anthocyanidin glucosyltransferase in ripening strawberry fruit. Plant Physiol 146:1528–1539. https://www.doi.org/10.1104/pp.107.114280

Guan Y, Peace C, Rudell D, et al (2015) QTLs detected for individual sugars and soluble solids content in apple. Mol Breed 35:135. https://www.doi.org/10.1007/s11032-015-0334-1

Guo M, Rupe MA, Dieter JA, et al (2010) Cell Number Regulator1 affects plant and organ size in maize: implications for crop yield enhancement and heterosis. Plant Cell 22:1057–1073. https://www.doi.org/10.1105/tpc.109.073676

Hackett CA, Milne L, Smith K, et al (2018) Enhancement of Glen Moy x Latham raspberry linkage map using GbS to further understand control of developmental processes leading to fruit ripening. BMC Genetics 19:59. https://www.doi.org/10.1186/s12863-018-0666-z

Halgren A, Tzanetakis IE, Martin RR (2007) Identification, characterization, and detection of black raspberry necrosis virus. Phytopathology 97:44–50. https://www.doi.org/10.1094/PHYTO-97-0044

Hancock RD, Petridis A, McDougall GJ (2018) Raspberry fruit chemistry in relation to fruit quality and human nutrition. In: Raspberries: breeding, challenges, and advances. Springer Science+Business Media, pp 89–120

119 Hummer KE, Peacock DN (1994) Seed dimension and weight of selected Rubus species. HortScience 29:1034–1036. https://www.doi.org/10.21273/HORTSCI.29.9.1034

Hyun TK, Lee S, Rim Y, et al (2014) De-novo RNA sequencing and metabolite profiling to identify genes involved in anthocyanin biosynthesis in Korean black raspberry (Rubus coreanus Miquel). PLoS ONE 9:e88292. https://www.doi.org/10.1371/journal.pone.0088292

Jennings DL (1988) Raspberries and blackberries: their breeding, diseases and growth. Academic Press, San Diego, CA

Jibran R, Dzierzon H, Bassil N, et al (2018) Chromosome-scale scaffolding of the black raspberry (Rubus occidentalis L.) genome based on chromatin interaction data. Hortic Res 5:8. https://www.doi.org/10.1038/s41438-017-0013-y

Khoo HE, Azlan A, Tang ST, Lim SM (2017) Anthocyanidins and anthocyanins: colored pigments as food, pharmaceutical ingredients, and the potential health benefits. Food Nutr Res 61:1361779. https://www.doi.org/10.1080/16546628.2017.1361779

Klee HJ (2010) Improving the flavor of fresh fruits: genomics, biochemistry, and biotechnology: Tansley review. New Phytol 187:44–56. https://www.doi.org/10.1111/j.1469-8137.2010.03281.x

Kresty LA, Mallery SR, Stoner GD (2016) Black raspberries in cancer clinical trials: past, present and future. J Berry Res 6:251–261. https://www.doi.org/10.3233/JBR- 160125

Krüger E, Dietrich H, Schöpplein E, et al (2011) Cultivar, storage conditions and ripening effects on physical and chemical qualities of red raspberry fruit. Postharvest Biol Tec 60:31–37. https://www.doi.org/10.1016/j.postharvbio.2010.12.001

Lee J, Dossett M, Finn CE (2012) Rubus fruit phenolic research: the good, the bad, and the confusing. Food Chem 130:785–796. https://www.doi.org/10.1016/j.foodchem.2011.08.022

Li H, Durbin R (2009) Fast and accurate short read alignment with Burrows–Wheeler transform. Bioinformatics 25:1754–1760. https://www.doi.org/10.1093/bioinformatics/btp324

Li H, Hearne S, Bänziger M, et al (2010) Statistical properties of QTL linkage mapping in biparental genetic populations. Heredity 105:257–267. https://www.doi.org/10.1038/hdy.2010.56

120 Li J, Ji L (2005) Adjusting multiple testing in multilocus analyses using the eigenvalues of a correlation matrix. Heredity 95:221–227. https://www.doi.org/10.1038/sj.hdy.6800717

Lodhi MA, Ye G-N, Weeden NF, Reisch BI (1994) A simple and efficient method for DNA extraction from grapevine cultivars and Vitis species. Plant Mol Biol Report 12:6–13. https://www.doi.org/10.1007/BF02668658

Maliepaard C, Jansen J, Van Ooijen JW (1997) Linkage analysis in a full-sib family of an outbreeding plant species: overview and consequences for applications. Genetical Research 70:237–250. https://www.doi.org/10.1017/S0016672397003005

Malosetti M, Ribaut JM, Vargas M, et al (2008) A multi-trait multi-environment QTL mixed model with an application to drought and nitrogen stress trials in maize (Zea mays L.). Euphytica 161:241–257. https://www.doi.org/10.1007/s10681- 007-9594-0

Malowicki SMM, Martin R, Qian MC (2008) Comparison of sugar, acids, and volatile composition in raspberry Bushy Dwarf Virus-resistant transgenic raspberries and the wild type ‘Meeker’ (Rubus Idaeus L.). J Agric Food Chem 56:6648–6655. https://www.doi.org/10.1021/jf800253e

Margarido GRA, Pastina MM, Souza AP, Garcia AAF (2015) Multi-trait multi- environment quantitative trait loci mapping for a sugarcane commercial cross provides insights on the inheritance of important traits. Mol Breed 35:175. https://www.doi.org/10.1007/s11032-015-0366-6

Martin RR, MacFarlane S, Sabanadzovic S, et al (2013) Viruses and virus diseases of Rubus. Plant Dis 97:168–182

Mathews KL, Malosetti M, Chapman S, et al (2008) Multi-environment QTL mixed models for drought stress adaptation in wheat. Theor Appl Genet 117:1077– 1091. https://www.doi.org/10.1007/s00122-008-0846-8

Mazur SP, Sønsteby A, Wold A-B, et al (2014a) Post-flowering photoperiod has marked effects on fruit chemical composition in red raspberry (Rubus idaeus). Ann Appl Biol 165:454–465. https://www.doi.org/10.1111/aab.12153

Mazur SP, Nes A, Wold A-B, et al (2014b) Quality and chemical composition of ten red raspberry (Rubus idaeus L.) genotypes during three harvest seasons. Food Chem 160:233–240. https://www.doi.org/10.1016/j.foodchem.2014.02.174

Mazzoni L, Perez-Lopez P, Giampieri F, et al (2016) The genetic aspects of berries: from field to health: The genetic aspects of berries. J Sci Food Agric 96:365–371. https://www.doi.org/10.1002/jsfa.7216

121 Miller AJ, Gross BL (2011) From forest to field: perennial fruit crop domestication. Am J Bot 98:1389–1414. https://www.doi.org/10.3732/ajb.1000522

Nguyên T, Yu J, Choi H-W, et al (2018) Optimization of genotyping-by-sequencing (GBS) in chrysanthemums: selecting proper restriction enzymes for GBS library construction. Hortic Sci Technol 36:108–114. https://www.doi.org/10.12972/kjhst.20180012

Ourecky DK, Slate GL (1973) Jewel black raspberry. New York’s Food and Life Sci Bul 35

Ozgen M, Wyzgoski FJ, Tulio AZ, et al (2008) Antioxidant capacity and phenolic antioxidants of Midwestern black raspberries grown for direct markets are influenced by production site. HortScience 43:2039–2047

Pastina MM, Malosetti M, Gazaffi R, et al (2012) A mixed model QTL analysis for sugarcane multiple-harvest-location trial data. Theor Appl Genet 124:835–849. https://www.doi.org/10.1007/s00122-011-1748-8

Paudel L, Wyzgoski FJ, Giusti MM, et al (2014) NMR-based metabolomic investigation of bioactivity of chemical constituents in black raspberry (Rubus occidentalis L.) fruit extracts. J Agric Food Chem 62:1989–1998. https://www.doi.org/10.1021/jf404998k

Perkins-Veazie P, Ma G, Fernandez GE, et al (2016) Black raspberry fruit composition over two years from seedling populations grown at four US geographic locations. Acta Hortic 1133:335–338. https://www.doi.org/10.17660/ActaHortic.2016.1133.52

Piepho HP (2005) Statistical tests for QTL and QTL-by-environment effects in segregating populations derived from line crosses. Theor Appl Genet 110:561– 566

Scalzo J, Battino M, Costantini E, Mezzetti B (2005) Breeding and biotechnology for improving berry nutritional quality. Biofactors 23:213–220. https://www.doi.org/10.1002/biof.5520230406

Sebesta B, Clark JR, Threlfall RT, Howard LR (2013) Characterization of seediness attributes of blackberry genotypes. Discovery, the student journal of Dale Bumpers College of Agricultural, Food and Life Sciences 14:72–79

Shulaev V, Sargent DJ, Crowhurst RN, et al (2011) The genome of woodland strawberry (Fragaria vesca). Nature Genet 43:109–116. https://www.doi.org/10.1038/ng.740

Silva NF, Goring DR (2001) Mechanisms of self-incompatibility in flowering plants. Cell Mol Life Sci 58:1988–2007

122 Simpson CG, Cullen DW, Hackett CA, et al (2017) Mapping and expression of genes associated with raspberry fruit ripening and softening. Theor Appl Genet 130:557–572. https://www.doi.org/10.1007/s00122-016-2835-7

Sonah H, Bastien M, Iquira E, et al (2013) An improved genotyping by sequencing (GBS) approach offering increased versatility and efficiency of SNP discovery and genotyping. PLoS ONE 8:e54603. https://www.doi.org/10.1371/journal.pone.0054603

Van Ooijen JW (2006) JoinMap® 4.0: software for the calculation of genetic linkage maps in experimental populations. Kyazma, B.V., Wageningen

Van Ooijen JW, Jansen J (2013) Genetic mapping in experimental populations. Cambridge University Press, New York

VanBuren R, Bryant D, Bushakra JM, et al (2016) The genome of black raspberry (Rubus occidentalis). Plant J 87:535–547. https://www.doi.org/10.1111/tpj.13215

VanBuren R, Wai CM, Colle M, et al (2018) A near complete, chromosome-scale assembly of the black raspberry (Rubus occidentalis) genome. GigaScience 7:1– 9. https://www.doi.org/10.1093/gigascience/giy094

Voillemot M, Pannell JR (2017) Inbreeding depression is high in a self-incompatible perennial herb population but absent in a self-compatible population showing mixed mating. Ecol Evol 7:8535–8544. https://www.doi.org/10.1002/ece3.3354

Vool E, Karp K, Noormets M, et al (2009) The productivity and fruit quality of the arctic bramble (Rubus arcticus ssp. arcticus) and hybrid arctic bramble (Rubus arcticus ssp. arcticus × Rubus arcticus ssp. stellatus). Acta Agr Scand Sect B 59:217– 224. https://www.doi.org/10.1080/09064710802022903

Voorrips RE (2002) MapChart: software for the graphical presentation of linkage maps and QTLs. J Hered 93:77–78. https://www.doi.org/10.1093/jhered/93.1.77

VSN International (2018) GenStat for Windows 19th Edition. VSN International Ltd, Hemel Hempstead

Weber CA (2003) Genetic diversity in black raspberry detected by RAPD markers. HortScience 38:4

Yang J, Benyamin B, McEvoy BP, et al (2010) Common SNPs explain a large proportion of the heritability for human height. Nature Genet 42:565–569. https://www.doi.org/10.1038/ng.608

Zorrilla-Fontanesi Y, Cabeza A, Domínguez P, et al (2011) Quantitative trait loci and underlying candidate genes controlling agronomical and fruit quality traits in octoploid strawberry (Fragaria × ananassa). Theor Appl Genet 123:755–778. https://www.doi.org/10.1007/s00122-011-1624-6

123 Figures

Chromosome 1 Chromosome 2 Chromosome 3

● ● ●● ●● ●● ● ●●●●●● ●● ●●●●●● ●●● ●● ●●● ●●●●● ●●● ● ● ● ●● ●● ● ●● ●●●● ● 150 ●● ●● ● 150 ●●● ● ● ●●●● ●● 100 ● ● ●● ●● ●● ● ● ●●●● ●● ●●● ●● ●● ●● ● ● ●●● ● ● ●● ●●● ● ● ● ●

100 ● ● ● 100 ● ● ● ●● ●● ●● ● 60 ● ● ● ●● ●●● ●● ● ●●● ●●● ●●● ●● ●●● ●●●● ● ●● ● ●● ● ● ● ●● ●●● 50 ● ● 50 ● ●● ● ●● ● ●● ● ● ●

20 ● ●● ●● ●● ●●●● ●●● ●● ● ●●● 0 ● 0 ● 0 ●

0 5 10 15 20 25 0 10 20 30 40 10 20 30 40

Chromosome 4 Chromosome 5 Chromosome 6

●● ● ●● ● ●● 200 ●●●●● ● ● ●● ●●● ●● ●●● ● ●● ●● ●●●● ●● ● ●● ●●● ●● 150 ● ● ● ●● ● ● ●●● ●● ●●● ● ● ● ●●● 150 ●●●● ● ●●● ● ● 150 ● ●●●●●● ● ●●● ●● ●●● ● ● ●● ●● ●● ●● ● ● ●●● ●● ●●● ● ● ●●●● ●●● ●● ●●● ● ● ● ● 100 ●● ● ●● ● 100 ● ● ● 100 ● ● ●● ● ● ●● ●●●● ●● ●●● ●●●● ●● ●● ●●●● ●● ●●● ●● ●● ●●● ●● 50 ● ● ●● 50

● 50 ● ● ●● ●● ●●● ●● ● ●●● ●●● ●● ● ● ●● ●● ●● ● ●● ●● ●● ●● ● ● ●● 0 0

● 0 ● ● Genetic position (cM) 0 10 20 30 0 10 20 30 40 0 10 20 30 40 50

Chromosome 7

●●● 200 ●● ●● ● ● ●●● ●● ●● ●● ●●● 150 ●●● ●● ● ●●● ●●●●● ●●●● ●●● ●● ●● ● 100 ●●● ●●● ●●●●● ●●●●● ●●●●●●● ●●● ●●●● 50 ●● ●● ●● ●● ●●● ● 0 ●

0 10 20 30 40

Physical position (Mbp)

Figure 3.1. Genetic position plotted against physical genomic position for 974 combined SNP and SSR markers. A reduced slope and gap between groups of markers suggest reduced recombination and few markers near the centromere. Discontinuous portions in chromosomes 1, 4, and 6 suggest misalignments within either the linkage maps or the current genome assembly.

124

Fruit Mass

NC_13 NC_14 15 10 5 5 0 0

0.0 0.5 1.0 1.5 2.0 0.5 1.0 1.5 2.0

NY_13 NY_14 NY_15 8 25 10 4 5 10 0 0 0

0.5 1.5 2.5 3.5 0.5 1.0 1.5 2.0 1.0 1.5 2.0 2.5 3.0

OH_13 OH_14 OH_15 Frequency (%) 12 20 8 10 10 5 4 0 0 0

1.0 1.5 2.0 2.5 3.0 1.0 1.4 1.8 2.2 1.2 1.4 1.6 1.8 2.0 2.2

OR_14 OR_15 20 15 10 5 0 0

1.0 1.5 2.0 2.5 0.8 1.2 1.6 2.0

Mass (g)

Figure 3.2. Distributions of average fruit mass observed in ten location-by-year environments.

125 Seed Mass

NC_13 NC_14 25 10 5 10 0 0

0.0 0.5 1.0 1.5 2.0 2.5 1.5 2.0 2.5

NY_13 NY_14 NY_15 14 20 20 8 10 10 4 0 0 0

1.6 1.8 2.0 2.2 2.4 1.2 1.6 2.0 2.4 1.0 1.5 2.0

OH_13 OH_14 OH_15 Frequency (%) 14 20 15 8 10 4 5 0 0 0

1.2 1.4 1.6 1.8 2.0 1.6 1.8 2.0 2.2 1.4 1.6 1.8 2.0

OR_14 OR_15 20 8 10 4 0 0

1.8 2.0 2.2 2.4 1.4 1.6 1.8 2.0 2.2

Mass (mg)

Figure 3.3. Distributions of average seed mass observed in ten location-by-year environments.

126 Drupelet Count

NC_13 NC_14 12 15 8 4 5 0 0

0 20 40 60 80 20 40 60 80 100

NY_13 NY_14 NY_15 14 15 10 8 5 5 4 0 0 0

40 60 80 100 40 60 80 100 60 80 100 120 140

OH_13 OH_14 OH_15 Frequency (%) 20 10 10 10 5 5 0 0 0

40 60 80 100 120 40 60 80 100 40 60 80 100 120

OR_14 OR_15 15 10 5 5 0 0

40 60 80 100 120 40 60 80 100

Drupelets per Fruit

Figure 3.4. Distributions of average drupelet count observed in ten location-by-year environments.

127 Seed Fraction

NC_13 NC_14 12 20 8 10 4 0 0

0 2 4 6 8 6 7 8 9

NY_13 NY_14 NY_15 20 15 15 10 5 5 0 0 0

4 6 8 10 12 14 10 15 20 25 6 8 10 12 14

OH_13 OH_14 OH_15 Frequency (%) 15 15 25 5 5 10 0 0 0

4 5 6 7 8 9 5 6 7 8 9 10 12 5.5 6.5 7.5 8.5

OR_14 OR_15 20 8 4 10 0 0

6 7 8 9 10 11 6 7 8 9 10 12

% by Mass

Figure 3.5. Distributions of seed fraction observed in ten location-by-year environments. Seed fraction was calculated as total seed mass / total fruit mass × 100%.

128 Soluble Solid Content

NC_13 NC_14 14 20 8 10 4 0 0

5 6 7 8 9 10 12 8 10 12 14 16

NY_13 NY_14 NY_15 20 30 25 10 15 10 0 0 0

5 10 15 20 8 10 12 14 16 18 20 6 8 10 12 14 16 18

OH_13 OH_14 OH_15 Frequency (%) 15 15 15 5 5 5 0 0 0

9.0 10.0 11.0 12.0 11 12 13 14 10 11 12 13 14

OR_13 OR_14 OR_15 20 20 25 10 10 10 0 0 0

8 9 10 11 12 8 10 12 14 16 10 14 18 22

% Brix

Figure 3.6. Distributions of soluble solid content observed in eleven location-by-year environments.

129 Titratable Acidity

NC_13 NC_14 20 15 10 5 0 0

0.4 0.8 1.2 1.6 0.4 0.8 1.2 1.6

NY_13 NY_14 NY_15 12 12 12 8 8 8 4 4 4 0 0 0

0.6 0.8 1.0 1.2 1.4 0.6 0.8 1.0 1.2 1.4 1.6 0.5 0.7 0.9 1.1

OH_13 OH_14 OH_15 Frequency (%) 12 20 15 8 10 4 5 0 0 0

0.8 1.0 1.2 1.4 1.6 0.7 0.9 1.1 1.3 0.8 1.0 1.2 1.4 1.6

OR_13 OR_14 OR_15 8 15 15 4 5 5 0 0 0

0.4 0.8 1.2 1.6 0.8 1.0 1.2 1.4 1.6 0.6 0.8 1.0 1.2 1.4

% citric acid

Figure 3.7. Distributions of titratable acidity observed in eleven location-by-year environments.

130 Anthocyanin Content

NC_13 NC_14 15 10 5 5 0 0

2000 3000 4000 5000 2000 3000 4000 5000

NY_13 NY_14 NY_15 25 15 10 5 10 5 0 0 0

2000 4000 6000 1000 3000 5000 7000 1500 2500 3500 4500

OH_13 OH_14 OH_15 Frequency (%) 20 25 15 10 10 5 0 0 0

2000 3000 4000 5000 2000 3000 4000 5000 1500 2500 3500 4500

OR_13 OR_14 OR_15 12 15 15 8 4 5 5 0 0 0

4000 6000 8000 1000 3000 5000 2000 4000 6000

mg/kg C3G

Figure 3.8. Distributions of anthocyanin content observed in eleven location-by-year environments. C3G=cyanadin 3-glucoside.

131 Phenolics Content

NC_13 NC_14 12 20 8 10 4 0 0

2500 3000 3500 4000 2500 3500 4500 5500

NY_13 NY_14 NY_15 14 20 20 8 10 10 4 0 0 0

3000 4000 4000 4500 5000 3000 4000 5000 6000

OH_13 OH_14 OH_15 Frequency (%) 20 25 20 10 10 10 0 0 0

2400 2800 3200 3600 2600 3000 3400 3800 3000 4000 5000 6000

OR_13 OR_14 OR_15 14 20 10 8 10 5 4 0 0 0

2800 3200 3600 4000 3000 3400 3800 4200 3000 4000 5000 6000 7000

mg/kg GAE

Figure 3.9. Distributions of phenolics content observed in eleven location-by-year environments. GAE=gallic acid equivalents.

132 Fruit Mass

qRoc-FrM1.1 qRoc-FrM1.2 qRoc-FrM2.1 qRoc-FrM6.1

Figure 3.10. Additive ORUS 3021-2 effects of four fruit mass QTL on linkage groups 1, 2 and 6. Positive or negative values indicate origin of the high value allele from either NC 84-10-3 or cv. Jewel, in no particular order. Heterogeneous effects indicate QTL-by- environment interaction.

133 Drupelet Count

qRoc-DCt1.1 qRoc-DCt1.2 qRoc-DCt4.1

Figure 3.11. Additive ORUS 3021-2 effects of three drupelet count QTL on linkage groups 1 and 4. Positive or negative values indicate origin of the high value allele from either NC 84-10-3 or cv. Jewel, in no particular order. Heterogeneous effects indicate QTL-by-environment interaction.

134 Seed Fraction

qRoc-SdF2.1 qRoc-SdF6.1 qRoc-SdF7.1

Figure 3.12. Additive ORUS 3021-2 effects of three seed fraction QTL on linkage groups 2, 6, and 7. Positive or negative values indicate origin of the high value allele from either NC 84-10-3 or cv. Jewel, in no particular order. Heterogeneous effects indicate QTL-by-environment interaction.

135 Titratable Acidity

qRoc-TAc3.1 qRoc-TAc4.1 qRoc-TAc6.1

Figure 3.13. Additive ORUS 3021-2 effects of three titratable acidity QTL on linkage groups 3, 4, and 6. Positive or negative values indicate origin of the high value allele from either NC 84-10-3 or cv. Jewel, in no particular order. Heterogeneous effects indicate QTL-by-environment interaction.

136 Anthocyanin Content

qRoc-AnC2.1 qRoc-AnC3.1

Figure 3.14. Additive ORUS 3021-2 effects of two anthocyanin content QTL on linkage groups 2 and 3. Positive or negative values indicate origin of the high value allele from either NC 84-10-3 or cv. Jewel, in no particular order. Heterogeneous effects indicate QTL-by-environment interaction.

137 Roc 1

0 5 10 15 20

25 DCt1.1 NY_14 DCt1.1 NY_13 FrM1.1 NC_13 FrM1.1 DCt1.1 OH_15 DCt1.1 NY_15 DCt1.1 OH_14 DCt1.1 NC_13 30 DCt1.1 OR_15 FrM1.1 OH_13 FrM1.1 DCt1.1 OH_13

35 DCt1.1 OR_14 40 OR_15 FrM1.1 45 50 55

60 FrM1.1 OH_14

65 DCt1.1 NC_14 FrM1.2 NC_14 FrM1.2

70 OR_15 FrM1.2 75 80

85 DCt1.2 OH_15

90 DCt1.2 NC_13 FrM1.2 NC_13 95

100 DCt1.2 NC_14 105 110 115 120 125

Figure 3.15. 95% Bayes credible intervals for linkage group 1 QTL. FrM = fruit mass; DCt = drupelet count.

Roc 2 FrM2.1 NC_14 FrM2.1 0 10 SdF2.1 OR_14 SdF2.1 OR_15 SdF2.1 OH_14 20 SdF2.1 NC_13

30 FrM2.1 OH_14 40 50

60 SdF2.1 NC_14 SdF2.1 NY_15 AnC2.1 OR_13 70 NC_13 FrM2.1 SdF2.1 OH_15 80 AnC2.1 OH_15 90 100

110 AnC2.1 OH_14 120 130 140 150 160 170

Figure 3.16. 95% Bayes credible intervals for linkage group 2 QTL. FrM = fruit mass; SdF = seed fraction; AnC = anthocyanin content.

138 Roc 3 TAc3.1 OR_13 TAc3.1 TAc3.1 NC_13 TAc3.1

0 NC_14 TAc3.1 10 TAc3.1 OH_13 TAc3.1 20

30 NY_13 TAc3.1 TAc3.1 OH_15 TAc3.1 40 50 TAc3.1 OR_14 TAc3.1

60 OR_15 TAc3.1

70 OH_14 TAc3.1 80 90

100 AnC3.1 NC_14 AnC3.1 OH_15

110 AnC3.1 OH_14 120 130 140 AnC3.1 OR_15 150 160 170

Figure 3.17. 95% Bayes credible intervals for linkage group 3 QTL. TAc = titratable acidity; AnC = anthocyanin content.

Roc 4 DCt4.1 NY_15

0 DCt4.1 OR_15 DCt4.1 NY_14 DCt4.1 OH_15 10 DCt4.1 OH_13 20 DCt4.1 OH_14

30 NY_13 TAc4.1 40 50 60

70 DCt4.1 NC_13 80

90 OH_15 TAc4.1 100 110 120 130 140 150 160 170

Figure 3.18. 95% Bayes credible intervals for linkage group 4 QTL. DCt = drupelet count; TAc = titratable acidity.

139 Roc 6 Roc 7 SdF6.1 OH_14 SdF6.1 OH_15 0 FrM6.1 NY_15 SdF6.1 NY_15 SdF6.1 OH_13 10 SdF6.1 NC_14 20 30 40 SdF7.1 OH_14 50 OH_15 TAc6.1 FrM6.1 OH_13 FrM6.1 60 FrM6.1 OH_14 FrM6.1 NC_13 FrM6.1 70 SdF6.1 OR_15 80 90

100 OR_15 TAc6.1 110 120

130 SdF7.1 OH_15 140

150 SdF7.1 NC_14 160 170 180 190

Figure 3.19. 95% Bayes credible intervals for linkage group 6 and 7 QTL. FrM = fruit mass; SdF = seed fraction; TAc = titratable acidity.

140 Tables

Table 3.1. SNP filtering criteria and counts for ORUS 4304 and ORUS 4305 linkage maps. 4304 4305 Minimum sequencing depth (per individual) 6 8 Maximum missing allele frequency 0.80 0.90 Minimum minor allele frequency 0.01 0.01 Number of individuals 140 109 Number of sites post-filtering 5224 4367

Table 3.2. ORUS 4305 linkage map marker counts and genetic lengths of seven linkage groups representing the seven chromosomes of the black raspberry genome. Linkage maps were constructed using JoinMap® 4.1. Linkage group Number of Length (cM) markers 1 96 127.8 2 146 174.6 3 151 176.2 4 123 174.7 5 136 186.8 6 151 194.2 7 171 196.4 Total 973 1230.7

Table 3.3. Number of ORUS 4305 individuals with phenotypic data for fruit size (FrM, SdM, DCt, SdF) and fruit chemistry (TAc, AnC, PhC) within trial environments. In NC_13, fruit size data were available for 57 individuals and fruit chemistry data were available for 51 of 103 individuals used to construct linkage maps. Among all environments, fruit size data were available for 84 individuals, and fruit chemistry data were available for 80 individuals of 103 individuals used to construct linkage maps.

OR_14 OR_15 All NY_13 NC_13 NC_14 NY_14 NY_15 OH_14 OH_15 OH_13 OR_13 Size 57 55 55 59 55 62 64 67 - 51 71 84 Chemistry 51 47 56 55 47 60 66 66 55 52 70 80

141 Table 3.4. Summary of QTL identified by single trait, multiple-environment analysis. FrM = fruit mass; SdM = seed mass; DCt = drupelet count; SdF = seed fraction; SSC = soluble solid content; TAc = titratable acidity; AnC = anthocyanin content; PhC = phenolics content. Trait # of QTL QTL Name LG Position Marker name -log10(P)

FrM 4 qRoc-FrM1.1 1 41.7 SRO01_5349433 9.17 qRoc-FrM1.2 1 106.5 SRO01_23458156 14.00

qRoc-FrM2.1 2 0 SRO02_329222 3.45

qRoc-FrM6.1 6 34.4 SRO06_21156325 8.64 SdM 0 - - - - DCt 3 qRoc-DCt1.1 1 41.7 SRO01_5349433 42.94

qRoc-DCt1.2 1 104.6 SRO01_23058326 16.24 qRoc-DCt4.1 4 3.2 SRO04_451654 7.93 SdF 3 qRoc-SdF2.1 2 46.6 SRO02_3650689 11.59

qRoc-SdF6.1 6 16.0 SRO06_251781 11.67 qRoc-SdF7.1 7 120.4 SRO07_28298811 6.96 SSC 0 - - - - TAc 3 qRoc-TAc3.1 3 6.6 Ri11795_SSR 6.06

qRoc-TAc4.1 4 70.6 Ro_CBEa0004G23 3.78

qRoc-TAc6.1 6 78.3 SRO06_13350974 3.93 AnC 2 qRoc-AnC2.1 2 124.8 SRO02_13181188 1.96

qRoc-AnC3.1 3 141.9 SRO03_38071080 4.17 PhC 0 - - - -

142 Table 3.5. Summary of QTL additive and dominance effects by trial environment. Table entries are ordered by trait, position, then environment. R2 is given as the change in % variance when that QTL is dropped from the full model. Env = location-by-year environment; LG = linkage group; LOD = log of odds; FrM = fruit mass; DCt = drupelet count; SdF= seed fraction; TAc = titratable acidity; AnC = anthocyanin content. QTL Genetic R2 Additive ORUS Additive ORUS Dominance Env LG LOD -log10(P) name position (%) 3021-2 (se) 4153-1 (se) (se) qRoc-FrM1.1 NC_13 1 41.7 6.6 2.31 14.2 -0.068 (0.016) 0.072 (0.015)

qRoc-FrM1.1 OH_13 1 41.7 4.7 1.45 20.3 -0.068 (0.016) 0.072 (0.015)

qRoc-FrM1.1 OH_14 1 41.7 3.4 1.41 12.1 -0.068 (0.016) 0.072 (0.015) qRoc-FrM1.1 OR_15 1 41.7 2.1 1.41 9.9% -0.068 (0.016) 0.072 (0.015)

qRoc-FrM1.2 NC_13 1 106.5 15.4 6.00 49.% 0.252 (0.037) -0.267 (0.037) 0.271 (0.037)

qRoc-FrM1.2 NC_14 1 106.5 2.7 1.91 13.0 0.187 (0.078) -0.199 (0.077) 0.269 (0.080) qRoc-FrM1.2 OR_15 1 106.5 2.1 1.87 9.7 0.078 (0.063) -0.090 (0.062) 0.147 (0.063)

qRoc-FrM2.1 NC_13 2 0 2.3 5.80 4.0 -0.040 (0.042) -0.013 (0.035) -0.044 (0.040)

qRoc-FrM2.1 NC_14 2 0 3.1 1.96 15.4 0.026 (0.048) -0.066 (0.040) -0.163 (0.047) qRoc-FrM2.1 OH_14 2 0 3.4 9.64 12.1 0.071 (0.033) -0.063 (0.027) 0.043 (0.031)

qRoc-FrM6.1 NC_13 6 34.39 2.3 4.70 4.0 -0.069 (0.015) -0.066 (0.014)

qRoc-FrM6.1 NY_15 6 34.39 4.8 7.31 26.6 -0.069 (0.015) -0.066 (0.014) qRoc-FrM6.1 OH_13 6 34.39 3.5 6.37 14.3 -0.069 (0.015) -0.066 (0.014)

qRoc-FrM6.1 OH_14 6 34.39 5.4 9.86 21.0 -0.069 (0.015) -0.066 (0.014)

qRoc-DCt1.1 NC_13 1 41.7 7.7 7.49 20.5 -5.34 (0.73) 7.86 (0.76) 2.04 (0.70)

qRoc-DCt1.1 NC_14 1 41.7 2.9 3.61 19.3 -5.34 (0.73) 7.86 (0.76) 2.04 (0.70) qRoc-DCt1.1 NY_13 1 41.7 12.6 7.75 57.2 -5.34 (0.73) 7.86 (0.76) 2.04 (0.70)

qRoc-DCt1.1 NY_14 1 41.7 6.3 11.79 31.8 -5.34 (0.73) 7.86 (0.76) 2.04 (0.70)

qRoc-DCt1.1 NY_15 1 41.7 9.7 1.37 46.8 -5.34 (0.73) 7.86 (0.76) 2.04 (0.70) qRoc-DCt1.1 OH_13 1 41.7 8.3 3.03 36.2 -5.34 (0.73) 7.86 (0.76) 2.04 (0.70)

qRoc-DCt1.1 OH_14 1 41.7 12.5 1.60 51.2 -5.34 (0.73) 7.86 (0.76) 2.04 (0.70)

qRoc-DCt1.1 OH_15 1 41.7 9.6 1.62 34.8 -5.34 (0.73) 7.86 (0.76) 2.04 (0.70) qRoc-DCt1.1 OR_14 1 41.7 5.1 2.73 33.8 -5.34 (0.73) 7.86 (0.76) 2.04 (0.70)

qRoc-DCt1.1 OR_15 1 41.7 9.8 2.87 35.0 -5.34 (0.73) 7.86 (0.76) 2.04 (0.70)

qRoc-DCt1.2 NC_13 1 104.61 15.1 1.83 56.3 11.54 (1.58) -12.45 (1.58) 9.50 (1.59) qRoc-DCt1.2 NC_14 1 104.61 2.1 4.88 13.6 3.90 (4.24) -7.51 (4.19) 1.54 (4.23)

qRoc-DCt1.2 OH_15 1 104.61 4.2 3.13 12.4 -1.09 (2.84) -1.07 (2.88) -5.82 (2.87)

qRoc-DCt4.1 NC_13 4 3.18 2.4 4.58 5.1 1.00 (1.46) 3.30 (1.70) qRoc-DCt4.1 NY_14 4 3.18 2.4 3.22 10.4 4.55 (1.64) -2.04 (1.89)

qRoc-DCt4.1 NY_15 4 3.18 3.9 2.23 14.5 4.35 (1.52) -4.56 (1.76)

qRoc-DCt4.1 OH_13 4 3.18 4.0 1.35 14.7 5.36 (1.58) 2.98 (1.80) qRoc-DCt4.1 OH_14 4 3.18 2.7 11.23 7.5 3.09 (1.20) -2.77 (1.37)

qRoc-DCt4.1 OH_15 4 3.18 6.4 1.64 20.8 6.12 (1.33) 1.30 (1.51)

qRoc-DCt4.1 OR_15 4 3.18 4.2 1.33 12.5 4.56 (1.13) 3.43 (1.30)

143 QTL Genetic R2 Additive ORUS Additive ORUS Dominance Env LOD -log10(P) name position (%) 3021-2 (se) 4153-1 (se) (se) qRoc-SdF2.1 NC_13 2 46.62 4.3 1.36 23.8 -0.24 (0.14) -0.58 (0.14) qRoc-SdF2.1 NC_14 2 46.62 2.8 1.95 14.2 -0.09 (0.10) -0.33 (0.10) qRoc-SdF2.1 NY_15 2 46.62 2.2 2.23 14.5 0.09 (0.15) -0.34 (0.15) qRoc-SdF2.1 OH_14 2 46.62 2.2 1.36 8.2 0.08 (0.08) -0.29 (0.08) qRoc-SdF2.1 OH_15 2 46.62 2.2 3.01 7.5 0.04 (0.06) -0.23 (0.06) qRoc-SdF2.1 OR_14 2 46.62 5.1 2.30 30.2 -0.20 (0.11) -0.59 (0.11) qRoc-SdF2.1 OR_15 2 46.62 6.5 3.79 30.2 0.22 (0.11) -0.66 (0.11) qRoc-SdF6.1 NC_14 6 16.2 5.4 3.08 30.0 0.09 (0.05) 0.37 (0.05) -0.10 (0.05) qRoc-SdF6.1 NY_15 6 16.2 3.4 1.90 24.3 0.09 (0.05) 0.37 (0.05) -0.10 (0.05) qRoc-SdF6.1 OH_13 6 16.2 2.7 1.36 13.8 0.09 (0.05) 0.37 (0.05) -0.10 (0.05) qRoc-SdF6.1 OH_14 6 16.2 6.1 1.45 26.4 0.09 (0.05) 0.37 (0.05) -0.10 (0.05) qRoc-SdF6.1 OH_15 6 16.2 8.3 1.50 34.7 0.09 (0.05) 0.37 (0.05) -0.10 (0.05) qRoc-SdF6.1 OR_15 6 16.2 2.0 3.62 7.9 0.09 (0.05) 0.37 (0.05) -0.10 (0.05) qRoc-SdF7.1 NC_14 7 120.43 2.4 4.96 11.5 0.20 (0.10) -0.08 (0.10) 0.10 (0.10) qRoc-SdF7.1 OH_14 7 120.43 2.9 3.89 10.9 -0.27 (0.08) -0.19 (0.09) -0.09 (0.09) qRoc-SdF7.1 OH_15 7 120.43 3.8 2.29 13.4 -0.05 (0.06) -0.15 (0.06) -0.21 (0.07) qRoc-TAc3.1 NC_13 3 6.59 3.7 1.82 24.1 -0.018 (0.025) 0.098 (0.027) qRoc-TAc3.1 NC_14 3 6.59 3.3 4.57 23.5 -0.054 (0.034) 0.109 (0.037) qRoc-TAc3.1 NY_13 3 6.59 4.4 6.45 22.8 -0.022 (0.018) 0.082 (0.019) qRoc-TAc3.1 OH_13 3 6.59 3.8 1.31 21.1 -0.039 (0.023) 0.098 (0.024) qRoc-TAc3.1 OH_14 3 6.59 2.4 1.53 13.5 -0.031 (0.017) 0.047 (0.018) qRoc-TAc3.1 OH_15 3 6.59 2.1 1.97 7.2 -0.047 (0.015) 0.009 (0.016) qRoc-TAc3.1 OR_13 3 6.59 3.5 2.72 19.3 -0.016 (0.029) 0.095 (0.032) qRoc-TAc3.1 OR_14 3 6.59 3.0 2.51 18.7 -0.022 (0.023) 0.083 (0.025) qRoc-TAc3.1 OR_15 3 6.59 2.8 2.18 11.7 -0.004 (0.017) 0.048 (0.018) qRoc-TAc4.1 NY_13 4 70.56 2.2 3.11 10.2 0.047 (0.019) 0.048 (0.018) qRoc-TAc4.1 OH_15 4 70.56 3.8 2.69 14.2 -0.046 (0.017) 0.034 (0.015) qRoc-TAc6.1 OH_15 6 78.3 5.1 1.63 19.7 0.042 (0.016) 0.038 (0.015) 0.041 (0.015) qRoc-TAc6.1 OR_15 6 78.3 4.0 1.37 17.4 0.001 (0.018) 0.036 (0.017) 0.062 (0.017) qRoc-AnC2.1 OH_14 2 124.8 3.1 2.40 12.9 144 (58) 45 (58) qRoc-AnC2.1 OH_15 2 124.8 2.1 1.98 11.8 -60 (83) 6 (83) qRoc-AnC2.1 OR_13 2 124.8 2.1 1.96 14.7 -5 (132) 351 (136) qRoc-AnC3.1 NC_14 3 141.9 2.1 1.40 18.4 -291 (109) 269 (109) qRoc-AnC3.1 OH_14 3 141.9 7.4 2.76 35.8 -201 (60) 108 (61) qRoc-AnC3.1 OH_15 3 141.9 2.6 3.76 15.2 -157 (86) 314 (88) qRoc-AnC3.1 OR_15 3 141.9 2.6 2.91 13.9 -265 (90) 131 (92)

144 Table 3.6. Summary of QTL intervals by trial environment. Multiple environment values were estimated in Genstat® by using a mixed model approach with genetic correlation modeled by a heterogeneous compound symmetry (CShet) variance-covariance matrix. Single environment values were estimated in R/qtl by using the refineqtl() function to optimize the position of the (multiple environment) QTL. Flanking marker positions were not defined for linkage group 6 due to discontinuity between linkage map and genome assembly (Figure 3-2). Table entries are ordered by trait, position, then environment. Env = location-by-year environment; LG = linkage group; LOD = log of odds; BCI = 95% Bayes credible interval; FrM = fruit mass; DCt = drupelet count; SdF= seed fraction; TAc = titratable acidity; AnC = anthocyanin content. Multiple Single environment environment Flanking marker Physical QTL Genetic Genetic Positions length Trait name Env LG position LOD position LOD 95% BCI (cM) (Mbp) (Mbp) FrM qRoc-FrM1.1 NC_13 1 41.7 6.6 41.0 7.4 36 - 48 5.0 6.6 1.6 FrM qRoc-FrM1.1 OH_13 1 41.7 4.7 48.0 7.9 42 - 51 5.3 7.7 2.4 FrM qRoc-FrM1.1 OH_14 1 41.7 3.4 117.0 2.8 18 - 128 2.9 28.4 25.5 FrM qRoc-FrM1.1 OR_15 1 41.7 2.1 41.7 2.0 0 - 105 0.1 23.0 22.9 FrM qRoc-FrM1.2 NC_13 1 106.5 15.4 105.7 15.8 104 - 107 24.5 13.7 10.8 FrM qRoc-FrM1.2 NC_14 1 106.5 2.7 109.0 3.5 40 - 128 5.0 28.4 23.4 FrM qRoc-FrM1.2 OR_15 1 106.5 2.1 104.6 1.8 42 - 128 5.3 28.4 23.1 FrM qRoc-FrM2.1 NC_13 2 0.0 2.3 0.0 1.8 0 - 175 0.3 38.2 37.8 FrM qRoc-FrM2.1 NC_14 2 0.0 3.1 14.0 4.4 0 - 22 0.3 3.5 3.2 FrM qRoc-FrM2.1 OH_14 2 0.0 3.4 56.0 7.3 16 - 84 2.1 10.6 8.5 FrM qRoc-FrM6.1 NC_13 6 34.4 2.3 78.3 3.0 9 - 167 - - - FrM qRoc-FrM6.1 NY_15 6 34.4 4.8 23.7 4.5 0 - 38 - - - FrM qRoc-FrM6.1 OH_13 6 34.4 3.5 18.0 5.0 2 - 148 - - - FrM qRoc-FrM6.1 OH_14 6 34.4 5.4 40.0 5.7 18 - 151 - - - DCt qRoc-DCt1.1 NC_13 1 41.7 7.7 41.7 8.4 41 - 47 5.0 6.3 1.3 DCt qRoc-DCt1.1 NC_14 1 41.7 2.9 81.0 6.0 72 - 83 10.7 11.7 1.0 DCt qRoc-DCt1.1 NY_13 1 41.7 12.6 41.7 11.7 39 - 44 5.0 5.8 0.8 DCt qRoc-DCt1.1 NY_14 1 41.7 6.3 43.0 6.8 24 - 54 2.8 7.7 4.9 DCt qRoc-DCt1.1 NY_15 1 41.7 9.7 42.0 10.1 39 - 46 5.0 6.3 1.3 DCt qRoc-DCt1.1 OH_13 1 41.7 8.3 47.2 10.6 45 - 49 6.3 6.6 0.3 DCt qRoc-DCt1.1 OH_14 1 41.7 12.5 41.7 12.0 40 - 48 5.0 6.6 1.6 DCt qRoc-DCt1.1 OH_15 1 41.7 9.6 40.0 10.1 38 - 47 5.0 6.3 1.3 DCt qRoc-DCt1.1 OR_14 1 41.7 5.1 41.7 5.3 40 - 59 5.0 8.1 3.1 DCt qRoc-DCt1.1 OR_15 1 41.7 9.8 41.7 4.1 41 - 48 5.0 6.6 1.6 DCt qRoc-DCt1.2 NC_13 1 104.6 15.1 104.0 14.9 103 - 106 24.5 23.5 1.1 DCt qRoc-DCt1.2 NC_14 1 104.6 2.1 112.0 5.6 102 - 127 24.5 28.4 3.9 DCt qRoc-DCt1.2 OH_15 1 104.6 4.2 103.0 4.5 91 - 108 12.8 13.7 0.9 DCt qRoc-DCt4.1 NC_13 4 3.2 2.4 47.2 2.6 0 - 174 0.3 24.2 23.9 DCt qRoc-DCt4.1 NY_14 4 3.2 2.4 0.0 3.4 0 - 43 0.3 4.5 4.2 DCt qRoc-DCt4.1 NY_15 4 3.2 3.9 3.2 3.6 0 - 19 0.3 0.7 0.4

145 DCt qRoc-DCt4.1 OH_13 4 3.2 4.0 44.3 6.5 0 - 68 0.3 7.9 7.6 DCt qRoc-DCt4.1 OH_14 4 3.2 2.7 40.0 2.9 0 - 83 0.3 10.1 9.8 DCt qRoc-DCt4.1 OH_15 4 3.2 6.4 36.7 8.1 3 - 43 0.3 4.5 4.2 DCt qRoc-DCt4.1 OR_15 4 3.2 4.2 16.0 4.1 0 - 34 0.3 2.7 2.4 SdF qRoc-SdF2.1 NC_13 2 46.6 4.3 46.6 3.5 20 - 54 2.1 5.2 3.0 SdF qRoc-SdF2.1 NC_14 2 46.6 2.8 70.4 2.5 0 - 157 0.3 35.4 35.1 SdF qRoc-SdF2.1 NY_15 2 46.6 2.2 144.7 2.3 8 - 153 0.5 34.5 34.1 SdF qRoc-SdF2.1 OH_14 2 46.6 2.2 51.0 2.8 9 - 64 0.5 7.2 6.7 SdF qRoc-SdF2.1 OH_15 2 46.6 2.2 29.6 3.0 19 - 165 5.1 36.9 31.8 SdF qRoc-SdF2.1 OR_14 2 46.6 5.1 29.6 6.8 22 - 47 2.1 4.3 2.1 SdF qRoc-SdF2.1 OR_15 2 46.6 6.5 27.0 5.8 23 - 48 2.1 4.3 2.1 SdF qRoc-SdF6.1 NC_14 6 16.2 5.4 31.0 4.9 14 - 48 - - - SdF qRoc-SdF6.1 NY_15 6 16.2 3.4 16.0 3.2 0 - 47 - - - SdF qRoc-SdF6.1 OH_13 6 16.2 2.7 18.0 3.9 2 - 58 - - - SdF qRoc-SdF6.1 OH_14 6 16.2 6.1 14.2 7.2 8 - 20 - - - SdF qRoc-SdF6.1 OH_15 6 16.2 8.3 17.0 9.4 15 - 20 - - - SdF qRoc-SdF6.1 OR_15 6 16.2 2.0 17.0 2.0 3 - 176 - - - SdF qRoc-SdF7.1 NC_14 7 120.4 2.4 153.0 4.0 150 - 196 35.0 41.3 6.3 SdF qRoc-SdF7.1 OH_14 7 120.4 2.9 86.0 4.8 44 - 94 5.7 22.4 16.6 SdF qRoc-SdF7.1 OH_15 7 120.4 3.8 117.0 4.5 112 - 196 28.6 41.3 12.7 TAc qRoc-TAc3.1 NC_13 3 6.6 3.7 9.0 3.1 0 - 32 1.7 8.8 7.0 TAc qRoc-TAc3.1 NC_14 3 6.6 3.3 8.0 3.6 0 - 39 1.7 10.1 8.3 TAc qRoc-TAc3.1 NY_13 3 6.6 4.4 9.3 5.3 5 - 89 2.4 31.2 28.8 TAc qRoc-TAc3.1 OH_13 3 6.6 3.8 7.0 3.8 1 - 67 2.8 26.2 23.5 TAc qRoc-TAc3.1 OH_14 3 6.6 2.4 96.0 3.5 0 - 176 1.7 42.2 40.5 TAc qRoc-TAc3.1 OH_15 3 6.6 2.1 19.8 3.6 11 - 99 5.0 32.8 27.8 TAc qRoc-TAc3.1 OR_13 3 6.6 3.5 13.0 4.3 4 - 22 2.4 5.7 3.3 TAc qRoc-TAc3.1 OR_14 3 6.6 3.0 2.0 3.7 0 - 149 1.7 39.5 37.7 TAc qRoc-TAc3.1 OR_15 3 6.6 2.8 143.0 4.5 0 - 158 1.7 40.5 38.8 TAc qRoc-TAc4.1 NY_13 4 70.6 2.2 64.1 2.2 0 - 94 0.3 11.3 11.0 TAc qRoc-TAc4.1 OH_15 4 70.6 3.8 56.8 5.3 43 - 173 4.1 24.2 20.1 TAc qRoc-TAc6.1 OH_15 6 78.3 5.1 51.2 5.3 46 - 91 - - - TAc qRoc-TAc6.1 OR_15 6 78.3 4.0 57.7 4.5 49 - 194 - - - AnC qRoc-AnC2.1 OH_14 2 124.8 3.1 124.7 3.1 84 - 172 10.6 38.4 27.8 AnC qRoc-AnC2.1 OH_15 2 124.8 2.1 130.0 3.1 57 - 150 6.1 32.6 26.5 AnC qRoc-AnC2.1 OR_13 2 124.8 2.1 139.0 2.4 0 - 174 0.3 38.8 38.5 AnC qRoc-AnC3.1 NC_14 3 141.9 2.1 160.0 3.3 64 - 176 23.6 42.2 18.7 AnC qRoc-AnC3.1 OH_14 3 141.9 7.4 121.0 8.0 112 - 149 34.7 39.5 4.7 AnC qRoc-AnC3.1 OH_15 3 141.9 2.6 128.4 3.8 89 - 166 31.2 41.9 10.7 AnC qRoc-AnC3.1 OR_15 3 141.9 2.6 158.1 4.6 148 - 176 39.5 42.2 2.8

146 Table 3.7. Black raspberry genomic regions with high synteny to known genes influencing organ size, fw2.2 and CNR01. Protein sequences were retrieved from NCBI (https://www.ncbi.nlm.nih.gov). BLAST was performed on the Genome Database for Rosaceae website (https://www.rosaceae.org/) using the Rubus occidentalis v3.0 genome. Start End % Gene query Chromosome position position E-value Bit score identity (bp) (bp) FW2.2 Ro04 69.9 1681057 1681275 1.22E-28 111.0 [Solanum 77.8 1680888 1680941 1.22E-28 36.2 lycopersicum] 57.4 4553237 4553539 9.77E-28 114.0 50.0 4553088 4553153 9.77E-28 30.0 71.2 1678936 1679154 2.61E-25 114.0 51.3 25127826 25128053 3.38E-15 81.5 50.0 9182055 9182270 1.43E-12 72.7 47.3 4546466 4546687 7.36E-12 70.5 48.8 25122572 25122811 1.44E-09 63.0 Ro02 66.7 36908635 36908877 2.71E-25 114.0 Ro05 47.7 3031923 3032243 4.43E-21 100.0 36.1 3039037 3039345 6.79E-10 63.9 CNR01 Ro04 62.5 1681060 1681275 1.57E-23 95.1 [Zea mays] 72.7 1680888 1680953 1.57E-23 35.3 57.3 4553240 4553485 8.01E-20 96.9 59.7 1678939 1679154 1.56E-17 89.4 40.3 9182055 9182270 5.56E-12 55.5 52.8 9182371 9182472 5.56E-12 35.7 50 25127841 25128063 2.02E-08 59.5 43.2 4546466 4546684 2.67E-08 59.1 50.6 25122572 25122793 9.51E-07 53.8 68.2 4546821 4546886 1.67E-05 39.7 62.5 4547004 4547051 1.67E-05 29.1 Ro02 59.7 36908662 36908877 3.63E-19 94.7 Ro05 49.4 3031947 3032189 7.71E-15 80.6 40.4 3039037 3039333 2.20E-08 59.5

147 Concluding Remarks

BR is a high-value specialty crop with great market potential due to its desirable flavor and reported human health benefits. However, fresh market consumption is currently limited by insufficient fruit quality of available cultivars arising from small fruit with highly discernable seeds. Further, both fresh and processing market production are limited by high susceptibility to disease and low tolerance to biotic and abiotic stress.

Renewed BR breeding efforts are combining cultivated and wild germplasm, following recent discovery of novel pest resistance and genetic diversity in wild BR accessions.

New BR cultivar development will need to target multiple traits related to both flavor and human health. These include morphological and biochemical traits, such as fruit size, seediness, soluble solid content and anthocyanin content.

Agricultural breeding involves identifying and consolidating genetic components influencing economically important traits. Genetic components are prioritized by their contribution to each trait as well as the stability of that contribution within target production environments. Past BR breeding efforts found limited success, owing to a reported lack of genetic diversity within elite germplasm. Current BR breeding approaches, which rely on combined elite and wild germplasm, may utilize recently available technologies, including next-generation sequencing, genome assemblies, and advanced statistical techniques. These newer technologies provide previously unavailable insights into BR genetics and may benefit forward selection and gene discovery.

The current study presents an assessment of genetic components and stability for several BR fruit quality traits derived from a multiple-environment trial of breeding

148 germplasm spanning four U.S. production regions. This assessment utilized genomic sequencing data, advanced statistical techniques, and a chromosome-scale BR genome assembly.

Genetic components contributing to fruit size and seediness were found to be highly heritable and fairly stable across environments. Further attention should be paid to the improvement of these traits (i.e. increased fruit size and reduced seediness), as they will be important contributors to any future success of the fresh market BR industry.

Genetic components contributing to fruit chemistry traits are also identified, though these components show low stability over U.S. production regions. Multiple acidity and anthocyanin content QTL were identified, confirming genetic variation for these traits are available.

Examination of QTL in this study indicate several QEI. The most striking of these occur between BR plants grown in NC and those grown in the other three research locations. This finding conforms with reported challenges to NC raspberry production generally attributed to high summer temperatures. If cultivars adapted to NC growing conditions are desired, they are unlikely to be developed in OH, NY, or OR. Further examination of QTL exhibiting interactions between NC and other trial locations is warranted to further explore the genetic basis of these interactions and any associations with environmental stress tolerance.

In addition to genetic characterization of fruit quality traits, successful domestication of BR will require improved disease and stress mitigation strategies, such as genetic resistance to soil-borne diseases and tolerance to abiotic stressors including high heat during anthesis and fruit development. These traits may be considered in

149 future BR studies. Assessments of consumer preferences and physiological bases for fruit quality traits such as mass and seediness would also be valuable in determining fresh market breeding goals and strategies. This study confirms genetic variation for these traits exists; however, acceptable breeding targets for these traits have not been established.

Though numerous significant findings are reported in this study, statistical power of this work was severely limited by several features of the available data. These include limited field observations, partially arising from mistaken population structure.

This highlights the importance population structure confirmation through genetic marker analysis as a quality control measure in experimental population studies.

Missing genetic marker data resulted from low sequencing depth and uneven distribution of sequence depth among study individuals. Future trait mapping efforts in

BR should consider further optimization of GBS protocol, including digestion enzyme selection and increased individual sequencing depth. Further, self-compatibility and low inbreeding depression associated with BR means that future genetic studies may derive experimental populations from highly-inbred individuals, thus reducing complexity of the ensuing analysis.

Next steps arising from this work include validation of QTLs and identification of candidate genes. Polymerase chain reaction (PCR) markers may be developed and used to track and validate the QTL in separate populations. These markers may also be used for marker assisted selection of FrM, DCt, SdM, TAc, and AnC. Comparative genomics may be used to explore homology of BR QTL regions to candidate genes discovered in related plant species; for example, overlap between fruit size QTL and

150 two such genes, FW2.2 and CNR01 are presented in this work. Further, methods described in this study, involving multiple environment trial analysis of an experimental outcrossing population, may be applied to additional outcrossing crop species.

151 References

Allan AC, Hellens RP, Laing WA (2008) MYB transcription factors that colour our fruit. Trends Plant Sci 13:99–102. https://www.doi.org/10.1016/j.tplants.2007.11.012

Altschul SF, Gish W, Miller W, et al (1990) Basic local alignment search tool. J Mol Bio 215:403–410. https://www.doi.org/10.1016/S0022-2836(05)80360-2

Amiot MJ, Riva C, Vinet A (2016) Effects of dietary polyphenols on metabolic syndrome features in humans: a systematic review. Obes Rev 17:573–586. https://www.doi.org/10.1111/obr.12409

Andrews S (2010) FastQC: a quality control tool for high throughput sequence data. Available online at: http://www.bioinformatics.babraham.ac.uk/projects/fastqc

Bates D, Mächler M, Bolker B, Walker S (2015) Fitting linear mixed-effects models using lme4. J Stat Softw 67:1. https://www.doi.org/10.18637/jss.v067.i01

Boer MP, Wright D, Feng L, et al (2007) A mixed-model quantitative trait loci (QTL) analysis for multiple-environment trial data using environmental covariables for QTL-by-environment interactions, with an example in maize. Genetics 177:1801– 1813. https://www.doi.org/10.1534/genetics.107.071068

Bolger AM, Lohse M, Usadel B (2014) Trimmomatic: a flexible trimmer for Illumina sequence data. Bioinformatics 30:2114–2120. https://www.doi.org/10.1093/bioinformatics/btu170

Bradish CM, Bushakra JM, Dossett M, et al (2016a) Genotyping and phenotyping heat tolerance in black raspberry (Rubus occidentalis L.). Acta Hortic 1127:321–324. https://www.doi.org/10.17660/ActaHortic.2016.1127.50

Bradish CM, Fernandez GE, Bushakra JM, et al (2016b) Evaluation of vigor and winter hardiness of black raspberry breeding populations (Rubus) grown in the southeastern US. Acta Hortic 1133:129–134. https://www.doi.org/10.17660/ActaHortic.2016.1133.19

Broman KW, Wu H, Sen S, Churchill GA (2003) R/qtl: QTL mapping in experimental crosses. Bioinformatics 19:889–890. https://www.doi.org/10.1093/bioinformatics/btg112

Burrows C, Moore PP (2002) Genotype × environment effects on raspberry fruit quality. Acta Hortic 585:467–473. https://www.doi.org/10.17660/ActaHortic.2002.585.75

Bushakra JM, Bassil NV, Weiland JE, et al (2016a) Comparative RNA-seq for the investigation of tolerance to Verticillium wilt in black raspberry. Acta Hortic 1133:103–114. https://www.doi.org/10.17660/ActaHortic.2016.1133.16

152 Bushakra JM, Bradish CM, Weber CA, et al (2016b) Toward understanding genotype × environment interactions in black raspberry (Rubus occidentalis L.). Acta Hortic 1117:25–30. https://www.doi.org/10.17660/ActaHortic.2016.1117.5

Bushakra JM, Bryant DW, Dossett M, et al (2015) A genetic linkage map of black raspberry (Rubus occidentalis) and the mapping of Ag 4 conferring resistance to the aphid Amphorophora agathonica. Theor Appl Genet 128:1631–1646. https://www.doi.org/10.1007/s00122-015-2541-x

Bushakra JM, Dossett M, Carter KA, et al (2018) Characterization of aphid resistance loci in black raspberry (Rubus occidentalis L.). Mol Breed 38:83. https://www.doi.org/10.1007/s11032-018-0839-5

Bushakra JM, Krieger C, Deng D, et al (2013) QTL involved in the modification of cyanidin compounds in black and red raspberry fruit. Theor Appl Genet 126:847– 865. https://www.doi.org/10.1007/s00122-012-2022-4

Bushakra JM, Stephens MJ, Atmadjaja AN, et al (2012) Construction of black (Rubus occidentalis) and red (R. idaeus) raspberry linkage maps and their comparison to the genomes of strawberry, apple, and peach. Theor Appl Genet 125:311–327. https://www.doi.org/10.1007/s00122-012-1835-5

Byrne DH (2012) Trends in fruit breeding. In: Fruit breeding. Springer, New York; London, pp 3–36

Cash CB (2016) Studies of linkage mapping, trait heritability, and pedigrees for breeding improvement of southeastern black and red raspberry. North Carolina State University

Chaves-Silva S, Santos AL dos, Chalfun-Júnior A, et al (2018) Understanding the genetic regulation of anthocyanin biosynthesis in plants – Tools for breeding purple varieties of fruits and vegetables. Phytochemistry 153:11–27. https://www.doi.org/10.1016/j.phytochem.2018.05.013

Collard BCY, Jahufer MZZ, Brouwer JB, Pang ECK (2005) An introduction to markers, quantitative trait loci (QTL) mapping and marker-assisted selection for crop improvement: The basic concepts. Euphytica 142:169–196. https://www.doi.org/10.1007/s10681-005-1681-5

Collard BCY, Mackill DJ (2008) Marker-assisted selection: an approach for precision plant breeding in the twenty-first century. Philos Trans Royal Soc B 363:557–572

Cullis BR, Jefferson P, Thompson R, Smith AB (2014) Factor analytic and reduced animal models for the investigation of additive genotype-by-environment interaction in outcrossing plant species with application to a Pinus radiata breeding programme. Theor Appl Genet 127:2193–2210

153 Danecek P, Auton A, Abecasis G, et al (2011) The variant call format and VCFtools. Bioinformatics 27:2156–2158. https://www.doi.org/10.1093/bioinformatics/btr330

Darrow GM, Sherwood H (1931) Seed and berry size of cane fruits. Proc Amer Soc Hort Sci 28:194–199

Daubeny HA (1996) Brambles. In: Janick J, Moore JN (eds) Fruit Breeding, Vine and Small Fruits. Wiley, New York, pp 109–190

De Franceschi P, Stegmeir T, Cabrera A, et al (2013) Cell number regulator genes in Prunus provide candidate genes for the control of fruit size in sweet and sour cherry. Mol Breed 32:311–326. https://www.doi.org/10.1007/s11032-013-9872-6

Des Marais DL, Hernandez KM, Juenger TE (2013) Genotype-by-environment interaction and plasticity: exploring genomic responses of plants to the abiotic environment. Annu Rev Ecol Evol S 44:5–29. https://www.doi.org/10.1146/annurev-ecolsys-110512-135806

Dobson P, Graham J, Stewart D, et al (2012) Over-seasons analysis of quantitative trait loci affecting phenolic content and antioxidant capacity in raspberry. J Agric Food Chem 60:5360–5366. https://www.doi.org/10.1021/jf3005178

Dossett M, Bassil NV, Lewers KS, Finn CE (2012a) Genetic diversity in wild and cultivated black raspberry (Rubus occidentalis L.) evaluated by simple sequence repeat markers. Genet Resour Crop Evol 59:1849–1865. https://www.doi.org/10.1007/s10722-012-9808-8

Dossett M, Finn CE (2011) Primocane-fruiting in black raspberry (Rubus occidentalis L.). J Am Pomol Soc 65:56–61

Dossett M, Finn CE (2010) Identification of resistance to the large raspberry aphid in black raspberry. J Am Soc Hortic Sci 135:438–444

Dossett M, Lee J, Finn CE (2012b) Anthocyanin content of wild black raspberry germplasm. Acta Hortic 946:43–48. https://www.doi.org/10.17660/ActaHortic.2012.946.3

Dossett M, Lee J, Finn CE (2008) Inheritance of phenological, vegetative, and fruit chemistry traits in black raspberry. J Am Soc Hortic Sci 133:408–417

Dossett M, Lee J, Finn CE (2010) Variation in anthocyanins and total phenolics of black raspberry populations. J Funct Foods 2:292–297. https://www.doi.org/10.1016/j.jff.2010.10.004

Dossett M, Lee J, Finn CE (2011) Characterization of a novel anthocyanin profile in wild black raspberry mutants: An opportunity for studying the genetic control of pigment and color. J Funct Foods 3:207–214. https://www.doi.org/10.1016/j.jff.2011.04.003

154 Drain BD (1953) Some inheritance data with black raspberries. Proc Amer Soc Hort Sci 61:231–234

Drain BD (1956) Inheritance in black raspberry species. Proc Amer Soc Hort Sci 68:169–170

Elshire RJ, Glaubitz JC, Sun Q, et al (2011) A robust, simple genotyping-by-sequencing (GBS) approach for high diversity species. PLoS ONE 6:e19379. https://www.doi.org/10.1371/journal.pone.0019379

Etienne A, Génard M, Lobit P, et al (2013) What controls fleshy fruit acidity? A review of malate and citrate accumulation in fruit cells. J Exp Biol 64:1451–1469. https://www.doi.org/10.1093/jxb/ert035

Etienne C, Rothan C, Moing A, et al (2002) Candidate genes and QTLs for sugar and organic acid content in peach [Prunus persica (L.) Batsch]. Theor Appl Genet 105:145–159. https://www.doi.org/10.1007/s00122-001-0841-9

Falconer DS, Mackay TFC (1996) Introduction to quantitative genetics, 4th ed. Burnt Mill, England: Longman

Fang J (2015) Classification of fruits based on anthocyanin types and relevance to their health effects. Nutrition 31:1301–1306. https://www.doi.org/10.1016/j.nut.2015.04.015

Finlay KW, Wilkinson GN (1963) The analysis of adaptation in a plant-breeding programme. Aust J Agric Res 14:742–754

Finn C, Knight VH (2002) What’s going on in the world of Rubus breeding? Acta Hortic 585:31–38. https://www.doi.org/10.17660/ActaHortic.2002.585.1

Finn C, Wennstrom K, Link J, Ridout J (2003) Evaluation of Rubus leucodermis populations from the Pacific Northwest. HortScience 38:1169–1172

Fisher RA (1918) The correlation between relatives on the supposition of Mendelian inheritance. Trans R Soc, Edinburgh 52:399–433

Fisher RA (1925) Statistical methods for research workers, 1st edn. Oliver and Boyd, Edinburgh and London

Frary A (2000) fw2.2: a quantitative trait locus key to the evolution of tomato fruit size. Science 289:85–88. https://www.doi.org/10.1126/science.289.5476.85

Fresnedo-Ramírez J, Bink MCAM, van de Weg E, et al (2015) QTL mapping of pomological traits in peach and related species breeding germplasm. Mol Breed 35:166. https://www.doi.org/10.1007/s11032-015-0357-7

155 Fresnedo-Ramírez J, Yang S, Sun Q, et al (2017) An integrative AmpSeq platform for highly multiplexed marker-assisted pyramiding of grapevine powdery mildew resistance loci. Mol Breed 37:145 https://doi.org/10.1007/s11032-017-0739-0

Fresnedo-Ramírez J, Yang S, Sun Q, et al (2019) Computational analysis of AmpSeq data for targeted, high-throughput genotyping of amplicons. Front Plant Sci. 10:599. https://www.doi.org/10.3389/fpls.2019.00599

Funt RC, Ross DS (2013) Soil and water management. In: Raspberries. CABI, pp 103– 120

Gabriel KR (1971) The biplot graphic display of matrices with application to principal component analysis. Biometrika 58:453–467

Gabriel KR (1978) Least squares approximation of matrices by additive and multiplicative models. J R Stat Soc B 40:186–196

Galleta GJ, Maas JL, Enns JM (1998) “Earlysweet” black raspberry. Fruit Var J 52:123– 124

Gao G, Draper E, Lewandowski R (2017) Raspberries for the home fruit planting. https://ohioline.osu.edu/factsheet/hyg-1421.

Gauch HG (1988) Model selection and validation for yield trials with interaction. Biometrics 44:705–715. https://www.doi.org/10.2307/2531585

Gauss KF (1809) Theoria motus corporum celestrium in sectionibus conics solem ambientium. Perthes and Besser: Hamburg

Gilmour AR, Gogel B, Cullis BR, et al (2015) ASReml User Guide Release 4.1. VSN International Ltd, Hemel Hempstead

Glaubitz JC, Casstevens TM, Lu F, et al (2014) TASSEL-GBS: a high capacity genotyping by sequencing analysis pipeline. PLoS ONE 9:2. https://www.doi.org/10.1371/journal.pone.0090346

Gollob HF (1968) A statistical model which combines features of factor analytic and analysis of variance techniques. Psychometrika 33:73–115. https://www.doi.org/10.1007/BF02289676

Graham J, Brennan R (eds) (2018) Raspberries: breeding, challenges, and advances. Springer Science+Business Media, New York, NY

Graham J, Hackett CA, Smith K, et al (2011) Towards an understanding of the nature of resistance to Phytophthora root rot in red raspberry. Theor Appl Genet 123:585– 601. https://www.doi.org/10.1007/s00122-011-1609-5

156 Graham J, Hackett CA, Smith K, et al (2009) Mapping QTLs for developmental traits in raspberry from bud break to ripe fruit. Theor Appl Genet 118:1143–1155. https://www.doi.org/10.1007/s00122-009-0969-6

Graham J, Jennings N (2008) Raspberry breeding. In: Jain S, Priyadarshan M (eds) Breeding tree crops. IBH & Science Publication Inc, Oxford

Graham J, Smith K, MacKenzie K, et al (2004) The construction of a genetic linkage map of red raspberry (Rubus idaeus subsp. idaeus) based on AFLPs, genomic- SSR and EST-SSR markers. Theor Appl Genet 109:740–749

Graham J, Smith K, Tierney I, et al (2006) Mapping gene H controlling cane pubescence in raspberry and its association with resistance to cane botrytis and spur blight, rust and cane spot. Theor Appl Genet 112:818–831

Griesser M, Hoffmann T, Bellido ML, et al (2008) Redirection of flavonoid biosynthesis through the down-regulation of an anthocyanidin glucosyltransferase in ripening strawberry fruit. Plant Physiol 146:1528–1539. https://www.doi.org/10.1104/pp.107.114280

Guan Y, Peace C, Rudell D, et al (2015) QTLs detected for individual sugars and soluble solids content in apple. Mol Breed 35:135. https://www.doi.org/10.1007/s11032-015-0334-1

Guo M, Rupe MA, Dieter JA, et al (2010) Cell Number Regulator1 affects plant and organ size in maize: implications for crop yield enhancement and heterosis. Plant Cell 22:1057–1073. https://www.doi.org/10.1105/tpc.109.073676

Hackett CA, Milne L, Smith K, et al (2018) Enhancement of Glen Moy x Latham raspberry linkage map using GbS to further understand control of developmental processes leading to fruit ripening. BMC Genetics 19:59. https://www.doi.org/10.1186/s12863-018-0666-z

Halgren A, Tzanetakis IE, Martin RR (2007) Identification, characterization, and detection of black raspberry necrosis virus. Phytopathology 97:44–50. https://www.doi.org/10.1094/PHYTO-97-0044

Hancock RD, Petridis A, McDougall GJ (2018) Raspberry fruit chemistry in relation to fruit quality and human nutrition. In: Raspberries: breeding, challenges, and advances. Springer Science+Business Media, pp 89–120

Hardner C (2017) Exploring opportunities for reducing complexity of genotype-by- environment interaction models. Euphytica 213:248. https://www.doi.org/10.1007/s10681-017-2023-0

Hardner C, Winks C, Stephenson R, Gallagher E (2001) Genetic parameters for nut and kernel traits in macadamia. Euphytica 117:151–161

157 Hardner CM, Hayes BJ, Kumar S, et al (2019) Prediction of genetic value for sweet cherry fruit maturity among environments using a 6K SNP array. Hortic Res 6:6. https://www.doi.org/10.1038/s41438-018-0081-7

Hardwick RC (1981) The analysis of genotype × environment interactions: what does it mean if varietal stability is linearly related to varietal performance? Euphytica 30:217–221

Hedrick UP (1925) The small fruits of New York. J.B. Lyon Company, Albany

Hummer KE, Peacock DN (1994) Seed dimension and weight of selected Rubus species. HortScience 29:1034–1036. https://www.doi.org/10.21273/HORTSCI.29.9.1034

Hyun TK, Lee S, Rim Y, et al (2014) De-novo RNA sequencing and metabolite profiling to identify genes involved in anthocyanin biosynthesis in Korean black raspberry (Rubus coreanus Miquel). PLoS ONE 9:e88292. https://www.doi.org/10.1371/journal.pone.0088292

Isik F, Holland J, Maltecca C (2017) Multienvironmental trials. In: Genetic data analysis for plant and animal breeding. Springer International Publishing

Jennings DL (1988) Raspberries and blackberries: their breeding, diseases and growth. Academic Press, San Diego, CA

Jibran R, Dzierzon H, Bassil N, et al (2018) Chromosome-scale scaffolding of the black raspberry (Rubus occidentalis L.) genome based on chromatin interaction data. Hortic Res 5:8. https://www.doi.org/10.1038/s41438-017-0013-y

Johnson DE, Graybill FA (1972) An analysis of a two-way model with interaction and no replication. J Am Stat Assoc 67:862. https://www.doi.org/10.2307/2284651

Kadir S, Sidhu G, Al-Khatib K (2006) Strawberry (Fragaria × ananassa Duch.) growth and productivity as affected by temperature. HortScience 41:1423–1430. https://www.doi.org/10.21273/HORTSCI.41.6.1423

Karaagac E, Vargas A, Andrés M, et al (2012) Marker assisted selection for seedlessness in table grape breeding. Tree Genet Genomes 8:1003–1015. https://www.doi.org/10.1007/s11295-012-0480-0

Kassim A, Poette J, Paterson A, et al (2009) Environmental and seasonal influences on red raspberry anthocyanin antioxidant contents and identification of quantitative traits loci (QTL). Mol Nutr Food Res 53:625–634. https://www.doi.org/10.1002/mnfr.200800174

Keep E (1989) Breeding red raspberry for resistance to diseases and pest. In: Janick J (ed) Plant Breeding Reviews. Timber Press, Portland, Oregon, pp 245–322

158 Keep E (1984) Breeding Rubus and Ribes crops at East Malling. Sci Hortic 35:54–71

Keep E, Knight RL (1967) A new gene from Rubus occidentalis L. for resistance to strains 1, 2, and 3 of the rubus aphid, Amphorophora rubi Kalt. Euphytica 16:209–214. https://www.doi.org/10.1007/BF00043456

Kempton RA (1984) The use of biplots in interpreting variety by environment interactions. J Agr Sci 103:123. https://www.doi.org/10.1017/S0021859600043392

Khoo HE, Azlan A, Tang ST, Lim SM (2017) Anthocyanidins and anthocyanins: colored pigments as food, pharmaceutical ingredients, and the potential health benefits. Food Nutr Res 61:1361779. https://www.doi.org/10.1080/16546628.2017.1361779

Klee HJ (2010) Improving the flavor of fresh fruits: genomics, biochemistry, and biotechnology: Tansley review. New Phytol 187:44–56. https://www.doi.org/10.1111/j.1469-8137.2010.03281.x

Kresty LA, Mallery SR, Stoner GD (2016) Black raspberries in cancer clinical trials: past, present and future. J Berry Res 6:251–261. https://www.doi.org/10.3233/JBR- 160125

Krüger E, Dietrich H, Schöpplein E, et al (2011) Cultivar, storage conditions and ripening effects on physical and chemical qualities of red raspberry fruit. Postharvest Biol Tec 60:31–37. https://www.doi.org/10.1016/j.postharvbio.2010.12.001

Kuhlman GW, Mumford DC (1949) Cost of producing black raspberries for processing in the Willamette Valley, Oregon. Agricultural Experiment Station, Oregon State College, Corvallis

Lawrence AA, Goldman DH, Macklin JA, Moore G (2014) Rubus. In: Flora of North America Editorial Committee, eds. 1993+. Flora of North America North of Mexico. 20+ vols. New York and Oxford. Vol. 9. On-line version: http://www.efloras.org/florataxon.aspx?flora_id=1&taxon_id=128837

Lee J, Dossett M, Finn CE (2012) Rubus fruit phenolic research: the good, the bad, and the confusing. Food Chem 130:785–796. https://www.doi.org/10.1016/j.foodchem.2011.08.022

Legendre AM (1806) Nouvelles methodes pour la determination des orbites des cometes; avec un supplement contenant divers perfectionnements de ces methodes et leur application aux deux cometes de 1805. Courcier, Paris

Li H, Durbin R (2009) Fast and accurate short read alignment with Burrows–Wheeler transform. Bioinformatics 25:1754–1760. https://www.doi.org/10.1093/bioinformatics/btp324

159 Li H, Hearne S, Bänziger M, et al (2010) Statistical properties of QTL linkage mapping in biparental genetic populations. Heredity 105:257–267. https://www.doi.org/10.1038/hdy.2010.56

Li J, Ji L (2005) Adjusting multiple testing in multilocus analyses using the eigenvalues of a correlation matrix. Heredity 95:221–227. https://www.doi.org/10.1038/sj.hdy.6800717

Li X, Guo T, Mu Q, et al (2018) Genomic and environmental determinants and their interplay underlying phenotypic plasticity. Proc Natl Acad Sci U S A 115:6679– 6684. https://www.doi.org/10.1073/pnas.1718326115

Lian L, de los Campos G (2016) FW: an R package for Finlay–Wilkinson regression that incorporates genomic/pedigree information and covariance structures between environments. G3 6:589–597. https://www.doi.org/10.1534/g3.115.026328

Lightle D, Dossett M, Ebert T, et al (2015) Effects of three novel resistant black raspberry selections on Amphorophora agathonica feeding behavior and performance. Arthropod Plant Interact 9:487–496. https://www.doi.org/10.1007/s11829-015-9390-z

Lodhi MA, Ye G-N, Weeden NF, Reisch BI (1994) A simple and efficient method for DNA extraction from grapevine cultivars and Vitis species. Plant Mol Biol Report 12:6–13. https://www.doi.org/10.1007/BF02668658

Maliepaard C, Jansen J, Van Ooijen JW (1997) Linkage analysis in a full-sib family of an outbreeding plant species: overview and consequences for applications. Genetical Research 70:237–250. https://www.doi.org/10.1017/S0016672397003005

Malosetti M, Ribaut J-M, van Eeuwijk FA (2013) The statistical analysis of multi- environment data: modeling genotype-by-environment interaction and its genetic basis. Front Physiol 4:44. https://www.doi.org/10.3389/fphys.2013.00044

Malosetti M, Ribaut JM, Vargas M, et al (2008) A multi-trait multi-environment QTL mixed model with an application to drought and nitrogen stress trials in maize (Zea mays L.). Euphytica 161:241–257. https://www.doi.org/10.1007/s10681- 007-9594-0

Malowicki SMM, Martin R, Qian MC (2008) Comparison of sugar, acids, and volatile composition in raspberry Bushy Dwarf Virus-resistant transgenic raspberries and the wild type ‘Meeker’ (Rubus Idaeus L.). J Agric Food Chem 56:6648–6655. https://www.doi.org/10.1021/jf800253e

Margarido GRA, Pastina MM, Souza AP, Garcia AAF (2015) Multi-trait multi- environment quantitative trait loci mapping for a sugarcane commercial cross provides insights on the inheritance of important traits. Mol Breed 35:175. https://www.doi.org/10.1007/s11032-015-0366-6

160 Martin RR, MacFarlane S, Sabanadzovic S, et al (2013) Viruses and virus diseases of Rubus. Plant Dis 97:168–182

Mathews KL, Malosetti M, Chapman S, et al (2008) Multi-environment QTL mixed models for drought stress adaptation in wheat. Theor Appl Genet 117:1077– 1091. https://www.doi.org/10.1007/s00122-008-0846-8

Mazur SP, Sønsteby A, Wold A-B., et al (2014a) Post-flowering photoperiod has marked effects on fruit chemical composition in red raspberry (Rubus idaeus). Ann Appl Biol 165:454–465. https://www.doi.org/10.1111/aab.12153

Mazur SP, Nes A, Wold A-B, et al (2014b) Quality and chemical composition of ten red raspberry (Rubus idaeus L.) genotypes during three harvest seasons. Food Chem 160:233–240. https://www.doi.org/10.1016/j.foodchem.2014.02.174

Mazzoni L, Perez-Lopez P, Giampieri F, et al (2016) The genetic aspects of berries: from field to health: The genetic aspects of berries. J Sci Food Agric 96:365–371. https://www.doi.org/10.1002/jsfa.7216

McCallum S, Woodhead M, Hackett CA, et al (2010) Genetic and environmental effects influencing fruit colour and QTL analysis in raspberry. Theor Appl Genet 121:611–627. https://www.doi.org/10.1007/s00122-010-1334-5

Miller AJ, Gross BL (2011) From forest to field: perennial fruit crop domestication. Am J Bot 98:1389–1414. https://www.doi.org/10.3732/ajb.1000522

Mooers CA (1921) The agronomic placement of varieties. J Am Soc Hortic Sci 13:337– 352

Muñoz PR, Resende MFR, Gezan SA, et al (2014) Unraveling additive from nonadditive effects using genomic relationship matrices. Genetics 198:1759–1768. https://www.doi.org/10.1534/genetics.114.171322

Nei M (1987) Genetic distance between populations. In: Molecular evolutionary genetics. Columbia University Press, New York, pp 208–253

Nguyên T, Yu J, Choi H-W, et al (2018) Optimization of genotyping-by-sequencing (GBS) in chrysanthemums: selecting proper restriction enzymes for GBS library construction. Hortic Sci Technol 36:108–114. https://www.doi.org/10.12972/kjhst.20180012

Ourecky DK (1975) Brambles. In: Janick J, Moore JN (eds) Advances in fruit breeding. Purdue University Press, West Lafayette, IN, pp 98–129

Ourecky DK, Slate GL (1973) Jewel black raspberry. New York’s Food and Life Sci Bul 35

161 Ourecky DK, Slate GL (1966) Hybrid vigor in Rubus occidentalis x Rubus leucodermis seedlings. In: Proceedings of the XVII International Horticultural Congress. Abstract 277

Ozgen M, Wyzgoski FJ, Tulio AZ, et al (2008) Antioxidant capacity and phenolic antioxidants of Midwestern black raspberries grown for direct markets are influenced by production site. HortScience 43:2039–2047

Pastina MM, Malosetti M, Gazaffi R, et al (2012) A mixed model QTL analysis for sugarcane multiple-harvest-location trial data. Theor Appl Genet 124:835–849. https://www.doi.org/10.1007/s00122-011-1748-8

Patterson HD, Silvey V, Talbot M, Weatherup STC (1977) Variability of yields of cereal varieties in U. K. trials. J Agr Sci 89:239. https://www.doi.org/10.1017/S002185960002743X

Paudel L, Wyzgoski FJ, Giusti MM, et al (2014) NMR-based metabolomic investigation of bioactivity of chemical constituents in black raspberry (Rubus occidentalis L.) fruit extracts. J Agric Food Chem 62:1989–1998. https://www.doi.org/10.1021/jf404998k

Pérez-Elizalde S, Cuevas J, Pérez-Rodríguez P, Crossa J (2015) Selection of the bandwidth parameter in a Bayesian kernel regression model for genomic- enabled prediction. J Agric Biol Environ Stat 20:512–532. https://www.doi.org/10.1007/s13253-015-0229-y

Perkins-Veazie P, Ma G, Fernandez GE, et al (2016) Black raspberry fruit composition over two years from seedling populations grown at four US geographic locations. Acta Hortic 1133:335–338. https://www.doi.org/10.17660/ActaHortic.2016.1133.52

Piepho HP (2005) Statistical tests for QTL and QTL-by-environment effects in segregating populations derived from line crosses. Theor Appl Genet 110:561– 566

Plummer M, Best N, Cowles K, Vines K (2006) CODA: convergence diagnosis and output analysis for MCMC. R News 6:7–11

Pritts M (2008) Primocane-fruiting raspberry production. HortScience 43:1640–1641. https://www.doi.org/10.21273/HORTSCI.43.6.1640

Pritts M, McDermott L, Demchak K, et al (2017) High tunnel raspberries and blackberries. Dept. of Horticulture, Cornell University

R Core Team (2018) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria

162 Remberg SF, Sønsteby A, Aaby K, Heide OM (2010) Influence of postflowering temperature on fruit size and chemical composition of Glen Ample raspberry (Rubus idaeus L.). J Agric Food Chem 58:9120–9128. https://www.doi.org/10.1021/jf101736q

Ritter E, Salamini F (1996) The calculation of recombination frequencies in crosses of allogamous plant species with applications to linkage mapping. Genet Res 67:55. https://www.doi.org/10.1017/S0016672300033474

Royo C, Torres-Pérez R, Mauri N, et al (2018) The major origin of seedless grapes is associated with a missense mutation in the MADS-box gene VviAGL11. Plant Physiol 177:1234–1253. https://www.doi.org/10.1104/pp.18.00259

Sanford JC, Ourecky DK, Reich JE (1985) ‘Titan’ red raspberry. HortScience 20:1133– 1134

Scalzo J, Battino M, Costantini E, Mezzetti B (2005) Breeding and biotechnology for improving berry nutritional quality. Biofactors 23:213–220. https://www.doi.org/10.1002/biof.5520230406

Searle SR (1989) Variance components - some history and a summary account of estimation methods. J Anim Breed Genet 106:1–29. https://www.doi.org/10.1111/j.1439-0388.1989.tb00209.x

Sebesta B, Clark JR, Threlfall RT, Howard LR (2013) Characterization of seediness attributes of blackberry genotypes. Discovery, the student journal of Dale Bumpers College of Agricultural, Food and Life Sciences 14:72–79

Shamaila M, Skura B, Daubeny H, Anderson A (1993) Sensory, chemical and gas chromatographic evaluation of five raspberry cultivars. Food Res Int 26:443–449. https://www.doi.org/10.1016/0963-9969(93)90090-6

Shulaev V, Sargent DJ, Crowhurst RN, et al (2011) The genome of woodland strawberry (Fragaria vesca). Nature Genetics 43:109–116. https://www.doi.org/10.1038/ng.740

Silva NF, Goring DR (2001) Mechanisms of self-incompatibility in flowering plants. Cell Mol Life Sci 58:1988–2007

Simmonds NW (1991) Selection for local adaptation in a plant breeding programme. Theor Appl Genet 82:363–367. https://www.doi.org/10.1007/BF02190624

Simpson CG, Cullen DW, Hackett CA, et al (2017) Mapping and expression of genes associated with raspberry fruit ripening and softening. Theor Appl Genet 130:557–572. https://www.doi.org/10.1007/s00122-016-2835-7

163 Singleton VL, Orthofer R, Lamuela-Raventos RM (1999) Analysis of total phenolics and other oxidation substrates and antioxidants by means of Folin-Ciocalteu reagent. Methods Enzymol 299:152–177

Slate GL, Klein LG (1952) Black raspberry breeding. Proc Amer Soc Hort Sci 59:266– 268

Smith AB, Cullis BR (2018) Plant breeding selection tools built on factor analytic mixed models for multi-environment trial data. Euphytica 214:143. https://www.doi.org/10.1007/s10681-018-2220-5

Smith AB, Cullis BR, Thompson R (2005) The analysis of crop cultivar breeding and evaluation trials: an overview of current mixed model approaches. J Agr Sci 143:449–462. https://www.doi.org/10.1017/S0021859605005587

Smith AB, Stringer JK, Wei X, Cullis BR (2007) Varietal selection for perennial crops where data relate to multiple harvests from a series of field trials. Euphytica 157:253–266. https://www.doi.org/10.1007/s10681-007-9418-2

Sonah H, Bastien M, Iquira E, et al (2013) An Improved Genotyping by Sequencing (GBS) Approach Offering Increased Versatility and Efficiency of SNP Discovery and Genotyping. PLoS ONE 8:e54603. https://www.doi.org/10.1371/journal.pone.0054603

Stanton MA, Scheerens JC, Funt RC, Clark JR (2007) Floral competence of primocane- fruiting blackberries Prime-Jan and Prime-Jim grown at three temperature regimens. HortScience 42:508–513. https://www.doi.org/10.21273/HORTSCI.42.3.508

Su G, Madsen P, Lund MS, et al (2006) Bayesian analysis of the linear reaction norm model with unknown covariates. J Anim Sci 84:1651–1657. https://www.doi.org/10.2527/jas.2005-517

Sullivan JA, Yan W, Privé JP (2002) Genotype/genotype × environment biplot analysis for cultivar evaluation and mega-environment investigation in primocane-fruiting red raspberry. J Am Soc Hortic Sci 127:776–780. https://www.doi.org/10.21273/JASHS.127.5.776

Tallman PH (2016) Black raspberry plant named ‘Niwot’. US PP27,131. http://patft.uspto.gov/netacgi/nph- Parser?Sect1=PTO2&Sect2=HITOFF&p=1&u=%2Fnetahtml%2FPTO%2Fsearch - bool.html&r=1&f=G&l=50&co1=AND&d=PTXT&s1=%22black+raspberry%22&s2 =Niwot&OS=

U.S. Dept. of Agr. (2017a) National statistics for raspberries. https://www.ers.usda.gov/data-products/fruit-and-tree-nut-data/data-by- commodity/ 164

U.S. Dept. of Agr. (2017b) Fruit and tree nut yearbook tables. https://www.ers.usda.gov/data-products/fruit-and-tree-nut-data/fruit-and-tree-nut- yearbook-tables/

U.S. Dept. of Commerce (2019) Local climatological data dataset. https://www.ncdc.noaa.gov/cdo- web/datasets/GHCND/stations/GHCND:USC00314464/detail van Eeuwijk FA (1995) Linear and bilinear models for the analysis of multi-environment trials: I. An inventory of models. Euphytica 84:1–7. https://www.doi.org/10.1007/BF01677551

Van Ooijen JW (2006) JoinMap® 4.0: software for the calculation of genetic linkage maps in experimental populations. Kyazma, B.V., Wageningen

Van Ooijen JW, Jansen J (2013) Genetic mapping in experimental populations. Cambridge University Press, New York

VanBuren R, Bryant D, Bushakra JM, et al (2016) The genome of black raspberry (Rubus occidentalis). Plant J 87:535–547. https://www.doi.org/10.1111/tpj.13215

VanBuren R, Wai CM, Colle M, et al (2018) A near complete, chromosome-scale assembly of the black raspberry (Rubus occidentalis) genome. GigaScience 7:1– 9. https://www.doi.org/10.1093/gigascience/giy094

VanRaden PM (2008) Efficient methods to compute genomic predictions. J Dairy Sci 91:4414–4423. https://www.doi.org/10.3168/jds.2007-0980

Vardi A, Levin I, Carmi N (2008) Induction of seedlessness in citrus: from classical techniques to emerging biotechnological approaches. J Am Soc Hortic Sci 133:117–126. https://www.doi.org/10.21273/JASHS.133.1.117

Villamor RR, Daniels CH, Moore PP, Ross CF (2013) Preference mapping of frozen and fresh raspberries. J Food Sci 78:S911–S919. https://www.doi.org/10.1111/1750- 3841.12125

Voillemot M, Pannell JR (2017) Inbreeding depression is high in a self-incompatible perennial herb population but absent in a self-compatible population showing mixed mating. Ecol Evol 7:8535–8544. https://www.doi.org/10.1002/ece3.3354

Vool E, Karp K, Noormets M, et al (2009) The productivity and fruit quality of the arctic bramble (Rubus arcticus ssp. arcticus) and hybrid arctic bramble (Rubus arcticus ssp. arcticus × Rubus arcticus ssp. stellatus). Acta Agr Scand Sect B 59:217– 224. https://www.doi.org/10.1080/09064710802022903

Voorrips RE (2002) MapChart: software for the graphical presentation of linkage maps and QTLs. J Hered 93:77–78. https://www.doi.org/10.1093/jhered/93.1.77

165 VSN International (2018) GenStat for Windows 19th Edition. VSN International Ltd, Hemel Hempstead

Weber C (2015) Performance of new berry varieties at the NYSAES in Geneva, NY. New York Fruit Quarterly 23:5–9

Weber CA (2003) Genetic diversity in black raspberry detected by RAPD markers. HortScience 38:4

Williams CF (1950) Influence of parentage in species hybridization of raspberries. Proc Amer Soc Hort Sci 56:149–156

Wimmer V, Albrecht T, Auinger H-J, Schön C-C (2012) synbreed: a framework for the analysis of genomic prediction data using R. Bioinformatics 28:2086–2087. https://www.doi.org/10.1093/bioinformatics/bts335

Woznicki TL, Heide OM, Remberg SF, Sønsteby A (2016) Effects of controlled nutrient feeding and different temperatures during floral initiation on yield, berry size and drupelet numbers in red raspberry (Rubus idaeus L.). Sci Hortic 212:148–154. https://www.doi.org/10.1016/j.scienta.2016.09.047

Wu R, Ma C-X, Painter I, Zeng Z-B (2002) Simultaneous maximum likelihood estimation of linkage and linkage phases in outcrossing species. Theor Popul Biol 61:349– 363. https://www.doi.org/10.1006/tpbi.2002.1577

Yang J, Benyamin B, McEvoy BP, et al (2010) Common SNPs explain a large proportion of the heritability for human height. Nat Genet 42:565–569. https://www.doi.org/10.1038/ng.608

Yates E, Cochran WG (1938) The analysis of groups of experiments. J Agric Sci Camb 28:556–580

Zorrilla-Fontanesi Y, Cabeza A, Domínguez P, et al (2011) Quantitative trait loci and underlying candidate genes controlling agronomical and fruit quality traits in octoploid strawberry (Fragaria × ananassa). Theor Appl Genet 123:755–778. https://www.doi.org/10.1007/s00122-011-1624-6

166