GENETIC VARIABILITY AND MAPPING OF COOKING TIME AND SENSORY ATTRIBUTES IN ANDEAN DRY BEANS By Amber Nichole Bassett A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Plant Breeding, Genetics, and Biotechnology – Crop and Soil Sciences – Doctor of Philosophy 2020 GENETIC VARIABILITY AND MAPPING OF COOKING TIME AND SENSORY ATTRIBUTES IN ANDEAN DRY BEANS ABSTRACT By Amber Nichole Bassett Cooking time, flavor, and texture of dry beans (Phaseolus vulgaris L.) are valued by consumers but are not major considerations of dry bean breeding programs. The aim of this research is 1) to investigate mechanisms underlying fast cooking times of select genotypes, 2) to characterize the genetic control of cooking time, flavor, and texture of cooked beans in a diversity panel and a recombinant inbred line population, and 3) to evaluate how fast-cooking bean genotypes process into canned products. The genetic mechanism of fast cooking time was assessed via the physical and compositional seed characteristics in a set of 8 genotypes. Faster cooking beans had thinner cotyledon cell walls and seed coat layers and lower seed coat percentage, seed weight, and total and insoluble fiber. To identify genomic loci underlying cooking time, flavor, and texture, genome-wide association (GWA) and quantitative trait loci (QTL) mapping approaches were used with 430 lines of the Andean Diversity Panel and 242 yellow recombinant inbred lines. Sensory attributes included total flavor, beany, vegetative, earthy, starchy, sweet, and bitter intensity as well as seed coat perception and cotyledon texture. SNPs and QTL were identified for most of the attributes, with QTL for earthy intensity having the most phenotypic variation explained. In both populations, sweet and starchy intensity were positively correlated and associated via PCA, but other trait associations were minimal. A subset of lines from the RIL population were evaluated for canning quality following different retort processing durations. For fast-cooking lines, canning quality improved with reduced retort processing time, revealing a potential cost-saving benefit to the canning industry. This information lays a foundation for targeting fast cooking times and specific sensory profiles in breeding programs. To Frodo and Sammie, two brave rabbits who took this journey with me. iv ACKNOWLEDGEMENTS Many people contributed to this work and to my personal and professional growth while at Michigan State University. In particular, I am very grateful to my adviser, Dr. Karen Cichy, who encouraged me to pursue my research interests, shared excitement for my work, and provided constructive feedback every step of the way. I am grateful to Dr. James Kelly, Dr. Dechun Wang, and Dr. Shi-You Ding, who served on my committee and were very helpful as advisers, editors, and educators. I also appreciate the friendships, working relationships, and collaborations shared with the current and former lab mates and other graduate students, staff, and faculty I have been fortunate to get to know. Outside of my program, I am grateful to my family for their support and encouragement. My parents Ben and Donna Bassett, my brother Eric Bassett, and my sister-in-law Jenna Bassett have always supported my education and served as great role models in their own pursuits. They were instrumental in this achievement, and I am forever grateful to them. Lastly, I want to thank Nolan Bornowski for his emotional support, scientific discussion, and appreciation of my cooking. This experience would not have been the same without him. v TABLE OF CONTENTS LIST OF TABLES ......................................................................................................................... ix LIST OF FIGURES ...................................................................................................................... xii INTRODUCTION .......................................................................................................................... 1 PROBLEM DEFINITION .......................................................................................................... 1 OBJECTIVES ............................................................................................................................. 2 DISSERTATION OUTLINE...................................................................................................... 2 REFERENCES ............................................................................................................................... 4 CHAPTER 1: GENETIC VARIABILITY OF COOKING TIME IN DRY BEANS (Phaseolus vulgaris L.) RELATED TO SEED COAT THICKNESS AND THE COTYLEDON CELL WALL ......................................................................................................................................................... 7 ABSTRACT ................................................................................................................................ 8 INTRODUCTION ...................................................................................................................... 9 MATERIALS AND METHODS .............................................................................................. 13 Germplasm............................................................................................................................. 13 Water uptake and Cooking Time Evaluation ........................................................................ 14 Scanning Electron Microscopy .............................................................................................. 15 Dietary Fiber Quantification .................................................................................................. 16 Seed Component Percentage and Cotyledon Cell Wall Isolation ......................................... 16 Statistical Analysis ................................................................................................................ 17 RESULTS ................................................................................................................................. 18 Cooking Time and Water Uptake .......................................................................................... 18 Seed and Physical Traits ........................................................................................................ 19 Seed Coat and Cotyledon Cell Wall Compositional Traits ................................................... 21 Principal Component Analysis and Correlations................................................................... 21 DISCUSSION ........................................................................................................................... 24 CONCLUSIONS....................................................................................................................... 27 ACKNOWLEDGEMENTS ...................................................................................................... 28 APPENDICES .............................................................................................................................. 29 APPENDIX A: CHAPTER 1 TABLES AND FIGURES ........................................................ 30 APPENDIX B: CHAPTER 1 SUPPLEMENTAL TABLES AND FIGURES ........................ 38 REFERENCES ............................................................................................................................. 48 CHAPTER 2: GENETIC VARIABILITY AND GENOME-WIDE ASSOCIATION ANALYSIS OF FLAVOR AND TEXTURE IN COOKED BEANS (Phaseolus vulgaris L.) ........................ 55 ABSTRACT ............................................................................................................................. 56 INTRODUCTION .................................................................................................................... 57 MATERIALS AND METHODS .............................................................................................. 59 Germplasm............................................................................................................................. 59 Cooking Time Evaluation ...................................................................................................... 60 vi Sensory Evaluation ................................................................................................................ 61 Panel Training and Assessment ............................................................................................. 62 Sample Preparation for Sensory Evaluation .......................................................................... 63 Statistics ................................................................................................................................. 64 Genotyping ............................................................................................................................ 65 Genome Wide Association .................................................................................................... 66 RESULTS ................................................................................................................................. 67 Sensory Extremes .................................................................................................................. 67 Sensory Evaluation ................................................................................................................ 68 Cooking Time Evaluation ...................................................................................................... 69 Correlations and PCA ............................................................................................................ 69 Genome-Wide Association Mapping .................................................................................... 72 DISCUSSION ........................................................................................................................... 75 CONCLUSION ......................................................................................................................... 80 ACKNOWLEDGMENTS ........................................................................................................ 81 APPENDICES .............................................................................................................................. 82 APPENDIX A: CHAPTER 2 TABLES AND FIGURES ........................................................ 83 APPENDIX B: CHAPTER 2 SUPPLEMENTAL TABLES AND FIGURES ........................ 94 REFERENCES ........................................................................................................................... 119 CHAPTER 3: QTL MAPPING OF SEED QUALITY TRAITS INCLUDING COOKING TIME, FLAVOR, AND TEXTURE IN YELLOW DRY BEANS (Phaseolus vulgaris L.) ................. 127 ABSTRACT ............................................................................................................................ 128 INTRODUCTION .................................................................................................................. 129 MATERIALS AND METHODS ............................................................................................ 132 Germplasm........................................................................................................................... 132 CIELAB Analysis and Seed Coat Postharvest Darkening .................................................. 133 Cooking Time Evaluation .................................................................................................... 133 Sensory Evaluation .............................................................................................................. 134 Panel Training...................................................................................................................... 135 Sample Preparation for Sensory Evaluation ........................................................................ 136 Statistics ............................................................................................................................... 136 Genotyping .......................................................................................................................... 137 RESULTS ............................................................................................................................... 138 Cooking Time Evaluation .................................................................................................... 138 Sensory Evaluation .............................................................................................................. 139 Color and Seed Coat Postharvest Darkening ....................................................................... 140 Seed Yield and Seed Weight ............................................................................................... 141 PCA ..................................................................................................................................... 141 QTL Mapping ...................................................................................................................... 142 DISCUSSION ......................................................................................................................... 147 CONCLUSION ....................................................................................................................... 152 ACKNOWLEDGMENTS ...................................................................................................... 152 APPENDICES ............................................................................................................................ 154 APPENDIX A: CHAPTER 3 TABLES AND FIGURES ...................................................... 155 APPENDIX B: CHAPTER 3 SUPPLEMENTAL TABLES AND FIGURES ...................... 166 vii REFERENCES ........................................................................................................................... 184 CHAPTER 4: REDUCED RETORT PROCESSING TIME IMPROVES CANNING QUALITY OF FAST-COOKING DRY BEANS (Phaseolus vulgaris L.) ................................................... 191 ABSTRACT ............................................................................................................................ 192 INTRODUCTION .................................................................................................................. 193 MATERIALS AND METHODS ............................................................................................ 196 Germplasm........................................................................................................................... 196 Cooking Time Determination .............................................................................................. 197 Canning Protocol ................................................................................................................. 197 Visual Evaluation ................................................................................................................ 198 Washed-drained Weight Determination and Image Analysis ............................................. 199 Texture Analysis .................................................................................................................. 199 Statistical Analysis .............................................................................................................. 200 RESULTS ............................................................................................................................... 200 Cooking Time and Water Uptake ........................................................................................ 200 Canned Bean Intactness ....................................................................................................... 201 Washed-drained Weight ...................................................................................................... 202 Texture Analysis .................................................................................................................. 203 CIELAB Values ................................................................................................................... 203 DISCUSSION ......................................................................................................................... 204 CONCLUSIONS..................................................................................................................... 207 ACKNOWLEDGEMENTS .................................................................................................... 207 APPENDICES ............................................................................................................................ 208 APPENDIX A: CHAPTER 4 TABLES AND FIGURES ...................................................... 209 APPENDIX B: CHAPTER 4 SUPPLEMENTAL TABLES AND FIGURES ...................... 220 REFERENCES ........................................................................................................................... 225 SUMMARY AND CONCLUSIONS ......................................................................................... 230 viii LIST OF TABLES Table 1.1 Means for all genotypes of seed weight; cotyledon, seed coat, and embryo percentage . ....................................................................................................................................................... 30 Table 1.2 Spearman correlations of all traits with cooking times from 0, 3, 6, 12, 18, and 24 hr soaked samples.............................................................................................................................. 31 Table S1.1 ANOVA results indicating the significance of the fixed effects genotype, soaking time, and genotype by soaking time for all traits. .................................................................................. 38 Table S1.2 Means for all genotypes of unsoaked and soaked (12 hr) macrosclereid-layer, osteosclereid-layer, and cotyledon cell wall thickness. ................................................................ 39 Table S1.3 Means for all genotypes (raw) of whole seed total, soluble, and insoluble fiber and total, soluble, and insoluble cotyledon cell wall isolate................................................................ 40 Table 2.1 Genotypes exhibiting extreme sensory attribute intensities identified from screening accessions of the Andean Diversity Panel grown in Hawassa, Ethiopia. ..................................... 83 Table 2.2 Least squares estimates, range, and coefficient of variation of sensory attribute intensities of the Andean Diversity Panel grown in three locations with ANOVA p-valuesa for genotype, location (Loc), and genotype by location indicated. .................................................... 84 Table 2.3 Mean, range, and coefficient of variation of raw seed weight, soak water uptake, cooking time, and total water uptake of the Andean Diversity Panel grown in three locations with ANOVA p-values for genotype, location (Loc), and genotype by location indicated. ................ 85 Table 2.4 GWAS significant markers associated with sensory attribute intensities with marker, chromosome (Chr), position, P-value, minor allele frequency (MAF), major and minor alleles (Maj/Min), significance (Sig), and method indicated. .................................................................. 86 Table S2.1 Genotype information. ............................................................................................... 94 Table S2.2 5-point sensory attribute intensity scales. ................................................................ 106 Table S2.3 P-values for the random effects from the sensory attribute intensity ANOVAs at the genotype level. ............................................................................................................................ 106 Table S2.4 Mean sensory attribute intensities across the 3 locations for the genotypes exhibiting extreme sensory attribute intensities. .......................................................................................... 107 Table S2.5 P-values for the fixed and random effects from the sensory attribute intensity ANOVAs at the seed type level. ................................................................................................. 107 ix Table S2.6 GWAS significant markers associated with sensory attribute intensities determined via BLINK with marker, chromosome (Chr), position, P-value, minor allele frequency (MAF), major and minor alleles (Maj/Min), significance (Sig), and location indicated. .................................. 108 Table S2.7 GWAS significant markers associated with cooking time, soak water uptake, raw seed weight, and total water uptake, with chromosome (Chr), position, R2, effect associated with the minor allele, major and minor alleles, minor allele frequency (MAF), major and minor alleles (Maj/Min), significance (Sig), and method indicated. ................................................................ 110 Table 3.1 Parental phenotypes, means, ranges, and broad-sense heritability (H2) for the RILs for both years combined with ANOVA p-values for genotype, year, and genotype by year indicated. ..................................................................................................................................................... 155 Table 3.2 Linkage map information for the 240 RILs. .............................................................. 156 Table 3.3 Quantitative trait loci identified in the RIL population (N = 240) grown in Entrican, MI in 2016 and 2017 for soak water uptake and cooking time. Linkage group (LG), peak position (Pos), year, logarithm of odds (LOD), R2, QTL effect (a), flanking markers, QTL range, and significancee of the QTL are indicated. ...................................................................................... 157 Table 3.4 Quantitative trait loci identified in the RIL population (N = 240) grown in Entrican, MI in 2016 and 2017 for sensory attributes. Linkage group (LG), peak position (Pos), year, logarithm of odds (LOD), R2, QTL effect (a), flanking markers, QTL range, and significance of the QTL are indicated. ..................................................................................................................................... 158 Table 3.5 Quantitative trait loci identified in the RIL population (N = 240) grown in Entrican, MI in 2017 for color and seed coat postharvest darkening. Linkage group (LG), peak position (Pos), year, logarithm of odds (LOD), R2, QTL effect (a), flanking markers, QTL range, and significance of the QTL are indicated. ............................................................................................................ 160 Table S3.1 Parental phenotypes, means, ranges, and broad-sense heritability (H2) for the RILs for both years combined with ANOVA p-values for genotype, year, and genotype by year indicated. ..................................................................................................................................................... 166 Table S3.2 Parental phenotypes and means and ranges for the RILs for 2016 and 2017. ..................................................................................................................................................... 167 Table S3.3 P-values for the random effects from the sensory attribute intensity ANOVAs at the genotype level. ............................................................................................................................ 168 Table S3.4 Quantitative trait loci identified in the RIL population (N = 240) grown in Entrican, MI in 2017 for seed weight, total water uptake, and yield. Linkage group (LG), peak position (Pos), year, logarithm of odds (LOD), R2, QTL effect (a), flanking markers, QTL range, and significance of the QTL are indicated. ............................................................................................................ 169 x Table 4.1 Average values of seed weight, soak water uptake, cooking time, and total water uptake for Ervilha, PI527538, and RILs. ................................................................................................ 209 Table 4.2 ANOVA results indicating the significance of the fixed effects genotype, retort time, and genotype by retort time for intactness, washed-drained weight, texture, and CIELAB color. .. ..................................................................................................................................................... 209 Table 4.3 Means and ranges of intactness (1–5 scale), washed-drained weight (W-D Wt.) (g), texture measurements (kg), and CIELAB values for Ervilha, PI527538, and the RILs at the five retort processing times. ............................................................................................................... 210 Table 4.4 ANOVA results indicating the significance of the fixed effects cooking group, retort time, and cooking group by retort time for intactness, washed-drained weight, texture, and CIELAB color. ............................................................................................................................ 211 Table 4.5 Means and ranges of intactness (1–5 scale), washed-drained weight, texture measurements, and CIELAB values for the fast-, medium-, and slow-cooking groups across all retort times. ................................................................................................................................. 211 Table 4.6 Means and ranges of intactness (1–5 scale), washed-drained weight, texture measurements, and CIELAB values for the fast-, medium-, and slow-cooking groups at the five retort times. ................................................................................................................................. 212 Table S4.1 Pearson correlation coefficients and P-values for correlations between cooking time and washed-drained weight, texture, intactness, and CIELAB color values at the five retort times. ..................................................................................................................................................... 220 xi LIST OF FIGURES Figure 1.1 Scatterplots of cooking time and water uptake vs soaking time and images of the genotypes used in this study. Circles indicate cooking time and squares indicate water uptake. 32 Figure 1.2 Bar plots of seed coat layer thickness for unsoaked and soaked (12 hr) beans with seed type and genotype indicated. Example SEM images (RedMottled-1) of the seed coat layers are presented with the measured layers indicated. MS = Macrosclereid layer; OS = osteosclereid layer. ....................................................................................................................................................... 33 Figure 1.3 Bar plots of cotyledon cell wall thickness for unsoaked and soaked (12 hr) beans with seed type and genotype indicated. Example SEM images (RedMottled-1) of cotyledon cells are presented with locations of measurements indicated by white arrows. ........................................ 34 Figure 1.4 Bar plots of soluble and insoluble whole seed dietary fiber and cotyledon cell wall isolate of raw beans with seed type and genotype indicated. ....................................................... 35 Figure 1.5 Principal component analysis biplot with each genotype indicated and loadings for cooking times across 0, 3, and 12 hr soaking times (CT0, CT3, and CT12); seed weight (SeedWt); seed coat (SeedCoat), cotyledon (Cotyledon), and embryo (Embryo) percentage; unsoaked and soaked (12 hr) macrosclereid-layer (MST0 and MST12), osteosclereid-layer thickness (OST0 and OST12), and cotyledon cell wall (CWT0 and CWT12); raw total (TCWI), soluble (SCWI), and insoluble (ISCWI) cotyledon cell wall isolate; and raw total (TFiber), soluble (SFiber), and insoluble (IFiber) whole seed dietary fiber. .................................................................................. 36 Figure 1.6 Diagram of a dry bean cross section indicating the traits associated with fast cooking time of unsoaked (U) or soaked (S) beans. MS = macrosclereid layer; OS = Osteosclereid layer. ....................................................................................................................................................... 37 Figure S1.1 Workflow depicting the steps for each objective. .................................................... 41 Figure S1.2 Means of cooking time and water uptake for all genotypes across 0, 3, 6, 12, 18, and 24 hr soaking times. Within seed type pairs, the faster cooking time for each soaking time is indicated in red. Mean separation of each trait (by row) is indicated by the letter superscript. ... 42 Figure S1.3 Pairwise comparison matrix of soaking time (Soak), cooking time (CT), and water uptake (WU) across 0, 3, 6, 12 18, and 24 hr soaking times. Spearman correlation coefficients are indicated in the lower left, and scatterplots for each pairwise comparison with LOWESS regression lines are shown in the upper right. P-values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. ......................................................................... 43 Figure S1.4 Pairwise comparison matrix of cooking times of unsoaked (CT0) and 12 hr soaked (CT12) beans, unsoaked (MST0) and 12 hr soaked (MST12) macrosclereid-layer thickness, unsoaked (OST0) and 12 hr soaked (OST12) osteoclereid-layer thickness, and unsoaked (CWT0) xii and 12 hr soaked (CWT12) cotyledon cell wall thickness. Spearman correlation coefficients are indicated in the lower left, and scatterplots for each pairwise comparison with LOWESS regression lines are shown in the upper right. P-values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. ......................................................................... 44 Figure S1.5 Pairwise comparison matrix of soaking time (Soak), cooking time (CT), macrosclereid-layer thickness (MST), osteoclereid-layer thickness (OST), and cotyledon cell wall thickness (CWT). Spearman correlation coefficients are indicated in the lower left, and scatterplots for each pairwise comparison with LOWESS regression lines are shown in the upper right. P- values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. .................................................................................................................................. 45 Figure S1.6 Pairwise comparison matrix of cooking time of unsoaked beans (CT0) and cooking time of 12 hr soaked beans (CT12), seed weight (SeedWt), and seed coat (SeedCoat), cotyledon (Cotyledon), and embryo (Embryo) percentage. Spearman correlation coefficients are indicated in the lower left, and scatterplots for each pairwise comparison with LOWESS regression lines are shown in the upper right. P-values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. .................................................................................................... 46 Figure S1.7 Pairwise comparison matrix of cooking times of unsoaked (CT0) and 12 hr soaked (CT12) beans; total (TFiber), soluble (SFiber), and insoluble (IFiber) whole seed dietary fiber; and total (TCWI), soluble (SCWI), and insoluble (ICWI) cotyledon cell wall isolate. Spearman correlation coefficients are indicated in the lower left, and scatterplots for each pairwise comparison with LOWESS regression lines are shown in the upper right. P-values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. .................. 47 Figure 2.1 Density plots of least squares estimates of sensory attribute intensities for the Andean Diversity Panel for all locations combined (C); Hawassa, ET (H); Kabwe, Zambia (K); and Lusaka, Zambia (L). ................................................................................................................................... 87 Figure 2.2 Boxplots of sensory attribute intensities separated by seed type. All boxplots are presented as least squares estimates averaged across all locations for seed types with N > 10, where “Other” includes the remaining seed types with N < 10. .............................................................. 88 Figure 2.3 Density plots of raw seed weight, soak water uptake, cooking time, and total water uptake for the Andean Diversity Panel for all locations combined (C); Hawassa, ET (H); Kabwe, Zambia (K); and Lusaka, Zambia (L). .......................................................................................... 89 Figure 2.4 Pairwise comparison matrix of cooking time (CT), total flavor intensity (TF), beany intensity (Beany), vegetative intensity (Veg), earthy intensity (Earthy), starchy intensity (Starchy), sweet intensity (Sweet), bitter intensity (Bitter), seed coat perception (SCP), and cotyledon texture (CTex). Pearson correlation coefficients were calculated using BLUPs and are indicated in the lower left, and scatterplots for each pairwise comparison with LOWESS regression lines are shown in the upper right. P-values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. ............................................................................................................... 90 xiii Figure 2.5 Principal component analysis biplot with each genotype colored by seed type and loadings for total flavor intensity (TF), beany intensity (Beany), vegetative intensity (Veg), earthy intensity (Earthy), starchy intensity (Starchy), sweet intensity (Sweet), bitter intensity (Bitter), seed coat perception (SCP), cotyledon texture (CTex), and cooking time (CT). ................................. 91 Figure 2.6 Manhattan and QQ plots for total flavor intensity, beany intensity, earthy intensity, seed coat perception, and cotyledon texture of the Andean Diversity Panel with mapping conducted using BLINK with BLUPs from all locations combined. The gray dashed line is the α = 0.05 FDR. .................................................................................................................................. 92 Figure 2.7 Phenotypic effects of carrying the indicated number of significant markers conferring a positive effect for each sensory attribute. Phenotypic values represent all locations combined as averages of least squares estimates from Hawassa, Ethiopia; Kabwe, Zambia; and Lusaka, Zambia. N is the number of individuals in each boxplot. ........................................................................... 93 Figure S2.1 Images of the genotypes exhibiting extreme sensory attribute intensities identified /from screening accessions of the Andean Diversity Panel grown in Hawassa, Ethiopia. ......... 112 Figure S2.2 Manhattan and QQ plots for total flavor intensity, beany intensity, earthy intensity, seed coat perception, and cotyledon texture of the Andean Diversity Panel with mapping conducted using MLM with BLUPs from all locations combined. The gray dashed line is the α = 0.05 Bonferroni correction based on the effective number of markers determined using the SimpleM algorithm. .................................................................................................................... 113 Figure S2.3 Manhattan and QQ plots for total flavor intensity of the Andean Diversity Panel with mapping conducted using BLINK with BLUPs for Hawassa, Ethiopia (H); Kabwe, Zambia (K); and Lusaka, Zambia (L). The gray dashed line is the α = 0.05 FDR. ......................................... 114 Figure S2.4 Manhattan and QQ plots for beany intensity of the Andean Diversity Panel with mapping conducted using BLINK with BLUPs for Hawassa, Ethiopia (H); Kabwe, Zambia (K); and Lusaka, Zambia (L). The gray dashed line is the α = 0.05 FDR. ......................................... 114 Figure S2.5 Manhattan and QQ plots for earthy intensity of the Andean Diversity Panel with mapping conducted using BLINK with BLUPs for Hawassa, Ethiopia (H); Kabwe, Zambia (K); and Lusaka, Zambia (L). The gray dashed line is the α = 0.05 FDR. ......................................... 115 Figure S2.6 Manhattan and QQ plots for seed coat perception of the Andean Diversity Panel with mapping conducted using BLINK with BLUPs for Hawassa, Ethiopia (H); Kabwe, Zambia (K); and Lusaka, Zambia (L). The gray dashed line is the α = 0.05 FDR. ......................................... 115 Figure S2.7 Manhattan and QQ plots for raw seed weight, soak water uptake, cooking time, and total water uptake of the Andean Diversity Panel with mapping conducted using BLINK with BLUPs from all locations combined. The gray dashed line is the α = 0.05 FDR. ...................... 116 Figure S2.8 Manhattan and QQ plots for raw seed weight, soak water uptake, cooking time, and total water uptake of the Andean Diversity Panel with mapping conducted using MLM with xiv BLUPs from all locations combined. The gray dashed line is the α = 0.05 Bonferroni correction based on the effective number of markers determined using the SimpleM algorithm. .............. 117 Figure S2.9 Phenotypic effects of carrying the indicated number of significant markers conferring a positive effect for raw seed weight, soak water uptake, and total water uptake and a negative effect for cooking time. Phenotypic values represent all locations combined as averages from Hawassa, Ethiopia; Kabwe, Zambia; and Lusaka, Zambia. N is the number of individuals in each boxplot. ....................................................................................................................................... 118 Figure 3.1 Images of Ervilha and PI527538 raw seeds.............................................................. 161 Figure 3.2 Density plots of soak water uptake and cooking time for the RILs from 2016, 2017, and both years combined (C). Means for Ervilha and PI527538 from both years combined are indicated in yellow and brown, respectively............................................................................... 161 Figure 3.3 Density plots of least squares estimates of sensory attribute intensities for the RILs from 2016, 2017, and both years combined (C). Attribute intensities for Ervilha and PI527538 from both years combined are indicated in yellow and brown, respectively.............................. 162 Figure 3.4 Density plots of CIELAB values for the RILs from 2016, 2017, and both years combined (C). Attribute intensities for Ervilha and PI527538 from both years combined are indicated in yellow and brown, respectively............................................................................... 163 Figure 3.5 Principal component analysis biplot with loadings for cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), and cotyledon texture (CTX). Ervilha and PI527538 are indicated in yellow and brown, respectively. ..................................................................................................................................................... 164 Figure 3.6 QTL map for soak water uptake, cooking time, total flavor intensity, beany intensity, vegetative intensity, earthy intensity, starchy intensity, sweet intensity, bitter intensity, seed coat perception, cotyledon texture, L*, a*, b*, and seed coat postharvest non-darkening in the RIL population. Size is in cM. Year is indicated for each QTL, where “C” is both years combined.165 Figure S3.1 Density plots of seed weight, total water uptake, and yield for the RILs from 2016, 2017, and both years combined (C). Means for Ervilha and PI527538 from both years combined (2017 for seed yield) are indicated in yellow and brown, respectively. ..................................... 171 Figure S3.2 QTL map for seed weight (SW), total water uptake (TWU), and seed yield (YLD) in the RIL population. Size is in cM. Year is indicated for each QTL, where “C” is both years combined. .................................................................................................................................... 172 Figure S3.3 Line graphs of LOD by Pv01 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat xv postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. ........... 173 Figure S3.4 Line graphs of LOD by Pv02 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. ........... 174 Figure S3.5 Line graphs of LOD by Pv03 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. ........... 175 Figure S3.6 Line graphs of LOD by Pv04 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. ........... 176 Figure S3.7 Line graphs of LOD by Pv05 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. ........... 177 Figure S3.8 Line graphs of LOD by Pv06 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. ........... 178 Figure S3.9 Line graphs of LOD by Pv07 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. ........... 179 xvi Figure S3.10 Line graphs of LOD by Pv08 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. ........... 180 Figure S3.11 Line graphs of LOD by Pv09 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. ........... 181 Figure S3.12 Line graphs of LOD by Pv10 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. ........... 182 Figure S3.13 Line graphs of LOD by Pv11 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. ........... 183 Figure 4.1 Histogram of the cooking times of Ervilha, PI527538, and the 242 RILs, determined using the Mattson cooker method following a 12 h soak. Seeds were grown at the Montcalm Research Farm in Michigan, USA in 2016. The nine fastest (in blue) and slowest cooking lines (in red) were selected for this study. ................................................................................................ 213 Figure 4.2 Pearson correlation matrix of seed weight, soak water uptake, cooking time, and total water uptake. Correlation coefficients are indicated in the lower left and represented by colored, directional ellipses in the upper right. *P < 0.05, **P < 0.01, ***P < 0.001. ............................ 214 Figure 4.3 Boxplots of seed weights, soak water uptake, cooking times, and total water uptake for Ervilha, PI527538, and selected RILs separated into fast-, medium-, and slow-cooking groups. Lines indicate Ervilha (yellow) and PI527538 (brown). Mean separation is indicated by letters above each boxplot. .................................................................................................................... 215 Figure 4.4 Pearson correlation matrix of retort time, cooking time, washed-drained weight, texture, intactness, and CIELAB color values across all genotypes and retort times. Correlation coefficients are indicated in the lower left and represented by colored, directional ellipses in the upper right. *P < 0.05, **P < 0.01, ***P < 0.001. ..................................................................... 216 xvii Figure 4.5 Scatterplots showing the relationship between cooking time and washed-drained weight, texture, intactness, and CIELAB color values separated by retort time. The five retort time series are indicated by colors and symbols as specified. ............................................................ 217 Figure 4.6 Boxplots of washed-drained weights, texture and intactness values across all retort times for Ervilha, PI527538, and selected RILs separated into fast-, medium-, and slow-cooking groups. Lines indicate Ervilha (yellow) and PI527538 (brown). Mean separation within each retort time is indicated by letters above each boxplot. ......................................................................... 218 Figure 4.7 Boxplots of CIELAB color values across all retort times for Ervilha, PI527538, and selected RILs separated into fast-, medium-, and slow-cooking groups. Lines indicate Ervilha (yellow) and PI527538 (brown). Mean separation within each retort time is indicated by letters above each boxplot. .................................................................................................................... 219 Figure S4.1 Images of the raw seed of Ervilha, PI527538, and the RILs selected for this study separated into fast-, medium-, and slow-cooking groups. .......................................................... 221 Figure S4.2 Images of the washed-drained canned samples for Ervilha and PI527538 after retort processing for 10, 15, 20, 30, and 45 minutes. ........................................................................... 222 Figure S4.3 Boxplots of the CIELAB values for the raw seed of Ervilha, PI527538, and the RILs selected for this study separated into fast, medium, and slow cooking groups. Lines indicate Ervilha (yellow) and PI527538 (brown). Mean separation is indicated by the letters above each boxplot. ....................................................................................................................................... 223 Figure S4.4 Pearson correlation matrix of retort time, cooking time, washed-drained weight, texture, intactness, and CIELAB color values (canned and raw) across all genotypes and retort times. Correlation coefficients are indicated in the lower left and represented by colored, directional ellipses in the upper right. P-values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. ....................................................................... 224 xviii INTRODUCTION PROBLEM DEFINITION Cooking time and sensory quality are important characteristics consumers consider when purchasing dry beans (Phaseolus vulgaris L.) (Castellanos et al., 1997; Scott and Maideni, 1998). Dry beans often require long cooking times, particularly when cooked without prior soaking. Many chemical and physical changes occur during the cooking process (Rockland and Jones, 1974; Cichy et al., 2015), but the mechanism of cooking time and the roles of the seed coat and cotyledon cell wall in cooking time are not well understood. Long cooking times deter consumers, particularly as convenience is increasingly valued concerning food and other aspects of modern life (Sloan, 2015). Taste is a major driver of food purchasing decisions among consumers (IFIC, 2019), and it is a common reason consumers choose not to eat beans (Leterme and Carmenza Muñoz, 2002; Eihusen and Albrecht, 2007; Winham et al., 2019). Food companies invest heavily in this aspect of product development (Banking, 2016), but limited research concerning dry bean flavor and associated volatiles is available (Vara-Ubol et al., 2004; Bott and Chambers, 2006; Oomah et al., 2007; Plans et al., 2014; Szczygiel et al., 2017). Cooking time and sensory quality have largely been overlooked by breeders, who have focused instead on seed yield, processing quality, disease resistance, architecture, agronomic adaptation, stress tolerance, and grower friendliness, which encompasses traits that reduce labor and inputs required by growers (Kelly and Cichy, 2012). Cooking time and sensory attributes can be costly in time and resources to evaluate, but a lack of focus on these consumer-valued traits may be limiting consumption of dry beans below their potential. Wide genotypic variability exists for both cooking time and sensory attributes, even within market class (Rivera et al., 2013; Cichy 1 et al., 2015), which can be targeted to improve new varieties and appeal to more consumers and product developers. In addition, fast cooking time may appeal to the canning industry by reducing energy costs and improving efficiency of production through shorter retort processing times. OBJECTIVES This study aims to explore the mechanism of cooking time as it relates to the seed coat and cell wall, identify genomic loci relevant for cooking time and sensory attributes using quantitative genetics approaches, and determine the relevance of cooking time to the canning industry. DISSERTATION OUTLINE Chapter 1 assesses the genetic variability of cooking time across different soaking times as it relates to physical and compositional traits of the seed coat and cell wall. The study was performed using eight genotypes across four seed types. The relationships between cooking time and soaking time, seed size, seed coat/cotyledon percent, seed coat layer thickness, cell wall thickness, cotyledon cell wall isolate, and dietary fiber were determined. Chapter 2 is a genome-wide association study aimed at understanding the genetic basis of cooking time and sensory attributes. Across three locations, 430 lines of the Andean Diversity Panel were evaluated for cooking time and sensory attributes intensities, including total flavor, beany, vegetative, earthy, starchy, sweet, bitter, seed coat perception, and cotyledon texture. Significant SNPs associated with these traits were identified and can be used for the development of molecular markers. Chapter 3 is a quantitative trait loci mapping study aimed at understanding the genetic basis of cooking time and sensory attributes. Across two years in Michigan, a recombinant inbred line 2 population of 242 yellow bean lines was developed and evaluated for cooking time and sensory attributes, including total flavor, beany, vegetative, earthy, starchy, sweet, bitter, seed coat perception, and cotyledon texture. QTL were identified for these traits and can be used for the development of molecular markers. Chapter 4 assesses canning quality as it relates to cooking time. Across five retort processing times, 20 yellow bean lines with varying cooking times were assessed for canning quality. The relationships between cooking time and intactness, washed-drained weight, texture, and color were determined. 3 REFERENCES 4 REFERENCES in Mexico. Arch. Latinoam. Nutr. 47, 163—167. Available Banking, W. B. I. (2016). Secular trends drive M & A interest in value-added ingredients and food solutions. Bott, L., and Chambers, E. (2006). Sensory characteristics of combinations of chemicals potentially associated with beany aroma in foods. J. Sens. Stud. 21, 308–321. doi:10.1111/j.1745- 459X.2006.00067.x. Castellanos, J. Z., Guzmán Maldonado, H., Jiménez, A., Mejía, C., Muñoz Ramos, J. J., Acosta Gallegos, J. A., et al. (1997). Preferential habits of consumers of common bean (Phaseolus vulgaris L.) at: http://europepmc.org/abstract/MED/9659433. Cichy, K. A., Wiesinger, J. A., and Mendoza, F. A. (2015). Genetic diversity and genome-wide association analysis of cooking time in dry bean (Phaseolus vulgaris L.). Theor. Appl. Genet. 128, 1555–1567. doi:10.1007/s00122-015-2531-z. Eihusen, J., and Albrecht, J. A. (2007). Dry bean intake of women ages 19-45. Rural. Rev. Undergrad. Res. Agric. Life Sci. 2. IFIC (2019). 2019 Food and Health Survey. Available at: https://foodinsight.org/2019-food-and- health-survey/. Kelly, J. D., and Cichy, K. A. (2012). Dry bean breeding and production technologies. Dry Beans Pulses Prod. Process. Nutr., 23–54. doi:10.1002/9781118448298.ch2. Leterme, P., and Carmenza Muñoz, L. (2002). Factors influencing pulse consumption in Latin America. Br. J. Nutr. 88, 251–254. doi:10.1079/bjn/2002714. Oomah, B. D., Liang, L. S. Y., and Balasubramanian, P. (2007). Volatile compounds of dry beans (Phaseolus vulgaris L.). Plant Foods Hum. Nutr. 62, 177–83. doi:10.1007/s11130-007-0059-3. Plans, M., Simó, J., Casañas, F., del Castillo, R. R., Rodriguez-Saona, L. E., and Sabaté, J. (2014). Estimating sensory properties of common beans (Phaseolus vulgaris L.) by near infrared spectroscopy. Food Res. Int. 56, 55–62. doi:10.1016/j.foodres.2013.12.003. Rivera, A., Fenero, D., Almirall, A., Ferreira, J. J., Simó, J., Plans, M., et al. (2013). Variability in sensory attributes in common bean (Phaseolus vulgaris L.): A first survey in the Iberian secondary diversity center. Genet. Resour. Crop Evol. 60, 1885–1898. doi:10.1007/s10722-013-9963-6. Rockland, L. B., and Jones, F. T. (1974). Scanning electron microscope studies on dry beans. Effects of cooking on the cellular structure of cotyledons in rehydrated large lima beans. J. Food Sci. 39, 342–346. doi:10.1111/j.1365-2621.1974.tb02890.x. 5 Scott, J., and Maideni, M. (1998). Socio-economic survey of three bean growing areas of Malawi. Kampala, Uganda. Sloan, A. E. (2015). The top ten food trends. Food Technol. 69, 24-+. Szczygiel, E. J., Harte, J. B., Strasburg, G. M., and Cho, S. (2017). Consumer acceptance and aroma characterization of navy bean (Phaseolus vulgaris) powders prepared by extrusion and conventional processing methods. J. Sci. Food Agric. 97, 4142–4150. doi:10.1002/jsfa.8284. Vara-Ubol, S., Chambers, E., and Chambers, D. H. (2004). Sensory characteristics of chemical compounds potentially associated with beany aroma in foods. J. Sens. Stud. 19, 15–26. doi:10.1111/j.1745-459X.2004.tb00133.x. Winham, D., Tisue, M., Palmer, S., Cichy, K., and Shelley, M. (2019). Dry bean preferences and attitudes among Midwest Hispanic and non-Hispanic White women. Nutrients 11, 178. doi:10.3390/nu11010178. 6 CHAPTER 1: GENETIC VARIABILITY OF COOKING TIME IN DRY BEANS (Phaseolus vulgaris L.) RELATED TO SEED COAT THICKNESS AND THE COTYLEDON CELL WALL [Submitted for publication in Food Research International] 7 Genetic variability of cooking time in dry beans (Phaseolus vulgaris L.) related to seed coat thickness and the cotyledon cell wall Amber Bassett1, Sharon Hooper1, Karen Cichy*12 1 Department of Plant, Soil and Microbial Sciences, Michigan State University, 1066 Bogue St, Plant and Soil Sciences Building, East Lansing, MI, USA 48824 2 Sugarbeet and Bean Research Unit, USDA-ARS, 1066 Bogue St, Plant and Soil Sciences Building, East Lansing, MI, USA 48824 ABSTRACT Dry beans are an affordable, nutritious food that often require long cooking times. Seed age, storage conditions, growing environment, and genotype influence cooking times. Little is known about underlying factors responsible for genetic variation for cooking time. Using fast and slow cooking genotypes from four different seed types (brown, cranberry, red mottled, yellow), the objectives of this study were to (1) characterize genetic variability for cooking time across multiple soaking time points; (2) determine the roles of the seed coat and cotyledon cell wall thickness in genetic variability in cooking time; and (3) identify seed coat and cotyledon cell wall composition differences associated with genetic variability in cooking time. Genotypes were evaluated for cooking time on unsoaked beans and beans soaked for 3, 6, 12, 18, and 24 hr. Cooking times were sharply reduced after 3 hr of soaking and plateaued after 6 hr of soaking. Soaking time influenced the cooking times differently across genotypes. Greater seed coat percentage, cotyledon cell wall thickness, total and insoluble whole seed dietary fiber, and insoluble cotyledon cell wall isolate were genotypic factors associated with longer cooking times of soaked beans. Thicker seed coat macrosclereid- and osteosclereid-layers were genotypic factors associated with longer cooking times of unsoaked beans. These findings suggest that cotyledon 8 cell wall thickness and composition has a significant role in genotypic variability for cooking time of soaked beans and seed coat layer thickness relates to the genetic variability for cooking time of unsoaked beans. INTRODUCTION Dry beans are an affordable, nutritious food incorporated in many cuisines with versatile preparations, but beans frequently require long cooking times when prepared from dry seed. Cooking dry beans can require anywhere from ~15 minutes to several hours depending on factors such as genotype, seed type, seed age, storage and harvest conditions, and pretreatments (Rockland and Jones, 1974; Hernandez-Unzon and Ortega-Delgado, 1989; Coelho et al., 2007; Cichy et al., 2015). As consumers spend less time preparing meals, there is rising demand for convenience in food preparation (Furst et al., 1996; Jabs and Devine, 2006; Hamrick et al., 2012; Monsivais et al., 2014). For consumers who rely on fuel sources such as firewood and charcoal, long cooking times can be prohibitive due to the increased resources required to prepare beans. Since wide genetic variability exists for bean cooking times, plant breeding is one approach that can be used to develop fast-cooking beans to meet the needs of the growing global population. Cooking dry beans by boiling is a hydrothermal process, and many physical and chemical changes occur during cooking. These physical and chemical changes include the uptake of water, denaturation of proteins, starch swelling and gelatinization, and partial solubilization of polysaccharides in the cell wall leading to separation of adjacent whole cells (Rockland and Jones, 1974). Soaking can greatly reduce cooking time and has been associated with changes in pectin content, starch gelatinization, and protein solubility (Bellido, Arntfield, Cenkowski, & Scanlon, 2006; Chigwedere, Njoroge, Van Loey, & Hendrickx, 2019; Martínez-Manrique et al., 2011). Seed 9 age and storage conditions also impact cooking time through changes in phytic acid content, enzyme activity, seed coat permeability, pectin solubility, cell wall content and thickness, oxidation of phenolic compounds, and membrane deterioration (Jackson and Varriano-Marston, 1981; Moscoso et al., 1984; Hincks and Stanley, 1987; Stanley, 1992; Garcia et al., 1994, 1998; Yousif and Deeth, 2003; Waldron et al., 2003; Galiotou-Panayotou et al., 2008; Shiga et al., 2009; Daher and Braybrook, 2015). Cooking time varies across genotypes. Within a screening of 206 bean genotypes across multiple seed types, cooking times ranged from 15 to 90 min as determined by the Mattson cooker method (Cichy et al., 2015). There was variability among seed types, such that white and yellow beans cook faster on average than red mottled beans. There was also variation within seed types such that one cranberry bean cooked in 15 minutes whereas another cooked in 90 min. The cooking time trait has been found to be highly to moderately heritable (Cichy et al., 2019; Katuuramu et al., 2020). The genetic variability for cooking time could be caused by multiple physical or chemical factors expressed at various stages of the cooking process. The first point of contact during cooking process is the seed coat, which serves as a physical barrier to water uptake (Jackson and Varriano- Marston, 1981). While the micropyle, hilum, strophiole, and raphe have been implicated as the primary means of water entry into the bean (Snyder, 1936; Powrie et al., 1960; Deshpande and Cheryan, 1986a; Agbo et al., 1987; Gargiulo et al., 2020), some studies have identified a relationship between water uptake and total seed coat thickness as well as the thickness of individual seed coat layers, including the macrosclereid, osteosclereid, and parenchyma, finding that thin seed coats increase water uptake rates (Deshpande and Cheryan, 1986a; Agbo et al., 1987). Studies have also shown that seed coat color influences water uptake, with darker beans taking up water more slowly (Marbach and Mayer, 1974, 1975; Tully et al., 1981; Valle et al., 10 1992). Therefore, to better understand genetic factors that contribute to cooking time, it is useful to compare fast- and slow-cooking genotypes both within and across seed types. When it comes to the prolonged cooking times exhibited by aged, improperly stored beans (i.e. hard-to-cook), the cotyledon has been shown to be more important than the seed coat in dictating cooking times for soaked beans (Chigwedere et al., 2018). The cotyledon cell walls are especially important, as they influence cooking time due to the physiological role they play during the soaking and cooking process. The cotyledon cell wall is made up of cellulose, hemicellulose, pectin, neutral sugars, proteins, glycoproteins, lignin, and phenolic compounds and presents a barrier surrounding a matrix of protein and starch (Ginzburg, 1961; Letham, 1962; Varriano- Marston and Jackson, 1981; Shiga and Lajolo, 2006; Yi et al., 2016). As beans cook, the cell wall partially solubilizes and the pectin-rich middle lamella breaks down, which allows separation of adjacent cells and adequate softening of the bean (Rockland and Jones, 1974; Shomer et al., 1990). Cell wall content, integrity, and thickness as well as dietary fiber content as a whole have been linked increased cooking time resulting from seed age and storage conditions (Moscoso et al., 1984; Hincks and Stanley, 1987; Yousif and Deeth, 2003; Shiga et al., 2009; Yi et al., 2016; Siqueira et al., 2018). As seeds age, they exhibit less soluble fiber and increased cell wall content and take longer to cook (Moscoso et al., 1984; Yousif and Deeth, 2003; Shiga et al., 2009). The hard-to-cook phenomenon associated with improperly stored beans has been linked to increased cotyledon cell wall thickness due to lignin deposition, increased prevalence of cotyledon cell wall ruptures, and increased covalent bonding of pectin and other cell wall polysaccharides as demonstrated by Chigwedere, Nkonkola, et al. (2019), which could potentially be hindering cell separation (Hincks and Stanley, 1987; Yi et al., 2016; Siqueira et al., 2018). Exploring differences in the seed coat layer thickness and cotyledon cell wall thickness and composition of fast- and 11 slow-cooking genotypes within different seed types will be useful to reveal heritable traits associated with genetic variability of cooking time. Evaluation of cooking time is frequently performed on soaked beans. However, consumers often prepare dry beans without soaking (unsoaked) or with variable soaking times. While previous studies have identified a relationship between cooking times of unsoaked and soaked beans, the correlation is only moderate (R = 0.67) (Mendoza et al., 2018). In addition, cooking time of unsoaked beans has been found to be less heritable as compared to cooking time of soaked beans (Cichy et al., 2019). Understanding the stability of the fast cooking trait across unsoaked and soaked treatments can inform phenotyping methods so that germplasm can be accurately assessed for cooking time. In addition, observing responses to different soaked treatments could help reveal genetic factors associated with the fast cooking trait. The overall goal of this study was to characterize genetic variability of cooking time as it relates to seed coat layers and cotyledon cell wall traits within fast- and slow-cooking genotypes across four seed types: brown, cranberry, red mottled, and yellow. The genotypes were identified and categorized as fast or slow cooking in a large germplasm screening evaluating cooking time following a 12 hr soak (Cichy et al., 2015). The three objectives of this study are (1) to determine cooking times and water uptake of the 8 genotypes across 6 soaking times (0, 3, 6, 12, 18, and 24 hr), (2) to determine physical differences in seed coat macrosclereid- and osteosclereid layers and cotyledon cell wall thickness as well as seed weight and percentage of cotyledon, seed coat, and embryo and how these factors relate to genetic variability for cooking time, and (3) to determine compositional differences in total, soluble, and insoluble whole seed dietary fiber content and cotyledon cell wall isolate and how these factors relate to genetic variability for cooking time (Figure S1.1). 12 MATERIALS AND METHODS Germplasm The Phaseolus vulgaris germplasm relevant to this study consists of 8 genotypes with a faster and slower cooking genotype represented from four seed types: brown, cranberry, red mottled, and yellow. Seed type refers to seed appearance in terms of color, pattern, and shape, with like seeds considered to have the same seed type. The genotypes include brown beans W616488, ADP0037 (Brown-1) and Incomparable, ADP0027 (Brown-2), cranberry beans G23086, ADP0367 (Cranberry-1) and Katarina Kibala, ADP0515 (Cranberry-2), red mottled beans JB-178, ADP0436 (RedMottled-1) and PR0737-1, ADP0434 (RedMottled-2), and yellow beans Ervilha (Yellow-1) and PI527538 (Yellow-2) (Figure 1.1). The selected germplasm were identified after screening 206 lines of the Andean Diversity Panel for cooking time (Cichy et al., 2015). The genotypes designated “-1” are considered faster cooking compared to those marked “-2” within a seed type pair when cooked after soaking for 12 hr. W616488 (Brown-1), Incomparable (Brown-2), and PI527538 (Yellow-2) are part of the US Phaseolus germplasm collection. G23086 (Cranberry-1) is part of the International Center of Tropical Agriculture germplasm collection. Katarina Kibala (Cranberry-2) and Ervilha (Yellow- 1) were originally collected in an Angolan marketplace in 2010. JB-178 (RedMottled-1) was developed and released by the Centro de Investigación Agrícolas del Suroeste, Ministry of Agriculture of the Dominican Republic in cooperation with the University of Puerto Rico and the University of Nebraska (Arnaud-Santana et al., 2000). PR0737-1 (RedMottled-2) was developed and released cooperatively by the University of Puerto Rico Agricultural Experiment Station, the USDA–ARS, the Instituto Dominicano de Investigaciones Agropecuarias y Forestales, and the 13 Ministry of Agriculture, Natural Resources and Rural Development of the Republic of Haiti (Prophete et al., 2014). All genotypes used in this study were planted in 2017 at the Montcalm Research Farm in MI, which has Eutric Glossoboralfs (coarse-loamy, mixed) and Alfic Fragiorthods (coarse-loamy, mixed, frigid) soil types. The genotypes were grown in duplicate in a randomized complete block design of two row plots 4.75 m long with 0.5 m spacing between rows. Standard agronomic practices were followed as described in the MSU SVREC 2017 Farm Research Report (Kelly et al., 2017). Plants were hand-pulled at maturity and threshed with a Hege 140 plot harvester. Following harvest, beans were cleaned by hand to remove field debris, off types, and damaged beans. All beans were stored at room temperature for 3 months following harvest. Seed weights in g/100 seeds were determined for each genotype with 6 technical replications per field replicate. Water uptake and Cooking Time Evaluation Cooking times were determined for each field replicate of each genotype using automated Mattson cookers (Wang and Daun, 2005). Thirty beans per sample at 10-14% moisture were soaked in 250 ml distilled water for 0, 3, 6, 12, 18, or 24 hr. The beans were then patted dry of excess water and weighed to determine water uptake. For each sample, twenty-five beans were loaded onto 25 well Mattson cookers (Michigan State University Machine Shop, East Lansing, MI) with weighted (65 g) 2 mm diameter pins positioned in the center of each bean. Loaded Mattson cookers were placed into 4 L stainless steel beakers with 1.8 L of boiling distilled water on Cuisinart CB-30 Countertop Single Burners. A low boil was maintained until 80% of the beans were pierced completely. The 80% cooking time was recorded, and samples were cooled to room temperature and weighed to determine water uptake during cooking. Cooked samples were frozen 14 at -80 °C and freeze dried using a in a VirTis Genesis 12EL freeze dryer (Figure S1.1- Objective 1). Scanning Electron Microscopy For each genotype, unsoaked and soaked beans (12 hr in distilled water) were imaged using a JEOL JSM-6610LV scanning electron microscope at the Center for Advanced Microscopy at Michigan State University with an accelerating voltage of 12kV, spot size of 30, and working distance of 11 mm. For unsoaked beans, four beans per field replicate were cut into cross sections using double-edged razor blades, with four halves per field replicate mounted on an aluminum stub using high vacuum carbon tabs. For soaked beans, four beans per field replicate were cut to obtain approximately 4 mm thick slices including the hilum. These slices were fixed at 4 °C for 1 hr in 4% glutaraldehyde buffered with 0.1 M sodium phosphate (pH 7.4). After fixation, the slices were soaked for 30 min in 0.1 M sodium phosphate buffer (pH 7.4) and dehydrated in an ethanol series (25%, 50%, 75%, 95%) for 40 min at each gradation followed by three 30 min soaks in 100% ethanol. The slices were then critical-point dried in a Leica Microsystems EM CPD300 critical point dryer using carbon dioxide as the transitional fluid. The dried slices were mounted on aluminum stubs using high vacuum carbon tabs. Following mounting, each sample was platinum coated to a thickness of 8 nm while rotating. Four micrographs per cross-section were collected: two of seed coat cells and two of cotyledon cells (Figure S1.2). Seed coat cells were imaged at 750X, and cotyledon cells were imaged at 1200X for unsoaked samples and 600X for soaked samples. Example images of seed coat and cotyledon cells are depicted in Figures 1.2 and 1.3. Measurements of macrosclereid thickness (5/micrograph), osteosclereid thickness (5/micrograph), and cotyledon cell wall thickness (10/micrograph) were collected using ImageJ version 1.51j8 (Schneider et al., 2012). Macrosclereid cells form the outer layer of the seed coat, 15 and osteosclereid cells are located between the macrosclereid layer and the parenchyma layer of the seed coat. For macrosclereid and osteosclereid layers, measurements were collected in regions with well-defined edges separating the seed coat components. Cotyledon cell wall thickness was recorded only for cell walls perpendicular to the plane of view with well-defined edges to ensure accurate measurements were collected (Figure S1.1- Objective 2). Dietary Fiber Quantification For each field replicate of each genotype, raw beans were ground in a Kinematica PX-MFC 90 D laboratory hammer mill and passed through a 0.5 mm sieve. Milled samples were submitted to Great Lakes Scientific (Stevensville, USA) for dietary fiber analysis. Soluble, and insoluble dietary fiber content of whole seeds by dry weight were determined using the enzymatic- gravimetric methods AOAC 993.19 and AOAC 991.42 (AOAC, 1995b, 1995a). The sum of soluble and insoluble fiber is total fiber (Figure S1.1- Objective 3). Seed Component Percentage and Cotyledon Cell Wall Isolation For each genotype, 25 raw beans were wrapped in moist paper towels for 2 hr prior to separating seed coats and embryos from the cotyledons with forceps. The cotyledons, seed coats, and embryos were then frozen at -80 °C, freeze-dried in a VirTis Genesis 12EL freeze dryer, and weighed, and for each component, the percentage of total weight was calculated. Cotyledons were ground in a Kinematica PX-MFC 90 D laboratory hammer mill and passed through a 0.5 mm sieve (Figure S1.1- Objective 2). Water soluble and insoluble cell wall components were isolated in triplicate from the cotyledon cells from one field replicate of each genotype using the method described in Shiga and Lajolo, 2006 (Shiga and Lajolo, 2006). In brief, milled samples were defatted and digested via α-amylase, protease, and amyloglucosidase. After centrifuging, supernatants were dialyzed for 48 hr, freeze dried, and weighed to determine soluble cell wall 16 content, and the residues were washed, treated with 0.5 M sodium phosphate buffer, and sonicated in dimethyl sulfoxide with thorough rinsing between steps. The final residues were rinsed, freeze dried, and weighed to determine insoluble cell wall content. The sum of soluble and insoluble cell wall isolate is total cell wall isolate. Isolates were normalized to g per 1g cotyledon (Figure S1.1- Objective 3). Statistical Analysis All analyses of variance (ANOVA) in this study were conducted using the MIXED procedure in SAS version 9.4 of the SAS System for Windows. For Objective 1, genotype, soaking time, and genotype by soaking time were included as fixed effects with field replicate as a random effect. For Objective 2, genotype was included as a fixed effect with field replicate as a random effect for seed coat percentage, cotyledon percentage, and embryo percentage. Genotype, soaking time, and genotype by soaking time were included as fixed effects and field replicate, technical replicate(field replicate), micrograph(technical replicate), and measurement(micrograph) as random effects for cotyledon cell wall, macrosclereid, and osteosclereid thickness ANOVAs. For Objective 3, genotype was included as a fixed effect and field replicate as a random effect for total, insoluble, and soluble dietary fiber of whole seeds. For total, insoluble, and soluble cell wall isolate, genotype was included as a fixed effect and technical replicate as a random effect. In each case, mean separation was determined using pdiff within the mixed procedure and a Tukey multiple comparison adjustment. Spearman correlation coefficients (rs) were determined in R using the Cor function. Spearman correlations were used rather than Pearson correlations due to the monotonic, non-linear relationship between soaking time and cooking time, the two fundamental variables studied in this work. Principal component analysis (PCA) was performed with averages of all traits using the Prcomp function in R. 17 RESULTS Cooking Time and Water Uptake Objective 1 was to determine cooking times and water uptake of the 8 genotypes across 6 soaking times (0, 3, 6, 12, 18, and 24 hr) (Figure S1.1). Cooking time and water uptake were influenced by genotype and soaking time (Table S1.1). Individual genotypes were affected differently by soaking time, resulting in differences cooking time curves, rates of water uptake, and significant genotype by soaking interactions. This was observed in some genotypes cooking relatively more quickly than others when not soaked prior to cooking, but those same genotypes taking relatively longer to cook than others when soaked. Within each seed type, the same genotype was not necessarily the fastest cooking across the different soaking treatments. The seed type pairs were originally selected after screening for cooking times of 12 hr soaked beans. The genotypes selected as faster cooking lines (indicated “-1”) based on this screening generally cooked faster following all soaked treatments longer than 6 hr. However, for the unsoaked treatment, the genotypes selected as slower cooking lines (indicated “-2”) cooked faster than their counterpart in every seed type except brown. Cooking times ranged 16.7 - 108.1 min across all soaking times (Figures 1.1, S1.2). For the unsoaked treatment, Cranberry-2 had the fastest cooking time (82.3 min) and RedMottled-1 had the slowest (108.1 min). For the 3 hr soaking time, Yellow-1 had the fastest cooking time (24.9 min) and Brown-2 had the slowest (49.7 min). Cooking times began to stabilize for all genotypes beginning at 6 hr of soaking when the beans were fully hydrated. For all soaking times 6 hr and longer, Cranberry-1 generally was the fastest cooking genotype (16.7 – 18.0 min) and RedMottled- 2 was the slowest (32.8 – 36.0 min). 18 The water uptake ranged 47.8 - 110.2 % across all soaking times (Figures 1.1, S1.2). For the 3 hr soaking time, Brown-1 and Brown-2 had the lowest water uptake (47.8% & 53.0%) and Cranberry-1 and Cranberry-2 had the highest (80.6% & 75.4%). For soaking times of 6 hr and longer, the water uptake varied less across genotypes as they approached full hydration. Yellow-2 exhibited the lowest (89.2 - 95.7%) and Yellow-1 the highest (100.4 – 110.2%) water uptake for all soaking times 6 hr and longer. Seed and Physical Traits Objective 2 to determine physical differences in seed coat macrosclereid- and osteosclereid layers and cotyledon cell wall thickness as well as seed weight and percentage of cotyledon, seed coat, and embryo and how these factors relate to genetic variability for cooking time (Figure S1.1). Genotype significantly affected seed weight and percentage of seed coat, cotyledon, and embryo (Table S1.1). Seed weights ranged 38.9 – 64.4 g/100 seeds (Table 1.1). Percentage of seed coat, cotyledon, and embryo percentage exhibited narrow ranges across genotypes (0.4 – 3.1%). Within each seed type, the genotypes that cooked faster when soaked had higher seed weight, lower seed coat percentage, and higher cotyledon percentage than their slower-cooking counterparts. Seed coat macrosclereid-layer, seed coat osteosclereid-layer, and cotyledon cell wall thicknesses varied by genotype and soaking time and there was a significant genotype by soaking time interaction for these traits (Table S1.1). The macrosclereid layer is the outer layer of the seed coat, and the osteosclereid layer is located between the macrosclereid layer and the parenchyma layer of the seed coat (Figure 1.2). Seed coat macrosclereid-layer thickness for unsoaked beans ranged 39.6 – 48.6 μm and for the soaked beans ranged 23.4 – 39.6 μm (Figure 1.2; Table S1.2). Soaking decreased 19 macrosclereid-layer thickness 14.7 – 49.9%. The brown genotypes had the thickest macrosclereid layers when unsoaked, and these layers remained thicker than most other genotypes after soaking. Within seed type pairs, no trends were identified between unsoaked macrosclereid-layer thickness and cooking time of unsoaked or soaked beans. However, genotypes that cook faster when unsoaked had macrosclereid layers that thinned to a greater extent following soaking than those of their slower-cooking counterparts within all seed type pairs. This conveys that macrosclereid-layer thickness is relevant for cooking time of unsoaked beans as they hydrate during the cooking process. For soaked beans, macrosclereid-layer thickness does not appear to relate to cooking time. Seed coat osteosclereid-layer thickness of unsoaked beans ranged 10.4 – 15.5 μm and for soaked beans ranged 7.4 – 13.7 μm (Figure 1.2; Table S1.2). Soaking decreased osteosclereid- layer thickness 2.1 – 37.2%. For all seed types except yellow, thicker unsoaked osteosclereid layers were observed in genotypes that cooked slower when unsoaked. For all seed types except cranberry, osteosclereid layers were thinner after soaking in the genotypes that cook faster when unsoaked as compared to their slower-cooking counterparts. These trends convey that osteosclereid-layer thickness is relevant for cooking time of unsoaked beans both before and after they are partially or fully hydrated during the cooking process. Cotyledon cell wall thickness for unsoaked beans ranged 0.96 – 1.41 μm and for soaked beans ranged 0.63 – 1.12 μm (Figure 1.3; Table S1.2). Soaking decreased cotyledon cell wall thickness 17.5 – 37.0%. Within all seed type pairs, the genotypes that cooked faster when soaked had thinner cotyledon cell walls in both unsoaked and soaked treatments as compared to their slow- cooking counterparts. 20 Seed Coat and Cotyledon Cell Wall Compositional Traits Objective 3 was to determine compositional differences in total, soluble, and whole seed dietary fiber content and insoluble cotyledon cell wall isolate and how these factors relate to genetic variability for cooking time (Figure S1.1). Total and insoluble whole seed fiber and insoluble cotyledon cell wall isolate varied by genotype (Table S1.1). Total whole seed dietary fiber ranged 13.6 – 21.8 g/100 g milled beans; soluble whole seed fiber ranged 3.1 – 6.9 g/100 g milled beans; and insoluble whole seed fiber ranged 8.3 – 16.8 g/100 g milled beans (Figure 1.4; Table S1.3). Total and insoluble whole seed dietary fiber were lower in the genotypes that cook faster when soaked within all seed types except cranberry, which showed no significant differences for insoluble whole seed dietary fiber. Total cotyledon cell wall isolate ranged 106.4 – 134.7 mg/g cotyledon; soluble cotyledon cell wall isolate ranged 27.6 – 44.1 mg/g cotyledon; and insoluble cotyledon cell wall isolate ranged 67.0 – 95.0 mg/g cotyledon (Figure 1.4; Table S1.3). There were no significant differences between genotypes for total or soluble cotyledon cell wall isolate, but there were significant genotypic differences for insoluble cotyledon cell wall isolate. Within all seed types except brown, insoluble cotyledon cell wall isolate was lower in the genotypes that cook faster when soaked compared to their slow-cooking counterparts, although only RedMottled-1 and RedMottled-2 had insoluble cotyledon cell wall isolate values that were significantly different from each other. Principal Component Analysis and Correlations Principal component analysis (PCA) was performed to relate genetic variability for cooking time with seed coat and cotyledon cell wall physical and compositional traits to support the overall goal of this study. The first two principal components (PCs) explained about 65% of the variation (Figure 1.5). The first principal component (PC) separated the genotypes 21 approximately by cooking time of soaked beans and seed coat/cotyledon percentage and represented almost half of the variation (44.4%). The second PC represented over a sixth of the variation (20.0%) and separated the genotypes loosely by cooking time of unsoaked beans. The remaining PCs accounted for 14.8, 11.4, 5.3, 3.6, 0.4, and 0% of the variance respectively (data not shown). For each seed type pair, the genotype that cooks faster when soaked separated toward the bottom right of the biplot with the slower cooking genotypes separating toward the top left (Figure 1.5). The loading for cooking time after a 3 hr soak is positioned between those for cooking time of unsoaked beans and 12 hr soaked beans, representing the transition point from cooking time patterns of unsoaked beans and of soaked beans. The separation of the unsoaked, 3 hr soaked, and 12 hr soaked time points is also an indication of distinct physical and compositional factors related to genetic variability for cooking time depending upon whether beans were soaked and for how long. Significant correlations were identified among soaking time, cooking time, and water uptake such that longer soaking times decreased cooking times (rs = -0.37, p-value < 0.0001) and increased water uptake (rs = 0.81, p-value < 0.0001). (Figure S1.3). Cooking time decreased with increased water uptake (rs = -0.42, p-value < 0.0001). The PCA biplot indicates that genotypes with fast cooking times when unsoaked had increased levels of soluble cell wall isolate and soluble whole seed dietary fiber and thinner macrosclereid- and osteosclereid-layers (Figure 1.5). However, soluble whole seed dietary fiber and soluble cotyledon cell wall isolates were highly variable with insignificant ANOVAs so there is insufficient evidence to suggest a relationship between these traits and cooking time for this study (Figure 1.4; Tables S1.1 & S1.3). Osteosclereid-layer thickness of unsoaked beans was the only physical characteristic that significantly correlated with cooking time of unsoaked beans, such 22 that cooking time of unsoaked beans increased with osteosclereid-layer thickness (rs = 0.93, p- value = 0.0081) (Figure S1.4; Table 1.2). However, longer cooking times were associated with thicker macrosclereid (rs = 0.73, p-value = 0.0002) and osteosclereid layers in general (rs = 0.80, p-value = 0.0019) (Figure S1.5), and unsoaked beans exhibited the longest cooking times (Figures 1.1, S1.2). Soaking decreased the thickness of both macrosclereid (rs = -0.84, p-value < 0.0001) and osteosclereid layers (rs = -0.68, p-value = 0.0069) (Figure S1.5). The PCA biplot also indicates that genotypes with fast cooking times when soaked had higher seed weight, higher cotyledon percentage, lower seed coat percentage, lower cotyledon cell wall thickness, lower total and insoluble whole seed dietary fiber, and lower insoluble cell wall isolate (Figure 1.5). Significant correlations support the relationships between these traits and soaked cooking time. Cooking time of 12 hr soaked beans decreased as seed weight increased (rs = -0.81, p-value = 0.0047); seed coat percentage decreased (rs = 0.95, p-value = 0.0021), and cotyledon percentage increased (rs = -0.95, p-value = 0.0019) (Figure S1.6). The same trend was observed for the 6, 18, and 24 hr soaking times (Table 1.2). Bean genotypes with greater seed weight had a lower seed coat percentage (rs = -0.83, p-value = 0.0174) and higher cotyledon percentage (rs = 0.93, p-value = 0.0084), and seed coat percentage was strongly and negatively correlated with cotyledon percentage (rs = -1.00, p-value < 0.0001) (Figure S1.6). Cooking time increased with cotyledon cell wall thickness (rs = 0.78, p-value = 0.0033), and soaking decreased cotyledon cell wall thickness (rs = -0.73, p-value = 0.0038) (Figure S1.5). Unsoaked and 12 hr soaked cotyledon cell wall thickness significantly correlated with cooking time of 6 hr soaked beans, but not with cooking times of other soak treatments despite high rs (Figure S1.4; Table 1.2). Total and insoluble whole seed dietary fiber correlated positively with cooking time of 6 hr soaked beans, but not with cooking times from other soaked treatments (Figure S1.7; Table 1.2). Insoluble 23 cotyledon cell wall isolate was correlated with longer cooking times after a 12 hr soak (rs = 0.60, p-value = 0.0160) (Figure S1.7; Table 1.2). Similar correlations were identified for the 6, 18, and 24 hr soaking times (Table 1.2). DISCUSSION For all genotypes, cooking time decreased as soaking time approached six or more hours. This negative association between cooking time and soaking time was expected, as it has been identified in prior studies (Bellido et al., 2006; Chigwedere et al., 2019a). Soaking has been shown to activate cell wall enzymes that change the polysaccharide arrangement in the cell wall, which increases the rate of pectic polysaccharide thermosolubility and therefore decreases cooking time (Martínez-Manrique et al., 2011). In addition, beans that are fully or almost fully hydrated prior to cooking require limited water uptake during cooking. Beans soaked for 6 or more hours in this study took up an additional 20-40% of their dry weight in water during cooking, while unsoaked beans took up about 120% of their dry weight during cooking (data not shown). The genetic variability of how soaking influenced cooking time was interesting because genotypes that cooked faster when soaked (for 6 or more hours) were not the same ones that cooked faster unsoaked as observed in the cranberry, red mottled, and yellow seed types. This suggests that different physical or compositional factors determine cooking times of unsoaked and soaked beans. The extent to which soaking reduces cooking time has been shown to vary by genotype (Martínez-Manrique et al., 2011), which could explain why some genotypes are fast-cooking when soaked but not unsoaked. In addition, water uptake rates vary among genotypes, and some genotypes experience a lag time for water uptake, which would be irrelevant for soaked beans but could be affecting cooking time of unsoaked beans (Ross et al., 2010). Combining the factors that 24 decrease cooking times of soaked and unsoaked beans into a single cultivar could be a worthwhile quality attribute to appeal to a broad base of consumers. Figure 1.6 summarizes the traits associated with fast cooking time of unsoaked and soaked beans as determined in this study. For unsoaked beans, fast cooking time is associated with thin macrosclereid and osteosclereid layers in the seed coat (Figure 1.6). Darker beans tend to have thicker macrosclereid layers, as was observed in the brown genotypes, in part due to polyphenol storage (Agbo et al., 1987; Smýkal et al., 2014). Previous studies showed that the thickness of seed coat layers affects water uptake, particularly in the early stages of hydration (Sefa-Dedeh and Stanley, 1979; Deshpande and Cheryan, 1986a; Agbo et al., 1987). Water uptake rate has been shown to impact cooking time (Deshpande and Cheryan, 1986b) and could explain the differences in cooking time observed for the unsoaked and 3 hr soaked treatments, which require a large amount of water to be taken up during cooking as compared to longer soaked treatments. Raw seed macrosclereid- and osteosclereid layer thickness were not correlated with water uptake for any soaking time, although both layers thinned in all genotypes following a 12 hr soak as the beans expanded to accommodate water uptake. Earlier time points may be needed to capture the impact of unsoaked seed coat layer thickness on water uptake in these genotypes. However, negative correlations were observed for 12 hr soaked macrosclereid- and osteosclereid-layer thickness and water uptake after a 3 hr soak. This finding could indicate a relationship between the extent to which the seed coat thins when hydrating and the rate of water uptake, which helps to explain why genotypes with thinner hydrated macrosclereid and osteosclereid layers had faster cooking times when cooked unsoaked. This is in line with the finding that thinner seed coats are associated with faster water uptake (Sefa-Dedeh and Stanley, 1979; Deshpande and Cheryan, 1986a; Agbo et al., 1987). 25 For soaked beans, fast cooking time is associated with high seed weight, low seed coat percentage/high cotyledon percentage, thin cotyledon cell walls, low total and insoluble fiber, and low insoluble cotyledon cell wall isolate (Figure 1.6). Larger beans have lower seed coat percentage and higher cotyledon percentage, which were found to associate with fast cooking time of soaked beans. The relationship between seed weight and cooking time of soaked beans has been previously identified (Cichy et al., 2015). Hydrated seeds coats allow free movement of water into the bean, and larger beans have more surface area, allowing for an increased rate of water movement into the bean during cooking (Deshpande and Cheryan, 1986a). Seed weight, seed coat percentage, and cotyledon percentage were not associated with cooking time of unsoaked beans. The surface area of an unsoaked bean may be less relevant than the permeability of the seed coat for water uptake during cooking, explaining a lack of correlation between cooking time of unsoaked beans and seed weight. Seed coat impermeability has been previously associated with hardshell, a textural defect with both genetic and environmental causes that results in seed hardness (Bourne, 1967; Jackson and Varriano-Marston, 1981; Stanley, 1992). In dry beans and cowpea, seed hardness was found to decrease with water uptake and increase with cooking time (Sefa- Dedeh et al., 1978, 1979; Castellanos et al., 1995; Marques Corrêa et al., 2010). The relationship between cotyledon cell wall thickness and cooking time of soaked beans is consistent with a prior study that found thicker cotyledon cell walls were associated with poor cell separation and longer cooking times in the context of the hard-to-cook phenomenon (Yousif and Deeth, 2003). Lignification and associations between hemicellulose and nitrogenous compounds have been associated with cell wall thickening during storage (Hincks and Stanley, 1987; Yousif and Deeth, 2003). Differences in cotyledon cell wall thickness and capacity for cell wall thickening during storage could be associated with genetic variability for cooking time. 26 Thicker cells walls translate to increased fiber, which is largely comprised of cell wall polysaccharides (Shiga et al., 2009), and cotyledon cell wall isolate. Total and insoluble whole seed dietary fiber and insoluble cotyledon cell wall isolate were all positively correlated with cooking time of soaked beans in this study. Softening of the cotyledon during cooking is mainly attributed to cell wall polysaccharide solubilization and pectin solubilization in the middle lamella (Chigwedere et al., 2018). An increase in insoluble cotyledon cell wall isolate and thereby insoluble whole seed dietary fiber could delay or prolong this solubilization, causing the observed increase in cooking time of soaked beans. In addition, crude fiber content has been found to increase resistance to water uptake (Saio, 1976; Deshpande and Cheryan, 1986a). A fiber- associated resistance to water uptake into cotyledon cells during cooking could be contributing to increased cooking time of soaked beans as an expression of the genetic variability for cooking time. A relationship has been previously identified between increased fiber content and the hard- to-cook phenomenon, which prevents cotyledons from taking up water and expanding (Gloyer, 1921; Agbo et al., 1987; de Godínez, 1990; Rodriguez and Mendoza, 1990; Gonzalez and Paredes- Lapez, 1993). CONCLUSIONS This study evaluated cooking time, soaking time, physical traits, and cell wall and seed coat compositional traits across four seed types of dry beans. The relationship between soaking time and cooking time was explored across these seed types to reveal that genetic factors related to the fast-cooking trait are not consistent across unsoaked and soaked treatments. Physical and compositional traits of the seed coat and cotyledon cell wall were identified that relate to cooking time for unsoaked or soaked beans via spearman correlation, PCA, and general trends. These 27 relationships help to reveal factors associated with fast cooking time in both unsoaked and soaked beans. Cooking time of soaked beans appears to be related to seed weight, cotyledon/seed coat percentage, cotyledon cell wall thickness, insoluble cell wall isolate, and total and insoluble whole seed dietary fiber. These traits affect cell separation, water uptake, and water transport during cooking. The thicknesses of seed coat layers appear to be related to cooking time of unsoaked beans. These traits also affect water uptake and transport, but at an earlier stage in the hydration process. Understanding the factors associated with genetic variability for cooking time in unsoaked and soaked beans is useful to direct progress in breeding fast-cooking beans as well as to recognize the potential consequences of faster-cooking germplasm, including trade-offs like reduced fiber or seed coat integrity. ACKNOWLEDGEMENTS This work was supported in part by funding from the USDA-ARS Pulse Crop Health Initiative 3060-21650-001-00D and the U.S. Department of Agriculture, Agricultural Research Service Projects 5050-21430-01000D (K.C.). We are grateful to the Michigan State University Center for Advanced Microscopy for their supportive staff and use of their scanning electron microscope. 28 APPENDICES 29 APPENDIX A: CHAPTER 1 TABLES AND FIGURES Table 1.1 Means for all genotypes of seed weight; cotyledon, seed coat, and embryo percentage. Brown Cranberry Red Mottled Yellow Trait 1 2 1 2 1 2 1 2 Seed Weight (g/100 seeds) 58.9b 49.4cd 64.4a 44.4e 48.9d 38.9f 56.3b 52.1c Seed Coat Percentage (%) 7.1cde 8.5ab 6.7de 7.6bcd 8.0abc 9.1a 6.0e 7.1cde Cotyledon Percentage (%) 92.0abc 90.3de 92.5ab 91.1cd 91.0cd 89.8e 92.9a 91.8bc Embryo Percentage (%) 1.0bcd 1.2ab 0.8d 1.2a 0.9cd 1.1abc 1.1abc 1.1abc Mean separation within seed type pairs is indicated by the letter superscript. 30 Table 1.2 Spearman correlations of all traits with cooking times from 0, 3, 6, 12, 18, and 24 hr soaked samples. Soaking Time Seed Weight Seed Coat Percentage Cotyledon Percentage Embryo Percentage Macrosclereid Layer (U)‡ Macrosclereid Layer (S) Osteosclereid Layer (U) 0 -0.24 0.36 -0.36 -0.11 -0.07 0.55 0.93** Osteosclereid Layer (S) 0.67 Cotyledon Cell Wall (U) Cotyledon Cell Wall (S) Total Whole Seed Dietary Fiber Soluble Whole Seed Dietary Fiber Insoluble Whole Seed Dietary Fiber Total Cotyledon Cell Wall Isolate Soluble Cotyledon Cell Wall Isolate Insoluble Cotyledon Cell Wall Isolate 0.12 0.29 -0.05 -0.36 0.43 -0.17 -0.40 0.07 Cooking Time 6 12 18 24 -0.64* -0.81** -0.83** -0.83** 0.93** 0.95** 0.93** 0.93** -0.93** -0.95** -0.93** -0.93** 0.18 0.38 0.07 0.74 0.12 0.74* 0.76* 0.71* -0.50 0.88** 0.69 -0.55 0.38 0.14 0.02 0.74 0.24 0.76 0.83 0.69 0.26 0.05 0.00 0.76 0.19 0.69 0.74 0.55 0.26 0.05 0.00 0.76 0.19 0.69 0.74 0.55 -0.33 -0.45 -0.45 0.79 0.74 0.71 0.67 0.71 0.67 -0.45 -0.52 -0.52 0.62* 0.6* 0.67** 0.67* 3 -0.31 0.67 -0.67 0.14 0.60 0.57 0.48 0.45 0.43 0.52 0.52 -0.14 0.55 0.48 -0.33 0.55 † P-values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. ‡ U and S indicate unsoaked and soaked (12 hr), respectively 31 Figure 1.1 Scatterplots of cooking time and water uptake vs soaking time and images of the genotypes used in this study. Circles indicate cooking time and squares indicate water uptake. 32 Figure 1.2 Bar plots of seed coat layer thickness for unsoaked and soaked (12 hr) beans with seed type and genotype indicated. Example SEM images (RedMottled-1) of the seed coat layers are presented with the measured layers indicated. MS = Macrosclereid layer; OS = osteosclereid layer. 33 Figure 1.3 Bar plots of cotyledon cell wall thickness for unsoaked and soaked (12 hr) beans with seed type and genotype indicated. Example SEM images (RedMottled-1) of cotyledon cells are presented with locations of measurements indicated by white arrows. 34 Figure 1.4 Bar plots of soluble and insoluble whole seed dietary fiber and cotyledon cell wall isolate of raw beans with seed type and genotype indicated. 35 Figure 1.5 Principal component analysis biplot with each genotype indicated and loadings for cooking times across 0, 3, and 12 hr soaking times (CT0, CT3, and CT12); seed weight (SeedWt); seed coat (SeedCoat), cotyledon (Cotyledon), and embryo (Embryo) percentage; unsoaked and soaked (12 hr) macrosclereid-layer (MST0 and MST12), osteosclereid-layer thickness (OST0 and OST12), and cotyledon cell wall (CWT0 and CWT12); raw total (TCWI), soluble (SCWI), and insoluble (ISCWI) cotyledon cell wall isolate; and raw total (TFiber), soluble (SFiber), and insoluble (IFiber) whole seed dietary fiber. 36 Figure 1.6 Diagram of a dry bean cross section indicating the traits associated with fast cooking time of unsoaked (U) or soaked (S) beans. MS = macrosclereid layer; OS = Osteosclereid layer. 37 APPENDIX B: CHAPTER 1 SUPPLEMENTAL TABLES AND FIGURES Table S1.1 ANOVA results† indicating the significance of the fixed effects genotype, soaking time, and genotype by soaking time for all traits. Trait Water Uptake Cooking Time Seed Weight Seed Coat Percentage Cotyledon Percentage Embryo Percentage Macrosclereid-layer Thickness Osteosclereid-layer Thickness Cotyledon Cell Wall Thickness Total Whole Seed Dietary Fiber Genotype <0.0001 <0.0001 <0.0001 0.0001 <0.0001 0.0008 <0.0001 <0.0001 <0.0001 0.0003 Soluble Whole Seed Dietary Fiber NS Insoluble Whole Seed Dietary Fiber 0.0006 Total Cotyledon Cell Wall Isolate NS Soluble Cotyledon Cell Wall Isolate NS Insoluble Cotyledon Cell Wall Isolate 0.0001 Soak Genotype by Soak <0.0001 <0.0001 . . . . <0.0001 <0.0001 <0.0001 . . . . . . <0.0001 <0.0001 . . . . <0.0001 <0.0001 <0.0001 . . . . . . † P-values, where NS indicates p-values that are not significant at α = 0.05 38 Table S1.2 Means for all genotypes of unsoaked and soaked (12 hr) macrosclereid-layer, osteosclereid-layer, and cotyledon cell wall thickness. Trait Soak Brown 1 Cranberry Red Mottled Yellow 2 1 2 1 2 1 2 Macrosclereid-layer Thickness (μm) 0 48.6a 48.5a 46.0b 40.2d 40.7cd 41.7c 39.6d 46.7b 12 33.0b 39.6a 26.6c 26.3c 34.7b 23.8c 31.5b 23.4c Osteosclereid-layer Thickness (μm) 0 12.4cd 14.0b 11.4de 10.4e 15.5a 13.9b 13.1bc 13.7b 12 10.2c 13.7a 7.4f 9.9cd 11.4b 9.0de 11.5b 8.6e Cotyledon Cell Wall Thickness (μm) 0 1.01ef 1.39b 1.00f 1.20d 1.03e 1.32c 0.96g 1.41a 12 0.81e 1.12a 0.63g 0.92c 0.85d 0.93c 0.68f 0.97b Mean separation within seed type pairs is indicated by the letter superscript. 39 Table S1.3 Means for all genotypes (raw) of whole seed total, soluble, and insoluble fiber and total, soluble, and insoluble cotyledon cell wall isolate. Brown Cranberry Red Mottled Yellow Trait 1 2 1 2 1 2 1 2 Total Whole Seed Dietary Fiber (g/100 g milled beans) 16.5bcd 21.8a 16.3bcd 18.0bc 16.1cd 19.3ab 13.6d 18.4bc Soluble Whole Seed Dietary Fiber (g/100 g milled beans) 4.2a 6.9a 4.6a 5.0a 3.1a 4.1a 5.2a 4.5a Insoluble Whole Seed Dietary Fiber (g/100 g milled beans) 11.9bcd 16.8a 12.1bc 11.2cd 13.0bc 15.2ab 8.3d 13.8abc Total Cotyledon Cell Wall Isolate (mg/g cotyledon) 109.7a 115.8a 109.4a 120.8a Soluble Cotyledon Cell Wall Isolate (mg/g cotyledon) 27.7a 41.8a 42.4a 44.1a 122.7a 134.7a 32.0a 27.6a 106.4a 112.6a 39.1a 36.0a Insoluble Cotyledon Cell Wall Isolate (mg/g cotyledon) 82.0ab 74.0bc 67.0c 76.7bc 77.4bc 95.0a 67.3c 76.6bc Mean separation (by row) is indicated by the letter superscript. 40 Figure S1.1 Workflow depicting the steps for each objective. 41 Figure S1.2 Means of cooking time and water uptake for all genotypes across 0, 3, 6, 12, 18, and 24 hr soaking times. Within seed type pairs, the faster cooking time for each soaking time is indicated in red. Mean separation of each trait (by row) is indicated by the letter superscript. 42 Figure S1.3 Pairwise comparison matrix of soaking time (Soak), cooking time (CT), and water uptake (WU) across 0, 3, 6, 12 18, and 24 hr soaking times. Spearman correlation coefficients are indicated in the lower left, and scatterplots for each pairwise comparison with LOWESS regression lines are shown in the upper right. P-values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. 43 Figure S1.4 Pairwise comparison matrix of cooking times of unsoaked (CT0) and 12 hr soaked (CT12) beans, unsoaked (MST0) and 12 hr soaked (MST12) macrosclereid-layer thickness, unsoaked (OST0) and 12 hr soaked (OST12) osteoclereid-layer thickness, and unsoaked (CWT0) and 12 hr soaked (CWT12) cotyledon cell wall thickness. Spearman correlation coefficients are indicated in the lower left, and scatterplots for each pairwise comparison with LOWESS regression lines are shown in the upper right. P-values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. 44 Figure S1.5 Pairwise comparison matrix of soaking time (Soak), cooking time (CT), macrosclereid-layer thickness (MST), osteoclereid-layer thickness (OST), and cotyledon cell wall thickness (CWT). Spearman correlation coefficients are indicated in the lower left, and scatterplots for each pairwise comparison with LOWESS regression lines are shown in the upper right. P- values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. 45 Figure S1.6 Pairwise comparison matrix of cooking time of unsoaked beans (CT0) and cooking time of 12 hr soaked beans (CT12), seed weight (SeedWt), and seed coat (SeedCoat), cotyledon (Cotyledon), and embryo (Embryo) percentage. Spearman correlation coefficients are indicated in the lower left, and scatterplots for each pairwise comparison with LOWESS regression lines are shown in the upper right. P-values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. 46 Figure S1.7 Pairwise comparison matrix of cooking times of unsoaked (CT0) and 12 hr soaked (CT12) beans; total (TFiber), soluble (SFiber), and insoluble (IFiber) whole seed dietary fiber; and total (TCWI), soluble (SCWI), and insoluble (ICWI) cotyledon cell wall isolate. Spearman correlation coefficients are indicated in the lower left, and scatterplots for each pairwise comparison with LOWESS regression lines are shown in the upper right. P-values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. 47 REFERENCES 48 REFERENCES Agbo, G. N., Hosfield, G. L., Uebersax, M. A., and Klomparens, K. (1987). Seed microstructure and its relationship to water uptake in isogenic lines and a cultivar of dry beans (Phaseolus vulgaris L.). Food Struct. 6, 91–102. AOAC (1995a). Official Method 991.42. 16th ed. Gaithersburg, MD, USA: AOAC INTERNATIONAL. AOAC (1995b). Official Method 993.19. 16th ed. Gaithersburg, MD, USA: AOAC INTERNATIONAL. Arnaud-Santana, E., Nin, J. C., Saladin, F., Godoy-Lutz, G., Beaver, J. S., Coyne, D. P., et al. (2000). Registration of “JB-178” red mottled bean. Crop Sci. 40, 857–858. Bellido, G., Arntfield, S. D., Cenkowski, S., and Scanlon, M. (2006). Effects of micronization pretreatments on the physicochemical properties of navy and black beans (Phaseolus vulgaris L.). LWT - Food Sci. Technol. 39, 779–787. doi:10.1016/j.lwt.2005.05.009. Bourne, M. C. (1967). Size, density, and hardshell in dry beans. Food Technol. 21, 335–338. Castellanos, J. Z., Guzman-maldonado, H., Acosta-gallegos, J. A., and Kelly, J. D. (1995). Effects of hardshell character on cooking time of common beans grown in the semiarid highlands of Mexico. J Sci Food Agric 69, 437–443. doi: 10.1002/jsfa.2740690406. Chigwedere, C. M., Njoroge, D. M., Van Loey, A. M., and Hendrickx, M. E. (2019a). Understanding the relations among the storage, soaking, and cooking behavior of pulses: A scientific basis for innovations in sustainable foods for the future. Compr. Rev. Food Sci. Food Saf. 18, 1135–1165. doi:10.1111/1541-4337.12461. Chigwedere, C. M., Nkonkola, C. M., Rai, S., Kyomugasho, C., Kermani, Z. J., Pallares Pallares, A., et al. (2019b). Cotyledon pectin molecular interconversions explain pectin solubilization during cooking of common beans (Phaseolus vulgaris). Food Res. Int. 116, 462–470. doi:10.1016/j.foodres.2018.08.062. Chigwedere, C. M., Olaoye, T. F., Kyomugasho, C., Jamsazzadeh Kermani, Z., Pallares Pallares, A., Van Loey, A. M., et al. (2018). Mechanistic insight into softening of Canadian wonder common beans 522–531. doi:10.1016/j.foodres.2018.01.016. Cichy, K. A., Wiesinger, J. A., Berry, M., Nchimbi‐Msolla, S., Fourie, D., Porch, T. G., et al. (2019). The role of genotype and production environment in determining the cooking time of dry beans (Phaseolus vulgaris L.). Legum. Sci. 1, 1–15. doi:10.1002/leg3.13. cooking. Food Res. (Phaseolus Int. 106, vulgaris) during 49 Cichy, K. A., Wiesinger, J. A., and Mendoza, F. A. (2015). Genetic diversity and genome-wide association analysis of cooking time in dry bean (Phaseolus vulgaris L.). Theor. Appl. Genet. 128, 1555–1567. doi:10.1007/s00122-015-2531-z. Coelho, C. M. M., de Mattos Bellato, C., Santos, J. C. P., Ortega, E. M. M., and Tsai, S. M. (2007). Effect of phytate and storage conditions on the development of the ‘hard-to-cook’ phenomenon in common beans. J. Sci. Food Agric. 87, 1237–1243. doi:10.1002/jsfa.2822. Daher, F. B., and Braybrook, S. a. (2015). How to let go: pectin and plant cell adhesion. Front. Plant Sci. 6, 1–8. doi:10.3389/fpls.2015.00523. de Godínez, C. M. (1990). Changes in hardness of stored beans in relation to phytate and dietary fibre fraction contents. Deshpande, S. S., and Cheryan, M. (1986a). Microstructure and water uptake of phaseolus and winged beans. J. Food Sci. 51, 1218–1223. doi:10.1111/j.1365-2621.1986.tb13089.x. Deshpande, S. S., and Cheryan, M. (1986b). Water uptake during cooking of dry beans (Phaseolus vulgaris L.). Qual. Plant. Plant Foods Hum. Nutr. 36, 157–165. doi:10.1007/BF01092032. Furst, T., Connors, M., Bisogni, C. A., Sobal, J., and Falk, L. W. (1996). Food choice: a conceptual model of the process. Appetite 26, 247–266. doi:10.1006/appe.1996.0019. Galiotou-Panayotou, M., Kyriakidis, N. B., and Margaris, I. (2008). Phytase – phytate – pectin hypothesis and quality of legumes cooked in calcium solutions. J. Sci. Food Agric. 88, 355–361. doi:10.1002/jsfa. Garcia, E., Filisetti, T. M. C. C., Udaeta, J. E. M., and Lajolo, F. M. (1998). Hard-To-cook beans (Phaseolus vulgaris): Involvement of phenolic compounds and pectates. J. Agric. Food Chem. 46, 2110–2116. doi:10.1021/jf970848f. Garcia, E., Lajolo, F. M., Paulo, U. D. S., and Paulo, S. (1994). Starch alterations in hard-to-cook beans (Phaseolus vulgaris). J. Agric. Food Chem. 42, 612–615. doi:10.1021/jf00039a002. Gargiulo, L., Sorrentino, G., and Mele, G. (2020). 3D imaging of bean seeds: Correlations between hilum Int. 134, 109211. doi:10.1016/j.foodres.2020.109211. Ginzburg, B. Z. (1961). Evidence for a protein gel structure cross-linked by metal cations in the intercellular cement of plant tissue. J. Exp. Bot. 12, 85–107. doi:10.1093/jxb/12.1.85. Gloyer, W. O. (1921). Sclerema and hardshell, two types of hardnes of the bean. in Proceedings of the Association of Official Seed Analysts of North America (The Association of Official Seed Analysts), 60. and hydration kinetics. Food Res. region structures 50 Gonzalez, E., and Paredes-Lapez, O. (1993). Hard-to-cook phenomenon in common beans — a review. Crit. Rev. Food Sci. Nutr. 33, 227–286. doi:10.1080/10408399309527621. Hamrick, K. S., Andrews, M., Guthrie, J., Hopkins, D., and McClelland, K. (2012). How much time do Americans spend on food? Am. Food Choices Sel. Res. Time Diet, 1–85. doi:10.22004/ag.econ.291940. Hernandez-Unzon, H. Y., and Ortega-Delgado, M. L. (1989). Phytic acid in stored common bean seeds (Phaseolus vulgaris L.). Plant Foods Hum. Nutr., 209–221. Hincks, M. J., and Stanley, D. W. (1987). Lignification: Evidence of a role in hard-to-cook beans. J. Food Biochem. 11, 41–58. doi:10.1111/j.1745-4514.1987.tb00111.x. Jabs, J., and Devine, C. M. (2006). Time scarcity and food choices: An overview. Appetite 47, 196–204. doi:10.1016/j.appet.2006.02.014. Jackson, G. M., and Varriano-Marston, E. (1981). Hard‐to‐cook phenomenon in beans: effects of accelerated storage on water absorption and cooking time. J. Food Sci. 46, 799–803. doi:10.1111/j.1365-2621.1981.tb15351.x. Katuuramu, D. N., Luyima, G. B., Nkalubo, S. T., Wiesinger, J. A., Kelly, J. D., and Cichy, K. A. (2020). On-farm multi-location evaluation of genotype by environment interactions for seed yield and cooking time in common bean. Sci. Rep. 10, 3628. doi:10.1038/s41598-020-60087-2. Kelly, J., Wright, E., and Bornowski, N. (2017). 2017 Dry bean yield trials, in 2017 Research report: Saginaw valley research and extension center. Letham, D. S. (1962). Separation of plant cells with hexametaphosphate and the nature of intercellular bonding. Exp. Cell Res. 27, 352–355. doi:10.1016/0014-4827(62)90244-6. Marbach, I., and Mayer, A. M. (1974). Permeability of seed coats to water as related to drying conditions and metabolism of phenolics. Plant Physiol. 54, 817–820. doi:10.1104/pp.54.6.817. Marbach, I., and Mayer, A. M. (1975). Changes in catechol oxidase and permeability to water in seed coats of Pisum elatius during seed development and maturation. Plant Physiol. 56, 93–96. doi:10.1104/pp.56.1.93. Marques Corrêa, M., Jaeger de Carvalho, L. M., Regini Nutti, M., Luiz Viana de Carvalho, J., Hohn Neto, A. R., and Gomes Ribeiro, E. M. (2010). Water absorption, hard shell and cooking time of common beans (Phaseolus vulgaris L.). African J. Food Sci. Technol. 1, 13–20. Available at: https://core.ac.uk/download/pdf/45489866.pdf. Martínez-Manrique, E., Jacinto-Hernández, C., Garza-García, R., Campos, A., Moreno, E., and Bernal-Lugo, I. (2011). Enzymatic changes in pectic polysaccharides related to the beneficial effect of soaking on bean cooking time. J. Sci. Food Agric. 91, 2394–2398. doi:10.1002/jsfa.4474. 51 on the navy bean J. 52, seed. J. Prev. Med. 47, Agron. indicators of eating. Am. Mendoza, F. A., Wiesinger, J. A., Lu, R., Nchimbi‐Msolla, S., Miklas, P. N., Kelly, J. D., et al. (2018). Prediction of cooking time for soaked and unsoaked dry beans (Phaseolus vulgaris L.) using hyperspectral imaging technology. Plant Phenome J. 1, 1–9. doi:10.2135/tppj2018.01.0001. Monsivais, P., Aggarwal, A., and Drewnowski, A. (2014). Time spent on home food preparation and 796–802. healthy doi:10.1016/j.amepre.2014.07.033. Moscoso, W., Bourne, M. C., and Hood, L. F. (1984). Relationships between the hard-to-cook phenomenon in red kidney beans and water absorption, puncture force, pectin, phytic acid, and minerals. J. Food Sci. 49, 1577–1583. doi:10.1111/j.1365-2621.1984.tb12848.x. Powrie, W. D., Adams, M. W., and Pflug, I. J. (1960). Chemical, anatomical, and histochemical studies 163–167. doi:10.2134/agronj1960.00021962005200030013x. Prophete, E., Demosthenes, G., Godoy-Lutz, G., Porch, T. G., and Beaver, J. S. (2014). Registration of PR0633-10 and PR0737-1 red mottled dry bean germplasm lines with resistance to BGYMV, BCMV, BCMNV, and common bacterial blight. J. Plant Regist. 8, 49–52. doi:10.3198/jpr2013.08.0046crg. Rockland, L. B., and Jones, F. T. (1974). Scanning electron microscope studies on dry beans. Effects of cooking on the cellular structure of cotyledons in rehydrated large lima beans. J. Food Sci. 39, 342–346. doi:10.1111/j.1365-2621.1974.tb02890.x. Rodriguez, F. M., and Mendoza, E. M. T. (1990). Physicochemical basis for hardseededness in mung bean J. Agric. Food Chem. 38, 29–32. doi:10.1021/jf00091a006. Ross, K. A., Arntfield, S. D., Cenkowski, S., and Fulcher, G. R. G. (2010). Relating mechanical changes at the glass transition with water absorption behavior of dry legume seeds. Int. J. Food Eng. 6, Article 13. doi:10.2202/1556-3758.1780. Saio, K. (1976). Soybeans resistant to water absorption. Cereal Foods World 21, 168–173. Schneider, C. A., Rasband, W. S., and Eliceiri, K. W. (2012). NIH Image to ImageJ: 25 years of image analysis. Nat. Methods 9, 671–675. doi:10.1038/nmeth.2089. Sefa-Dedeh, S., and Stanley, D. (1979). The relationship of microstructure of cowpeas to water absorption and dehulling properties. Cereal Chem. 56, 379–386. Sefa-Dedeh, S., Stanley, D. W., and Voisey, P. W. (1978). Effects of soaking time and cooking conditions on texture and microstructure of cowpeas (Vigna unguiculata). J. Food Sci. 43, 1832– 1838. doi:10.1111/j.1365-2621.1978.tb07426.x. (L.) Wilczek). (Vigna radiata 52 of Carbohydr. Polym. 76, and beans. Polym. 63, hard-to-cook structure. Carbohydr. Sefa-Dedeh, S., Stanley, D. W., and Voisey, P. W. (1979). Effect of storage time and conditions on the hard-to-cook defect in cowpeas (Vigna unguiculata). J. Food Sci. 44, 790–796. doi:10.1111/j.1365-2621.1979.tb08503.x. Shiga, T. M., Cordenunsi, B. R., and Lajolo, F. M. (2009). Effect of cooking on non-starch polysaccharides 100–109. doi:10.1016/j.carbpol.2008.09.035. Shiga, T. M., and Lajolo, F. M. (2006). Cell wall polysaccharides of common beans (Phaseolus vulgaris L.) - Composition 1–12. doi:10.1016/j.carbpol.2005.06.025. Shomer, I., Paster, N., Lindner, P., and Vasiliver, R. (1990). The role of cell wall structure in the hard-to-cook phenomenon in beans (Phaseolus vulgaris L.). Food Struct 9, 139–149. Siqueira, B. S., Fernandes, K. F., Brito, P. V. A., and Santos, F. C. A. (2018). Histochemical and ultrastructural characterization of easy-to-cook and hard-to-cook carioca bean genotypes. LWT 97, 117–123. doi:10.1016/j.lwt.2018.06.048. Smýkal, P., Vernoud, V., Blair, M. W., Soukup, A., and Thompson, R. D. (2014). The role of the testa during development and in establishment of dormancy of the legume seed. Front. Plant Sci. 5, 1–19. doi:10.3389/fpls.2014.00351. Snyder, E. B. (1936). Some factors affecting the cooking quality of the pea and great northern types of dry beans. Hist. Res. Bullitins Nebraska Agric. Res. Stn. 85, 145. Stanley, D. W. (1992). Hard beans—A problem for growers, processors, and consumers. Horttechnology 2, 370–378. doi:10.21273/HORTTECH.2.3.370. Tully, R. E., Musgrave, M. E., and Leopold, A. C. (1981). The seed coat as a control of imbibitional chilling injury. Crop Sci. 21, 312–317. doi:10.2135/cropsci1981.0011183X002100020026x. Valle, J. M., Stanley, D. W., and Bourne, M. C. (1992). Water absorption and swelling in dry bean seeds. J. Food Process. Preserv. 16, 75–98. doi:10.1111/j.1745-4549.1992.tb00193.x. Varriano-Marston, E., and Jackson, G. M. (1981). Hard-to-cook phenomenon in beans: Structural changes during storage and imbibition. J. Food Sci. 46, 1379–1385. doi:10.1111/j.1365- 2621.1981.tb04179.x. Waldron, K. W., Parker, M. L., and Smith, A. C. (2003). Plant cell walls and food quality. Compr. Rev. Food Sci. Food Saf. 2, 128–146. doi:10.1111/j.1541-4337.2003.tb00019.x. Wang, N., and Daun, J. K. (2005). Determination of cooking times of pulses using an automated Mattson cooker apparatus. J. Sci. Food Agric. 85, 1631–1635. doi:10.1002/jsfa.2134. 53 beans (Phaseolus Yi, J., Njoroge, D. M., Sila, D. N., Kinyanjui, P. K., Christiaens, S., Bi, J., et al. (2016). Detailed analysis of seed coat and cotyledon reveals molecular understanding of the hard-to-cook defect of common 481–490. doi:10.1016/j.foodchem.2016.05.018. Yousif, A. M., and Deeth, H. C. (2003). Effect of storage time and conditions on the cotyledon cell wall of the adzuki bean (Vigna angularis). Food Chem. 81, 169–174. doi:10.1016/S0308- 8146(02)00409-0. vulgaris L.). Food Chem. 210, 54 CHAPTER 2: GENETIC VARIABILITY AND GENOME-WIDE ASSOCIATION ANALYSIS OF FLAVOR AND TEXTURE IN COOKED BEANS (Phaseolus vulgaris L.) [Submitted for publication in Theoretical and Applied Genetics] 55 Genetic variability and genome-wide association analysis of flavor and texture in cooked beans (Phaseolus vulgaris L.) Amber Bassett1, Kelvin Kamfwa2, Daniel Ambachew34, Karen Cichy15* 1 Department of Plant, Soil and Microbial Sciences, Michigan State University, East Lansing, MI, USA 2 Department of Plant Science, University of Zambia, Lusaka, Zambia 3 Southern Agricultural Research Institute, Hawassa, Ethiopia 4 Department of Agricultural and Environmental Sciences, Tennessee State University, Nashville, Tennessee 5 Sugarbeet and Bean Research Unit, USDA-ARS, East Lansing, MI, USA ABSTRACT Dry beans are a nutritious food recognized as a staple globally, but consumption is low in the US. Improving dry bean flavor and texture through breeding has the potential to improve consumer acceptance and suitability for new end-use products. Little is known about the genetic variability and inheritance of bean sensory characteristics. A total of 430 genotypes of the Andean Diversity Panel of 20 seed types were grown in three locations, and cooked seeds were evaluated by a trained sensory panel for flavor and texture attribute intensities, including total flavor, beany, vegetative, earthy, starchy, sweet, bitter, seed coat perception, and cotyledon texture. Extensive variation in sensory attributes was found across and within seed types. A set of genotypes was identified that exhibit extreme attribute intensities generally stable across all three environments. Seed coat perception and total flavor intensity had the highest broad-sense heritability (0.39 and 0.38 respectively), while earthy and vegetative intensities exhibited the lowest (0.14 and 0.15 56 respectively). Starchy and sweet flavors were positively correlated and highest in white bean genotypes according to PCA. SNPs associated with total flavor intensity (6 SNPs across three chromosomes), beany (5 SNPs across 4 chromosomes), earthy (3 SNPs across two chromosomes), starchy (1 SNP), bitter (1 SNP), seed coat perception (3 SNPs across 2 chromosomes), and cotyledon texture (2 SNPs across 2 chromosomes) were detected. These findings lay a foundation for incorporating flavor and texture in breeding programs for the development of new varieties that entice growers, consumers, and product developers alike. INTRODUCTION Dry beans (Phaseolus vulgaris L.) are a nutritious food that serve as a staple in many majority-world countries (Akibode and Maredia, 2011). Despite their global pervasiveness, they have limited consumption in the US, with only 2.2 kg per capita consumed in 2019 (Parr and Lucier, 2020). In the US, primary breeding goals for dry beans include yield, processing quality, disease resistance, architecture, agronomic adaptation, stress tolerance, and grower friendliness, which encompasses traits that reduce labor and inputs required by growers (Kelly and Cichy, 2012). Quality characteristics such as flavor and texture, however, have largely been overlooked in breeding programs. Quality is most commonly addressed through processing and the addition of sauces and flavors, especially to canned beans and bean products, often at the expense of nutritional value (Borchgrevink, 2013; Roland et al., 2017; Gilham et al., 2018). Taste is a primary factor driving consumer purchasing decisions of food, which motivates food companies to invest heavily in this aspect of product development (William Blair, 2016; IFIC, 2019). Consumers are also very interested in clean labels and food products with few additives (Asioli et al., 2017). Therefore improving dry bean flavor and texture through breeding has the potential to increase 57 consumer acceptance and utilization of beans and inclusion of beans as ingredients in products while appealing to consumers’ interest in flavor without many additives. Along with cooking time and price, flavor and texture are important characteristics that consumers consider when purchasing dry beans, influencing their decisions regarding market class and product type (Castellanos et al., 1997; Scott and Maideni, 1998; Leterme and Carmenza Muũoz, 2002; Eihusen and Albrecht, 2007; Winham et al., 2019). However, for many consumers, beans are not palatable, and the beany flavor they impart when used as ingredients is often perceived as undesirable (Nachay, 2017; Dougkas et al., 2019). Flavor and texture are not typically evaluated prior to variety release in the U.S., and this lack of focus on sensory quality may be limiting consumption of dry beans below their potential. A breeding approach to address flavor and texture in beans has not been explored in part due to the complexity and cost associated with sensory evaluations. Protocols have been developed for the preparation and evaluation of cooked bean samples as well as the training and maintenance of sensory panels (Koehler et al., 1987; Sanz-Calvo and Atienza-del-Rey, 1999; Romero del Castillo et al., 2008; Romero del Castillo et al., 2012), but these protocols are designed for few samples with plentiful seed and are not feasible to implement in breeding programs. The application of these sensory methods have identified genetic variability for texture and flavor acceptability (Koehler et al., 1987) and attribute intensities, including seed-coat perception, roughness, mealiness, and beany flavor (Rivera et al., 2013). This indicates that sensory quality can be addressed by harnessing the genetic variability present through breeding, provided appropriate phenotyping methods are available. There is a need for further evaluation of genetic variability for sensory attributes within P. vulgaris to understand the full range of attribute 58 intensities available and to assess the genetic control of these attributes. These are important steps to develop a breeding program that incorporates flavor and texture. For this study, a modified quantitative descriptive analysis approach was developed and applied to the screening of 1,940 samples for cooked bean flavor and texture. This approach was used to address three objectives: (1) to evaluate nine sensory attributes in 430 genotypes of a dry bean diversity panel grown in three locations, (2) to examine the relationships among sensory attributes, seed types, and cooking time, and (3) to identify genetic markers associated with sensory attributes across multiple locations. MATERIALS AND METHODS Germplasm Subsets of the Andean Diversity Panel were grown and evaluated across three locations for this study. The genetic composition and germplasm origin of the ADP is described by Cichy et al. (2015) and included in Table S2.1. Only Andean genotypes were included in statistical and GWAS analyses. The Southern Agricultural Research Institute provided seeds from 373 Andean genotypes grown in Hawassa, Ethiopia in Fall 2015, and the University of Zambia provided seeds from 251 Andean genotypes grown in Kabwe, Zambia and 356 Andean genotypes grown in Lusaka, Zambia in Spring 2018. Combined, a total of 430 genotypes were represented covering 20 seed types. Raw seed weights were recorded for each field rep as grams per 100 seeds. In Hawassa, the ADP was grown during the main cropping season (July to October) in 2015 at the Hawassa Research Station, which has soil classified as Eutric Fluvisol with a pH of 7.0. The ADP genotypes were planted using an augmented design with genotypes arranged in 21 blocks, which each contained 13 test entries and 5 standard checks randomly allocated. Each 59 genotype was planted in two-row plots with 0.4 m and 0.1m inter-row and intra-row spacing, respectively. Each block was spaced 1 m apart. Fertilizer in the forms of urea (46% N, 0% P2O5, 0% K2O) and DAP (8% N, 46% P2O5, 0% K2O) were applied at a rate of 100 kg/ha. In Kabwe, the ADP was grown at the Zambia Agricultural Research Institute Farm, which has soil classified as Ultisol and had a pH of 5.0. In Lusaka, the ADP was grown in the field during the rainy season in 2017 at the University of Zambia Research Farm, which has soil classified as fine loamy Isohyperthermic Paleustalf with a pH of 5.5. During the 2017 rainy season a total of 850 mm of rain was received at the experimental site at the University Farm. In both Zambia locations, the ADP genotypes were planted using a randomized complete block designs with two replications. In each replication a genotype was planted in a single-row plot that was 4 M long with 0.60 M inter-row spacing. A compound fertilizer (10P: 20P: 10K) was applied to the experimental site at a rate of 100 Kg Ha-1 just before planting. Genotypes exhibiting extreme attribute intensities along with Red Hawk (dark red kidney) and Etna (cranberry) were grown at the Montcalm Research Farm in MI in 2018. The soil type is Eutric Glossoboralfs (coarse-loamy, mixed) and Alfic Fragiorthods (coarse-loamy, mixed, frigid). Two row plots 4.75 m long with 0.5 m spacing between rows were arranged in a randomized complete block design with two replications per genotype. Standard agronomic practices were followed as described in the MSU SVREC 2018 Farm Research Report (Kelly et al., 2018). Cooking Time Evaluation For each location, two replicates of 30 seed per genotype were equilibrated to 10-14% moisture in a 4 °C humidity chamber prior to evaluating for cooking time. For the seed from both locations in Zambia, each replicate corresponded to a field replicate. For the seed from Hawassa, Ethiopia, the single field replicate for each genotype was split to create two replicates. Each 30 60 seed sample was soaked for 12 hours in distilled water and weighed prior to cooking time evaluation using an automated Mattson cooker method (Wang and Daun, 2005). Genotypes were cooked in a random order to minimize seed aging effects. Mattson cookers were loaded with soaked seeds and placed in boiling distilled water to cook. The Mattson cookers (Michigan State University Machine Shop, East Lansing, MI) use twenty-five 65g stainless steel rods with 2mm diameter pins to pierce beans as they finish cooking in each well. As the pins drop, a custom software reports the cooking time associated with each pin. The cooking times were recorded, with the 80% cooking time regarded as the time required to fully cook each sample. Cooked samples were weighed and total water uptake following cooking was calculated. Sensory Evaluation The ADP subsets from each location were evaluated in duplicate by four panelists each using a Quantitative Descriptive Analysis (QDA) approach (Stone et al., 1974) in which each panelist independently evaluated samples using a non-consensus approach to limit group bias. QDA has been found to yield reproducible measurements with small differences for boiled dry beans, although it is typically applied to small numbers of samples due to the substantial time and personnel commitment it requires (McTigue et al., 1989). For the purposes of this study, the QDA approach was modified to make it feasible to screen hundreds of samples with replication using a small number of panelists, which is necessary for implementation in public breeding programs with limited resources. For each location, seeds were prepared for sensory evaluation in the same order as for cooking time evaluation. Four panelists were present at each sensory evaluation session, scheduled according to their availability. Sensory evaluation sessions were held daily until each genotype had been evaluated twice for each location. For the Ethiopia location, twenty genotypes were evaluated at each session. For the Zambia locations, twelve genotypes including 61 cranberry (Etna) and dark red kidney (Red Hawk) bean controls grown at the Montcalm Research Center were evaluated at each session. Each sample was evaluated using 5-point attribute intensity scales (low → high intensity) for total, beany, vegetative, earthy, starchy, bitter, and sweet flavor intensities as well as seed coat perception and cotyledon texture. The scale for seed coat perception ranged from imperceptible (1) to tough and lingering (5). For cotyledon texture, the scale ranged from mushy (1) to very gritty/firm (5) (Table S2.2). This sensory evaluation protocol was approved by the Institutional Review Board of Michigan State University (IRB# x16-763e Category: Exempt 6). Panel Training and Assessment Panelists were recruited from the USDA-ARS (East Lansing, MI) and Michigan State University Dry Bean Breeding programs due to their familiarity with dry beans and their availability for long term sensory evaluation projects. An initial training session was conducted with eight panelists using a consensus approach to determine which attributes to evaluate and how to evaluate them. A diverse set of dry bean genotypes was selected from the USDA and MSU dry bean programs with the intention of exposing panelists to a wide range of attribute intensities. This initial set included black, cranberry, dark red kidney, great northern, Jacob’s cattle, navy, pink, pinto, small red, and yellow beans. Following screening of the ADP grown in Hawassa, Ethiopia, a training set of genotypes exhibiting extreme attribute intensities was developed (Table 2.1, Figure S2.1). This set was used to train eleven panelists to rate the selected attributes prior to evaluating the ADP grown in the Zambia locations. For the sensory evaluation of the ADP from both Zambia locations, Red Hawk and Etna were used as controls. Red Hawk (Kelly et al., 1998), a dark red kidney bean, is a variety 62 released by the Michigan State University dry bean breeding program. Etna (PI 546490), a cranberry bean, is a private variety developed by Seminis of Monsanto Vegetable Seeds. Panelists were trained over multiple sessions using a non-consensus approach to improve their familiarity with the selected scales and their sensory evaluation skills. Panelist performance was assessed via ANOVA with FGenotype (p-value < 0.05) indicating ability to discriminate and Frep (p-value > 0.05) indicating consistency (Meilgaard et al., 1999; Armelim et al., 2006). Sensory evaluation of each location commenced after successful training of each panelist. Following screening of the ADP from each location, panel performance was assessed as during training. Sample Preparation for Sensory Evaluation A standardized method for preparing boiled dry beans for sensory evaluation was previously developed (Romero del Castillo et al., 2012), but could not be applied in this study due to limited seed per genotype. Instead, the preparation method used by Mkanda et al. (2007) was modified to suit smaller seed volumes and a larger number of samples, as well as maintain consistent soaking time with the cooking evaluation method. In preparation for each sensory evaluation session, large tea bags filled with 12 hour soaked seeds were boiled in distilled water for the cooking time determined by the Mattson cooker method, timed so they all finished cooking together. No salt was added. The cooked samples were poured into preheated (105 °C) ceramic ramekins, covered with aluminum foil, and placed in a chafing dish to maintain temperature. Samples were given a random letter code to mask their identity. Panelists were asked to refrain from wearing strong scents or eating during the hour before each session. Samples were served out of the ceramic ramekins with a plastic spoon onto paper plates. Lemon water was made available as a palette cleanser, and panelists were asked to drink water between samples. 63 Statistics PROC MIXED in SAS version 9.4 of the SAS System for Windows (SAS Institute Inc. Cary, NC, USA) was used to conduct ANOVAs for each recorded trait. For raw seed weight, soak water uptake, cooking time, and total water uptake traits, the fixed effects were genotype, location, and genotype by location with replicate as a random effect. For the sensory attribute intensity traits, the fixed effects were genotype, location, and genotype by location with rep, panelist(location), and session(location) as random effects. Least squares estimates (LSEs) for sensory traits were calculated via the LSMeans statement in PROC MIXED for visualization of trait distributions with outliers excluded. To evaluate differences among seed types, ANOVAs were also performed with the seed type, location, and seed type by location as fixed effects and rep, panelist(location), and session(location) as random effects. To analyze all locations combined while minimizing environmental effects, best linear unbiased predictors (BLUPs) were generated for each trait using the lme4 package (Bates et al., 2015) in R (R Core Team, 2017) with genotype, location, genotype by location, and rep nested in location as random effects. For sensory traits, panelist nested in location and session nested in location were also included as random effects. For analysis within individual locations, BLUPs were calculated for sensory traits with genotype, rep, panelist, and session included as random effects. Broad sense heritability (H2) was calculated on a family mean basis for each trait using the equation var(G)/(var(G)+(var (G*L)/no. loc)+(var(error)/no. loc * rep), where var is variance, G is genotype, and G*L is genotype by location, and no. loc is number of locations. Variance components were calculated using PROC VARCOMP in SAS version 9.4 with method = restricted maximum likelihood method (reml) (Holland et al., 2003). Pearson correlation coefficients among 64 traits were determined with BLUPs from all locations combined using the Cor function in R. Principle component analysis among traits was conducted with BLUPs from all locations combined using the prcomp function in R. Genotyping The ADP has been genotyped previously via genotyping-by-sequencing (GBS), and associated data including hapmaps are available at the Feed the Future – Development and Characterization of the Common Bean Diversity Panel (ADP) website (http://arsftfbean.uprm.edu/bean/) (Katuuramu et al., 2018). In brief, two GBS libraries were constructed at 364-plex and 137-plex as described by Elshire et al. (2011) with modifications described by Hart and Griffiths (2015). The raw sequencing data are available in association with BioProject accession number PRJNA290028 in the NCBI BioProject database (https://www.ncbi.nlm.nih.gov/bioproject/). For this study, the raw sequence data were cleaned of adapters and trimmed for quality score ≥ 30 and minimum length ≥ 30 via Cutadapt (Martin, 2011) and evaluated via FastQC (Andrews, 2010). Cleaned reads were demultiplexed using the Next Generation Sequencing Eclipse Plugin (NGSEP) pipeline with NGSEP version 3.0.2 (Duitama et al., 2014; Perea et al., 2016), aligned to the Phaseolus vulgaris v2.1 genome (DOE-JGI and USDA-NIFA, http://phytozome.jgi.doe.gov/) using Bowtie 2 (Langmead and Salzberg, 2012), and sorted using Picard (http://www.bioinformatics.babraham.ac.uk/projects/fastqc). Variant calling and annotation were performed via NGSEP. Raw SNPs were filtered to eliminate those with more than 90% missing data, and remaining missing data were imputed using FILLIN in Tassel 5.2.31 (Bradbury et al., 2007a; Swarts et al., 2014). 65 Genome Wide Association Genome-wide association analyses were performed with Bayesian-information and Linkage-disequilibrium Iteratively Nested Keyway (BLINK) (Huang et al., 2018) in R. BLINK has increased statistical power as compared to other methods and better controls for false negatives and false positives (Liu et al., 2016; Huang et al., 2018). Instead of using kinship, BLINK uses iterations to select a set of markers associated with a trait of interest, which are fitted as covariates. The first 3 principle components were determined using prcomp in R and included in each analysis to control for population structure. Single nucleotide polymorphisms (SNPs) with MAF < 0.05 or with more than two alleles were excluded from analysis. BLUPs were used in genome-wide association analyses for all locations combined and for sensory traits for individual locations, and means were used for analyses of all other traits for individual locations. BLINK does not report R2 for identified SNPs. To support the BLINK findings, additional genome-wide association analyses were performed using a mixed linear model (MLM) approach in TASSEL v 5.2.31 (Bradbury et al., 2007b). Kinship was calculated using normalized IBS (Yang et al., 2011), and the first 3 PCs were included to control for population structure. SNPs with MAF < 0.05 or with more than two alleles were excluded from analysis. Manhattan plots and QQ plots were generated using the CMPlot R package (https://github.com/YinLiLin/R-CMplot), and significance levels were established using the False Discovery Rate (Benjamini and Hochberg, 1995) for the BLINK analyses and using a Bonferroni correction based on the effective number of markers tested determined via SimpleM for the MLM analyses (Gao et al., 2008). When reporting significant SNPs from each GWAS analysis, the SNP with the lowest p-value was chosen to represent each locus of interest. 66 RESULTS Sensory Extremes Twelve genotypes were identified which exhibited extreme sensory attributes (Table 2.1, Figure S2.1). These genotypes include Zawadi (ADP0106), a purple speckled variety from Tanzania with low total flavor intensity; Bellagio (ADP0681), a cranberry variety from the United States (Kelly et al., 2010) with high total flavor intensity; USDK-4 (ADP0654), a dark red kidney germplasm line from the United States (Miklas et al., 2004) with high beany intensity; SELIAN94 (ADP0530), a red speckled variety from Tanzania with high vegetative intensity; Kijivu, W616460 (ADP0057), a dark red kidney landrace from Tanzania with high earthy intensity; Perry Marrow, G4499 (ADP0206), a white variety from the United States with high starchy intensity; Baetao- Manteiga 41, G1678 (ADP0190), a purple speckled landrace from Brazil with high sweet intensity; Carioca, Kibala (ADP0517), a carioca landrace from Angola with high bitter intensity; Kabuku, W616464 (ADP0005), a small red landrace from Tanzania with low seed coat perception; Blanco Belén, INIAP422 (ADP0450), a white variety from Ecuador (Minchala et al., 2003) with high seed coat perception; PR1146-123 (ADP0791), a yellow germplasm line and sibling of the germplasm release PR1146-138 (Beaver et al., 2016) from Puerto Rico with smooth cotyledon texture; and Kijivu, W616491 (ADP0044), a purple speckled landrace from Tanzania with grainy cotyledon texture. These genotypes were selected for training panelists because they exhibited the range of attributes likely present in the entire sample set. While the attribute intensities of these genotypes varied somewhat across the three locations, they collectively represented a large portion of the attribute intensity ranges that were observed, reflected by their averages across locations (Table 2.2, S2.4). Significant genotype effects for each sensory attribute and insignificant rep effects 67 indicated that the panelists were trained sufficiently to detect differences among genotypes and were consistent across reps despite significant panelist and session effects (Table 2.2, S2.3). Sensory Evaluation The pin drop Mattson cooker was used to determine cooking times of the beans used in the sensory evaluation. The trained panel rated doneness of each cooked bean sample based on mouthfeel and concluded that the cooking times determined via the Mattson cooker equated to fully cooked samples (data not shown). Least squares estimates for all sensory attribute intensities exhibited approximately normal distributions (Figure 2.1). Genotype significantly affected all sensory attributes (p-value < 0.05) (Table 2.2). Location significantly affected total flavor intensity and cotyledon texture (p-value < 0.05), but was not significant for other sensory attributes. Genotype by location significantly affected total flavor intensity, vegetative intensity, sweet intensity, seed coat perception, and cotyledon texture (p-value < 0.05). Across all three locations, least squares estimates ranged 1.6 – 4.5 for total flavor intensity, 1.5 – 5.0 for beany intensity, 1.1 – 4.0 for vegetative intensity, 1.2 – 3.4 for earthy intensity, 2.1 – 4.4 for starchy intensity, 0.8 – 3.5 for sweet intensity, 0.5 – 3.5 for bitter intensity, 1.6 – 4.4 for seed coat perception, and 1.1 – 4.2 for cotyledon texture. While panelists were able to differentiate among genotypes using 5-point scales, sensory attribute ranges did not exceed 3.2 in any single location, suggesting panelists did not make full use of the scales. Twenty seed types were represented in the ADP, and seed type significantly affected all sensory attribute intensities (p-value < 0.0001) (Table S2.5). However, large ranges of attribute intensities are observed for each seed type (Figure 2.2), indicating variability of flavor and texture within a seed type. Brown genotypes (N = 10) tended to vary the least across sensory attributes 68 followed by light red kidney (N = 41), with cranberry (N = 63) and red mottled/red speckled (N = 80) varying the most. Earthy intensity followed by bitter intensity had the least variability across all seed types, and seed coat perception and cotyledon texture had the most. Broad-sense heritability for sensory attribute intensities was low, ranging from 0.14 to 0.39 (Table 2.2). Seed coat perception and total flavor intensity exhibited the highest broad-sense heritability (0.39 and 0.38), while earthy intensity and vegetative intensity exhibited the lowest (0.14 and 0.15). Cooking Time Evaluation Genotype, location, and genotype by location significantly affected raw seed weight, soak water uptake, cooking time, and total water uptake (Table 2.3). The means and ranges of raw seed weight, soak water uptake, cooking time, and total water uptake varied across locations (Figure 2.3). Across all 3 locations, raw seed weight ranged from 20.7 – 72.2 g per 100 seeds; soak water uptake ranged from 29.5 – 140.4%; cooking time ranged from 16.7 – 85.8 min; and total water uptake ranged from 100.4 – 169.7% (Table 2.3). Raw seed weight, soak water uptake, cooking time, and total water uptake exhibited approximately normal distributions (Figure 2.3). Broad- sense heritability was moderate to high for raw seed weight (0.90), soak water uptake (0.85), cooking time (0.73), and total water uptake (0.65). Correlations and PCA Significant correlations among sensory attribute intensities and cooking time were observed (Figure 2.4). Total flavor intensity correlated with all other sensory attributes such that earthy (R = 0.44, p-value < 0.0001), beany (R = 0.39, p-value < 0.0001), sweet (R = 0.38, p-value < 0.0001), vegetative (R = 0.33, p-value < 0.0001), bitter (R = 0.27, p-value < 0.0001), and starchy (R = 0.17, p-value = 0.0004) intensity all increased with total flavor intensity. The correlations 69 between total flavor intensity and seed coat perception (R = 0.17, p-value = 0.0003) and cotyledon texture (R = 0.14, p-value = 0.0050) were weak, but indicate that more flavor is associated with tougher, lingering seed coats and grittier, firmer cotyledons in fully cooked seeds. Total flavor intensity was negatively correlated with cooking time (R = -0.16, p-value = 0.0009), suggesting that genotypes with shorter cooking times have more total flavor, potentially due to less time for leaching during the cooking process. Individual sensory attributes also correlated with one another, suggesting that some attributes tend to be observed together. Genotypes with high beany intensity tended to be somewhat earthy (R = 0.27, p-value < 0.0001) and bitter (R = 0.25, p-value < 0.0001) and less starchy (R = -0.13, p-value = 0.0073). Genotypes with high vegetative intensity also tended to be somewhat earthy (R = 0.21, p-value < 0.0001) and bitter (R = 0.27, p-value < 0.0001). Genotypes with high earthy intensity were bitter (R = 0.36, p-value < 0.0001) as well as beany and vegetative as already noted. Genotypes with high starchy intensity were notably sweet (R = 0.48, p-value < 0.0001), less bitter (R = -0.26, p-value < 0.0001), and less beany as already mentioned. Genotypes with high sweet intensity were also observed as being less bitter (R = -0.18, p-value = 0.0002). Genotypes with high bitter intensity were somewhat beany, vegetative, and earthy and less starchy or sweet as previously noted. Genotypes with tougher seed coats were beany (R = 0.22, p-value < 0.0001) and bitter (R = 0.10, p-value = 0.0386) and less starchy (R = -0.17, p-value = 0.0003) or sweet (R = -0.10, p-value = 0.0343). Genotypes with grittier/firmer cotyledon texture were vegetative (R = 0.15, p-value = 0.0024), earthy (R = 0.24, p-value < 0.0001), and bitter (R = 0.12, p-value = 0.0147) and less beany (R = -0.12, p-value = 0.0167). Many of these correlations are relatively weak, suggesting that these tendencies are not always observed and that these attributes can be packaged together in multiple ways. 70 Cooking time also correlated with individual sensory attributes. Faster-cooking genotypes were starchy (R = -0.36, p-value < 0.0001) and sweet (R = -0.34, p-value < 0.0001) and had smoother cotyledon texture (R = -0.12, p-value = 0.0123), while slower cooking genotypes were beany (R = 0.23, p-value < 0.0001) and bitter (R = 0.12, p-value = 0.0167) and had tougher seed coats (R = 0.2, p-value < 0.0001). These correlations were relatively weak, indicating that fast cooking time can be packaged with target sensory profiles. For the PCA, the first two principal components (PCs) explained about 45% of the variation (Figure 2.5). The first PC separated the genotypes approximately by total flavor, vegetative, earthy, beany, and bitter intensity as well as cotyledon texture and somewhat seed coat perception and represented almost a quarter of the variation (22.8%). The second PC represented a similar amount of the variation (20.9%) and separated the genotypes by starchy and sweet intensity and cooking time. The remaining PCs accounted for 13.1, 8.9, 8.4, 6.7, 6.2, 5.5, 4.4, and 3.1% of the variance respectively (data not shown). The PCA highlights a positive relationship among total flavor, vegetative, earthy, beany, and bitter intensity as well as seed coat perception and cotyledon texture. Total flavor, vegetative, and earthy intensity and cotyledon texture are positioned closer together as are beany and bitter intensity and seed coat perception, indicating stronger relationships within each group. A positive relationship was also observed between starchy and sweet and sweet intensity, which appear to be negatively associated with cooking time. Each genotype within the PCA is colored by seed type, which reveals substantial variation within seed type. All seed types are spaced somewhat evenly across the biplot with the exception of the white seed type. White genotypes tend to cluster near starchy and sweet and away from cooking time and seed coat perception, indicates that white genotypes tend to be starchy and sweet 71 with shorter cooking times. Dark red kidney, light red kidney, and red mottled genotypes are distributed somewhat closer toward loadings for total flavor intensity, vegetative intensity, earthy intensity, and cotyledon texture, and purple speckled genotypes are distributed somewhat away, but the clustering is very loose. Genome-Wide Association Mapping Across the 430 Andean genotypes evaluated in this study, 31,273 SNPs remained after imputing and filtering. For each location, a similar number of SNPs were used in GWAS: 29,926 SNPs from Hawassa, Ethiopia (N = 373), 29,545 SNPs from Kabwe, Zambia (N = 251), and 31,484 SNPs from Lusaka, Zambia (N = 356). Across all locations combined, significant SNPs were identified using BLINK and MLM for several sensory attributes, including total flavor intensity, beany intensity, earthy intensity, starchy intensity, bitter intensity, seed coat perception, and cotyledon texture (Figure 2.6, S2.2). Significant SNPs detected for sensory traits were not consistent across the BLINK and MLM analyses methods, except for cotyledon texture (Table 2.4). MLM identified fewer significant SNPs overall, as expected due to its lower power and poor control of false negatives as compared to BLINK (Liu et al., 2016; Huang et al., 2018). For each sensory attribute with significant marker associations, an increase in the number of alleles conferring positive effects corresponded to an increase in mean attribute intensity (Figure 2.7). For total flavor intensity, 6 significant SNPs were identified on Pv01, Pv02, Pv05, and Pv09 (Table 2.4). MLM identified S01_5952237 on Pv01, which had no significant SNPs detected by BLINK. S02_34288083, S02_38579748, and S09_235919 were most significant. Genotypes with 5 significant SNPs conferring positive effects had a mean total flavor intensity rating 1.2 higher than those with no positive significant SNPs (Figure 2.7). There were no genotypes with all 72 6 positive significant SNPs. For beany intensity, 5 significant SNPs were identified on Pv02, Pv06, Pv07, and Pv10 (Table 2.4). S02_47727086, S06_5174714, and S10_42475118 were most significant. Genotypes with all 5 significant SNPs conferring positive effects had a mean beany intensity rating 0.8 higher than those with no positive significant SNPs (Figure 2.7). For earthy intensity, 3 significant SNPs were identified on Pv04 and Pv11, with S04_528286 being most significant (Table 2.4). Genotypes with all 3 significant SNPs conferring positive effects had a mean earthy intensity rating about equal to those with no positive significant SNPs when presented as means of least squares estimates (Figure 2.7) and slightly increased (0.1) when presented as means of BLUPs (data not shown). Starchy intensity had 1 significant marker on Pv01 (S01_42652564), which was detected by MLM and not BLINK (Table 2.4). Genotypes with the significant marker conferring a positive effect had a mean starchy intensity rating 0.1 higher than those without the positive significant marker (Figure 2.7). Bitter intensity also had 1 significant marker on Pv01 (S01_51119029), which was detected by MLM and not BLINK (Table 2.4). Genotypes with the significant marker conferring a positive effect had a mean bitter intensity rating 0.2 higher than those without the positive significant marker (Figure 2.7). For seed coat perception, 3 significant SNPs were detected on Pv02 and Pv08 (Table 2.4). All three were highly significant. Genotypes with all 3 significant SNPs conferring positive effects had a mean seed coat perception rating 0.7 higher than those with no positive significant SNPs (Figure 2.7). For cotyledon texture, 2 significant SNPs were detected on Pv03 and Pv08, which were detected by both BLINK and MLM (Table 2.4). Both SNPs were highly significant. Genotypes with both significant SNPs conferring positive effects had a mean cotyledon texture rating 0.4 higher than those with no positive significant SNPs (Figure 2.7). 73 For each individual location, significant SNPs were also identified using BLINK for total flavor intensity, beany intensity, earthy intensity, and seed coat perception (Table S2.6). MLM was not performed for individual locations. The identified SNPs somewhat reflect the findings for all locations combined, but largely point to different SNPs relevant for specific locations. For total flavor intensity, a total of 15 significant SNPs were identified on Pv02, Pv03, Pv04, Pv05, and Pv11 in the samples from Hawassa Ethiopia; Pv03, Pv08, Pv09, Pv10, and Pv11 in the samples from Kabwe, Zambia; and Pv05, Pv06, and Pv10 in the samples from Lusaka, Zambia (Table S2.6, Figure S2.3). For beany intensity, a total of 6 significant SNPs were identified on Pv10 and Pv11 in the samples from Kabwe, Zambia and Pv02, Pv06, Pv10, and Pv11 in the samples from Lusaka, Zambia (Table S2.6, Figure S2.4). For earthy intensity, a total of 3 significant SNPs were identified on Pv04 in the samples from Kabwe, Zambia and Pv02 and Pv11 in the samples from Lusaka, Zambia (Table S2.6, Figure S2.5). For seed coat perception, a total of 5 significant SNPs were identified on Pv02 and Pv05 in the samples from Hawassa, Ethiopia; Pv05 in the samples from Kabwe, Zambia; and Pv02 and Pv07 in the samples from Lusaka, Zambia (Table S2.6, Figure S2.6). Across all locations combined, significant SNPs were identified using BLINK and MLM for raw seed weight, soak water uptake, cooking time, and total water uptake (Figure S2.7, S2.8). Both methods identified different SNPs, with some overlap for raw seed weight and soak water uptake (Table S2.7). MLM identified fewer significant SNPs overall, as was the case for the sensory attributes. For raw seed weight, 15 significant SNPs were identified on Pv01, Pv02, Pv03, Pv04, Pv05, Pv06, Pv08, Pv09, and Pv11 (Table S2.7). MLM identified S03_41895570, which was also detected by BLINK, and S05_1138961, which was not. Genotypes with 13 significant SNPs 74 conferring positive effects had a mean raw seed weight 31 grams per 100 seeds higher than those with only 3 positive significant SNPs (Figure S2.9). There were no genotypes with fewer than 3 or more than 13 positive significant SNPs. For soak water uptake, 17 significant SNPs were identified on Pv02, Pv03, Pv04, Pv05, Pv07, Pv08, Pv10, and Pv11 (Table S2.7). MLM identified 6 of those SNPs, of which 1 was also detected by BLINK. Genotypes with 15 significant SNPs conferring positive effects had a mean soak water uptake 64% higher than those with only 4 positive significant SNPs (Figure S2.9). There were no genotypes with fewer than 4 or more than 15 positive significant SNPs. For cooking time, 11 significant SNPs were identified on Pv03, Pv04, Pv06, Pv07, Pv08, and Pv11 (Table S2.7). MLM identified S04_3957256 and S08_62659170, which were not detected by BLINK. Genotypes with 9 significant SNPs conferring negative effects had a mean cooking time 23 min faster than those with 3 or fewer negative significant SNPs (Figure S2.9). There were no genotypes with 0, 2, or more than 9 negative significant SNPs, and there was a single genotype with only 1 negative significant marker. For total water uptake, 5 significant SNPs were identified on Pv03, Pv04, Pv09, and Pv11 (Table S2.7). No SNPs were identified by MLM for total water uptake. S04_30764016 was associated with both soak water uptake and total water uptake. Genotypes with all 5 significant SNPs conferring positive effects had a mean total water uptake 10% higher than those with 1 or fewer positive significant SNPs (Figure S2.9). There was only 1 genotype with no positive significant SNPs. DISCUSSION The modified QDA approach used in this study successfully detected differences among genotypes for the purposes of identifying extremes, evaluating the relationships among sensory attributes and seed type, and performing genome-wide association analyses to reveal SNPs 75 associated with sensory attributes. Although significant panelist effects were identified (Table S2.3), these effects are not concerning because QDA does not rely on consensus among panelists. However, limited use of the scales by the panelists prevents detection of small differences between samples. This can be remedied by increasing the size of the scales or using line scales that allow for continuous rather than discrete ratings. As for panelists, differences among sessions are expected and can be accounted for in the ANOVAs and by using BLUPs where appropriate. Genotypes exhibiting extreme attribute intensities were identified (Table 2.1) and successfully used for training panelists for sensory evaluation (Tables 2.2, S2.3). These genotypes could serve as a training set for future sensory research or for training sensory panels for germplasm evaluation in breeding programs. Location of production and crop management practices have previously been identified as factors affecting sensory quality (Mkanda et al., 2007; Ferreira et al., 2012), which complicates efforts to understand and breed for sensory quality in beans. The location and genotype by location effects were significant for many of the sensory attributes in this study (Table 2.2), supporting these findings. Differences among locations were also apparent in density plots for some flavor and texture attributes (Figure 2.1). Despites small fluctuations in sensory profile across locations, the genotypes exhibiting extreme sensory attribute intensities remained extreme for their attribute of interest in each location (Table S2.4). This suggests that differences across location affect magnitude of sensory attribute intensities, but do not substantially alter sensory attribute intensities relative to each other. Many significant correlations were identified among flavor, texture, and cooking time, although correlation coefficients were generally weak, suggesting that traits can combine in multiple ways (Figure 2.4). Sweet and starchy intensity were the two most strongly correlated 76 attributes, and the loadings for these attributes were positioned near each other in the PCA, away from other attributes (Figure 2.5). White seeds were generally sweet and starchy, but otherwise, few trends were identified in regard to seed type, which indicates that seed type does not define the sensory profile of a genotype (Figure 2.2, 2.5). This supports a previous study that found similarities in morphology and genetic background do not indicate similarity of sensory attributes among genotypes (Rivera et al., 2013). The genetic variability existing within seed type could be harnessed to achieve a target sensory profile and ensure greater consistency and uniformity of flavor and texture. In addition, fast cooking time could be targeted without substantially influencing sensory profile, which would address another major factor influencing consumer purchasing decisions (Leterme and Carmenza Muũoz, 2002; Eihusen and Albrecht, 2007; Winham et al., 2019). Many SNPs significantly associated with flavor and texture were identified using BLINK and MLM, and they appear to confer minor effects, highlighting the complexity of the genetics underlying these traits. (Table 2.4, Figure 2.6, S2.2). Significant SNPs varied for each individual location (Table S2.6, Figure S2.3-S2.6), emphasizing the importance of location in expression of genetic variability for sensory attributes. The significant SNPs identified have not been previously associated with sensory attributes as this is the first study of its kind in beans. No significant SNPs were associated with vegetative or sweet intensity, but alternative approaches such as QTL mapping or genomic prediction with a population of related individuals may provide increased power to detect relevant loci for these traits. Other studies in fruits have successfully used volatiles and instrumental measures in GWAS as proxies for flavor and texture, allowing for easier phenotyping and in some cases higher heritability than traits evaluated via descriptive panels (Zhang et al., 2015; Amyotte et al., 2017; Bauchet et al., 2017; Zhao et al., 2019). However, 77 volatiles and instrumental measures do not always successfully predict flavor and texture as it is perceived by a descriptive panel (Amyotte et al., 2017), and for dry beans, little is known about how volatiles or other measures relate to flavor and texture. The screening of the ADP performed in this study provides a resource for future population development to further understanding of the genetic control of sensory attributes and how volatiles and instrumental measurements relate to sensory attributes. One of the unique flavor characteristics found in dry beans and other legumes consumed as seeds is the “beany” flavor, which has proven a challenge to define and is often described as an “off” flavor in products using beans as ingredients (Kinsella, 1979; Bott and Chambers, 2006; Hooper et al., 2019). One study defined the flavor as undesirable, with multiple contributing volatiles (Vara-Ubol et al. 2004). In soybean, significant SNPs have been associated with volatiles contributing to beany flavor, and some of these SNPs are present in regions syntenic with dry bean chromosomes where SNPs associated with beany flavor were identified in this study (Schmutz et al., 2014; Xia et al., 2019b; Xia et al., 2019a; Wang et al., 2020). In particular, the end of Pv02 where S02_47727086 and S02_49605939 are located is syntenic with soybean chromosomes 5 and 8 (Schmutz et al., 2014). Using Minimap2 (Li, 2018) and the soybean reference genome (Williams 82) from SoyBase (Grant et al., 2009), the 50 kb regions around S02_47727086 and S02_49605939 align near rs39728576 and rs4039554, respectively, markers on soybean chromosome 5 and 8 associated with hexanal content in soybean (Wang et al., 2020). Off-flavors in soy products are generated by lipoxygenases, primarily Lipoxygenase-2, or the oxidative rancidity of unsaturated fatty acids (Wolf et al., 1971; Kim et al., 2004). Markers linked to Lipoxygenase-2 are available and in use for breeding efforts targeting the reduction of beany flavor in soybean (Lenis et al., 2010; Talukdar and Shivakumar, 2016). Several 78 lipoxygenase genes are located within a megabase of S07_28996873 and S10_42475118 (http://phytozome.jgi.doe.gov/). In addition, a single lipoxygenase is located within three megabases of S06_5174714. While some lipoxygenases are present on Pv02, they are not close to S02_47727086 or S02_49605939. It is not yet understood whether beany flavor in a boiled beans translates to off-flavor in products made using beans as ingredients. In addition, consumer preference as it relates to sensory attribute intensities has not been explored for boiled beans beyond a general preference for beans that are sweet and soft when fully cooked (Mkanda et al., 2007). Further research relating consumer preference to attribute intensities in boiled beans as well as products using beans as ingredients could allow breeders to identify target sensory profiles for different seed types or varieties intended for use as ingredients. In regard to raw seed weight, soak water uptake, cooking time, and total water uptake, many significant SNPs were identified in association with these traits as well via BLINK and MLM (Table S2.7, Figure S2.7, S2.8). Most of the SNPs identified were novel, but some were proximal to QTL and markers identified in previous studies. Of particular interest, S11_10805992, which was significantly associated with cooking time (Table S2.7), is near a QTL identified for cooking time by Berry et al. (2020). S02_47837868, S03_50652595, S03_51140861, S04_30764016, S07_3919560, S10_37637761, which were significantly associated with soak water uptake (Table S2.7), appear to be supported by hydration coefficient and water absorption QTL previously identified (Pérez-Vega et al., 2010; Cichy et al., 2014; Kelly and Bornowski, 2018; Sandhu et al., 2018). While broad-sense heritability for each sensory attribute was generally low (Table 2.2), heritability could be improved in the context of a breeding program by screening only promising 79 lines with greater replication. This could allow for better understanding of panelists and session effects and a balanced statistical design while maintaining a manageable time and personnel commitment. If fewer samples are evaluated each session, sensory fatigue could be reduced, allowing for better detection of small differences between samples. Potential alternative methods for screening sensory attributes could also be explored, including screening volatile profiles via GC-MS and collecting NIR spectra. NIR spectra of both raw seeds and cooked and dried seeds have been analyzed for their ability to predict beany flavor, mealiness, seed-coat roughness, and seed-coat brightness, although correlations between NIR spectra and these attributes were poor for raw beans (Plans et al., 2014). Using alternative methods for screening sensory attributes could increase the throughput of sensory profile characterization, but more research is needed to identify predictive measurements. CONCLUSION This study lays a foundation for incorporating sensory quality traits into dry bean breeding programs. The broad range of sensory attribute intensities observed across and within seed types indicates a lack of uniformity within seed type, but also a wealth of genetic variability for sensory quality. This presents an opportunity for specific sensory profiles to be defined for each seed type. The limited correlation among sensory attributes indicates that they can combine in multiple ways, suggesting it is feasible to target specific sensory profiles according to consumer preference. Using the modified QDA approach to screen materials and the significant genetic SNPs identified for flavor and texture attributes, breeders could continue to improve agronomic traits without sacrificing desirable sensory quality. The set of genotypes exhibiting extreme sensory attribute intensities identified during this study can be used for panel training as well as future work 80 exploring sensory attributes and consumer preference. In addition, further understanding of sensory profiles suitable for bean products would allow varieties to be developed for use as ingredients, increasing the chance of success for bean products on the market. Improving flavor and texture in dry beans can ensure they are appreciated as a delicious and tasteful component of a healthful diet in all the versatile ways consumers choose to eat them. ACKNOWLEDGMENTS This research was funded by the U.S. Department of Agriculture, Agricultural Research Service Project 3060-21650-001-00D (Pulse Crop Health Initiative), 5050-21430-010-00D (K.A.C.). The authors would like to thank members of the USDA-ARS and MSU Dry Bean Breeding programs for their contributions as panelists. Dr. Janice Harte, Dr. Sungeun Cho, and Ed Szczygiel are appreciated for their advice in designing the sensory evaluation method described. Additional thanks go to Francisco Santos, Hannah Cooperider, Hannah Peplinski, Anna Peter, Gasana Elyvine, and Queen Iribagiza for their assistance with cooking samples and managing seeds. The IRB of Michigan State University granted exempt status for the sensory evaluation component of this study (IRB# x16-763e Category: Exempt 6). 81 APPENDICES 82 APPENDIX A: CHAPTER 2 TABLES AND FIGURES Table 2.1 Genotypes exhibiting extreme sensory attribute intensities identified from screening accessions of the Andean Diversity Panel grown in Hawassa, Ethiopia. Genotype Sensory Attribute Region of origin ADP ID Seed Type Zawadi Bellagio USDK-4 SELIAN94 ADP0106 Purple speckled Tanzania Low total flavor intensity ADP0681 Cranberry United States High total flavor intensity ADP0654 Dark red kidney United States High beany intensity ADP0530 Red speckled Tanzania High vegetative intensity Kijivu (W616460) ADP0057 Dark red kidney Tanzania High earthy intensity Perry Marrow (G4499) ADP0206 White United States High starchy intensity Baetao-Manteiga 41 (G1678) ADP0190 Purple speckled Brazil High sweet intensity Carioca,Kibala ADP0517 Carioca Kabuku (W616463) ADP0005 Small red ADP0450 White INIAP422 PR1146-123 Kijivu (W616491) Angola Tanzania Ecuador High bitter intensity Low seed coat perception High seed coat perception ADP0791 Yellow Puerto Rico Smooth cotyledon texture ADP0044 Purple speckled Tanzania Grainy cotyledon texture 83 Table 2.2 Least squares estimates, range, and coefficient of variation of sensory attribute intensities of the Andean Diversity Panel grown in three locations with ANOVA p-valuesa for genotype, location (Loc), and genotype by location indicated. Trait Location LSE Range CV ( %)b Genotype Loc Genotype x Loc H2c Total Flavor Intensity Hawassa, ET 2.8 1.6 - 3.7 14.4 <.0001 <.0001 <.0001 0.38 Kabwe, ZM 3.4 2.2 - 4.4 12.6 Lusaka, ZM 3.4 2.0 - 4.5 13.3 Beany Intensity Hawassa, ET 2.8 1.7 - 3.8 13.3 <.0001 NS NS 0.30 Kabwe, ZM 2.9 1.5 - 4.1 14.9 Lusaka, ZM 3.4 1.8 - 5.0 16.1 Vegetative Intensity Hawassa, ET 2.0 1.1 - 3.4 17.8 <.0001 NS 0.0013 0.15 Kabwe, ZM 2.4 1.3 - 3.7 16.0 Lusaka, ZM 2.6 1.6 - 4.0 16.4 Earthy Intensity Hawassa, ET 2.0 1.2 - 3.0 15.7 <.0001 NS NS 0.14 Kabwe, ZM 2.1 1.2 - 3.2 17.0 Lusaka, ZM 2.1 1.2 - 3.4 18.6 Starchy Intensity Hawassa, ET 3.2 2.2 - 4.4 10.4 <.0001 NS NS 0.21 Kabwe, ZM 3.2 2.1 - 4.0 11.7 Lusaka, ZM 3.2 2.2 - 4.1 12.2 Sweet Intensity Hawassa, ET 1.7 1.0 - 3.5 21.2 <.0001 NS <.0001 0.26 Kabwe, ZM 1.9 0.9 - 3.2 21.2 Lusaka, ZM 1.8 0.8 - 3.1 21.2 Bitter Intensity Hawassa, ET 1.6 0.8 - 3.5 22.0 <.0001 NS NS 0.22 Kabwe, ZM 1.5 0.8 - 3.0 22.0 Lusaka, ZM 1.4 0.5 - 2.8 24.6 Seed Coat Perception Hawassa, ET 3.0 1.6 - 4.4 13.3 <.0001 NS <0.0001 0.39 Kabwe, ZM 3.1 2.2 - 4.1 13.1 Lusaka, ZM 3.0 1.6 - 4.1 13.8 Cotyledon Texture Hawassa, ET 2.7 1.4 - 4.0 16.1 <.0001 0.0025 <.0001 0.31 Kabwe, ZM 2.3 1.4 - 4.2 15.2 Lusaka, ZM 2.2 1.1 - 3.4 14.2 84 Table 2.2 (cont’d) a NS indicates non-significant p-values at α = 0.05 b Coefficient of variation c Broad sense heritability Table 2.3 Mean, range, and coefficient of variation of raw seed weight, soak water uptake, cooking time, and total water uptake of the Andean Diversity Panel grown in three locations with ANOVA p-values for genotype, location (Loc), and genotype by location indicated. Trait Location Mean Range CV (%)a Genotype Loc Genotype x Loc H2b Raw Seed Weight (g per 100 seed) Hawassa, ET 37.2 20.7 – 54.0 16.4 <.0001 <.0001 <.0001 Kabwe, ZM 44.8 25.9 – 62.0 15.6 Lusaka, ZM 45.1 24.3 - 72.2 17.0 Soak Water Uptake (%) 0.90 Hawassa, ET 112.1 51.9 - 140.4 8.9 <.0001 <.0001 <.0001 0.85 Kabwe, ZM 100.3 54.0 - 118.6 9.3 Lusaka, ZM 101.0 29.5 - 128.1 8.7 Cooking Time (min) Hawassa, ET 31.5 16.7 - 68.9 Kabwe, ZM 31.5 17.8 - 75.5 Lusaka, ZM 33.8 21.0 - 85.8 22.8 23.8 24.9 Total Water Uptake (%) <.0001 <.0001 <.0001 0.73 Hawassa, ET 139.5 100.4 - 165.2 5.7 <.0001 <.0001 <.0001 0.65 Kabwe, ZM 134.8 110.7 - 156.2 5.1 Lusaka, ZM 135.0 105.0 - 169.7 5.6 a Coefficient of variation b Broad sense heritability 85 Table 2.4 GWAS significant markers associated with sensory attribute intensities with marker, chromosome (Chr), position, P-value, minor allele frequency (MAF), major and minor alleles (Maj/Min), significance (Sig), and method indicated. Trait Marker Chr Positiona P-value MAF Maj/Minb Sigc Method Total Flavor Intensity S01_5952237 1 5952237 1.87E-05 0.06 G/T * MLM S02_34288083 2 34288083 1.94E-07 0.27 A/G *** BLINK S02_38579748 2 38579748 2.31E-07 0.07 T/A *** BLINK S05_36225444 5 36225444 1.91E-06 0.15 C/T ** BLINK S05_39325999 5 39325999 1.23E-05 0.28 C/T * BLINK S09_235919 Beany Intensity 9 235919 6.53E-07 0.10 C/T *** BLINK S02_47727086 2 47727086 3.67E-08 0.22 G/C *** BLINK S02_49605939 2 49605939 2.48E-06 0.06 C/T ** BLINK S06_5174714 6 5174714 6.15E-07 0.14 G/T *** BLINK S07_28996873 7 28996873 6.66E-06 0.37 G/T ** BLINK 4 4 S10_42475118 10 Earthy Intensity S04_528286 S04_4661131 S11_47172346 11 Starchy Intensity S01_42652564 1 Bitter Intensity S01_51119029 1 Seed Coat Perception S02_34629777 2 42475118 5.51E-09 0.15 T/C *** BLINK 528286 8.63E-08 0.07 C/T *** BLINK 4661131 1.98E-06 0.19 G/A ** BLINK 47172346 1.23E-06 0.30 A/T 42652564 5.42E-06 0.30 G/A 51119029 1.47E-05 0.20 C/T ** BLINK ** MLM * MLM 34629777 2.43E-07 0.10 A/C *** BLINK S02_48936819 2 48936819 9.06E-11 0.26 C/T *** BLINK S08_60104671 8 Cotyledon Texture S03_31659572 3 60104671 4.90E-07 0.23 C/G *** BLINK 31659572 9.43E-11 0.18 G/T *** BLINK, MLM S08_2356200 8 2356200 3.32E-07 0.08 A/G *** BLINK, MLM a Position is based on the P. vulgaris v2.1 reference genome (DOE-JGI and USDA-NIFA, http://phytozome.jgi.doe.gov/) b Alleles in bold confer a positive effect on the indicated trait c Significance is indicated by asterisks, such that *, **, *** indicate significance at α = 0.1, α = 0.05, α = 0.01 using the false discovery rate for the BLINK method and a Bonferroni correction based on the effective number of markers determined using the SimpleM algorithm for the MLM method 86 Figure 2.1 Density plots of least squares estimates of sensory attribute intensities for the Andean Diversity Panel for all locations combined (C); Hawassa, ET (H); Kabwe, Zambia (K); and Lusaka, Zambia (L). 87 Figure 2.2 Boxplots of sensory attribute intensities separated by seed type. All boxplots are presented as least squares estimates averaged across all locations for seed types with N > 10, where “Other” includes the remaining seed types with N < 10. 88 Figure 2.3 Density plots of raw seed weight, soak water uptake, cooking time, and total water uptake for the Andean Diversity Panel for all locations combined (C); Hawassa, ET (H); Kabwe, Zambia (K); and Lusaka, Zambia (L). 89 Figure 2.4 Pairwise comparison matrix of cooking time (CT), total flavor intensity (TF), beany intensity (Beany), vegetative intensity (Veg), earthy intensity (Earthy), starchy intensity (Starchy), sweet intensity (Sweet), bitter intensity (Bitter), seed coat perception (SCP), and cotyledon texture (CTex). Pearson correlation coefficients were calculated using BLUPs and are indicated in the lower left, and scatterplots for each pairwise comparison with LOWESS regression lines are shown in the upper right. P-values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. 90 Figure 2.5 Principal component analysis biplot with each genotype colored by seed type and loadings for total flavor intensity (TF), beany intensity (Beany), vegetative intensity (Veg), earthy intensity (Earthy), starchy intensity (Starchy), sweet intensity (Sweet), bitter intensity (Bitter), seed coat perception (SCP), cotyledon texture (CTex), and cooking time (CT). 91 Figure 2.6 Manhattan and QQ plots for total flavor intensity, beany intensity, earthy intensity, seed coat perception, and cotyledon texture of the Andean Diversity Panel with mapping conducted using BLINK with BLUPs from all locations combined. The gray dashed line is the α = 0.05 FDR. 92 Figure 2.7 Phenotypic effects of carrying the indicated number of significant markers conferring a positive effect for each sensory attribute. Phenotypic values represent all locations combined as averages of least squares estimates from Hawassa, Ethiopia; Kabwe, Zambia; and Lusaka, Zambia. N is the number of individuals in each boxplot. 93 APPENDIX B: CHAPTER 2 SUPPLEMENTAL TABLES AND FIGURES Table S2.1 Genotype information. ID Genotype Seed Type Grown in Hawassa, ET Kabwe, ZA Lusaka, ZA ADP0001 ROZIKOKO ADP0002 W616444 ADP0003 KIDUNGU ADP0004 KILOMBERO ADP0005 KABUKU ADP0006 W616465 ADP0007 BUKOBA ADP0008 Nyayo ADP0009 Maalasa ADP0010 CANADA ADP0011 KIBOROLONI ADP0012 W616489 ADP0013 KIBUMBULA ADP0014 KIANGWE ADP0015 W616495 ADP0016 GOLOLI ADP0017 W616529 ADP0018 SODAN ADP0019 KASUKANYWELE ADP0020 KIGOMA ADP0021 MBULAMTWE ADP0022 KISAPURI ADP0023 MSHORONYLONI ADP0024 YELLOW red mottled yes purple speckled no red striped red DRK yellow red mottled red mottled DRK red DRK DRK yellow DRK red DRK DRK striped yellow yellow red red yellow yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes no yes yes yes yes ADP0025 RUHONDELA purple speckled yes ADP0026 BlackWonder ADP0027 Incomparable ADP0028 Sisi ADP0030 RHNo.6 ADP0031 RHNo.11 ADP0032 RHNo.21 yes yes yes yes yes yes black brown yellow black DRK DRK 94 yes yes yes no no no no no yes yes yes yes yes yes no no no yes no yes yes yes yes yes yes no yes yes yes yes yes yes yes yes yes no no yes no yes yes yes yes yes yes yes no yes yes yes yes yes yes yes yes yes no yes yes yes yes yes Table S2.1 (cont’d) ID Genotype Seed Type Grown in Hawassa, ET Kabwe, ZA Lusaka, ZA ADP0033 KIJIVU ADP0034 KIJIVU ADP0035 Kokola ADP0036 Lyamungu85 ADP0037 W616488 ADP0038 Moono ADP0039 RoziKoko ADP0041 MRONDO ADP0042 MKOKOLA ADP0043 BWANASHAMBA ADP0044 KIJIVU ADP0045 RHNo.12 ADP0047 MSOLINI ADP0048 W616534 ADP0049 W616546 ADP0050 SALUNDE ADP0051 RHNo.3 ADP0052 RHNo.9 purple speckled yes purple speckled yes red mottled red mottled brown DRK red mottled DRK DRK DRK yes yes no yes no yes yes yes purple speckled yes purple speckled yes brown red DRK yellow yes yes yes yes purple speckled yes purple speckled yes ADP0053 MAHARAGEMAKUBWA DRK yes yes no yes yes yes yes no no yes yes no no yes no yes yes yes no no no no yes yes no yes yes yes no yes yes yes no no yes yes yes yes no yes yes yes yes yes no yes yes yes yes yes yes yes yes yes no yes no no no yes no yes yes yes no yes yes yes yes yes yes yes ADP0054 W616447 ADP0055 KABUKU ADP0056 SOYA ADP0057 KIJIVU ADP0058 CANADA ADP0059 Poto ADP0060 CANADA ADP0061 Maulasi ADP0062 MAULASI ADP0063 Soya ADP0064 W616500 ADP0065 W616501 ADP0066 NJANO ADP0067 NJANO ADP0068 Soya ADP0070 Msafiri ADP0071 NJANO-DOLEA yes yes yes cranberry red speckled purple speckled yes DRK DRK yes yes purple speckled yes DRK cranberry red speckled yes yes yes purple speckled yes yellow red yellow yellow yes yes yes yes purple speckled no yes yes DRK yellow 95 Table S2.1 (cont’d) ID Genotype Seed Type Grown in Hawassa, ET Kabwe, ZA Lusaka, ZA ADP0072 MASUSU ADP0073 MASUSU brown brown yes no ADP0074 KABLANKETI purple speckled yes ADP0075 MABUKU ADP0076 KABLANKETI ADP0077 NAMWANGA ADP0080 KABLANKETI ADP0081 KABLANKETI ADP0082 KABLANKETI ADP0083 W616547 brown yes purple speckled yes purple speckled yes purple speckled yes purple speckled yes purple speckled yes purple speckled yes ADP0084 KABLANKETINDEFU purple speckled yes ADP0085 KABLANKETI purple speckled yes ADP0086 NYAMHONGAMWEKUNDU purple speckled yes ADP0087 KABLANKETI ADP0088 KABLANKETI purple speckled yes purple speckled yes ADP0089 KABLANKETI purple speckled no ADP0090 KASUKANYWELE ADP0092 MORO ADP0093 MORO ADP0094 LUSHALA ADP0095 CANADA ADP0096 Rojo ADP0098 Selian97 ADP0099 BwanaShamba ADP0100 EG21 ADP0101 Witrood ADP0102 Jesca ADP0103 Pesa ADP0105 Sewani97 ADP0106 Zawadi ADP0107 Mishindi ADP0108 Njano ADP0109 Kablanketi ADP0110 SUG-131 ADP0111 Uyole98 ADP0113 OPS-RS4 striped yellow yellow yellow striped DRK DRK DRK yes yes yes yes yes yes yes yes purple speckled no white yes purple speckled yes red DRK yes yes purple speckled yes purple speckled yes yellow yes purple speckled yes yes no no cranberry yellow cranberry 96 no no no no no no no no no no no yes yes no no no no yes no yes no yes yes yes yes no yes yes yes yes yes yes no yes yes yes no yes no no yes yes no yes yes no no yes yes yes no yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes Table S2.1 (cont’d) ID Genotype Seed Type Grown in Hawassa, ET Kabwe, ZA Lusaka, ZA ADP0114 OPS-RS1 ADP0117 A483 ADP0118 Werna ADP0119 A193 ADP0120 Tygerberg ADP0121 KranskopHR-1 ADP0122 Kranskop ADP0123 Jenny ADP0124 Mani ADP0127 SELIAN06 ADP0166 NABE4 ADP0168 KANYEBWA ADP0180 G433 ADP0186 G1368 ADP0190 G1678 ADP0191 G1939 ADP0196 G2875 ADP0199 G3452 ADP0205 G4494 ADP0206 G4499 ADP0207 G4564 ADP0208 G4644 ADP0211 G4780 ADP0212 G4970 ADP0213 G5034 ADP0214 G5087 ADP0220 G5625 ADP0224 G6239 ADP0225 G6415 ADP0232 G7930 ADP0242 G9013 ADP0247 G9975 ADP0255 G10994 ADP0267 G12689 ADP0269 G13092 ADP0271 G13167 cranberry purple mottled cranberry red mottled cranberry cranberry cranberry cranberry yellow pink red mottled red speckled cranberry DRK yes yes yes no yes yes no yes no yes yes yes yes yes purple speckled yes cranberry cranberry pink red mottled white Jacob's cattle red mottled red mottled yellow gray black DRK yellow eye LRK white cranberry cranberry yellow no no yes yes yes yes yes yes yes yes yes no yes yes yes yes yes yes purple cranberry yes yes yes white white 97 yes yes yes yes yes yes yes yes yes no no no no yes yes yes yes no yes yes yes yes yes yes yes yes yes yes yes no yes yes yes no no no yes yes yes yes yes yes yes yes yes no yes no no yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes no Table S.1 (cont’d) ID Genotype Seed Type Grown in Hawassa, ET Kabwe, ZA Lusaka, ZA ADP0272 G13336 ADP0276 G13654 ADP0277 G13778 ADP0279 G14423 ADP0280 G14440 ADP0303 G17913 ADP0310 G18356 ADP0324 G20729 ADP0336 G21210 ADP0345 G22147 ADP0346 G22246 ADP0350 G22365 ADP0353 G22455 ADP0354 G22502 ADP0366 G23070 ADP0367 G23086 ADP0368 G23093 ADP0376 PI189408 ADP0383 PI209486 ADP0390 PI307808 ADP0391 PI308894 ADP0392 PI309701 ADP0395 PI310511 ADP0417 PI451906 ADP0427 Badillo ADP0428 ColoradodelPais ADP0429 PR9920-171 ADP0430 PR1013-3 ADP0431 Gurabo5 ADP0432 PR0637-134 ADP0433 PR9745-232 ADP0434 PR0737-1 ADP0436 JB-178 ADP0437 PC-50 ADP0438 46-1 ADP0442 LargaComercial cranberry yellow purple mottled Jacob's cattle white yellow cranberry DRK red mottled DRK red mottled red mottled brown yes no no yes no yes yes yes no yes yes no no purple speckled yes no no yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes no yes yes yes red cranberry pink DRK red mottled DRK LRK cranberry red mottled DRK LRK red speckled red speckled red speckled red speckled red mottled red mottled red mottled red mottled red mottled red mottled red mottled 98 no yes no no no no yes yes no no no no yes no no yes yes no yes yes yes no no yes yes yes yes yes yes yes yes yes no yes no no no yes yes no yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes no no yes yes yes yes yes yes yes yes yes no yes yes yes Table S.1 (cont’d) ID Genotype Seed Type Grown in Hawassa, ET Kabwe, ZA Lusaka, ZA ADP0449 INIAP 420 ADP0450 INIAP422 ADP0452 INIAP425 ADP0453 INIAP428 ADP0454 INIAP429 ADP0455 INIAP430 ADP0456 INIAP480 ADP0457 INIAP481 ADP0458 INIAP483 ADP0459 PI331356-C ADP0460 PI331356-B ADP0462 IZ 117 ADP0464 G39308 ADP0465 PI321094-D ADP0466 PI449430 ADP0467 PI209808 ADP0468 PI527538 ADP0469 PI527521 ADP0470 PI527508 ADP0471 IZ 102 ADP0472 IZ 102 ADP0474 PI527519 ADP0475 PI319706 ADP0476 Hutterite ADP0477 PI527512 ADP0478 PI353536 ADP0479 PI527530 ADP0480 PI209804 ADP0481 PI449428 ADP0482 PI209802 ADP0483 PI209815 ADP0508 Calembe ADP0509 Fernando yellow white white yellow red mottled red mottled yellow red mottled red mottled cranberry purple mottled yellow DRK yellow yes yes yes yes yes yes yes yes yes yes yes yes yes yes purple speckled yes purple speckled yes yellow white red speckled yellow brown red mottled yellow yellow yes yes yes yes yes yes yes yes purple speckled no brown yellow purple mottled red mottled purple mottled yellow yellow yellow no yes no no no yes no yes no yes yes no yes yes yes no no yes yes no yes yes yes yes yes no no yes yes no yes no yes yes yes yes no yes yes no yes yes yes yes yes yes yes yes yes yes yes no yes yes yes no yes yes yes yes yes yes yes yes yes no yes yes yes yes yes yes yes yes yes yes yes yes yes no yes no yes ADP0510 Ohliodeperdiz Jacob's cattle ADP0511 Canario ADP0512 Ervilha yellow yellow 99 Table S.1 (cont’d) ID Genotype Seed Type Grown in Hawassa, ET Kabwe, ZA Lusaka, ZA ADP0513 Canario ADP0514 MantegaAmarela ADP0515 Katarina,Kibala ADP0516 Mantega, kibala ADP0518 Mantegablanca,Kibala ADP0519 Katarina,Cela ADP0520 Chumbo,Cela ADP0521 Cebo,Cela ADP0523 Canario,Cela ADP0524 KATB1 ADP0525 KATB9 ADP0526 CAL143 ADP0527 POA2 ADP0528 LYAMUNGO85 ADP0530 SELIAN94 ADP0531 AND620 ADP0532 A197 ADP0534 G22501 ADP0535 ARA4 ADP0536 CAL96 ADP0537 AFR619 ADP0538 RWR221 ADP0540 AFR708 ADP0541 CIM9314-36 ADP0543 G16157 ADP0544 PVA773 yellow yellow cranberry yellow yellow cranberry yellow yellow yellow yellow red red mottled red mottled red mottled red speckled red mottled yellow yellow red mottled red mottled red mottled LRK red mottled red mottled LRK red mottled ADP0546 REDCANADIANWONDER DRK ADP0549 RWR10 ADP0551 AFR612 ADP0553 AND277 ADP0554 AND279 ADP0555 BRB191 ADP0556 BRB194 ADP0557 COS16 ADP0558 DAB528 ADP0559 DAB555 DRK red mottled red mottled red mottled red mottled red cranberry red yellow 100 no yes yes no yes yes yes no yes yes no yes yes yes yes yes yes yes yes yes yes yes yes no yes yes yes yes yes yes yes yes yes yes yes yes yes no yes no no yes no yes no no yes yes yes yes yes yes yes no no yes yes yes yes yes yes yes yes yes yes no no yes no no no yes yes yes yes yes yes yes no yes yes yes yes yes yes yes yes yes yes no yes yes yes yes yes yes yes yes no yes yes no no yes yes no yes yes Table S.1 (cont’d) ID Genotype Seed Type Grown in Hawassa, ET Kabwe, ZA Lusaka, ZA ADP0560 DAB230 ADP0561 DAB246 ADP0562 DAB387 ADP0564 G5164 ADP0566 G5686 ADP0567 G4523 ADP0569 MDRK ADP0570 NATALSUGAR ADP0571 NUA45 ADP0572 NUA56 red mottled red mottled cranberry red speckled yellowmottled red mottled DRK cranberry red mottled red mottled ADP0574 RADICALCERINZA red ADP0575 SAB259 ADP0576 SAB618 ADP0577 SAB620 ADP0579 SAB623 ADP0580 SAB626 ADP0581 SAB629 ADP0582 SAB630 ADP0583 SAB650 ADP0584 SAB659 ADP0585 SAB686 ADP0586 SAB691 ADP0587 SAB712 ADP0588 SAP1 ADP0590 SEQ11 cranberry red mottled DRK DRK cranberry cranberry cranberry red mottled red mottled cranberry cranberry white red mottled purple mottled ADP0591 VELAZCOLARGO LRK ADP0592 AND1005 ADP0595 G13094 ADP0598 Charlevoix ADP0599 Isles ADP0601 Camelot ADP0603 Wallace773-V98 ADP0604 1062-V98 ADP0605 1132-V96 ADP0606 NY104 ADP0607 NY105 red mottled yellow DRK DRK DRK LRK LRK LRK LRK LRK 101 yes yes yes yes yes yes no yes yes yes yes yes yes yes yes yes yes no yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes no no yes no yes yes yes yes no yes yes yes no no no yes no yes yes no yes yes no yes yes yes no no yes yes yes yes yes no yes no no yes yes yes yes yes yes yes no yes yes yes no no yes yes yes yes yes yes yes yes yes yes yes yes no yes yes yes yes yes yes no yes no Table S.1 (cont’d) ID Genotype Seed Type Grown in Hawassa, ET Kabwe, ZA Lusaka, ZA ADP0608 UI-51 ADP0609 K-407 ADP0610 G122 ADP0611 PompadourB ADP0612 ICAQuimbaya ADP0615 Litekid ADP0616 OACLyrick ADP0617 RedRider ADP0618 ACElk ADP0619 UCD0906 ADP0620 UCD0405 ADP0621 JaloEEP558 ADP0622 UCD0701 ADP0623 Drake ADP0624 Dolly ADP0625 Micran ADP0627 H9659-21-1 ADP0628 H9659-27-7 ADP0629 H9659-27-10 ADP0630 H9659-23-1 ADP0631 OACInferno ADP0632 TARSHT1 ADP0633 TARS-HT2 ADP0635 OACRedstar ADP0636 Montcalm ADP0637 Isabella ADP0638 RedHawk ADP0639 Chinook2000 ADP0640 Beluga ADP0641 Capri ADP0642 TaylorHort ADP0643 Cardinal ADP0644 FoxFire ADP0645 Lassen ADP0646 Myasi ADP0647 RedKanner yes yes yes yes yes yes no yes no yes yes yes yes yes yes yes no yes yes yes no yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes cranberry DRK cranberry red mottled DRK LRK LRK cranberry LRK Jacob's cattle red speckled yellow Jacob's cattle DRK cranberry cranberry LRK LRK LRK LRK LRK DRK LRK DRK DRK LRK DRK LRK white cranberry cranberry cranberry LRK white yellow LRK 102 no yes yes no no yes yes yes no yes yes yes yes no yes yes yes yes yes yes no no yes yes yes no yes yes yes no yes yes yes yes no no no yes yes yes yes yes yes yes yes yes yes yes yes no yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes no yes yes yes yes yes no Table S.1 (cont’d) ID Genotype Seed Type Grown in Hawassa, ET Kabwe, ZA Lusaka, ZA ADP0648 RedKloud ADP0649 Kamiakin ADP0650 K-42 ADP0651 K-59 ADP0652 Lisa ADP0653 USDK-CBB-15 ADP0654 USDK-4 ADP0655 Fiero ADP0656 RoyalRed ADP0657 Kardinal ADP0658 Blush ADP0659 USLK-1 ADP0660 Krimson ADP0662 USCR-9 ADP0663 USCR-CBB-20 ADP0664 SilverCloud ADP0665 USWK-CBB-17 ADP0666 USWK-6 ADP0667 VA-19 ADP0668 Cran-09 ADP0669 OACLyrick ADP0670 ACCalmont ADP0671 ACElk ADP0672 CDRK ADP0673 UCNichols ADP0674 UCD0704 ADP0675 UCD0801 ADP0676 CELRK ADP0677 Etna ADP0678 Hooter ADP0679 RedRover ADP0680 Clouseau ADP0682 UI-686 ADP0683 IJR ADP0684 Majesty ADP0687 PinkPanther yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes no no no yes yes yes yes yes yes yes yes yes yes yes no no LRK LRK LRK LRK white DRK DRK DRK DRK LRK LRK LRK cranberry cranberry cranberry white white white LRK cranberry LRK DRK LRK DRK DRK white cranberry LRK cranberry cranberry DRK LRK cranberry red speckled DRK LRK 103 no yes yes yes no yes no no no yes no no no no yes no no no no yes yes yes yes yes yes no yes yes yes yes yes no no yes yes yes yes yes yes yes yes yes yes no yes yes no no no no yes yes yes yes no yes yes yes yes yes yes no yes yes yes yes yes no no yes yes yes Table S.1 (cont’d) ID Genotype Seed Type Grown in Hawassa, ET Kabwe, ZA Lusaka, ZA ADP0701 1. Bola 60 Dias ADP0704 4. Canela ADP0705 5. Cerrillos ADP0708 8. Gordo ADP0710 10. Dore de Kirundo ADP0711 11. Lingua de Fuego ADP0716 MW-1 ADP0717 MW-2 ADP0719 MW-4 ADP0720 MW-5 ADP0721 MW-6 ADP0722 MW-7 ADP0724 MW-9 ADP0725 MW-10 ADP0726 MW-11 ADP0727 MW-12 ADP0728 MW-13 ADP0729 MW-14 ADP0730 MW-15 ADP0731 MW-16 ADP0732 MW-17 ADP0733 MW-18 ADP0734 MW-19 ADP0735 MW-20 ADP0736 MW-21 ADP0737 MW-22 ADP0738 MW-23 ADP0739 MW-24 ADP0740 MW-25 ADP0741 PI638823 ADP0742 PI661755 ADP0743 PI638811 ADP0744 PI638818 ADP0745 W616496 ADP0746 PI661774 ADP0747 PI638816 no no no no no no yes yes yes no yes yes yes yes yes yes no yes yes yes yes yes no yes yes yes yes yes yes yes no yes yes yes yes no yellow yellow white yellow yellow cranberry cranberry cranberry red mottled cranberry DRK cranberry DRK cranberry cranberry yellow LRK striped red mottled red mottled red mottled red mottled red mottled red mottled red mottled red mottled red striped red mottled brown DRK DRK red speckled DRK DRK DRK 104 no no no yes no no no yes no yes no yes yes yes yes yes yes yes yes yes yes no yes yes yes yes no no yes no yes yes no no yes no yes yes yes no yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes yes no yes yes yes yes yes yes no yes yes Table S.1 (cont’d) ID Genotype Seed Type Grown in Hawassa, ET Kabwe, ZA Lusaka, ZA ADP0748 PI661756 ADP0750 W616493 ADP0751 W616550 ADP0752 PI321119 ADP0753 PI146757 ADP0754 PI661779 ADP0757 Pasi ADP0758 Kabinima ADP0759 Urafiki ADP0760 Wanja ADP0761 Uyole-04 ADP0762 Uyole-84 ADP0763 Punda ADP0764 Kalubungula ADP0765 KK25/KK73/3/666/5-L7 ADP0767 ACUG10-D3 ADP0768 ACUG12-D1 ADP0769 ACUG12-C1 ADP0770 ACUG12-C2 ADP0771 ACUG13-C1 ADP0773 ACUG13-L1 ADP0774 ACUG13-L2 ADP0775 HR202-4973 ADP0776 Dynasty ADP0777 AC-Darkid ADP0778 . ADP0779 CDC-Sol ADP0780 L11YL002 ADP0781 L11YL012 ADP0782 L12LK007 ADP0783 PS03-001-5-1-B3 ADP0784 PS11-006C-8-B ADP0785 PS11-006C-1-B ADP0788 Snowdon ADP0789 PR0313-3 ADP0791 PR1146-123 red speckled DRK red mottled DRK yes yes yes yes black and white yes DRK yellow red mottled DRK yellow yellow yellow yes yes yes yes yes no no purple speckled no no no no no yes yes yes yes yes no yes yes yes yes yes yes yes yes yes yes no no yes red DRK DRK DRK cranberry cranberry cranberry LRK LRK cranberry DRK DRK white yellow yellow yellow LRK cranberry cranberry cranberry white red speckled yellow 105 no yes no no yes no no no no yes no no yes yes yes no no yes no yes no yes no no no no no no no yes no no no no yes no no yes no yes yes no no yes no yes yes yes yes yes yes yes yes yes yes yes no yes yes no no yes yes yes no yes yes no yes yes yes yes Table S.1 (cont’d) ID Genotype Seed Type ADP0792 PR1146-124 ADP0794 Sederberg ADP0796 RS 6 yellow cranberry cranberry Table S2.2 5-point sensory attribute intensity scales. Trait Scale Description Total Flavor Intensity 1-5, bland to strongly flavored Grown in Hawassa, ET Kabwe, ZA Lusaka, ZA yes no no no yes no no yes yes Beany Intensity 1-5, no/very little beany flavor to very strong beany flavor Vegetative Intensity 1-5, no vegetative flavor to very strong vegetative flavor Earthy Intensity 1-5, no earthy flavor to very strong earthy flavor Starchy Intensity 1-5, no starchy taste to very strong starchy taste Sweet Intensity 1-5, no sweet taste to very strong sweet taste Bitter Intensity 1-5, no bitter taste to very strong bitter taste Seed Coat Perception 1-5, imperceptible seed coat to very tough/lingering seed coat Cotyledon Texture 1-5: mushy to very gritty/firm Table S2.3 P-valuesa for the random effects from the sensory attribute intensity ANOVAs at the genotype level. Trait Rep Panelist(Loc) Session(Loc) Total Flavor Intensity Beany Intensity Vegetative Intensity Earthy Intensity Starchy Intensity Sweet Intensity Bitter Intensity NS NS NS NS NS NS NS <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 Seed Coat Perception <0.0001 <0.0001 Cotyledon Texture NS <0.0001 <0.0001 <0.0001 <0.0001 0.0004 <0.0001 <0.0001 0.0002 <0.0001 <0.0001 a NS indicates non-significant p-values at α = 0.05 106 Table S2.4 Mean sensory attribute intensities across the 3 locations for the genotypes exhibiting extreme sensory attribute intensitiesa. ADP ID ADP0106 ADP0681 ADP0654 ADP0530 ADP0057 ADP0206 ADP0190 ADP0517 ADP0005 ADP0450 ADP0791 ADP0044 Total Flavor 2.64 3.98 3.75 3.74 2.51 3.03 3.60 3.64 2.43 2.73 2.16 2.78 Beany Vegetative Earthy Starchy Sweet Bitter Seed Coat Perception Cotyledon Texture 2.96 3.42 3.61 3.04 2.74 2.00 2.59 2.56 2.90 2.44 2.17 2.04 2.00 2.44 1.67 3.39 1.97 2.03 1.20 2.71 1.69 2.38 1.98 1.63 1.65 2.18 2.22 2.83 2.99 2.59 1.82 2.01 1.54 1.97 1.83 1.72 3.10 3.2 2.93 3.48 3.49 4.38 3.63 2.1 2.79 2.94 3.35 2.83 1.25 2.19 1.71 2.01 1.85 2.19 3.46 1.23 1.37 1.48 1.40 1.50 1.74 1.28 2.06 1.45 1.68 1.08 1.03 3.18 1.13 1.28 1.28 2.64 2.80 3.82 3.02 2.68 2.53 2.67 2.34 3.3 1.64 4.43 2.53 2.78 2.86 2.32 2.56 2.86 2.44 3.19 2.82 2.54 2.64 3.07 2.00 3.17 a Extreme attributes exhibited by each genotype are indicated with boxes Table S2.5 P-valuesa for the fixed and random effects from the sensory attribute intensity ANOVAs at the seed type level. Trait Seed Type Loc Seed Type x Loc Rep Reviewer(Loc) Session(Loc) Total Flavor Intensity <.0001 0.0001 <0.0001 0.0008 <0.0001 Beany Intensity <.0001 Vegetative Intensity <.0001 Earthy Intensity Starchy Intensity Sweet Intensity Bitter Intensity <.0001 <.0001 <.0001 <.0001 Seed Coat Perception <.0001 NS NS NS NS NS NS NS 0.0002 0.0006 0.0009 NS <0.0001 0.0032 0.0011 0.0408 <0.0001 NS NS NS <0.0001 <0.0001 <0.0001 0.0314 <0.0001 NS <0.0001 <0.0001 <0.0001 Cotyledon Texture <.0001 0.0009 <0.0001 NS <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 a NS indicates non-significant p-values at α = 0.05 107 Table S2.6 GWAS significant markers associated with sensory attribute intensities determined via BLINK with marker, chromosome (Chr), position, P-value, minor allele frequency (MAF), major and minor alleles (Maj/Min), significance (Sig), and location indicated. Trait Marker Total Flavor Intensity S02_37932341 2 Chr Positiona P-value MAF Maj/Minb Sigc Locd 37932341 1.04E-08 0.25 G/A S03_36213088 3 36213088 4.32E-06 0.18 T/A S03_51252684 3 51252684 1.57E-07 0.45 G/A S04_47465212 4 47465212 8.95E-09 0.12 T/C S05_8530078 5 8530078 1.84E-06 0.19 T/A S05_35951411 5 35951411 8.83E-08 0.09 G/A S05_40598752 5 40598752 7.14E-06 0.10 G/T S06_6583452 S08_4550936 6 8 6583452 1.92E-07 0.42 T/A 4550936 9.09E-09 0.06 C/T S09_10273671 9 10273671 3.39E-06 0.32 G/A S10_42515259 10 42515259 2.38E-08 0.06 T/A S10_42798266 10 42798266 4.04E-06 0.11 A/G *** H ** K *** H *** H *** H *** H * L *** L *** K ** K *** K * L *** H *** K ** K * L ** L ** L *** K *** L *** L *** K ** L *** L *** H *** H * K *** L S10_42528848 10 42528848 6.12E-09 0.12 G/A *** K 4 11 S11_726776 S11_1465049 11 11 S11_46750806 11 Beany Intensity S02_48688740 2 726776 6.73E-07 0.25 G/A 1465049 1.12E-07 0.24 C/T 46750806 1.72E-06 0.25 G/T 48688740 1.29E-05 0.07 A/G S06_5391064 6 5391064 1.93E-06 0.07 T/G S10_44117615 10 44117615 4.69E-06 0.06 C/T S11_44125952 11 44125952 2.57E-09 0.07 C/A S11_46580267 11 Earthy Intensity S02_48899330 2 S04_448769 S11_8151131 Seed Coat Perception S02_34387999 2 46580267 2.21E-08 0.16 A/C 48899330 4.00E-11 0.12 A/C 448769 1.56E-09 0.08 C/T 8151131 7.34E-07 0.09 A/G 34387999 6.35E-07 0.06 G/T S02_49203869 2 49203869 5.40E-10 0.28 G/A S05_1034657 S05_2198768 S07_3664145 5 5 7 1034657 7.00E-09 0.37 T/A 2198768 2.29E-06 0.44 C/G 3664145 2.50E-07 0.49 G/C a Position is based on the P. vulgaris v2.1 reference genome (DOE-JGI and USDA-NIFA, http://phytozome.jgi.doe.gov/) 108 Table S2.6 (cont’d) b Alleles in bold confer a positive effect on the indicated trait c Significance is indicated by asterisks, such that *, **, *** indicate significance at α = 0.1, α = 0.05, α = 0.01 using the false discovery rate d H is Hawassa, Ethiopia; K is Kabwe, Zambia; and L is Lusaka, Zambia 109 Table S2.7 GWAS significant markers associated with cooking time, soak water uptake, raw seed weight, and total water uptake, with chromosome (Chr), position, R2, effect associated with the minor allele, major and minor alleles, minor allele frequency (MAF), major and minor alleles (Maj/Min), significance (Sig), and method indicated. Trait Marker Raw Seed Weight Chr Positiona P-value MAF Maj/Minb Sigc Method S01_47840887 1 47840887 9.04E-07 0.14 G/A *** BLINK S01_49584124 1 49584124 3.14E-06 0.47 C/G *** BLINK S02_33254640 2 33254640 5.72E-18 0.06 T/A *** BLINK S03_40318649 3 40318649 6.57E-06 0.30 G/A ** BLINK S03_41895570 3 41895570 3.55E-07 0.31 A/C *** BLINK, MLM S04_1769598 S05_1069847 S05_1138961 4 5 5 1769598 1.36E-06 0.12 G/A *** BLINK 1069847 1.79E-15 0.49 A/G *** BLINK 1138961 6.10E-07 0.36 C/T ** MLM S05_36225413 5 36225413 1.05E-11 0.15 T/C *** BLINK S06_18456447 6 18456447 2.59E-09 0.08 G/A *** BLINK S07_1842933 7 1842933 1.91E-06 0.18 C/A *** BLINK S07_25513414 7 25513414 8.08E-06 0.18 T/C ** BLINK S08_61954787 8 61954787 7.27E-12 0.39 A/G *** BLINK S09_33770475 9 33770475 1.27E-06 0.07 A/G *** BLINK S11_46634045 11 Soak Water Uptake S02_47837868 2 46634045 4.36E-07 0.13 G/A *** BLINK 47837868 8.17E-08 0.25 G/A *** BLINK S03_25546920 3 25546920 5.30E-06 0.05 C/T * MLM S03_50652595 3 50652595 3.36E-06 0.07 A/T ** MLM S03_51140861 3 51140861 1.89E-11 0.07 G/A *** BLINK S04_30764016 4 30764016 1.41E-05 0.24 C/T ** BLINK S04_47654443 4 47654443 1.26E-05 0.08 G/A * MLM S05_37924556 5 37924556 4.53E-06 0.07 C/G ** MLM S07_3919560 7 3919560 2.66E-06 0.06 C/G *** BLINK S07_18212326 7 18212326 1.34E-06 0.09 A/G *** BLINK S07_27774103 7 27774103 4.22E-10 0.19 C/T *** BLINK S07_38497123 7 38497123 3.20E-06 0.38 A/G *** BLINK S07_39390008 7 39390008 1.10E-09 0.42 A/G *** BLINK S08_59981977 8 59981977 1.49E-06 0.05 T/A *** BLINK S08_60478317 8 60478317 1.14E-07 0.34 C/T *** BLINK S10_37637761 10 37637761 8.68E-08 0.12 T/C *** BLINK,MLM S10_43391440 10 43391440 3.25E-07 0.06 A/G ** MLM S11_5714496 11 5714496 1.83E-11 0.05 C/A *** BLINK 110 Table S2.7 (cont’d) Cooking Time S03_4885990 3 3 4885990 1.03E-07 0.05 T/G *** BLINK S03_5243893 5243893 2.04E-06 0.07 A/G S03_51292502 3 51292502 2.07E-05 0.06 A/T * * BLINK BLINK S04_3957256 4 3957256 2.24E-05 0.24 C/G * MLM S04_47068842 4 47068842 2.93E-08 0.08 A/G *** BLINK S06_19636517 6 19636517 2.02E-05 0.08 T/G * BLINK S07_3009718 7 3009718 6.05E-08 0.07 T/C *** BLINK S07_30919254 7 30919254 1.54E-05 0.35 T/C * BLINK S08_60104796 8 60104796 2.31E-06 0.27 C/A ** BLINK S08_62659170 8 62659170 3.52E-06 0.16 A/G ** MLM S11_10805992 11 Total Water Uptake S03_2580077 S03_7619818 10805992 2.22E-07 0.10 C/T *** BLINK 2580077 1.11E-06 0.07 G/A ** BLINK 7619818 1.17E-06 0.27 T/C ** BLINK 3 3 S04_30764016 4 30764016 4.25E-07 0.24 C/T ** BLINK S09_37046204 9 37046204 8.29E-06 0.08 C/T * BLINK S11_48753729 11 48753729 3.44E-06 0.12 T/A ** BLINK a Position is based on the P. vulgaris v2.1 reference genome (DOE-JGI and USDA-NIFA, http://phytozome.jgi.doe.gov/) b Alleles in bold confer a positive effect on the indicated trait c Significance is indicated by asterisks, such that *, **, *** indicate significance at α = 0.1, α = 0.05, α = 0.01 using the false discovery rate for the BLINK method and a Bonferroni correction based on the effective number of markers determined using the SimpleM algorithm for the MLM method 111 Figure S2.1 Images of the genotypes exhibiting extreme sensory attribute intensities identified from screening accessions of the Andean Diversity Panel grown in Hawassa, Ethiopia. 112 Figure S2.2 Manhattan and QQ plots for total flavor intensity, beany intensity, earthy intensity, seed coat perception, and cotyledon texture of the Andean Diversity Panel with mapping conducted using MLM with BLUPs from all locations combined. The gray dashed line is the α = 0.05 Bonferroni correction based on the effective number of markers determined using the SimpleM algorithm. 113 Figure S2.3 Manhattan and QQ plots for total flavor intensity of the Andean Diversity Panel with mapping conducted using BLINK with BLUPs for Hawassa, Ethiopia (H); Kabwe, Zambia (K); and Lusaka, Zambia (L). The gray dashed line is the α = 0.05 FDR. Figure S2.4 Manhattan and QQ plots for beany intensity of the Andean Diversity Panel with mapping conducted using BLINK with BLUPs for Hawassa, Ethiopia (H); Kabwe, Zambia (K); and Lusaka, Zambia (L). The gray dashed line is the α = 0.05 FDR. 114 Figure S2.5 Manhattan and QQ plots for earthy intensity of the Andean Diversity Panel with mapping conducted using BLINK with BLUPs for Hawassa, Ethiopia (H); Kabwe, Zambia (K); and Lusaka, Zambia (L). The gray dashed line is the α = 0.05 FDR. Figure S2.6 Manhattan and QQ plots for seed coat perception of the Andean Diversity Panel with mapping conducted using BLINK with BLUPs for Hawassa, Ethiopia (H); Kabwe, Zambia (K); and Lusaka, Zambia (L). The gray dashed line is the α = 0.05 FDR. 115 Figure S2.7 Manhattan and QQ plots for raw seed weight, soak water uptake, cooking time, and total water uptake of the Andean Diversity Panel with mapping conducted using BLINK with BLUPs from all locations combined. The gray dashed line is the α = 0.05 FDR. 116 Figure S2.8 Manhattan and QQ plots for raw seed weight, soak water uptake, cooking time, and total water uptake of the Andean Diversity Panel with mapping conducted using MLM with BLUPs from all locations combined. The gray dashed line is the α = 0.05 Bonferroni correction based on the effective number of markers determined using the SimpleM algorithm. 117 Figure S2.9 Phenotypic effects of carrying the indicated number of significant markers conferring a positive effect for raw seed weight, soak water uptake, and total water uptake and a negative effect for cooking time. Phenotypic values represent all locations combined as averages from Hawassa, Ethiopia; Kabwe, Zambia; and Lusaka, Zambia. N is the number of individuals in each boxplot. 118 REFERENCES 119 REFERENCES Akibode, S., and Maredia, M. (2011). Global and regional trends in production, trade and consumption of food legume crops. Dep. Agric. Food Resour. Econ. Michigan State Univ., 87. Available at: http://impact.cgiar.org/sites/default/files/images/Legumetrendsv2.pdf. Amyotte, B., Bowen, A. J., Banks, T., Rajcan, I., and Somers, D. J. (2017). Mapping the sensory perception of apple using descriptive sensory evaluation in a genome wide association study. PLoS One 12. doi:10.1371/journal.pone.0171710. Andrews, S. (2010). FastQC: a quality control tool for high throughput sequence data. Available at: http://www.bioinformatics.babraham.ac.uk/projects/fastqc. Armelim, J. M., Canniatti-Brazaca, S. G., Spoto, M. H. F., Arthur, V., and Piedade, S. M. S. (2006). Quantitative descriptive analysis of common bean (Phaseolus vulgaris L.) under gamma radiation. J. Food Sci. 71, 8–12. doi:10.1111/j.1365-2621.2006.tb12410.x. Asioli, D., Aschemann-Witzel, J., Caputo, V., Vecchio, R., Annunziata, A., Næs, T., et al. (2017). Making sense of the “clean label” trends: A review of consumer food choice behavior and discussion of industry implications. Food Res. Int. 99, 58–71. doi:10.1016/j.foodres.2017.07.022. Banking, W. B. I. (2016). Secular trends drive M & A interest in value-added ingredients and food solutions. Bates, D., Mächler, M., Bolker, B. M., and Walker, S. C. (2015). Fitting linear mixed-effects models using lme4. J. Stat. Softw. 67. doi:10.18637/jss.v067.i01. Bauchet, G., Grenier, S., Samson, N., Segura, V., Kende, A., Beekwilder, J., et al. (2017). Identification of major loci and genomic regions controlling acid and volatile content in tomato fruit: implications for flavor improvement. New Phytol. 215, 624–641. doi:10.1111/nph.14615. Beaver, J. S., Prophete, E., Démosthène, G., and Porch, T. G. (2016). Registration of PR1146-138 yellow bean germplasm line. J. Plant Regist. 10, 145–148. doi:10.3198/jpr2015.09.0058crg. Benjamini, Y., and Hochberg, Y. (1995). Controlling the false discovery rate: A practical and powerful approach to multiple testing. J. R. Stat. Soc. Ser. B 57, 289–300. Berry, M., Izquierdo, P., Jeffery, H., Shaw, S., Nchimbi-Msolla, S., and Cichy, K. A. (2020). QTL analysis of cooking time and quality traits in dry bean (Phaseolus vulgaris L.). Theor. Appl. Genet., Accepted with minor revision. doi:10.1007/s00122-020-03598-w. Borchgrevink, C. P. (2013). “Culinary perspective of dry beans and pulses,” in Dry beans and pulses production, processing and nutrition, eds. M. Siddiq and M. A. Uebersax (Ames, Iowa: John Wiley & Sons, Inc.), 313–334. 120 in Mexico. Arch. Latinoam. Nutr. 47, 163—167. Available Bott, L., and Chambers, E. (2006). Sensory characteristics of combinations of chemicals potentially associated with beany aroma in foods. J. Sens. Stud. 21, 308–321. doi:10.1111/j.1745- 459X.2006.00067.x. Bradbury, P. J., Zhang, Z., Kroon, D. E., Casstevens, T. M., Ramdoss, Y., and Buckler, E. S. (2007a). TASSEL: software for association mapping of complex traits in diverse samples. Bioinformatics 23, 2633–2635. doi:10.1093/bioinformatics/btm308. Bradbury, P. J., Zhang, Z., Kroon, D. E., Casstevens, T. M., Ramdoss, Y., and Buckler, E. S. (2007b). TASSEL: Software for association mapping of complex traits in diverse samples. Bioinformatics 23, 2633–2635. doi:10.1093/bioinformatics/btm308. Castellanos, J. Z., Guzmán Maldonado, H., Jiménez, A., Mejía, C., Muñoz Ramos, J. J., Acosta Gallegos, J. A., et al. (1997). Preferential habits of consumers of common bean (Phaseolus vulgaris L.) at: http://europepmc.org/abstract/MED/9659433. Cichy, K. A., Fernandez, A., Kilian, A., Kelly, J. D., Galeano, C. H., Shaw, S., et al. (2014). QTL analysis of canning quality and color retention in black beans (Phaseolus vulgaris L.). Mol. Breed. 33, 139–154. doi:10.1007/s11032-013-9940-y. Cichy, K. A., Porch, T. G., Beaver, J. S., Cregan, P., Fourie, D., Glahn, R. P., et al. (2015). A Phaseolus vulgaris diversity panel for Andean bean improvement. Crop Sci. 55, 2149–2160. doi:10.2135/cropsci2014.09.0653. Dougkas, A., Vannereux, M., and Giboreau, A. (2019). The impact of herbs and spices on increasing the appreciation and intake of low-salt legume-based meals. Nutrients 11, 2901. doi:10.3390/nu11122901. Duitama, J., Quintero, J. C., Cruz, D. F., Quintero, C., Hubmann, G., Foulquié-Moreno, M. R., et al. (2014). An integrated framework for discovery and genotyping of genomic variants from high- throughput sequencing experiments. Nucleic Acids Res. 42, e44–e44. doi:10.1093/nar/gkt1381. Eihusen, J., and Albrecht, J. A. (2007). Dry bean intake of women ages 19-45. Rural. Rev. Undergrad. Research Agric. Life Sci. 2. Elshire, R. J., Glaubitz, J. C., Sun, Q., Poland, J. A., Kawamoto, K., Buckler, E. S., et al. (2011). A robust, simple genotyping-by-sequencing (GBS) approach for high diversity species. PLoS One 6. doi:10.1371/journal.pone.0019379. Ferreira, J. J., del Castillo, R. R., Perez-Vega, E., Plans, M., Simó, J., and Casañas, F. (2012). Sensory changes related to breeding for plant architecture and resistance to viruses and anthracnose (Phaseolus vulgaris L.). Euphytica 186, 687–696. in bean market class Fabada doi:10.1007/s10681-011-0540-9. 121 and genomics database. Nucleic Acids Res. Gao, X., Starmer, J., and Martin, E. R. (2008). A multiple testing correction method for genetic association studies using correlated single nucleotide polymorphisms. Genet. Epidemiol. 32, 361– 369. doi:10.1002/gepi.20310. Gilham, B., Hall, R., and Woods, J. L. (2018). Vegetables and legumes in new Australasian food launches: how are they being used and are they a healthy choice? Nutr. J. 17, 1–9. doi:10.1186/s12937-018-0414-2. Grant, D., Nelson, R. T., Cannon, S. B., and Shoemaker, R. C. (2009). SoyBase, the USDA-ARS soybean genetics 38, 843–846. doi:10.1093/nar/gkp798. Hart, J. P., and Griffiths, P. D. (2015). Genotyping-by-sequencing enabled mapping and marker development for the potyvirus resistance allele in common bean. Plant Genome 8. doi:10.3835/plantgenome2014.09.0058. Holland, J. B., Nyquist, W. E., and Cervantes-Martínez, C. T. (2003). “Estimating and interpreting heritability for plant breeding: An update,” in Plant Breeding Reviews, ed. J. Janick (John Wiley & Sons, Inc.), 9–112. Hooper, S. D., Glahn, R. P., and Cichy, K. A. (2019). Single Varietal Dry Bean (Phaseolus vulgaris L.) Pastas: Nutritional Profile and Consumer Acceptability. Plant Foods Hum. Nutr. 74, 342–349. doi:10.1007/s11130-019-00732-y. Huang, M., Liu, X., Zhou, Y., Summers, R. M., and Zhang, Z. (2018). BLINK: A package for the next level of genome-wide association studies with both individuals and markers in the millions. Gigascience 8, 1–12. doi:10.1093/gigascience/giy154. IFIC (2019). 2019 Food and Health Survey. Available at: https://foodinsight.org/2019-food-and- health-survey/. Katuuramu, D. N., Hart, J. P., Porch, T. G., Grusak, M. A., Glahn, R. P., and Cichy, K. A. (2018). Genome-wide association analysis of nutritional composition-related traits and iron bioavailability in cooked dry beans (Phaseolus vulgaris L.). Mol. Breed. 38. doi:10.1007/s11032-018-0798-x. Kelly, J. D., and Bornowski, N. (2018). “Marker-assisted breeding for economic traits in common bean,” in Biotechnologies of Crop Improvement, Volume 3: Genomic Approaches, eds. S. S. Gosal and S. H. Wani (Cham: Springer International Publishing), 211–238. doi:10.1007/978-3-319- 94746-4_10. Kelly, J. D., and Cichy, K. A. (2012). Dry bean breeding and production technologies. Dry Beans Pulses Prod. Process. Nutr., 23–54. doi:10.1002/9781118448298.ch2. Kelly, J. D., Varner, G. V., and Wright, E. M. (2010). Registration of ‘Bellagio’ cranberry bean. J. Plant Regist. 4, 171–174. doi:10.3198/jpr2010.02.0069crc. 122 Kelly, J. D., Wright, E. M., and Wiersma, A. (2018). 2018 Dry Bean Breeding Yield Trials. Kim, M. Y., Ha, B. K., Jun, T. H., Hwang, E. Y., Van, K., Kuk, Y. I., et al. (2004). Single nucleotide polymorphism discovery and linkage mapping of lipoxygenase-2 gene (Lx2) in soybean. Euphytica 135, 169–177. doi:10.1023/B:EUPH.0000014909.40136.20. Kinsella, J. E. (1979). Functional properties of soy proteins. J. Am. Oil Chem. Soc. 56, 242–258. doi:10.1007/BF02671468. Koehler, H. H., Chang, C.-H., Scheier, G., and Burke, D. W. (1987). Nutrient composition, protein quality, and sensory properties of thirty-six cultivars of dry beans (Phaseolus vulgaris L.). J. Food Sci. 52, 1335–1340. doi:10.1111/j.1365-2621.1987.tb14076.x. Langmead, B., and Salzberg, S. L. (2012). Fast gapped-read alignment with Bowtie 2. Nat. Methods 9, 357–359. doi:10.1038/nmeth.1923. Lenis, J. M., Gillman, J. D., Lee, J. D., Shannon, G. G., and Bilyeu, K. D. (2010). Soybean seed lipoxygenase genes: Molecular characterization and development of molecular marker assays. Theor. Appl. Genet. 120, 1139–1149. doi:10.1007/s00122-009-1241-9. Leterme, P., and Carmenza Muũoz, L. (2002). Factors influencing pulse consumption in Latin America. Br. J. Nutr. 88, 251–254. doi:10.1079/bjn/2002714. Li, H. (2018). Minimap2: Pairwise alignment for nucleotide sequences. Bioinformatics 34, 3094– 3100. doi:10.1093/bioinformatics/bty191. Liu, X., Huang, M., Fan, B., Buckler, E. S., and Zhang, Z. (2016). Iterative usage of fixed and random effect models for powerful and efficient genome-wide association studies. PLOS Genet. 12, e1005767. doi:10.1371/journal.pgen.1005767. Martin, M. (2011). Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal 17, 10. doi:10.14806/ej.17.1.200. McTigue, M. C., Koehler, H. H., and Silbernagel, M. J. (1989). Comparison of four sensory evaluation methods for assessing cooked dry bean flavor. J. Food Sci. 54, 1278–1283. doi:10.1111/j.1365-2621.1989.tb05973.x. Meilgaard, M. C., Civille, G. V., and Carr, B. T. (1999). Sensory evaluation techniques. 3rd ed. Boca Raton, Fla: CRC Press. Miklas, P. N., Hang, A. N., Kelly, J. D., Strausbaugh, C. A., and Forster, R. L. (2004). Registrations of three kidney bean germplasm lines resistant to bean common mosaic and necrosis potyviruses: USLK-2 light red kidney, USDK-4 dark red kidney, and USWK-6 white kidney. Crop Sci. 44, 343–344. 123 Minchala, L., Murillo, Á., Peralta, E., Guamán, M., and Pinzón, J. (2003). INIAP 422 Blanco Belén: new variety of bush beans for the sourthern highlands of Ecuador. Mkanda, A. V, Minnaar, A., and de Kock, H. L. (2007). Relating consumer preferences to sensory and physicochemical properties of dry beans (Phaseolus vulgaris). J. Sci. Food Agric. 87, 2868– 2879. doi:10.1002/jsfa.3046. Nachay, K. (2017). The power of pulses. Food Technol. Mag. 71. Parr, B., and Lucier, G. (2020). Dry pulse crops: U.S. per capita availability, 1970-2019. Perea, C., De La Hoz, J. F., Cruz, D. F., Lobaton, J. D., Izquierdo, P., Quintero, J. C., et al. (2016). Bioinformatic analysis of genotype by sequencing (GBS) data with NGSEP. BMC Genomics 17. doi:10.1186/s12864-016-2827-7. Pérez-Vega, E., Pañeda, A., Rodríguez-Suárez, C., Campa, A., Giraldez, R., and Ferreira, J. J. (2010). Mapping of QTLs for morpho-agronomic and seed quality traits in a RIL population of common bean (Phaseolus vulgaris L.). Theor. Appl. Genet. 120, 1367–1380. doi:10.1007/s00122- 010-1261-5. Plans, M., Simó, J., Casañas, F., del Castillo, R. R., Rodriguez-Saona, L. E., and Sabaté, J. (2014). Estimating sensory properties of common beans (Phaseolus vulgaris L.) by near infrared spectroscopy. Food Res. Int. 56, 55–62. doi:10.1016/j.foodres.2013.12.003. R Core Team (2017). R: A Language and Environment for Statistical Computing. Available at: https://www.r-project.org/. Rivera, A., Fenero, D., Almirall, A., Ferreira, J. J., Simó, J., Plans, M., et al. (2013). Variability in sensory attributes in common bean (Phaseolus vulgaris L.): A first survey in the Iberian secondary diversity center. Genet. Resour. Crop Evol. 60, 1885–1898. doi:10.1007/s10722-013-9963-6. Roland, W. S. U., Pouvreau, L., Curran, J., Van De Velde, F., and De Kok, P. M. T. (2017). Flavor aspects of pulse ingredients. Cereal Chem. 94, 58–65. doi:10.1094/CCHEM-06-16-0161-FI. Romero del Castillo, R., Costell, E., Plans, M., Simó, J., and Casañas, F. (2012). A standardized method of preparing common beans (Phaseolus vulgaris L.) for sensory analysis. J. Sens. Stud. 27, 188–195. doi:10.1111/j.1745-459X.2012.00381.x. Romero del Castillo, R., Valero, J., Casañas, F., and Costell, E. (2008). Training, Validation and Maintenance of a Panel to Evaluate the Texture of Dry Beans (Phaseolus vulgaris L.). J. Sens. Stud. 23, 303–319. Sandhu, K. S., You, F. M., Conner, R. L., Balasubramanian, P. M., and Hou, A. (2018). Genetic analysis and QTL mapping of the seed hardness trait in a black common bean (Phaseolus vulgaris) recombinant inbred line (RIL) population. Mol. Breed. 38. doi:10.1007/s11032-018-0789-y. 124 Agron. société 3, 201–204. Environ. Available Sanz-Calvo, M., and Atienza-del-Rey, J. (1999). Sensory analysis of beans (Phaseolus vulgaris). Biotechnol. at: http://www.doaj.org/doaj?func=abstract&id=94222&recNo=3&toc=1&uiLanguage=en. Schmutz, J., McClean, P. E., Mamidi, S., Wu, G. A., Cannon, S. B., Grimwood, J., et al. (2014). A reference genome for common bean and genome-wide analysis of dual domestications. Nat. Genet. 46, 707–13. doi:10.1038/ng.3008. Scott, J., and Maideni, M. (1998). Socio-economic survey of three bean growing areas of Malawi. Kampala, Uganda. Stone, H., Sidel, J., Oliver, S., Woolsey, A., and Singleton, R. C. (1974). Sensory evaluation by quantitative descriptive analysis. Food Technol. 28, 51. Swarts, K., Li, H., Romero Navarro, J. A., An, D., Romay, M. C., Hearne, S., et al. (2014). Novel methods to optimize genotypic imputation for low-coverage, next-generation sequence data in crop plants. Plant Genome 7, plantgenome2014.05.0023. doi:10.3835/plantgenome2014.05.0023. Talukdar, A., and Shivakumar, M. (2016). Genetic improvement of food-grade soybean in India: Current status and future prospects. Indian J. Genet. Plant Breed. 76, 626–630. doi:10.5958/0975- 6906.2016.00077.8. Wang, N., and Daun, J. K. (2005). Determination of cooking times of pulses using an automated Mattson cooker apparatus. J. Sci. Food Agric. 85, 1631–1635. doi:10.1002/jsfa.2134. Wang, Z., Bao, G., Yang, C., Yang, M., Zhao, X., Shao, Y., et al. (2020). A genome-wide association study of hexanal content related to soymilk off-flavours in seed of soybean (Glycine max). Crop Pasture Sci. doi:10.1071/cp20068. Winham, D., Tisue, M., Palmer, S., Cichy, K., and Shelley, M. (2019). Dry bean preferences and attitudes among Midwest Hispanic and non-Hispanic White women. Nutrients 11, 178. doi:10.3390/nu11010178. Wolf, W. J., Cowan, J. C., and Wolff, H. (1971). Soybeans as a food source. C R C Crit. Rev. Food Technol. 2, 81–158. doi:10.1080/10408397109527117. Xia, N., Yan, W. bing, Wang, X. qi, Shao, Y. peng, Yang, M. ming, Wang, Z. kun, et al. (2019a). Genetic dissection of hexanol content in soybean seed through genome-wide association analysis. J. Integr. Agric. 18, 1222–1229. doi:10.1016/S2095-3119(19)62674-2. Xia, N., Yang, M., Zhao, J., Shao, Y., Shi, Y., Yan, W., et al. (2019b). Genome-wide association analysis of 1-octen-3-ol content related to soymilk off-flavor in soybean seed. Crop Pasture Sci. 70, 133–139. doi:10.1071/CP18423. Yang, J., Lee, S. H., Goddard, M. E., and Visscher, P. M. (2011). GCTA: A tool for genome-wide complex trait analysis. Am. J. Hum. Genet. 88, 76–82. doi:10.1016/j.ajhg.2010.11.011. 125 Zhang, J., Zhao, J., Xu, Y., Liang, J., Chang, P., Yan, F., et al. (2015). Genome-wide association mapping for tomato volatiles positively contributing to tomato flavor. Front. Plant Sci. 6, 1–13. doi:10.3389/fpls.2015.01042. Zhao, J., Sauvage, C., Zhao, J., Bitton, F., Bauchet, G., Liu, D., et al. (2019). Meta-analysis of genome-wide association studies provides insights into genetic control of tomato flavor. Nat. Commun. 10, 1–12. doi:10.1038/s41467-019-09462-w. 126 CHAPTER 3: QTL MAPPING OF SEED QUALITY TRAITS INCLUDING COOKING TIME, FLAVOR, AND TEXTURE IN YELLOW DRY BEANS (Phaseolus vulgaris L.) 127 QTL mapping of seed quality traits including cooking time, flavor, and texture in yellow dry beans (Phaseolus vulgaris L.) Amber Bassett1, Dennis Katuuramu12, and Karen Cichy13 1 Department of Plant, Soil and Microbial Sciences, Michigan State University, East Lansing, MI 2 U.S. Vegetable Laboratory, USDA-ARS, Charleston, SC 3 Sugarbeet and Bean Research Unit, USDA-ARS, East Lansing, MI ABSTRACT Manteca yellow beans have many quality traits that appeal to consumers, including fast cooking times, creamy texture, and sweet, buttery flavor. These beans are native to Chile and consumed in regions in South America and Africa, but are largely unfamiliar to U.S. consumers. While cooking time, flavor, and texture have not been a focus in U.S. dry bean breeding programs, genetic variability exists, which could allow consumer preferences for these traits to be addressed through breeding. In this study, a recombinant inbred line (RIL) population was developed from Ervilha and PI527538, Manteca and Njano yellow beans with contrasting cooking time and sensory attributes. The population and parents were grown for two years in Michigan and evaluated for cooking time and sensory attribute intensities, including total flavor, beany, vegetative, earthy, starchy, sweet, bitter, seed coat perception, and cotyledon texture. Cooking time ranged from 19 to 34 minutes and exhibited a high broad-sense heritability of 0.76. Sensory attribute intensities also exhibited variation among the RILs, although broad-sense heritability was low, with beany and total flavor intensity exhibiting the highest (0.33 and 0.27). A linkage map of 973 SNP markers was developed for QTL mapping, which revealed important loci for soak water uptake, cooking time, sensory attribute intensities, color, seed coat postharvest non-darkening, seed weight, total 128 water uptake, and seed yield. Co-localization was identified for total flavor, beany, starchy, bitter, seed coat perception, cotyledon texture, and color (Pv03); vegetative, earthy, sweet, and cotyledon texture (Pv07); and color and non-darkening (Pv10). INTRODUCTION Dry beans (Phaseolus vulgaris L.) are widely regarded as a nutritious and affordable food (Akibode and Maredia, 2011). The species encompasses many different market classes grown and consumed around the world with many regional preferences (Siddiq and Uebersax, 2012). There is variability not just for seed size, color, and shape, but also end-use quality attributes, including cooking time, color, and flavor (Bassett et al., 2020). Some market classes may be of particular interest to modern consumers looking to incorporate beans into their diets for their nutritional benefits and also looking for convenience not typically associated with dry beans considering their often long cooking times (Sloan, 2015). The Manteca yellow bean market class has multiple quality traits of value to consumers (Leakey, 2000; Wiesinger et al., 2016, 2018). Manteca are pale yellow with a grey hilum. They are Andean beans native to Chile (Leakey, 1992) and currently consumed in South America and Africa (Wiesinger et al., 2018). Manteca are appreciated for their sweet, buttery flavor (Leakey, 2000) as well as fast cooking time and high iron bioavailability (Wiesinger et al., 2016, 2018). U.S. American consumers are largely unfamiliar with this yellow market class, making it easy to set apart from familiar market classes and highlight its positive attributes in new varieties. Current dietary guidelines in the US recommend ¼ cup of pulse per day, but less than 50% of the population meets that recommendation (Britten et al., 2012). There is an opportunity to increase utilization of dry beans by addressing consumer preferences for convenience and flavor 129 as well as developing bean products to reach new consumers (IPSOS, 2010; Karlsen et al., 2016; Hooper et al., 2019; Winham et al., 2019). While U.S. dry bean breeders have always prioritized quality traits, they primarily have focused on seed size, shape, color, and canning quality and production-related traits with minor if any consideration for cooking time and flavor (Kelly and Cichy, 2012). As a result, genetic variability exists for cooking time, flavor, and texture in modern cultivars as well as the breeding lines used for their development (Bassett et al., 2020b). There is an opportunity to address these consumer-valued traits through breeding to increase dry bean consumption, and Manteca beans make a prime target for this effort, as they already excel in these traits and provide additional novelty to those unfamiliar with them. Cooking time has been reported to be controlled by few genes and have moderate to high heritability, with narrow sense heritability values estimated between 0.74 and 0.90 (Elia et al., 1997; Jacinto-Hernandez et al., 2003). Genotypic cooking time patterns are stable across environments (Cichy et al., 2019). Following screening of 206 accessions of the Andean Diversity Panel (ADP), several significant single nucleotide polymorphisms (SNPs) associated with cooking time were identified on Pv02, Pv03 and Pv06 (Cichy et al., 2015b). A more recent screening of 430 accessions of the ADP revealed additional significant SNPs on Pv03, Pv04, Pv06, Pv07, Pv08, and Pv11 (Bassett et al., 2020). In addition, a recent quantitative trait loci (QTL) mapping study using a recombinant inbred line (RIL) population developed from two ADP accessions revealed QTL for cooking time on Pv01, Pv02, Pv03, Pv05, Pv06, Pv10, and Pv11 (Berry et al., 2020). With further study, marker-assisted selection may be a feasible method for breeding faster cooking beans, which could reduce the need to phenotype for cooking time and allow greater incorporation of the fast cooking trait in breeding programs. 130 Flavor is a major influence on consumer food choices (Glanz, Basil et al. 1998), but evaluating flavor and texture is time consuming and requires trained panelists. As it stands, little is understood about consumer preference in regard to flavor and texture in dry beans apart from a general preference for beans that are sweet and soft and for bean products without a beany “off” flavor (Kinsella, 1979; Bott and Chambers, 2006; Mkanda et al., 2007; Hooper et al., 2019). A few studies have identified genetic variability for sensory attributes, including flavor and texture acceptability, seed coat perception, seed coat roughness, cotyledon mealiness, and beany flavor intensity (Koehler et al., 1987; Rivera et al., 2013). A recent study identified genetic variability in the Andean Diversity Panel (ADP) for total, beany, vegetative, earthy, starchy, bitter, and sweet flavor intensities as well as seed coat perception and cotyledon texture (Bassett et al., 2020b). Using a genome-wide association approach, significant SNPs were identified for many of these traits. As for cooking time, the potential for marker-assisted selection could reduce the need for extensive phenotyping and allow breeders to incorporate flavor and texture into their breeding programs more easily. With a greater understanding of consumer preference for flavor and texture, new varieties could be developed that appeal to consumers and are suitable for use as ingredients in products. In this study, a yellow bean RIL population developed from two ADP accessions with contrasting cooking time and sensory characteristics was screened for cooking time and sensory attribute intensities to elucidate their genetic control and aid in the development of molecular markers for these traits. 131 MATERIALS AND METHODS Germplasm A RIL population of 240 F5:F7-F8 lines was developed from two yellow bean genotypes of the Andean gene pool: Ervilha (ADP0512) and PI527538 (ADP0468) (Figure 3.1) (Bassett and Cichy, 2020). The RILs were developed by advancing F2 seed via single seed descent to the F5 generation and then bulking seeds from individual plants to form RILs. Ervilha is a pale yellow Manteca seed type with a gray hilum that was collected at a marketplace in Angola in 2010 (Cichy et al., 2015a). PI527538 is a yellow-green Njano seed type with hints of purple and a black hilum that was collected in Burundi in 1985 (Cichy et al., 2015a). Both genotypes are likely members of race Nueva Granada. These genotypes were selected to develop a RIL population after a screening of 206 lines of the Andean Diversity Panel (ADP) for cooking time (Cichy et al., 2015b). Ervilha cooks faster than PI527538, and this relative difference in cooking time is stable across environments (Cichy et al., 2015b; Katuuramu et al., 2020). The genotypes were grown at the Montcalm Research Farm in MI in 2016 and 2017. The soil type is Eutric Glossoboralfs (coarse-loamy, mixed) and Alfic Fragiorthods (coarse-loamy, mixed, frigid). Two row plots 4.75 m long with 0.5 m spacing between rows were arranged in a randomized complete block design with two replications per genotype. In 2016, 100 seeds were planted per plot due to limited seed, and in 2017, 160 seeds were planted per plot. Standard agronomic practices were followed as described in the MSU SVREC 2017 Farm Research Report (Kelly et al., 2017). Plants were hand-pulled at maturity and threshed with a Hege 140 plot harvester (Wintersteiger, Utah, USA). Following harvest, seeds were cleaned by hand to remove field debris, off types, and damaged seed. Seed weights (g per 100 seeds) and seed yield (kg per ha) were recorded for each field replicate. 132 CIELAB Analysis and Seed Coat Postharvest Darkening For both years, images were collected for one field replicate of each genotype using a custom machine vision system as described in Mendoza et al., 2017. For each image, a 60 x 15 mm petri dish was filled with representative seeds cleaned of debris and damaged seeds. The EOS Rebel T3i software settings were consistent across each image as follows: lens aperture f = 5.6, shutter speed 1/125, white balanced, and ISO = 100. Following image collection, each image was cropped to center the petri dish and minimize background. To examine the relationship among color, cooking time, and sensory attributes, CIELAB values were obtained using a custom batch macro in ImageJ that applies a gamma correction of 0.5, excludes background pixels outside the petri dish, and measures each slice of the LAB stack. CIELAB uses three values to describe color: L* for black (0) to white (100), a* for green (−) to red (+), and b* for blue (−) to yellow (+). These values were collected relative to the imaging conditions and reflect average color of seeds without calibration for the purpose of observing differences among lines rather than determining absolute color. Variability in seed coat postharvest darkening among genotypes was observed after the first year, so the potential presence of the non-darkening trait in this population was explored. Genotypes grown in 2017 were stored for approximately two years in opaque paper bags in a cool, dry barn prior to evaluation for seed coat postharvest darkening in January 2020. Samples that appeared visibly darkened after this storage period were given a score of 1 and those that remained light were given a score of 0. Cooking Time Evaluation For each year, two field replicates of 30 seed per genotype were equilibrated to 10-14% moisture in a 4 °C humidity chamber prior to evaluating for cooking time. Each 30 seed sample 133 was soaked for 12 hours in distilled water prior to cooking time evaluation using an automated Mattson cooker method (Wang and Daun, 2005). Genotypes were cooked in a random order to minimize seed aging effects. Seed weights after soaking were recorded for each sample to determine soak water uptake. Mattson cookers loaded with soaked seeds were placed into 4 L stainless steel beakers with 1.8 L of boiling distilled water on Cuisinart CB-30 Countertop Single Burners to cook. The Mattson cookers (Michigan State University Machine Shop, East Lansing, MI) use twenty-five 65g stainless steel rods with 2mm diameter pins to pierce beans as they finish cooking in each well. As the pins drop, a custom software reports the cooking time associated with each pin. A low boil was maintained during cooking, and the 80% cooking times were recorded and regarded as the time required to fully cook each sample. Final cooked seed weights were recorded, and the total water uptake following cooking was calculated. Sensory Evaluation Ervilha, PI527538, and the RILs were evaluated in duplicate using a modified Quantitative Descriptive Analysis (QDA) approach (Stone et al., 1974), in which four panelists per session independently evaluated samples using a non-consensus approach to limit group bias. For the purposes of this study, the QDA approach was modified as described by Bassett et al. (2020b) to increase suitability for implementation in public breeding programs with limited resources. In brief, seeds from each field replicate were prepared for sensory evaluation in the same order that they were cooked for cooking time evaluation to minimize seed aging effects. Sensory evaluation sessions were held daily with four panelists per session until each genotype had been evaluated twice for each year. Twelve genotypes were evaluated at each session including Ervilha and PI527538 as controls. Each sample was evaluated using 5-point attribute intensity scales (low → high intensity) for total, beany, vegetative, earthy, starchy, bitter, and sweet flavor intensities as 134 well as seed coat perception and cotyledon texture. The scale for seed coat perception ranged from imperceptible (1) to tough and lingering (5). For cotyledon texture, the scale ranged from mushy (1) to very gritty/firm (5) (Bassett et al., 2020b). This sensory evaluation protocol was approved by the Institutional Review Board of Michigan State University (IRB# x16-763e Category: Exempt 6). Panel Training Panelists were recruited from the USDA (East Lansing, MI) and Michigan State University dry bean breeding programs due to their familiarly with dry beans and their availability for long term sensory evaluation projects. Initially, seven panelists were trained using a diverse set of dry bean genotypes selected from the USDA and MSU dry bean programs with the intention of exposing panelists to a wide range of attribute intensities. This initial set included dark red kidney, Jacob’s cattle, white kidney, and yellow beans. A training set of genotypes exhibiting extreme attribute intensities identified in the ADP (Bassett et al., 2020b) was used to train eleven panelists for the second year. This training set was grown at the MSU Montcalm Research Center in Lakeview, MI alongside the RIL population. Panelists were trained over multiple sessions using a non-consensus approach to improve their familiarity with the selected scales and their sensory evaluation skills. Panelist performance was assessed via ANOVA with FGenotype (p-value < 0.05) indicating ability to discriminate and Frep (p-value > 0.05) indicating consistency (Meilgaard et al., 1999; Armelim et al., 2006). Sensory evaluation commenced after successful training of each panelist. Following screening of the parents and RILs from both years, panel performance was assessed as during training. 135 Sample Preparation for Sensory Evaluation Samples were prepared as described in Bassett et al. (2020b). Prior to each session, 4 seeds per panelist of each genotype scheduled for evaluation were soaked for 12 hours in distilled water prior to cooking. Large tea bags filled with the soaked samples were boiled in distilled water for the cooking time determined by the Mattson cooker method, timed so they all finished cooking together. The cooked samples were poured into preheated (105 °C) ceramic ramekins, covered with aluminum foil, and placed in a chafing dish to maintain temperature prior to evaluation. Samples were given a random letter code to mask their identity. Panelists were asked to refrain from wearing strong scents or eating during the hour before each session. Samples were served out of the ceramic ramekins with a plastic spoon onto paper plates. Lemon water was made available as a palette cleanser, and panelists were asked to drink water between samples. Statistics PROC MIXED in SAS version 9.4 of the SAS System for Windows (SAS Institute Inc. Cary, NC, USA) was used to conduct analyses of variance (ANOVAs) for each recorded trait. For seed weight, soak water uptake, cooking time, and total water uptake, the fixed effects were genotype, year, and genotype by year with replicate as a random effect. For L*, a*, and b* color values, the fixed effects were genotype and year with no random effects. For the sensory attribute intensity traits, the fixed effects were genotype, year, and genotype by year with replicate, panelist(year), and session(year) as random effects. Least squares estimates for sensory traits were calculated via the LSMeans statement in PROC MIXED for visualization of trait distributions. Mean separation of parents was determined using pdiff in PROC MIXED. To analyze both years combined while minimizing environmental effects, best linear unbiased predictors (BLUPs) were generated for each trait using the lme4 package (Bates et al., 136 2015) in R (R Core Team, 2017) with genotype, year, genotype by year, and rep nested in year as random effects. For sensory traits, panelist nested in year and session nested in year were also included as random effects. For analysis within individual years, BLUPs were calculated for sensory traits with genotype, rep, panelist, and session included as random effects. Broad sense heritability (H2) was calculated on a family mean basis for each trait using the equation var(G)/(var(G) + (var (G*Y)/no. Y) + (var(error)/no. Y * rep), where var is variance, G is genotype, and G*Y is genotype by year, and no. Y is number of years. Variance components were calculated using PROC VARCOMP in SAS version 9.4 with method = restricted maximum likelihood method (reml) (Holland et al., 2003). Principle component analysis among traits was conducted with BLUPs from both years combined using the Prcomp function in R. Genotyping DNA was extracted from young trifoliate leaf tissue from three plants each for the 240 RILs and the two parental lines (Ervilha and PI527538) using a Macherey-Nagel NucleoSpin Plant II kit. Three genotyping-by-sequencing (GBS) libraries were constructed at 96-plex as described by Elshire et al. (2011) with the parental lines prepared in quintuplicate. Fragment sizes were evaluated using the Agilent Bioanalyzer High Sensitivity DNA Kit (Bioanalyzer 2100, Agilent). Single-end sequencing (50 bp reads) of one 96-plex library per flowcell channel was performed on an Illumina HiSeq 4000. The raw sequence data were cleaned of adapters and trimmed for quality score ≥ 30 and minimum length ≥ 30 via Cutadapt (Martin, 2011) and evaluated via FastQC (Andrews, 2010). Cleaned reads were demultiplexed using the Next Generation Sequencing Eclipse Plugin (NGSEP) pipeline with NGSEP version 3.0.2 (Duitama et al., 2014; Perea et al., 2016), aligned to the Phaseolus vulgaris v2.1 genome (DOE-JGI and USDA-NIFA, http://phytozome.jgi.doe.gov/) using Bowtie 2 (Langmead and Salzberg, 2012), and then sorted 137 using Picard (http://broadinstitute.github.io/picard). Variant calling and annotation were performed via NGSEP. Raw SNPs were filtered to eliminate repetitive regions, markers with more than 50% missing data, and markers that were not polymorphic in the parents. Linkage and QTL Mapping Linkage mapping was performed using MapDisto version 2.1.7 (Heffelfinger et al., 2017). Genotyping error candidates meeting the 1e-4 threshold were replaced with missing data, and missing data was filled with flanking genotypes. Markers exhibiting segregation distortion (p- value < 1e-10) or causing excessive map length were excluded. A fixed order genetic map of 1567 cM was generated using the Kosambi function with 973 markers. QTL mapping was performed using QTL Cartographer version 2.5 (Wang et al., 2005). The composite interval mapping (CIM) procedure was performed with the parameters set to 10 cM window size and 1 cM walkspeed with forward and backward regression. BLUPs were used in QTL mapping for both years combined and for sensory traits for individual years, and means were used for analyses of all other traits for individual years. The LOD thresholds for each trait in each year and across years were determined using 1000 permutations in scanone from rQTL with the extended Haley-Knott method (p-value < 0.05) (Broman et al., 2003; Feenstra et al., 2006). The constructed linkage maps with QTL overlaid were visualized using Mapchart 2.32 (Voorrips, 2002). RESULTS Cooking Time Evaluation Genotype significantly affected soak water uptake, cooking time, and total water uptake (p-value < 0.05) (Table 3.1, S3.1). Year significantly affected soak water uptake and cooking time 138 (p-value < 0.05), and genotype by year significantly affected cooking time (p-value < 0.05). For the parents Ervilha and PI527538 respectively averaged across both years, the soak water uptakes were 109.3 and 98.8 percent; the cooking times were 21.0 and 29.7 min; and the total water uptakes were 138.2 and 146.3 percent. Soak water uptake, cooking time, and total water uptake for the RILs varied minimally across years and exhibited approximately normal distributions (Table S3.2; Figures 3.1, S3.1). Averaged across both years, soak water uptake ranged 69.2 – 117.4%; cooking time ranged 19.1 – 33.9 min; and total water uptake ranged 109.9 – 148.0% (Table 3.1). Broad-sense heritability varied greatly across traits, with cooking time (0.76) exhibiting high heritability and soak water uptake (0.34) and total water uptake (0.23) exhibiting low heritability. Sensory Evaluation Genotype significantly affected all sensory attributes (p-value < 0.05) (Table 3.1). Year did not significantly affect any sensory attributes, and genotype by year only significantly affected cotyledon texture (p-value < 0.05). Rep effects were insignificant for all sensory attributes, which indicates panelists were consistent across reps, although significant panelist and session effects were observed (Table S3.3). For the parents Ervilha and PI527538 respectively with least squares estimates averaged across both years, the total flavor intensities were 3.1 and 3.2; beany intensities were 2.2 and 3.3; vegetative intensities were 2.7 and 2.5; earthy intensities were 2.0 and 2.2; starchy intensities were 3.6 and 3.0; sweet intensities were 2.3 and 1.8; bitter intensities were 1.4 and 1.9; seed coat perceptions were 2.8 and 3.4; and cotyledon textures were 2.4 and 2.0 (Table 3.1). Least squares estimates for all sensory attribute intensities varied minimally across years and exhibited approximately normal distributions (Table S3.2, Figure 3.3). Across both years, least 139 squares estimates ranged 2.2 – 4.1 for total flavor intensity, 1.5 – 3.9 for beany intensity, 1.7 – 3.4 for vegetative intensity, 1.5 – 3.1 for earthy intensity, 2.5 – 3.9 for starchy intensity, 1.3 – 3.2 for sweet intensity, 1.1 – 2.3 for bitter intensity, 2.4 – 3.9 for seed coat perception, and 1.4 – 3.0 for cotyledon texture (Table 3.1). While panelists were able to differentiate among genotypes using 5-point scales, sensory attribute ranges did not exceed 2.4, suggesting panelists did not make full use of the scales. This could reflect the limited differences in sensory attribute intensities observed between the parents. Broad-sense heritability for sensory attribute intensities were low, ranging from 0.05 to 0.33 (Table 3.1). Beany intensity and total flavor intensity exhibited the highest broad-sense heritability (0.33 and 0.27), while vegetative intensity, earthy intensity, and cotyledon texture exhibited the lowest (0.05, 0.06, and 0.06). Color and Seed Coat Postharvest Darkening Genotype significantly affected L*, a*, b*, and seed coat postharvest darkening (p-value < 0.05) (Table 3.1). Year significantly affected L*, a*, and b* (p-value < 0.05). For the parents Ervilha and PI527538 respectively averaged across both years, L* values were 64.8 and 54.1; a* values were -0.7 and 3.5; b* values were 22.3 and 14.6; and seed coat postharvest darkening values were 0 (non-darkening) and 1 (darkening). The L*, a*, and b* for the RILs varied minimally across years and exhibited approximately normal distributions (Table S3.2, Figure 3.4). Averaged across both years, L* ranged from 40.3 – 67.3; a* ranged from -3.2 – 5.9; and b* ranged from 8.5 – 34.4 (Table 3.1). Seed coat postharvest darkening was only determined for seeds from one year (2017), and progeny exhibiting both non- darkening and darkening were observed. Broad-sense heritability was high for L* (0.86), a* (0.86), b* (0.78), and seed coat postharvest darkening (1.00). 140 Seed Yield and Seed Weight Genotype, year, and genotype by year significantly affected seed weight and seed yield (p- value < 0.05) (Table S3.1). For the parents Ervilha and PI527538 respectively averaged across both years, the seed weights were 52.8 and 48.0 g per 100 seeds. Seed yield data for Ervilha is not available for 2016 (Table S3.2), and fewer seeds were planted per plot in 2016, making averages across years misleading. In 2017, the seed yields for Ervilha and PI527538 respectively were 1731.4 and 2384.4 kg/ha. The seed weight for the RILs varied minimally across years and exhibited approximately normal distributions (Table S3.2, Figure S3.1). Seed yield for the RILs varied substantially across years due to reduced seeds planted per plot in 2016 but exhibited approximately normal distributions. Averaged across both years, seed weight ranged 39.1 – 68.4 g per 100 seeds and seed yield ranged 751.0 – 3283.9 kg per ha (Table S3.2). Broad-sense heritability for seed weight (0.89) was high and for seed yield was moderate (0.57) (Table S3.1). PCA For the PCA, the first two principal components (PCs) explained approximately 52% of the variance (Figure 3.5). The first PC separates the genotypes approximately by beany, earthy, and bitter intensities as well as L*, a*, b*, and seed coat postharvest non-darkening and represents over a third of the variation (38.5%). The second PC separates the genotypes approximately by cooking time; total flavor, vegetative, starchy, and sweet intensities; and cotyledon texture and seed coat perception. The second PC represents over an eighth of the variance (13.0%). The remaining PCs accounted for 11.1, 7.4, 6.0, 5.4, 4.2, 3.2, 2.9, 2.5, 2.2, 1.6, 1.1, 0.9% of the variance respectively (data not shown). 141 The PCA biplot highlights distinct groupings of traits that tend to be observed together. Loadings that group together highlight strong positive relationships within each group, and groups of loadings opposite of each other highlight strong negative relationships between groups. Loadings for starchy intensity, sweet intensity, and cotyledon texture are positioned close to each other and opposite cooking time and seed coat perception. Loadings for beany intensity and bitter intensity also group together and are somewhat opposite starchy intensity, sweet intensity, and cotyledon texture. The loadings for total flavor intensity earthy intensity, a*, and seed coat postharvest non-darkening group together, opposite of loadings for L* and b*. The loading for vegetative intensity does not appear to group with or opposite of other loadings, but lies in between loadings for total flavor intensity and sweet intensity. The genotypes are fairly evenly spread across the biplot, with Ervilha and PI527538 positioned opposite each other. QTL Mapping A linkage map was developed with 973 SNPs spread across eleven chromosomes for a total map length of 1,567 cM with a marker density of 1.61 cM per SNP (Table 3.2). Significant QTL were identified for soak water uptake, cooking time, total flavor intensity, beany intensity, vegetative intensity, earthy intensity, starchy intensity, sweet intensity, bitter intensity, seed coat perception, cotyledon texture, L*, a*, b*, seed coat postharvest darkening, seed weight, total water uptake, and seed yield (Tables 3.3-5, S3.4; Figure 3.5, S3.2-13). For soak water uptake, four QTL were identified on Pv03, Pv06, Pv10, and Pv11 (Table 3.3, Figures 3.6, S3.5, S3.8, S3.12-13). WU.3.1 and WU.10.1 were identified in both years combined; WU.6.1 was only identified in 2016; and WU.11.1 was only identified in 2017 (Table 3.3). The total proportion of variance explained by the two QTL identified in both years combined 142 was 11.0%. For WU3.1 and WU.10.1, alleles contributed by Ervilha conferred positive effects (Tables 3.3). For cooking time, two QTL were identified on Pv03 and Pv11 (Table 3.3, Figures 3.6, S3.5, S3.13). CT.3.1 and CT.11.1 were identified in both years combined (Table 3.3). The total proportion of variance explained by the two QTL identified in both years combined was 17.5%. For CT.3.1 and CT.11.1, alleles contributed by Ervilha conferred both negative and positive effects, respectively, despite Ervilha cooking significantly faster cooking than PI527538 (Tables 3.1, 3.3). For total flavor intensity, one QTL was identified on Pv03 (Table 3.4, Figures 3.6, S3.5). TFI.3.1 was only identified in 2016 (Table 3.4). The proportion of variance explained by TFI.3.1 was 5.7%, and Ervilha contributed an allele conferring a negative effect, reflecting its lower total flavor intensity compared to PI527538 (Tables 3.1, 3.4). For beany intensity, one QTL was identified on Pv03 (Table 3.4, Figures 3.6, S3.5). BFI.3.1 was identified in both years combined (Table 3.4). The proportion of the variance explained by BFI.3.1 in both years combined was 6.8%, and Ervilha contributed an allele conferring a negative effect, reflecting its lower beany intensity compared to PI527538 (Tables 3.1, 3.4). For vegetative intensity, three QTL were identified on Pv02, and Pv07 (Table 3.4, Figures 3.6, S3.4, S3.9). VFI.7.1 was identified in both years combined; VF.7.2 was only identified in 2016, and VF.2.1 was only identified in 2017 (Table 3.4). The proportion of variance explained by VFI.7.1 in both years combined was 5.6%. Across all vegetative intensity QTL, most alleles contributed by Ervilha conferred negative effects despite its higher vegetative intensity compared to PI527538. For VFI.7.1, the allele contributed by Ervilha conferred a positive effect. 143 For earthy intensity, three QTL were identified on Pv07 and PV10 (Table 3.4, Figures 3.6, S3.9, S3.12). EFI.10.1 was identified in both years combined; EFI.7.1 was only identified in 2016; and EFI.10.2 was only identified in 2017 (Table 3.4). The proportion of variance explained by EFI.10.1 in both years combined was 12.3%. Across all earthy intensity QTL, most alleles contributed by Ervilha conferred positive effects despite its lower earthy intensity compared to PI527538. For starchy intensity, two QTL were identified on Pv03 and Pv11 (Table 3.4, Figures 3.6, S3.5, S3.13). STI.3.1 was identified in both years combined, and STI.11.1 was only identified in 2016 (Table 3.4). The proportion of variance explained by STI.3.1 in both years combined was 6.9%. Across both starchy intensity QTL, alleles contributed by Ervilha conferred positive effects, reflecting its higher starchy intensity as compared to PI527538 (Tables 3.1, 3.4). For sweet intensity, two QTL were identified on Pv02 and Pv07 (Table 3.4, Figures 3.6, S3.4, S3.9). SWI.2.1 was identified in both years combined, and SWI.7.1 was only identified in 2016 (Table 3.4). The proportion of variance explained by SWI.2.1 in both years combined was 5.9%. Across both sweet intensity QTL, alleles contributed by Ervilha conferred positive effects, reflecting its higher sweet intensity as compared to PI527538 (Tables 3.1, 3.4). For bitter intensity, two QTL were identified on Pv01 and Pv03 (Table 3.4, Figures 3.6, S3.3, S3.5). BI.1.1 was identified in both years combined, and BI.3.1 was only identified in 2017 (Table 3.4). The proportion of variance explained by BI.1.1 in both years combined was 5.9%. Across both bitter intensity QTL, alleles contributed by Ervilha conferred negative effects, reflecting its lower bitter intensity as compared to PI527538 (Tables 3.1, 3.4). For seed coat perception, one QTL was identified on Pv03 (Table 3.4, Figures 3.6, S3.5). SPE.3.1 was identified in both years combined (Table 3.4). The proportion of variance explained 144 by SPE.3.1 in both years combined was 6.8%, and Ervilha contributed an allele conferring a negative effect, reflecting its lower seed coat perception compared to PI527538 (Tables 3.1, 3.4). For cotyledon texture, two QTL were identified on Pv05 and Pv07 (Table 3.4, Figures 3.6, S3.7, S3.9). CTX.7.1 was only identified in 2016, and CTX.5.1 was only identified in 2017 (Table 3.4). Across both bitter intensity QTL, alleles contributed by Ervilha conferred negative effects despite its higher cotyledon texture as compared to PI527538 (Tables 3.1, 3.4). For L*, two QTL were identified on Pv03 and Pv10 (Table 3.5, Figures 3.6, S3.5, S3.12). SL*.3.1 and SL*.10.1 were identified in both years combined (Table 3.5). The total proportion of variance explained by two QTL identified in both years combined was 12.1%. Across both QTL, Ervilha contributed alleles conferring positive effects, reflecting its higher L* as compared to PI527538 (Tables 3.1, 3.5). For a*, two QTL were identified on Pv03 and Pv10 (Table 3.5, Figures 3.6, S3.5, S3.12). Sa*.3.1 and Sa*.10.1 were identified in both years combined (Table 3.5). The total proportion of variance explained by the two QTL identified in both years combined was 15%. Across both QTL, Ervilha contributed alleles conferring negative effects, reflecting its lower a* as compared to PI527538 (Tables 3.1, 3.5). For b*, four QTL were identified on Pv01, Pv05, and Pv10 (Table 3.5, Figures 3.6, S3.3, S3.7, S3.12). Sb*.5.1, Sb*.5.2, and Sb*.10.1 were identified in both years combined, and Sb*.1.1 was only identified in 2017 (Table 3.5). The total proportion of variance explained by the three QTL identified in both years combined was 21%. Across the b* QTL, Ervilha contributed alleles conferring mostly positive effects, reflecting its higher b* as compared to PI527538 (Tables 3.1, 3.5). 145 For seed coat postharvest darkening, one QTL was identified on Pv10 (Table 3.5, Figures 3.6, S3.12). Seed coat postharvest darkening was only evaluated for 2017 seeds. The proportion of variance explained by ND.10.1 was 10.4%, and Ervilha contributed an allele conferring a negative effect, reflecting its lack of darkening over time (Tables 3.1, 3.5). While seed weight, total water uptake, and seed yield were not central to this study, several QTL were identified for these traits as well. Additional information is available in the supplemental material (Figures S3.2, Table S3.4). Several QTL co-localized on Pv03, Pv07, and Pv10. On Pv03, QTL for soak water uptake (WU.3.1), beany intensity (BFI.3.1), bitter intensity (BI.3.1), seed coat perception (SPE.3.1), L* (SL*.3.1), and a* (Sa*.3.1) co-localized. Alleles from Ervilha conferred positive effects for WU.3.1 and SL*.3.1 and negative effects for BFI.3.1, SPE.3.1, and Sa*.3.1. QTL for cooking time (CT.3.1), total flavor intensity (TFI.3.1), and starchy intensity (STI.3.1) also co-localized on Pv03. Alleles from Ervilha conferred a positive effect for STI.3.1 and negative effects for CT.3.1 and TFI.3.1. On Pv07, QTL for vegetative intensity (VFI.7.1) and sweet intensity (SWI.7.1) co- localized. Alleles from Ervilha conferred positive effects for both VFI.7.1 and SWI.7.1. QTL for vegetative intensity (VFI.7.2), earthy intensity (EFI.7.1), and cotyledon texture (CTX.7.1) also co- localized on Pv07. Alleles from Ervilha conferred a positive effect for EFI.7.1 and negative effects for VFI.7.2 and CTX.7.1. On Pv10, QTL for L*, a*, b*, and seed coat postharvest darkening (ND.10.1) co-localized. Alleles from Ervilha conferred positive effects for SL*.10.1 and Sb*.10.1 and negative effects for Sa*.10.1 and ND.10.1. 146 DISCUSSION The broad-sense heritability for cooking time was moderately high in this study, as was the case for previous reports looking at both broad-sense and narrow-sense heritability (Elia et al., 1997; Jacinto-Hernandez et al., 2003; Cichy et al., 2019; Bassett et al., 2020b). This supports the idea that marker-assisted selection for fast cooking time may be feasible with few molecular markers. Using marker-assisted selection as opposed to phenotyping could save breeding programs time and prevent the need to purchase specialized machinery specific for the evaluation of cooking time. It could also allow for early generation screening that would otherwise not be feasible due to limited seed and the large number of lines to be evaluated for cooking time. Differences in sensory attribute intensities among genotypes were successfully detected, allowing the relationship among attributes in this population to be determined and for significant QTL to be identified for the evaluated sensory attributes. While significant panelist and session effects were identified (Table S3.2), QDA does not rely on consensus among panelists, and these effects can be accounted for by using least squares estimates and BLUPs where appropriate. Although broad-sense heritability for sensory attributes tended to be low to very low, it is clear that genotype is important for flavor and texture. In the context of a breeding program, heritability can be improved by screening fewer lines with greater replication to better account for panelist and session effects while managing limited seed and personnel resources. As has been previously noted, panelists tend not to use the full range of the rating scales, which prevents detection of small differences between samples (Bassett et al., 2020b). In the case of this population, it is unlikely that this RIL population exhibited a full range of sensory attribute intensities, especially for traits with limited differences in the parents, so incomplete use of the scales likely reflects a lack of extreme differences among genotypes. However, increasing the size of the scales or using line 147 scales that allow for continuous ratings may better reflect the diversity of attribute intensities exhibited in a population in future studies, which might return higher heritability for sensory traits. Year and genotype by year effects were not significant for sensory traits, apart from cotyledon texture, which had a significant genotype by year effect. This is encouraging because location of production and crop management practices have previously been identified as factors affecting sensory quality (Mkanda et al., 2007; Ferreira et al., 2012). This indicates that flavor and texture traits do not change across years in the same production environment, which is useful for meeting expectations of consistency for consumers and for product developers, who need consistent ingredients over time for their products to be successful. There did not appear to be distinct groupings of genotypes based on cooking time and attribute intensity in the PCA biplot, indicating that there was a general mixing of these traits in the progeny (Figure 3.4). This suggests that extensive efforts at breaking linkages among traits are not needed to combine desired traits and achieve a target cooking time and sensory profile. Developing new yellow bean varieties with both fast cooking time and desirable flavor and texture would address two major factors influencing consumer purchasing decisions regarding dry beans and provide novelty for the many consumers unfamiliar with the yellow seed type (Leterme and Carmenza Muñoz, 2002; Eihusen and Albrecht, 2007; Winham et al., 2019). Many QTL were identified in this study, with those for cooking time and sensory attribute intensities of particular interest. Both cooking time QTL (CT.3.1 and CT.11.1) were located in close proximity to significant SNPs previously identified via genome-wide association in the ADP (Bassett et al., 2020b). While the LODs and R2 values were not particularly high for CT.3.1 and CT.11.1, they have potential for use in marker-assisted selection due to their consistently detectable effects in this study and their support in a previous study. Other recent studies have 148 identified QTL or significant SNPs related to cooking time on Pv03 and Pv11 as well, but the physical positions were not proximal to CT.3.1 or CT.11.1 (Cichy et al., 2015b; Berry et al., 2020). The genetic control of sensory attributes is a new area of research in dry beans with limited study (Bassett et al., 2020b). For total flavor intensity, TFI.3.1 was located in relatively close proximity to SNP S03_51252684, which was identified in association with total flavor intensity for the ADP grown in Hawassa, Ethiopia. Otherwise, the QTL identified for sensory attributes in this study were novel. While several QTL including BFI.3.1, VFI.7.1, EFI.10.1, STI.3.1, SWI.2.1, BI.1.1, and SPE.3.1 were consistent across environments, further validation would be beneficial before use in marker-assisted selection. For certain traits, including vegetative intensity, earthy intensity, and cotyledon texture, the alleles contributed by Ervilha conferred effects that would seem more likely to come from PI527538. These traits also had the lowest heritability in this study, indicating that evaluating them was particularly challenging. Many QTL were identified for soak water uptake and CIELAB values. Some soak water uptake and CIELAB QTL were proximal to QTL and genetic markers identified in previous studies (Cichy et al., 2014, 2015b; Erfatpour et al., 2018; Bassett et al., 2020b; Berry et al., 2020). WU.3.1 was near SNPs for water uptake identified by Cichy et al. (2015b) and a QTL identified by Berry et al. (2020) for water uptake. SL*.10.1 and Sa*.10.1 overlapped with the J-locus associated with postharvest non-darkening (Erfatpour et al., 2018). Sb*5.1 and Sb*5.2 were near QTL identified for anthocyanin content, L*, and b* of canned beans (Cichy et al., 2014). Most of the QTL identified for these traits were novel and may be useful for research central to these traits. Seed coat postharvest darkening was detected in PI527538 and half of the RILs. Seed coat postharvest darkening describes the tendency of some genotypes to darken in color over time due to the presence of proanthocyanidin precursors in the seed coat (Beninger et al., 2005; Chen et al., 149 2015). This phenomenon has been most studied in pinto and cranberry beans but can be observed in other market classes. Lighter seed coats are perceived by consumers as indications of freshness or quality, so seeds exhibiting postharvest darkening have reduced market value (Nasar-Abbas et al., 2009; Erfatpour and Pauls, 2020). The J locus was previously identified on Pv10, and genotypes that are homozygous recessive at J do not exhibit postharvest darkening (Bassett, 2007; Elsadr et al., 2011; Erfatpour et al., 2018). The QTL identified for the non-darkening trait in this study overlaps with a previously identified QTL for non-darkening located between 40,164,667 bp and 40,295,580 bp on Pv10 (Table S3.5) (Erfatpour et al., 2018). Flavan-3-ols, which include proanthocyandidins, have been previously associated with bitterness and astringency depending on their degree of polymerization (Robichaud and Noble, 1990; Peleg et al., 1999), so seed coat postharvest darkening may alter flavor over time. The relationship between seed coat postharvest darkening and flavor after beans have darkened was not examined in this study, but it remains practical to select against darkening when developing new varieties to ensure greater visual appeal to consumers, which would bypass flavor changes caused by darkening altogether. A SNP-based marker has been developed to allow marker-assisted selection for this trait (Erfatpour and Pauls, 2020). As there is still much to be understood regarding flavor and texture in dry beans, other methods for assessing these sensory traits like GC-MS and texture measurements should be explored. Volatile concentrations and texture measurements have been used successfully as proxies for flavor and texture in studies looking at genetic control of sensory traits in other crops, and these measurements can be cheaper and easier to obtain than those generated by a descriptive panel (Zhang et al., 2015; Amyotte et al., 2017; Bauchet et al., 2017; Zhao et al., 2019). Apart from beany intensity (Vara-Ubol et al., 2004; Bott and Chambers, 2006), however, the contribution 150 of volatiles to perceived flavors in dry beans is not well understood, and texture measurements have not been well explored outside of their use in the evaluation of firmness in canned samples (Kelly and Cichy, 2012). In addition, research assessing consumer preference for flavor and texture in dry beans is needed to define breeding targets for sensory attributes. Understanding which traits are most important for consumer preference and what the expectations are for different seed types will help breeders address flavor and texture with a focused, efficient approach. Dry beans in the U.S. are sold as market classes rather than variety preserved. Variation exists within market classes for consumer-valued traits like cooking time, flavor, and texture so consumers are not able to make informed purchasing decisions taking these traits into account (Cichy et al., 2015b; Bassett et al., 2020b; Berry et al., 2020). In addition, the canning industry cannot receive the benefits of reduced energy costs and higher efficiency associated with fast- cooking genotypes if slow-cooking genotypes are present in the same cans (Bassett et al., 2020a). Because yellow beans are largely unfamiliar to U.S. consumers, there is an opportunity to develop new yellow bean varieties that prioritize these traits so that the yellow color can serve as a marker for convenience and culinary quality to consumers and the canning industry can produce quality canned products with yellow beans while benefitting from shorter processing times. Consumers are already seeking out unique flavors, textures, seed patterns, and colors from heirloom beans (Bullard, 2016), but heirlooms are not suited to modern farming practices, which makes them more expensive and less widely available than more familiar market classes. Yellow beans, the Manteca market class in particular, could serve this consumer interest while addressing grower needs. 151 CONCLUSION This work adds to the currently limited pool of resources available for dry bean breeders to target fast cooking time, flavor, and texture in their breeding programs. The QTL identified in this work, in particular CT.3.1 and CT.11.1, can be used to develop molecular markers for the incorporation of fast cooking time into new bean varieties to benefit both consumers and the canning industry. For sensory attributes, several QTL including BFI.3.1, VFI.7.1, EFI.10.1, STI.3.1, SWI.2.1, BI.1.1, and SPE.3.1 were consistent across years and show potential for use in marker-assisted selection following identification of breeding targets for sensory attributes informed by consumer preference. Consumers are seeking bean products with improved culinary characteristics and unique appearance. Yellow beans like those used in this study are unfamiliar to U.S. consumers, but they tend to be fast cooking with desirable sensory attributes. With the recent increased interest in plant-based proteins, now is an opportune time to address consumer preference in dry beans to remain competitive with other pulses, and yellow beans might be an ideal vehicle to a fast-cooking, flavorful, and flourishing future of dry beans. ACKNOWLEDGMENTS This material is based upon work that is supported by the National Institute of Food and Agriculture, U.S. Department of Agriculture, under award number 2017-67013-26212. Any opinions, findings, conclusions, or recommendations expressed in this publication are those of the author(s) and do not necessarily reflect the view of the U.S. Department of Agriculture. The authors would like to thank members of the USDA-ARS and MSU dry bean breeding programs for their contributions as panelists. Additional thanks go to Hannah Cooperider, Hannah Peplinski, 152 Anna Peter, Gasana Elyvine, and Queen Iribagiza for their assistance with cooking samples and managing seeds. 153 APPENDICES 154 APPENDIX A: CHAPTER 3 TABLES AND FIGURES Table 3.1 Parental phenotypes, meansa, ranges, and broad-sense heritability (H2) for the RILs for both years combined with ANOVA p-valuesb for genotype, year, and genotype by year indicated. Trait PI527538 Mean Genotype x Year Genotype Year Ervilha Range H2 Soak Water Uptake (%) 109.3a ± 3.5 98.8a ± 1.2 101.64 ± 0.3 69.2 - 117.4 0.25 < 0.0001 < 0.0001 NS Cooking Time (min) 21.0b ± 1.5 29.7a ± 2.4 25.25 ± 0.2 19.1 - 33.9 0.68 < 0.0001 < 0.0001 0.0054 Total Flavor Intensity 3.1b ± 0.1 3.2a ± 0.1 Beany Intensity 2.2b ± 0.2 Vegetative Intensity 2.7a ± 0.1 Earthy Intensity 2.0b ± 0.0 Starchy Intensity 3.6a ± 0.0 Sweet Intensity 2.3a ± 0.1 Bitter Intensity 1.4b ± 0.0 3.3a ± 0.1 2.5b ± 0.1 2.2a ± 0.0 3.0b ± 0.1 1.8b ± 0.1 1.9a ± 0.1 Seed Coat Perception 2.8b ± 0.1 3.4a ± 0.0 Cotyledon Texture 2.4a ± 0.1 3.28 ± 0.0 2.2 - 4.1 2.87 ± 0.0 1.5 - 3.9 2.59 ± 0.0 1.7 - 3.4 2.23 ± 0.0 1.5 - 3.1 3.17 ± 0.0 2.5 - 3.9 2.05 ± 0.0 1.3 - 3.2 1.7 ± 0.0 1.1 - 2.3 3.05 ± 0.0 2.4 - 3.9 0.27 0.33 0.05 0.06 0.13 0.19 0.14 0.21 < 0.0001 NS < 0.0001 NS 0.0020 0.0010 NS NS < 0.0001 NS < 0.0001 NS < 0.0001 NS < 0.0001 NS NS NS NS NS NS NS NS NS 2.0b ± 0.1 2.29 ± 0.0 1.4 - 3.0 0.06 < 0.0001 NS < 0.0001 64.8a ± 0.2 54.1b ± 1.8 58.8 ± 0.3 40.3 - 67.3 0.86 < 0.0001 < 0.0001 . -0.7b ± 0.6 3.5a ± 0.2 1.4 ± 0.1 -3.2 - 5.9 0.86 < 0.0001 < 0.0001 . L* a* b* 22.3a ± 0.9 14.6b ± 2.3 20.2 ± 0.3 8.5 - 34.4 0.78 Seed Coat Postharvest Darkening (0 = non-darkening; 1 = darkening) 0.5 ± 0.0 0 - 1 0b ± 0 1a ± 0 1.00 < 0.0001 < 0.0001 . < 0.0001 . . a Mean separation is indicated by letter superscript. Least squares estimates are presented for sensory attribute intensities instead of means 155 Table 3.1 (cont’d) b NS indicates non-significant p-values at α = 0.05 Table 3.2 Linkage map information for the 240 RILs. Chromosome Number of Markers Size (cM) 1 2 3 4 5 6 7 8 9 10 11 Total: 41 104 115 141 64 46 70 105 51 182 54 973 112.88 138.56 112.46 158.81 115.83 103.26 163.81 159.35 104.61 188.83 208.80 1567.20 Marker Density (cM) 2.75 1.33 0.98 1.13 1.81 2.24 2.34 1.52 2.05 1.04 3.87 1.61 156 Table 3.3 Quantitative trait loci identified in the RIL population (N = 240) grown in Entrican, MI in 2016 and 2017 for soak water uptake and cooking time. Linkage group (LG), peak position (Pos), yeara, logarithm of odds (LOD), R2, QTL effectb (a), flanking markersc, QTL ranged, and significancee of the QTL are indicated. Trait QTL Name LG Pos (cM) Year LOD R2 (%) a Flanking Markers QTL Range (cM) Sig Soak Water Uptake WU.3.1 Pv03 48.6 C 2.92 4.8 + 33177106 - 33854971 48.36 - 49.56 WU.6.1 Pv06 43.8 2016 3.26 6.9 - 19164538 - 19553914 40.77 - 44.77 WU.10.1 Pv10 93.4 C 3.26 6.2 + 30125056 - 31195987 92.85 - 95.67 Pv10 95.4 2016 3.47 6.7 + 31195987 - 31195987 93.37 - 95.67 WU.11.1 Pv11 151.7 2017 2.82 4.8 Cooking Time 46682849 - 47215098 150.81 - 152.65 ** ** ** ** * * ** ** - - - CT.3.1 Pv03 111.1 C 4.07 7.7 Pv03 102.1 2016 3.18 5.6 51423691 - 51934861 101.05 - 112.07 ** 51291118 - 51934861 95.46 - 112.07 CT.11.1 Pv11 37.8 C 3.46 9.8 + 11733856 - 16663857 36.84 - 40.15 Pv11 38.8 2016 3.64 9.8 + 11733856 - 16663857 35.84 - 41.15 The largest LOD and R2 within the QTL are reported a “C” indicates both years combined b + and – indicate positive and negative effects on the mean as conferred by alleles from Ervilha in the QTL region. c Flanking markers indicate the physical positions of the nearest markers upstream and downstream d Region where LOD scores are significant at the indicated significance level e Significance at α = 0.1 and α = 0.05 are indicated by * and **, respectively, based on 1000 permutations 157 Table 3.4 Quantitative trait loci identified in the RIL population (N = 240) grown in Entrican, MI in 2016 and 2017 for sensory attributes. Linkage group (LG), peak position (Pos), yeara, logarithm of odds (LOD), R2, QTL effectb (a), flanking markersc, QTL ranged, and significancee of the QTL are indicated. Trait QTL Name LG Pos (cM) Year LOD Total Flavor Intensity R2 (%) TFI.3.1 Pv03 102.9 2016 3 5.7 Beany Intensity BFI.3.1 Pv03 27.3 C 4.3 6.8 Pv03 41.8 2017 5.52 8.8 Vegetative Intensity VFI.2.1 Pv02 86.0 2017 3.91 6.7 - - - - a Flanking Markers QTL Range (cM) Sig 51433552 - 51495521 102.06 - 103.80 ** 5101738 - 11872699 26.67 - 29.81 12630923 - 33254421 30.30 – 48.56 41385822 - 41401290 85.88 - 86.48 VFI.7.1 Pv07 47.2 C 3.2 5.6 + 7714725 - 8753083 46.96 - 48.21 VFI.7.2 Pv07 132.4 2016 3.39 8.1 Earthy Intensity - 30492292 - 38224460 130.75 - 132.84 EFI.7.1 Pv07 136.8 2016 3.12 5.9 + 39018976 - 39082640 136.29 - 137.80 EFI.10.1 Pv10 72.3 C 4.41 12.3 + 9922603 - 28143113 68.12 - 77.33 Pv10 71.3 2016 2.95 7.9 + 28038978 - 28143113 70.33 - 72.33 Pv10 71.3 2017 3.7 9.9 + 9922603 - 28143113 69.12 - 74.33 EFI.10.2 Pv10 180.6 2017 4.68 7.7 ** ** ** ** ** * ** * ** ** ** ** ** ** * * ** ** ** ** ** ** * * Starchy Intensity STI.3.1 Pv03 89.0 C 4.0 Pv03 88.1 2016 3.7 Pv03 89.3 2017 3.3 STI.11.1 Pv11 100.5 2016 3.5 Sweet Intensity SWI.2.1 Pv02 55.4 C 3.2 Pv02 55.4 2016 2.8 SWI.7.1 Pv07 37.4 2016 3.9 Bitter Intensity BI.1.1 BI.3.1 Pv01 44.8 C 3.3 Pv03 37.8 2017 3.9 Seed Coat Perception SPE.3.1 Pv03 46.1 C 3.6 Pv03 46.1 2016 4.4 Pv03 39.2 2017 4.5 Cotyledon Texture CTX.5.1 Pv05 79.4 2017 3.1 CTX.7.1 Pv07 140.4 2016 3.0 - 42840998 - 43205231 180.04 - 186.99 + 51161323 - 51220644 88.07 - 89.30 + 51140633 - 51171573 87.24 - 89.00 + 51171573 - 51224571 89.00 - 90.12 + 32258466 - 40017638 99.48 - 103.25 + 31959417 - 33225006 54.53 - 55.72 + 31959417 - 33225006 54.53 - 55.72 + 7175486 - 7438454 37.03 - 37.9 - - - - - - - 14308851 - 32577126 44.51 - 45.76 18090483 - 21241672 36.84 - 38.43 32484452 - 32983892 46.02 - 48.07 29498694 - 32983892 45.33 - 47.12 18090483 - 33254421 35.84 - 48.56 25938962 - 39102354 77.60 - 81.43 39116608 - 39166109 139.37 - 140.94 6.9 6.6 5.4 9.1 5.9 4.9 7.2 5.9 7.9 6.8 8.1 8.7 6.5 5.9 158 Table 3.4 (cont’d) The largest LOD and R2 within the QTL are reported a “C” indicates both years combined b + and – indicate positive and negative effects on the mean as conferred by alleles from Ervilha in the QTL region. c Flanking markers indicate the physical positions of the nearest markers upstream and downstream d Region where LOD scores are significant at the indicated significance level e Significance at α = 0.1 and α = 0.05 are indicated by * and **, respectively, based on 1000 permutations 159 Table 3.5 Quantitative trait loci identified in the RIL population (N = 240) grown in Entrican, MI in 2017 for color and seed coat postharvest darkening. Linkage group (LG), peak position (Pos), year, logarithm of odds (LOD), R2, QTL effecta (a), flanking markersb, QTL rangec, and significanced of the QTL are indicated. Trait QTL Name LG Pos (cM) Year LOD R2 (%) a Flanking Markers L* QTL Range (cM) C SL*.3.1 Pv03 28.2 3.3 5.5 + 6264322 - 7619818 27.33 - 28.57 Pv03 28.2 2016 3.3 5.4 + 6264322 - 7619818 27.33 - 28.57 Pv03 37.8 2017 3.0 5.8 + 1872699 - 18827421 29.81 - 37.84 SL*.10.1 Pv10 114.7 C 4.2 6.6 + 40970486 - 41609109 112.42 - 116.97 ** Pv10 114.7 2016 4.2 6.6 + 40970486 - 41609109 112.42 - 116.97 ** Sig ** ** * ** ** ** ** ** ** * ** ** ** a* b* C 3.4 Sa*.3.1 Pv03 29.6 Pv03 29.6 2016 3.4 Pv03 29.6 2017 4.2 Sa*.10.1 Pv10 114.47 C 6.2 Pv10 114.47 2016 6.2 Pv10 114.47 2017 4.6 Sb*.1.1 Pv01 0.01 2017 3 Sb*.5.1 Pv05 1.21 C Sb*.5.2 Pv05 12.96 C 4.1 3.7 5.3 5.2 6.6 9.7 9.6 7.1 4.8 6.7 - - - - - - - - 7619818 - 11872699 29.57 - 29.81 7619818 - 11872699 28.57 - 29.81 6514180 - 11872699 28.17 - 29.81 35799233 - 41965681 105.38 - 123.1 35799233 - 41965681 105.38 - 123.1 40970486 - 41965681 111.42 - 123.1 703233 - 787996 0.01 - 1.01 1077531 - 1091909 0.91 - 1.21 6.3 + 1960475 - 1963046 12.96 - 13.96 Pv05 12.96 2016 3.7 6.3 + 1960475 - 1963046 12.96 - 13.96 Sb*.10.1 Pv10 114.69 C 4.6 8 + 35799233 - 41088499 105.38 - 115.97 ** Pv10 109.38 2016 4.2 8.9 + 35799233 - 41070183 104.38 - 115.28 ** Pv10 114.69 2017 6.3 10.2 + 35799233 - 41922512 104.38 - 121.95 ** Seed Coat Postharvest Darkening ND.10.1 Pv10 115.3 2017 5.9 10.4 - 35799233 - 41931454 108.38 - 122.23 ** The largest LOD and R2 within the QTL are reported a “C” indicates both years combined b + and – indicate positive and negative effects on the mean as conferred by alleles from Ervilha in the QTL region. c Flanking markers indicate the physical positions of the nearest markers upstream and downstream d Region where LOD scores are significant at the indicated significance level e Significance at α = 0.1 and α = 0.05 are indicated by * and **, respectively, based on 1000 permutations 160 Figure 3.1 Images of Ervilha and PI527538 raw seeds. Figure 3.2 Density plots of soak water uptake and cooking time for the RILs from 2016, 2017, and both years combined (C). Means for Ervilha and PI527538 from both years combined are indicated in yellow and brown, respectively. 161 Figure 3.3 Density plots of least squares estimates of sensory attribute intensities for the RILs from 2016, 2017, and both years combined (C). Attribute intensities for Ervilha and PI527538 from both years combined are indicated in yellow and brown, respectively. 162 Figure 3.4 Density plots of CIELAB values for the RILs from 2016, 2017, and both years combined (C). Attribute intensities for Ervilha and PI527538 from both years combined are indicated in yellow and brown, respectively. 163 Figure 3.5 Principal component analysis biplot with loadings for cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), and cotyledon texture (CTX). Ervilha and PI527538 are indicated in yellow and brown, respectively. 164 Figure 3.6 QTL map for soak water uptake, cooking time, total flavor intensity, beany intensity, vegetative intensity, earthy intensity, starchy intensity, sweet intensity, bitter intensity, seed coat perception, cotyledon texture, L*, a*, b*, and seed coat postharvest non-darkening in the RIL population. Size is in cM. Year is indicated for each QTL, where “C” is both years combined. 165 APPENDIX B: CHAPTER 3 SUPPLEMENTAL TABLES AND FIGURES Table S3.1 Parental phenotypes, meansa, ranges, and broad-sense heritability (H2) for the RILs for both years combined with ANOVA p-values for genotype, year, and genotype by year indicated. Trait Ervilha PI527538 Mean Range H2 Genotype Year Genotype x Year Seed Weight(g per 100 seeds) 52.8a ± 0.1 48.0b ± 1.2 51.35 ± 0.3 39.1 - 68.4 0.84 < 0.0001 < 0.0001 < 0.0001 Total Water Uptake (%) 138.2a ± 7.7 146.3a ± 9.0 131.85 ± 0.3 109.9 – 148.0 0.16 0.0012 NS Seed Yield (kg/ha) NS 1891.6a ± 403.9 1731.4a ± 639.2 2072.9 ± 23.8 751.0 – 3283.9 0.57 < 0.0001 < 0.0001 < 0.0001 a Mean separation is indicated by letter superscript. 166 Table S3.2 Parental phenotypes and means and ranges for the RILs for 2016 and 2017. Trait Year Ervilha Soak Water Uptake (%) 2016 104.3 ± 1.9 2017 114.2 ± 4.8 Cooking Time (min) 2016 23.1 ± 0.6 2017 18.8 ± 0.3 Total Flavor Intensity 2016 3.2 ± 0.1 2017 2.9 ± 0.0 Beany Flavor Intensity 2016 2.4 ± 0.2 2017 1.9 ± 0.1 Vegetative Flavor Intensity 2016 2.9 ± 0.1 2017 2.5 ± 0.2 Earthy Flavor Intensity 2016 2.0 ± 0.0 2017 2.0 ± 0.0 Starchy Flavor Intensity 2016 3.6 ± 0.1 2017 3.6 ± 0.0 Sweet Flavor Intensity 2016 2.2 ± 0.2 2017 2.5 ± 0.0 Bitter Flavor Intensity 2016 1.4 ± 0.1 2017 1.3 ± 0.0 Seed Coat Perception 2016 2.9 ± 0.0 2017 2.6 ± 0.0 Cotyledon Texture 2016 2.3 ± 0.0 L* 2017 2.5 ± 0.2 2016 65.2 2017 64.5 PI527538 Mean Range 97.0 ± 0.1 99.0 ± 0.5 38.6 - 114.9 100.5 ± 4.8 103.9 ± 0.3 90.2 - 135.7 33.2 ± 1.4 27.2 ± 0.3 19.7 - 40.3 26.3 ± 1.1 23.6 ± 0.2 17.8 - 33.0 3.4 ± 0.0 2.1 - 4.2 3.2 ± 0.0 2.1 - 4.2 3.0 ± 0.0 1.8 - 4.1 3.0 ± 0.0 1.5 - 3.9 2.7 ± 0.0 1.7 - 3.9 2.5 ± 0.0 1.5 - 3.8 2.3 ± 0.0 1.2 - 4.0 2.2 ± 0.0 1.3 - 3.0 3.2 ± 0.0 2.5 - 4.1 3.1 ± 0.0 2.2 - 4.0 2.0 ± 0.0 1.1 - 3.1 2.1 ± 0.0 1.2 - 3.1 1.7 ± 0.0 0.9 - 2.5 1.7 ± 0.0 0.9 - 2.8 3.1 ± 0.0 2.3 - 3.9 3.0 ± 0.0 2.0 4.1 2.3 ± 0.0 1.4 - 3.3 2.3 ± 0.0 1.4 - 3.3 59.9 ± 0.3 49.3 - 68.0 58.0 ± 0.3 40.3 - 68.8 3.4 ± 0.1 3.1 ± 0.0 3.5 ± 0.1 3.1 ± 0.1 2.6 ± 0.0 2.4 ± 0.2 2.2 ± 0.1 2.2 ± 0.0 3.1 ± 0.0 2.9 ± 0.1 1.7 ± 0.1 1.9 ± 0.1 2.0 ± 0.1 1.8 ± 0.0 3.4 ± 0.1 3.4 ± 0.1 1.8 ± 0.1 2.2 ± 0.0 51.6 56.6 167 Table S3.2 (cont’d) a* b* 2016 -1.5 2017 0.1 2016 21.0 2017 23.5 3.7 3.2 11.4 17.8 Darkening (0 = Nondarkening; 1 = Darkening) 2016 . 1.4 ± 0.1 -3.2 - 5.9 1.1 ± 0.1 -2.4 - 4.7 19.9 ± 0.3 8.4 - 34.4 23.6 ± 0.2 . 0.5 - 0.0 12.8 - 34.6 . 0.0 - 1.0 . 1 2017 0 Seed Weight(g) 2016 53.0 ± 1.8 2017 52.7 ± 2.3 Total Water Uptake (%) 2016 127.3 ± 9.0 2017 149.0 ± 11.7 Seed Yield (kg/ha) 2016 . 46.3 ± 1.4 49.9 ± 0.3 39.5 - 64.7 49.7 ± 0.9 52.7 ± 0.3 39.1 - 72.1 133.6 ± 0.4 131.4 ± 0.5 89.8 - 160.6 159.0 ± 15.6 132.3 ± 0.4 102.0 ± 159.0 1399.2 ± 23.1 1448.3 ± 32.0 301.0 - 2876.5 2017 1731.5 ± 639.2 2384.4 ± 591.0 2623.2 ± 32.4 592.1 - 3912.0 Table S3.3 P-valuesa for the random effects from the sensory attribute intensity ANOVAs at the genotype level. Trait Rep Panelist(Year) Session(Year) Total Flavor Intensity Beany Intensity Vegetative Intensity Earthy Intensity Starchy Intensity Sweet Intensity Bitter Intensity Seed Coat Perception Cotyledon Texture NS NS NS NS NS NS NS NS NS <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.0145 <0.0001 NS 0.0136 <0.0001 0.005 <0.0001 <0.0001 a NS indicates non-significant p-values at α = 0.05 168 Table S3.4 Quantitative trait loci identified in the RIL population (N = 240) grown in Entrican, MI in 2017 for seed weight, total water uptake, and yield. Linkage group (LG), peak position (Pos), year, logarithm of odds (LOD), R2, QTL effecta (a), flanking markersb, QTL rangec, and significanced of the QTL are indicated. Trait QTL Name LG Pos (cM) Year LOD R2 (%) a Flanking Markers QTL Range (cM) Sig Seed Weight (g per 100 seeds) SW.6.1 Pv06 6.1 2017 2.89 4.8 + 5685389 - 6728096 5.1 - 7.1 SW.8.1 Pv08 5.9 C 4.10 6.7 + 5255874 - 7162048 5.73 - 14.04 Pv08 5.9 2016 3.40 5.7 + 5255874 - 7162048 4.73 - 14.04 Pv08 5.9 2017 4.36 7.1 + 5255874 - 7162048 5.73 - 14.14 SW.10.1 Pv10 11.8 2016 3.34 5.5 - 861914 - 881899 10.78 - 11.79 SW.10.2 Pv10 42.7 2016 3.43 6.5 + 2459869 - 2484767 41.73 - 43.73 SW.10.3 Pv10 114.7 C 5.54 9.2 Pv10 114.7 2016 3.42 5.7 Pv10 114.7 2017 4.48 7.5 - - - 40970486 - 41070183 112.42 - 115.28 40984982 - 41070183 114.47 - 115.28 40971012 - 41070183 113.47 - 115.28 SW.10.4 Pv10 174.6 2016 3.95 6.1 + 42732517 - 42758310 174.16 - 175.05 Total Water Uptake TWU.3.1 Pv03 28.6 C 3.79 5.8 + 6514180 - 11872699 28.17 - 29.57 Pv03 28.6 2016 3.56 5.5 + 6514180 - 11872699 28.17 - 29.57 TWU.8.1 Pv08 95.9 2017 4.66 8.0 - 41088373 - 44547580 94.79 - 96.85 TWU.9.1 Pv09 25.5 C 4.01 6.6 + 10763996 - 12942768 22.98 - 26.39 Pv09 25.5 2016 4.17 7.0 + 10449643 - 12942768 22.37 - 26.39 TWU.10.1 Pv10 114.7 C 3.09 4.8 + 40971012 - 41070183 114.47 - 115.28 Pv10 114.7 2016 3.10 4.8 + 40971012 - 41070183 114.47 - 115.28 TWU.10.2 Pv10 159.3 2017 3.91 6.7 + 42515259 - 42521192 158.68 - 160.27 TWU.10.3 Pv10 176.0 2017 4.30 7.2 + 42758310 - 42791961 175.05 - 177.01 TWU.10.4 Pv10 184.9 2017 5.11 8.7 + 42853098 - 42872742 183.21 - 185.99 Seed Yield (kg/ha) YLD.2.1 Pv02 12.5 C 3.35 5.8 + 3095899 - 5589384 12.37 - 13.45 Pv02 12.5 2016 3.21 6.1 + 3095899 - 5589384 12.37 - 13.45 YLD.3.1 Pv03 98.3 2016 4.23 7.2 YLD.7.1 Pv07 152.1 2017 3.79 6.3 YLD.10.1 Pv10 118.9 2016 4.6 8.2 - - - 51291118 - 51376970 95.46 - 99.42 39309171 - 39372454 151.07 - 153.07 41088499 - 41965681 116.97 - 123.1 YLD.11.1 Pv11 64.5 2017 3.26 5.9 + 21357068 - 21564226 64.37 - 64.49 YLD.11.2 Pv11 112.3 2017 4.11 9.4 - 38773018 - 41950354 107.25 - 116.31 * ** ** ** ** ** ** ** ** ** ** ** ** ** ** * * ** ** ** ** * ** ** ** ** ** The largest LOD and R2 within the QTL are reported a + and – indicate positive and negative effects on the mean as conferred by alleles from Ervilha in the QTL region. b Flanking markers indicate the physical positions of the nearest markers upstream and downstream 169 Table 3.4 (cont’d) c Region where LOD scores are significant at the indicated significance level d Significance at α = 0.1 and α = 0.05 are indicated by * and **, respectively, based on 1000 permutations 170 Figure S3.1 Density plots of seed weight, total water uptake, and yield for the RILs from 2016, 2017, and both years combined (C). Means for Ervilha and PI527538 from both years combined (2017 for seed yield) are indicated in yellow and brown, respectively. 171 Figure S3.2 QTL map for seed weight (SW), total water uptake (TWU), and seed yield (YLD) in the RIL population. Size is in cM. Year is indicated for each QTL, where “C” is both years combined. 172 Figure S3.3 Line graphs of LOD by Pv01 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. 173 Figure S3.4 Line graphs of LOD by Pv02 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. 174 Figure S3.5 Line graphs of LOD by Pv03 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. 175 Figure S3.6 Line graphs of LOD by Pv04 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. 176 Figure S3.7 Line graphs of LOD by Pv05 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. 177 Figure S3.8 Line graphs of LOD by Pv06 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. 178 Figure S3.9 Line graphs of LOD by Pv07 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. 179 Figure S3.10 Line graphs of LOD by Pv08 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. 180 Figure S3.11 Line graphs of LOD by Pv09 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. 181 Figure S3.12 Line graphs of LOD by Pv10 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. 182 Figure S3.13 Line graphs of LOD by Pv11 position for both years combined, 2016, and 2017 for soak water uptake (WU), cooking time (CT), total flavor intensity (TFI), beany intensity (BFI), vegetative intensity (VFI), earthy intensity (EFI), starchy intensity (STI), sweet intensity (SWI), bitter intensity (BI), seed coat perception (SPE), cotyledon texture (CTX), L*, a*, b*, seed coat postharvest non-darkening (ND), seed weight (SW), total water uptake (TWU), and seed yield (YLD). The gray dashed line approximates the LOD threshold (α = 0.05) for all traits. 183 REFERENCES 184 REFERENCES Akibode, S., and Maredia, M. (2011). Global and regional trends in production, trade and consumption of food legume crops. Dep. Agric. Food Resour. Econ. Michigan State Univ., 87. Available at: http://impact.cgiar.org/sites/default/files/images/Legumetrendsv2.pdf. Amyotte, B., Bowen, A. J., Banks, T., Rajcan, I., and Somers, D. J. (2017). Mapping the sensory perception of apple using descriptive sensory evaluation in a genome wide association study. PLoS One 12. doi:10.1371/journal.pone.0171710. Andrews, S. (2010). FastQC: a quality control tool for high throughput sequence data. Available at: http://www.bioinformatics.babraham.ac.uk/projects/fastqc. Armelim, J. M., Canniatti-Brazaca, S. G., Spoto, M. H. F., Arthur, V., and Piedade, S. M. S. (2006). Quantitative descriptive analysis of common bean (Phaseolus vulgaris L.) under gamma radiation. J. Food Sci. 71, 8–12. doi:10.1111/j.1365-2621.2006.tb12410.x. Bassett, A., and Cichy, K. (2020). QTL analysis of a yellow Phaseolus vulgaris recombinant inbred line population for a fast-cooking, flavorful, and flourishing future of dry beans. Annu. Rep. Bean Improv. Coop. 63, 95–96. Bassett, A., Dolan, K. D., and Cichy, K. (2020a). Reduced retort processing time improves canning quality of fast‐cooking dry beans (Phaseolus vulgaris L.). J. Sci. Food Agric. 100, 3995–4004. doi:10.1002/jsfa.10444. Bassett, A. N., Kamfwa, K., Ambachew, D., and Cichy, K. A. (2020b). Genetic variability and genome-wide association analysis of flavor and texture in cooked beans (Phaseolus vulgaris L.). In review. Bassett, M. J. (2007). “Genetics of seed coat color and pattern in common bean,” in Plant Breeding Reviews, ed. J. Janick (John Wiley & Sons, Inc.), 239–315. Bates, D., Mächler, M., Bolker, B. M., and Walker, S. C. (2015). Fitting linear mixed-effects models using lme4. J. Stat. Softw. 67. doi:10.18637/jss.v067.i01. Bauchet, G., Grenier, S., Samson, N., Segura, V., Kende, A., Beekwilder, J., et al. (2017). Identification of major loci and genomic regions controlling acid and volatile content in tomato fruit: implications for flavor improvement. New Phytol. 215, 624–641. doi:10.1111/nph.14615. Beninger, C. W., Gu, L., Prior, R. L., Junk, D. C., Vandenberg, A., and Bett, K. E. (2005). Changes in polyphenols of the seed coat during the after-darkening process in pinto beans (Phaseolus vulgaris L.). J. Agric. Food Chem. 53, 7777–7782. doi:10.1021/jf050051l. 185 Berry, M., Izquierdo, P., Jeffery, H., Shaw, S., Nchimbi-Msolla, S., and Cichy, K. A. (2020). QTL analysis of cooking time and quality traits in dry bean (Phaseolus vulgaris L.). Theor. Appl. Genet., Accepted with minor revision. doi:10.1007/s00122-020-03598-w. Bott, L., and Chambers, E. (2006). Sensory characteristics of combinations of chemicals potentially associated with beany aroma in foods. J. Sens. Stud. 21, 308–321. doi:10.1111/j.1745- 459X.2006.00067.x. Britten, P., Cleveland, L. E., Koegel, K. L., Kuczynski, K. J., and Nickols-Richardson, S. M. (2012). Updated US Department of Agriculture Food Patterns meet goals of the 2010 Dietary Guidelines. J. Acad. Nutr. Diet. 112, 1648–1655. doi:10.1016/j.jand.2012.05.021. Broman, K. W., Wu, H., Sen, Ś., and Churchill, G. A. (2003). R/qtl: QTL mapping in experimental crosses. Bioinformatics 19, 889–890. doi:10.1093/bioinformatics/btg112. Bullard, G. (2016). Heirloom beans are a thing, but do they really taste better than bulk beans? The Plate. Chen, P. X., Tang, Y., Marcone, M. F., Pauls, P. K., Zhang, B., Liu, R., et al. (2015). Characterization of free, conjugated and bound phenolics and lipophilic antioxidants in regular- and non-darkening cranberry beans (Phaseolus vulgaris L.). Food Chem. 185, 298–308. doi:10.1016/j.foodchem.2015.03.100. Cichy, K. A., Fernandez, A., Kilian, A., Kelly, J. D., Galeano, C. H., Shaw, S., et al. (2014). QTL analysis of canning quality and color retention in black beans (Phaseolus vulgaris L.). Mol. Breed. 33, 139–154. doi:10.1007/s11032-013-9940-y. Cichy, K. A., Porch, T. G., Beaver, J. S., Cregan, P., Fourie, D., Glahn, R. P., et al. (2015a). A Phaseolus vulgaris diversity panel for Andean bean improvement. Crop Sci. 55, 2149–2160. doi:10.2135/cropsci2014.09.0653. Cichy, K. A., Wiesinger, J. A., Berry, M., Nchimbi‐Msolla, S., Fourie, D., Porch, T. G., et al. (2019). The role of genotype and production environment in determining the cooking time of dry beans (Phaseolus vulgaris L.). Legum. Sci. 1, 1–15. doi:10.1002/leg3.13. Cichy, K. A., Wiesinger, J. A., and Mendoza, F. A. (2015b). Genetic diversity and genome-wide association analysis of cooking time in dry bean (Phaseolus vulgaris L.). Theor. Appl. Genet. 128, 1555–1567. doi:10.1007/s00122-015-2531-z. Duitama, J., Quintero, J. C., Cruz, D. F., Quintero, C., Hubmann, G., Foulquié-Moreno, M. R., et al. (2014). An integrated framework for discovery and genotyping of genomic variants from high- throughput sequencing experiments. Nucleic Acids Res. 42, e44–e44. doi:10.1093/nar/gkt1381. Eihusen, J., and Albrecht, J. A. (2007). Dry bean intake of women ages 19-45. Rural. Rev. Undergrad. Res. Agric. Life Sci. 2. 186 Elia, F. M., Hosfield, G. L., Kelly, J. D., and Uebersax, M. A. (1997). Genetic analysis and interrelationships between traits for cooking time, water absorption, and protein and tannin content of Andean dry beans. J. Am. Soc. Hortic. Sci. 122, 512–518. doi:10.21273/JASHS.122.4.512. Elsadr, H. T., Wright, L. C., Pauls, K. P., and Bett, K. E. (2011). Characterization of seed coat post harvest darkening in common bean (Phaseolus vulgaris L.). Theor. Appl. Genet. 123, 1467–1472. doi:10.1007/s00122-011-1683-8. Elshire, R. J., Glaubitz, J. C., Sun, Q., Poland, J. A., Kawamoto, K., Buckler, E. S., et al. (2011). A robust, simple genotyping-by-sequencing (GBS) approach for high diversity species. PLoS One 6. doi:10.1371/journal.pone.0019379. Erfatpour, M., Navabi, A., and Pauls, K. P. (2018). Mapping the non-darkening trait from ‘Wit- in bean (Phaseolus vulgaris). Theor. Appl. Genet. 131, 1331–1343. rood boontje’ doi:10.1007/s00122-018-3081-y. Erfatpour, M., and Pauls, K. P. (2020). A R2R3-MYB gene-based marker for the non-darkening seed coat trait in pinto and cranberry beans (Phaseolus vulgaris L.) derived from ‘Wit-rood boontje.’ Theor. Appl. Genet. 133, 1977–1994. doi:10.1007/s00122-020-03571-7. Feenstra, B., Skovgaard, I. M., and Broman, K. W. (2006). Mapping quantitative trait loci by an extension of the Haley-Knott regression method using estimating equations. Genetics 173, 2269– 2282. doi:10.1534/genetics.106.058537. Ferreira, J. J., del Castillo, R. R., Perez-Vega, E., Plans, M., Simó, J., and Casañas, F. (2012). Sensory changes related to breeding for plant architecture and resistance to viruses and anthracnose (Phaseolus vulgaris L.). Euphytica 186, 687–696. in bean market class Fabada doi:10.1007/s10681-011-0540-9. Heffelfinger, C., Fragoso, C. A., and Lorieux, M. (2017). Constructing linkage maps in the genomics 2224–2225. doi:10.1093/bioinformatics/btx177. Holland, J. B., Nyquist, W. E., and Cervantes-Martínez, C. T. (2003). “Estimating and interpreting heritability for plant breeding: An update,” in Plant Breeding Reviews, ed. J. Janick (John Wiley & Sons, Inc.), 9–112. Hooper, S. D., Glahn, R. P., and Cichy, K. A. (2019). Single Varietal Dry Bean (Phaseolus vulgaris L.) Pastas: Nutritional Profile and Consumer Acceptability. Plant Foods Hum. Nutr. 74, 342–349. doi:10.1007/s11130-019-00732-y. IPSOS in Canada. Available at: https://www1.agric.gov.ab.ca/$Department/deptdocs.nsf/all/sis13117/$FILE/v4_factors_influenc ing_pulse_consumption_summary_feb24_2010.pdf. influencing pulse consumption (2010). Factors Bioinformatics 33, era with MapDisto 2.0. 187 Jacinto-Hernandez, C., Azpiroz-Rivero, S., Acosta-Gallegos, J. A., Hernandez-Sanchez, H., and Bernal-Lugo, I. (2003). Genetic analysis and random amplified polymorphic DNA markers associated with cooking time in common bean. Crop Sci. 43, 329. doi:10.2135/cropsci2003.3290. Karlsen, M. C., Ellmore, G. S., and McKeown, N. (2016). Seeds-health benefits, barriers to incorporation, and strategies for practitioners in supporting consumption among consumers. Nutr. Today 51, 50–59. doi:10.1097/NT.0000000000000135. Katuuramu, D. N., Luyima, G. B., Nkalubo, S. T., Wiesinger, J. A., Kelly, J. D., and Cichy, K. A. (2020). On-farm multi-location evaluation of genotype by environment interactions for seed yield and cooking time in common bean. Sci. Rep. 10, 3628. doi:10.1038/s41598-020-60087-2. Kelly, J. D., and Cichy, K. A. (2012). Dry bean breeding and production technologies. Dry Beans Pulses Prod. Process. Nutr., 23–54. doi:10.1002/9781118448298.ch2. Kelly, J., Wright, E., and Bornowski, N. (2017). 2017 Dry bean yield trials. Kinsella, J. E. (1979). Functional properties of soy proteins. J. Am. Oil Chem. Soc. 56, 242–258. doi:10.1007/BF02671468. Koehler, H. H., Chang, C.-H., Scheier, G., and Burke, D. W. (1987). Nutrient composition, protein quality, and sensory properties of thirty-six cultivars of dry beans (Phaseolus vulgaris L.). J. Food Sci. 52, 1335–1340. doi:10.1111/j.1365-2621.1987.tb14076.x. Langmead, B., and Salzberg, S. L. (2012). Fast gapped-read alignment with Bowtie 2. Nat. Methods 9, 357–359. doi:10.1038/nmeth.1923. Leakey, C. L. A. (1992). Breeding on the C and J and B loci for modification of bean seedcoat flavonoids with the objective of improving food acceptibility. Annu. Rep. Bean Improv. Coop. 35, 13–17. Leakey, C. L. A. (2000). Progress in developing tannin-free dry Phaseolus beans. Annu. Rep. Bean Improv. Coop. 43, 18–20. Leterme, P., and Carmenza Muñoz, L. (2002). Factors influencing pulse consumption in Latin America. Br. J. Nutr. 88, 251–254. doi:10.1079/bjn/2002714. Martin, M. (2011). Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal 17, 10. doi:10.14806/ej.17.1.200. Meilgaard, M. C., Civille, G. V., and Carr, B. T. (1999). Sensory evaluation techniques. 3rd ed. Boca Raton, Fla: CRC Press. Mendoza, F. A., Kelly, J. D., and Cichy, K. A. (2017). Automated prediction of sensory scores for color and appearance in canned black beans (Phaseolus vulgaris L.) using machine vision. Int. J. Food Prop. 20, 83–99. doi:10.1080/10942912.2015.1136939. 188 Mkanda, A. V, Minnaar, A., and de Kock, H. L. (2007). Relating consumer preferences to sensory and physicochemical properties of dry beans (Phaseolus vulgaris). J. Sci. Food Agric. 87, 2868– 2879. doi:10.1002/jsfa.3046. Nasar-Abbas, S. M., Siddique, K. H. M., Plummer, J. A., White, P. F., Harris, D., Dods, K., et al. (2009). Faba bean (Vicia faba L.) seeds darken rapidly and phenolic content falls when stored at higher temperature, moisture and light intensity. LWT - Food Sci. Technol. 42, 1703–1711. doi:10.1016/j.lwt.2009.05.013. Peleg, H., Gacon, K., Schlich, P., and Noble, A. C. (1999). Bitterness and astringency of flavan- 3-ol monomers, dimers and trimers. J. Sci. Food Agric. 79, 1123–1128. doi:10.1002/(SICI)1097- 0010(199906)79:8<1123::AID-JSFA336>3.0.CO;2-D. Perea, C., De La Hoz, J. F., Cruz, D. F., Lobaton, J. D., Izquierdo, P., Quintero, J. C., et al. (2016). Bioinformatic analysis of genotype by sequencing (GBS) data with NGSEP. BMC Genomics 17. doi:10.1186/s12864-016-2827-7. R Core Team (2017). R: A Language and Environment for Statistical Computing. Available at: https://www.r-project.org/. Rivera, A., Fenero, D., Almirall, A., Ferreira, J. J., Simó, J., Plans, M., et al. (2013). Variability in sensory attributes in common bean (Phaseolus vulgaris L.): A first survey in the Iberian secondary diversity center. Genet. Resour. Crop Evol. 60, 1885–1898. doi:10.1007/s10722-013-9963-6. Robichaud, J. L., and Noble, A. C. (1990). Astringency and bitterness of selected phenolics in wine. J. Sci. Food Agric. 53, 343–353. doi:10.1002/jsfa.2740530307. Siddiq, M., and Uebersax, M. A. (2012). Dry beans and pulses production and consumption--An overview, in Dry Beans and Pulses Production, Processing and Nutrition, 1st ed., eds. M. Siddiq and M. A. Uebersax, John Wiley & Sons, Inc. doi:10.1002/9781118448298. Stone, H., Sidel, J., Oliver, S., Woolsey, A., and Singleton, R. C. (1974). Sensory evaluation by quantitative descriptive analysis. Food Technol. 28, 51. Vara-Ubol, S., Chambers, E., and Chambers, D. H. (2004). Sensory characteristics of chemical compounds potentially associated with beany aroma in foods. J. Sens. Stud. 19, 15–26. doi:10.1111/j.1745-459X.2004.tb00133.x. Voorrips, R. E. (2002). MapChart: Software for the graphical presentation of linkage maps and QTLs. J. Hered. 93, 77–78. doi:10.1093/jhered/93.1.77. Wang, N., and Daun, J. K. (2005). Determination of cooking times of pulses using an automated Mattson cooker apparatus. J. Sci. Food Agric. 85, 1631–1635. doi:10.1002/jsfa.2134. Wang, S., Basten, C. J., and Zeng, Z. B. (2005). “Windows QTL cartographer version 2.5,” in Statistical genetics (Raleigh, NC: North Carolina State University). 189 Wiesinger, J. A., Cichy, K. A., Glahn, R. P., Grusak, M. A., Brick, M. A., Thompson, H. J., et al. (2016). Demonstrating a nutritional advantage to the fast-cooking dry bean (Phaseolus vulgaris L.). J. Agric. Food Chem. 64, 8592–8603. doi:10.1021/acs.jafc.6b03100. Wiesinger, J. A., Cichy, K. A., Tako, E., and Glahn, R. P. (2018). The fast cooking and enhanced iron bioavailability properties of the Manteca yellow bean (Phaseolus vulgaris L.). Nutrients 10. doi:10.3390/nu10111609. Winham, D., Tisue, M., Palmer, S., Cichy, K., and Shelley, M. (2019). Dry bean preferences and attitudes among Midwest Hispanic and non-Hispanic White women. Nutrients 11, 178. doi:10.3390/nu11010178. Zhang, J., Zhao, J., Xu, Y., Liang, J., Chang, P., Yan, F., et al. (2015). Genome-wide association mapping for tomato volatiles positively contributing to tomato flavor. Front. Plant Sci. 6, 1–13. doi:10.3389/fpls.2015.01042. Zhao, J., Sauvage, C., Zhao, J., Bitton, F., Bauchet, G., Liu, D., et al. (2019). Meta-analysis of genome-wide association studies provides insights into genetic control of tomato flavor. Nat. Commun. 10, 1–12. doi:10.1038/s41467-019-09462-w. 190 CHAPTER 4: REDUCED RETORT PROCESSING TIME IMPROVES CANNING QUALITY OF FAST-COOKING DRY BEANS (Phaseolus vulgaris L.) [Published in: Journal of the Science of Food and Agriculture 100(10):3995-40004] 191 Reduced retort processing time improves canning quality of fast-cooking dry beans (Phaseolus vulgaris L.) Amber Bassett1, Kirk D. Dolan2, and Karen Cichy13* 1 Department of Plant, Soil and Microbial Sciences, Michigan State University 2 Department of Food Science and Human Nutrition, Michigan State University 3 Sugarbeet and Bean Research Unit, USDA-ARS ABSTRACT While it is generally accepted that fast‐cooking germplasm benefits consumers, benefits to the canning industry have not been established. Genotypes with good canning quality withstand the canning process while remaining intact with good appearance, but canning protocols used by breeders typically involve long processing times that may overcook some genotypes. The goal of this study was to identify whether cooking time influences canning quality in dry beans and whether reducing processing time could improve canning quality of fast‐cooking genotypes. A set of 20 yellow bean genotypes including Ervilha, PI527538 and 18 derived recombinant inbred lines were selected for their varied cooking times. By comparing the genotypes processed across five retort times, differences in canning quality were identified. All genotypes performed better when processed for less time than the standard 45 min, but canning quality was highest at 10 min for fast‐ and medium‐cooking genotypes and 15 min for slow‐cooking genotypes. Cooking time was correlated positively with texture and intactness and negatively with washed‐drained weights, indicating that slower cooking beans have higher canning quality. Color changed with retort processing such that longer times produced darker beans with more red and yellow. While fast‐ cooking beans exhibited lower canning quality at standard processing times, reduced retort 192 processing time allowed them to meet quality standards while still maintaining food safety. By accounting for cooking time as a component of canning quality, breeders can develop varieties that are convenient and cost efficient for preparation for both consumers and the canning industry. INTRODUCTION Dry beans are an affordable protein source with additional nutritional benefits, including soluble and insoluble dietary fiber, folate, and mineral content (Hornick and Weiss, 2011). In order to be edible, dry beans require hydrothermal processing to render inactive lectins and other anti- nutrients (Liener, 1979; Deshpande et al., 1984). Typically, dry beans are prepared by cooking in boiling water with or without prior soaking. The time required to cook beans can be significant, requiring hours of boiling to sufficiently soften the cotyledons. Genotype, age of the seeds, storage conditions, and moisture content all impact the ability of dry beans to take up water and cook in an acceptable amount of time (Hernandez-Unzon and Ortega-Delgado, 1989; Liu et al., 1995; Coelho et al., 2007; Cichy et al., 2015b). As a more convenient option, many consumers purchase canned beans. Canned beans are fully cooked and safe to eat without further processing, and they are more accessible to the average modern consumer with limited time for food preparation. However, some limitations to canned beans included their lack of prevalence in majority-world countries and their increased cost and negative health perceptions as compared to dry beans (Povey et al., 1998; Cichy et al., 2015b; Winham et al., 2019). While canned beans provide a partial solution to the inconvenience of long cooking times, fast-cooking beans are valuable to consumers who purchase dry beans. In addition, fast-cooking genotypes could benefit the canning industry by reducing the processing time required to prepare canned beans, resulting in energy savings (Deshpande et al., 1984). 193 The canning of beans is a hydrothermal process consisting of several steps that are modified depending on seed type: cleaning, soaking, blanching, filling, adding brine or sauce, sealing, and retort processing (Deshpande et al., 1984; Matella et al., 2013). Quality of canned bean products is evaluated on seed coat splitting, seed clumping, broth viscosity, extruded starch, or undesirable seed shape, color or size (Hosfield et al., 1995). Quality can be variable and is impacted by seed quality, canning protocol, and genotype (Hosfield and Uebersax, 1980; Hosfield et al., 1984; Hosfield, 1991; Ghasemlou et al., 2013; Matella et al., 2013). Retort temperature and duration are also important to prevent under- or over-cooking and to ensure safety for consumption. Due to the impact of genotype on canning quality, some dry bean breeding programs incorporate canning quality as a selection criterion for germplasm development. A five-point scale is used by trained evaluators to indicate canning quality based on appearance, and it incorporates several factors including overall appearance, prevalence of splits, presence of clumps, viscosity of cooking broth, extent of extruded starch, and conformity of seed shape, color, and size to the relevant market class (Hosfield et al., 1995; Mendoza et al., 2017). A score of 3 for appearance is considered acceptable, although expectations vary depending on the market class. Appearance requires many trained evaluators and is difficult to rate accurately and consistently, but image analysis may be a suitable alternative for canning quality evaluation in the near future (Long et al., 2019). After being evaluated by trained panelists, samples are rinsed and weighed to determine water uptake during canning. Washed-drained weights indicate whether the samples were under- or over-hydrated following canning. Subsamples (100 g) of each washed-drained sample are evaluated for texture using peak force measurements recorded with a Kramer shear cell. Low peak force values indicate mushy, overcooked samples or those that could not withstand canning 194 without splitting. High peak force indicates firm, intact samples or those that are undercooked. Ideal peak force ranges from 50 to 75 kg depending on the market class (Hosfield and Uebersax, 1980). Appearance, washed-drained weights, and texture comprise the primary measurements for evaluating canning quality in a dry bean breeding program (Hosfield, 1991; Kelly and Cichy, 2012). Dry bean breeding programs use a small-scale canning protocol that approximates industrial canning on limited sample sizes (Hosfield and Uebersax, 1980; Kelly and Cichy, 2012). Large numbers of genotypes and limited seed availability dictate that the canning protocol be standardized regarding retort time and temperature, preventing multiple processing methods from being applied to different samples. Breeding programs have generally relied on a 45 min retort time to evaluate canning quality. This practice may introduce bias against fast-cooking genotypes as they would be over-processed at 45 min and appear mushy with lower texture scores (Nordstrom and Sistrunk, 1977, 1979; Davis et al., 1980; Junek et al., 1980; Santoro et al., 2010). Cooking time measurements have not traditionally been part of the canning quality evaluation pipeline. However, if canning quality of fast-cooking lines is improved with reduced retort time, genotypes that are both fast-cooking and suitable for canning could be identified, benefitting the canning industry with reduced energy costs. The goal of the study reported here was to characterize the relationship between cooking time and canning quality in dry beans. A set of 20 bean genotypes, including Ervilha (fast- cooking), PI527538 (slow-cooking), and 18 derived recombinant inbred lines (RILs) were processed in a stationary retort at 121 °F for 10, 15, 20, 30, and 45 min and evaluated for intactness, washed-drained weight, texture, and color. These parents and RILs were selected for their varied cooking times so that the relationship between cooking time and canning quality could be assessed 195 with minimal confounding genetic variation. The prevalence of splits was evaluated with other attributes of the appearance scale excluded to improve accuracy of the scores and target the trait with the largest impact on canning quality. This study explored whether canning quality of fast- cooking genotypes is improved in samples with reduced retort processing duration to determine whether cooking time should be considered as a component of canning quality. MATERIALS AND METHODS Germplasm The germplasm relevant to this study consists of 18 yellow F5:8 RILs and their parental lines Ervilha and PI527538 (Figure S4.1). Ervilha has a pale yellow Manteca seed type with a gray hilum. It was originally collected from a marketplace in Angola in 2010 and is part of the Andean Diversity Panel (ADP0512) (Cichy et al., 2015a). PI527538 is also part of the Andean Diversity Panel (ADP0468) and exhibits a greenish brown seed coat with hints of purple and a black hilum. It was collected in Burundi in 1985. Ervilha cooks faster than PI527538, and this relative difference in cooking time is stable across environments (Cichy et al., 2015b; Katuuramu et al., 2020). The RILs were developed by advancing F2 seed via single seed descent to the F5 generation and then bulking seeds from individual plants to form RILs. The RILs evaluated in this study are a subset from 242 lines selected for their varying cooking times (Bassett and Cichy, 2020). All RILs were evaluated for cooking time in 2017, and the nine fastest and nine slowest were selected for this study (Figure 4.1). The genotypes were grown at the Montcalm Research Farm in MI in 2017. The soil type is Eutric Glossoboralfs (coarse-loamy, mixed) and Alfic Fragiorthods (coarse-loamy, mixed, frigid). Two row plots 4.75 m long with 0.5 m spacing between rows were arranged in a 196 randomized complete block design with two replications per genotype. Standard agronomic practices were followed to control biotic and abiotic stresses and ensure adequate growing conditions. Plants were hand-pulled at maturity and threshed with a plot Hege 140 plot harvester (Wintersteiger, Utah, USA). Seeds were cleaned by hand following harvest to remove field debris, off types, and damaged seed. All seeds were stored at room temperature and low humidity for 5 months following harvest. Cooking Time Determination Cooking times were determined using pre-soaked seeds with two replicates per genotype with automated Mattson cookers (Wang and Daun, 2005). Prior to cooking, 30 seeds per replicate were sorted into coin envelopes and equilibrated to 10-14% moisture in a humid cold room. Moisture content was checked using a moisture meter (Moisture Check Plus, Deere and Company, Moline, IL). After the desired moisture had been achieved, samples were weighed and soaked for 12 h in 250 ml distilled water in preparation for cooking. The seeds were then blotted dry and weighed to determine soaked weight. Twenty-five soaked seeds for each replicate were loaded onto 25-well Mattson cookers (Michigan State University Machine Shop, East Lansing, MI) with weighted (65 g) 2 mm diameter pins positioned in the center of each seed. Loaded Mattson cookers were placed into 4 L stainless steel beakers with 1.8 L of boiling distilled water. A low boil was maintained using a Cuisinart Countertop Burner (Cuisinart, Stamford, CT), and each replicate was cooked until 80% of the seeds were pierced completely. Cooking time was recorded, and samples were cooled for up to 10 min at room temperature and then weighed to determine cooked weight. Canning Protocol Each genotype was processed in duplicate across five retort times for a total of 10 samples per genotype. Prior to canning, seed moisture was increased to 14-17% moisture using a moisture 197 chamber. Once the seeds reached the desired moisture, 90 g dry weight of seeds per sample were placed in mesh bags and soaked for 12 h in 0.0028% CaCl2 solution prior to canning. The soaked samples were placed into 300 x 407 tin cans, which were then filled with brine (1.5% sucrose, 1.25% NaCl, 0.03% CaCl2). Filled cans were transported via a 5.6 m metal-tiled conveyor belt moving 2.15 cm/s through an exhaust box to facilitate water uptake and removal of air bubbles. The cans were heated to approximately 75 °C upon exiting the exhaust box, at which point they were sealed using a Dixie Double Seamer (Dixie Canner Co., Athens, GA). Cans were then placed in a Melco Steel Steam Sterilizer (Melco Steel Inc., Azusa, CA) and processed stationary at 121 °C for 10, 15, 20, 30, or 45 min. The come-up time to reach 121 °C in the retort was 15 min. All process times exceeded minimum safety requirements for the production of sterile canned bean products (F0 > 6 min) (Matella et al., 2013) and destruction of anti-nutritional factors including lectins and protease inhibitors (Dhurandhar and Chang, 1990; Lajolo and Genovese, 2002; Nciri and Cho, 2018; Thompson, 2019). Following processing, cans were cooled to 40 °C via the addition of cold water to the retort. Once cooled, water was drained from the retort, and the cans were removed, dried, and left to equilibrate at room temperature for 1 week prior to opening. Visual Evaluation To prepare for visual evaluation, cans were opened and poured into paper food trays, with samples gently stirred and evenly distributed across each tray. Samples were randomly arranged to minimize bias. Trained reviewers then evaluated each sample using a five-point scale for intactness (1: 0-20% intact, 2: 21-40% intact, 3: 41-60% intact, 4: 61-80% intact, 5: 81-100% intact). Intactness is defined as an absence of splits. There were 14 total evaluators with a minimum of seven observations per sample due to absent evaluators during select can opening sessions. 198 Washed-drained Weight Determination and Image Analysis Following visual evaluations, samples were rinsed to remove brine and randomly arranged in a large weigh boat to determined washed-drained weight. Samples were then imaged in the weigh boat for downstream image analysis (Figure S4.2). Images of canned samples were collected using the custom machine vision system described in Mendoza et al., 2017. The camera settings were as follows: manual exposure, auto focusing, lens aperture f = 5.6, shutter speed of 1/125, white balanced, ISO 100, and flash off. A custom macro in ImageJ software developed for canned bean color analysis was used to determine CIELAB values for each sample (Bornowski, 2018). Each image was preprocessed to brighten samples and minimize reflections using a constant gamma correction and the noise reduction feature in the ImageJ software. To obtain CIELAB values, images were partitioned into L*, a*, and b* slices, and the mean value for each slice was recorded. L* measures black (0) to white (100); a* measure green (-) to red (+), and b* measures blue (-) to yellow (+). Texture Analysis After imaging each sample, texture was determined for two replicates of 100 g subsamples per can. The samples were evenly distributed in a 10 blade TA-91X Kramer shear cell attachment. Using a TA.XTPlus100 texture analyzer (Texture Technologies Corp., Hamilton, MA) with a 100 kg load cell, the samples were completely compressed for the 105 cm length of the Kramer shear cell at 20 mm/s. Peak force measurements were recorded in kilograms using Exponent version 6.1.4.0 (Stable Micro Systems, Godalming, UK). 199 Statistical Analysis All analyses of variance (ANOVAs) in this study were conducted using the MIXED procedure in SAS version 9.4 of the SAS System for Windows (SAS Institute Inc. Cary, NC). For seed weight, water uptake, and cooking time phenotypes, the model included genotype as a fixed effect and replicate as a random effect. For intactness, washed-drained weight, texture, and CIELAB values, the model included genotype, retort time, genotype by retort time as fixed effects and replicate as a random effect. Intactness included evaluator and genotype by evaluator as random effects. Least squares estimates were reported in place of means for intactness to account for evaluator effects. Texture included subsample as a random effect. For ANOVAs within retort times, retort time and genotype by retort time were excluded from the fixed effects. Mean separation was determined using pdiff within the MIXED procedure and a Tukey multiple comparison adjustment. Pearson correlation coefficients were determined in R using the Cor function. Genotypes were separated into fast (18-20 min), medium (20-26 min), and slow (26-28 min) cooking groups determined via least squares differences to evaluate the relationship between cooking time and canning quality traits. Analyses for these groups were performed using PROC MIXED with the models including cooking group in place of genotype. RESULTS Cooking Time and Water Uptake The genotypes selected for this study exhibited significant differences in seed weight, soak water uptake, and cooking time, although the parents did not differ in seed weight (Table 4.1). Significant correlations between seed weight and cooking time and between water uptake after soaking and cooking time were detected (Figure 4.2). Similar correlations have been identified in 200 a previous study (Cichy et al., 2015b). No significant differences in total water uptake were identified among genotypes (Table 4.1). The raw 100 seed weights of Ervilha and PI527538 were 52.84 and 47.99 g respectively. The seed weights of the 18 RILs ranged from 45.32 to 64.47 g. After the 12 h soak, water uptake of Ervilha was 114.1% and of PI527538 was 100.6%. The water uptake of the 18 RILs ranged from 98.4% to 116.4%. The cooking time for Ervilha and PI527538 were 18.8 and 26.3 min respectively. The cooking times of the RILs ranged from 18 to 28 min. The fast, medium, and slow groups were not significantly different for seed weight or total water uptake, but the fast group had a significantly higher water uptake as compared to the medium and slow groups (Figure 4.3). The fast, medium, and slow groups exhibited significant differences in their cooking times as intended when the groups were defined (Figure 4.3). Canned Bean Intactness Genotype, retort time, and genotype by retort time significantly affected intactness as rated by trained evaluators (Table 4.2). In addition, evaluator and genotype by evaluator were significant effects (data not shown). The least squares estimates for intactness were correlated positively with texture and negatively with retort time, washed-drained weight, L*, and b* (Figure 4.4). Cooking time was positively correlated with intactness ratings across all retort times (Figures 4.4-5, Table S4.1). The intactness ratings for Ervilha were 2.5, 2.8, 2.2, 2.3, and 2.8 and for PI527538 were 3.8, 3.4, 3.1, 3.2, and 3.0 for 10, 15, 20, 30, and 45 min processing times respectively (Table 4.3). For each retort time, the intactness ranges of the RILs are as presented in Table 4.3. Ervilha, PI527538 and the RILs showed a decrease in intactness as retort time increased, except for the 45 min retort time for Ervilha (Table 4.3). For all retort times except 45 min, Ervilha had a significantly lower intactness value than PI527538 (Table 4.3). 201 Cooking group and retort time for the fast, medium, and slow groups had a significant effect on intactness, but cooking group by retort time did not (Table 4.4). In addition, evaluator and cooking group by evaluator were significant effects (data not shown). The fast-cooking group had lower intactness scores than the slow-cooking group overall and across retort times (Table 4.5, Figure 4.6). In each group, a significant decline in intactness was observed as retort time increased (Figure 4.6, Table 4.6). Washed-drained Weight Genotype, retort time, and genotype by retort time significantly affected washed-drained weights (Table 4.2). The RILs increased in washed-drained weight as retort time increased (Table 4.3). For all retort times, Ervilha had a significantly higher washed-drained weight than PI527538 (Table 4.3). The washed-drained weights ranged from 271.1 to 280.7 g for Ervilha and 246.0 to 256.3 g for PI527538 across all retort times (Table 4.3). For each retort time, the washed-drained weight ranges of the RILs are as presented in Table 4.3. Washed-drained weight was correlated positively with retort time, L* and b* and negatively with intactness, texture, and a* (Figure 4.4). Cooking time was negatively correlated with washed-drained weight across all processing times (Figures 4.4-5, Table S4.1). For the fast, medium, and slow groups, cooking group and retort time had a significant effect on washed-drained weight, but cooking group by retort time did not (Table 4.4). The fast group had a higher washed-drained weight as compared to the medium and slow groups for each retort time and overall (Table 4.5, Figure 4.6). In each group, a significant increase in washed- drained weights can be observed as retort time is increased (Figure 4.6, Table 4.6). 202 Texture Analysis Genotype, retort time, and genotype by retort time significantly affected texture (Table 4.2). Ervilha, PI527538, and the RILs texture values decreased significantly as retort time increased (Table 4.3). For all retort times, Ervilha had a significantly lower peak force measurement than PI527538 (Table 4.3). The measurements ranged from 25.9 to 56.8 kg for Ervilha and 38.9 to 84.6 kg for PI527538 across all retort times (Table 4.3). For each retort time, the measurement ranges of the RILs are presented in Table 4.3. Texture was correlated positively with intactness and negatively with retort time, washed-drained weight, L*, and b* (Figure 4.4). Cooking time was positively correlated with texture across all processing times such that beans that take longer to cook in boiling water as determined with a Mattson cooker also have firmer texture when canned (Figures 4.4-5, Table S4.1). Cooking group, retort time, and cooking group by retort time significantly affected texture for the fast, medium and slow groups (Table 4.4). The fast, medium, and slow groups had significantly different texture overall such that measurements increased from fast to medium to slow (Table 4.5, Figure 4.6). Within each retort time, the fast group had significantly softer texture than the medium and slow groups (Figure 4.6). The medium group had lower texture measurements than the slow group for the 10 min retort time, but otherwise was equivalent to the slow group (Figure 4.6). In each group, a significant decrease in firmness was observed as retort time increased (Figure 4.6, Table 4.6). CIELAB Values Genotype and retort time significantly affected L*, a*, and b*, but genotype by retort time was only significant for a* and b* (Table 4.2). For each retort time, the L*, a*, and b* ranges of the RILs are presented in Table 4.3. For Ervilha and PI527538 respectively, L* ranged from 62.4 203 to 66.7 and 44.3 to 48.8; a* ranged from 5.5 to 7.3 and 8.4 to 14.6, and b* ranged from 23.6 to 25.9 and 18.3 to 22.9 across all retort times. L*, a*, and b* values for Ervilha and PI527538 were significantly different such that Ervilha had higher L* and b* and lower a*. L* decreased and a* and b* increased as retort time increased for Ervilha, PI527538, and the RILs, although L* was not significantly different across retort times for Ervilha (Table 4.3). L* was correlated positively with washed-drained weight and b* and negatively with retort time, cooking time, intactness, texture, and a*; a* was correlated positively with retort time and cooking time and negatively with washed-drained weight and L*; and b* was correlated positively with retort time, washed-drained weight, and L* and negatively with cooking time, intactness, and texture (Figure 4.4). Cooking time was correlated positively with a* and negatively with L* and b* across all retort times (Figures 4.4-5, Table S4.1). For the fast, medium and slow groups, cooking group and retort time significantly affected L*, a*, and b* and cooking group by retort time significantly affected a* and b* (Table 4.4). The fast, medium, and slow groups had significantly different CIELAB values such that the fast group had the highest L* and b* and lowest a* values and the slow group had the highest a* and lowest L* and b* values (Table 4.5, Figure 4.7). Within each retort time, the groups were distinct for L* and a* and the fast and slow groups were distinct for b* with the medium group falling in between the two other groups (Figure 4.7). In each group, a significant decrease in firmness was observed as retort time increased (Figure 4.6, Table 4.6). L* decreased and a* and b* increased as retort time increased for all groups (Table 4.6). DISCUSSION The genotypes in this study exhibited below-acceptable values for both intactness and 204 texture under normal retort processing conditions (45 min) (Tables 4.3, 4.6). However, intactness and texture improved and washed-drained weight decreased with reduced processing time, indicating that too long processing times negatively impact canning quality. While faster cooking genotypes displayed poorer canning quality overall, reducing retort time to as low as 10 min improved canning quality in faster cooking genotypes as indicated by decreased washed-drained weight and increased firmness and intactness approaching and in some cases exceeding quality standards. The 10 min retort time resulted in the best canning quality for the fast and medium cooking groups, but the slow group performed best with a 15 min retort time, as 10 min was insufficient to achieve texture within the ideal range of 50 – 75 kg. These results indicate that current small-scale canning protocols used for germplasm screening are biased toward slow cooking genotypes that can withstand longer processing times, preventing genotypes with acceptable canning quality at lower retort processing times from being identified. Energy costs from heat processing are a significant expense for canning companies (Featherstone, 2015a, 2015b), and retort time can be reduced substantially while maintaining safety of the canned product (F0 > 6 min) (Matella et al., 2013). Previous research found that anti-nutritional factors including lectins and protease inhibitors can be deactivated by cooking for 10 min at 100 °C or pressure cooking for 7.5 min and that beans cooked to acceptable texture have minimal residual anti- nutrient activity (Dhurandhar and Chang, 1990; Lajolo and Genovese, 2002; Nciri and Cho, 2018; Thompson, 2019). Evaluating canning quality under shorter retort times would bias selection toward fast-cooking genotypes, which could allow the advancement of germplasm that is more convenient to prepare for consumers and requires less energy to process for canning companies. While reduced retort processing time appears to improve canning quality in fast-cooking genotypes, it also affects the color of the canned product such that longer retort processing times 205 lead to darker beans exhibiting more red and yellow color. Fast-cooking time was associated with lighter (+L*) canned products exhibiting more green (-a*) and yellow (+b*) color, and this is also seen in the raw seed prior to processing (Figures S4.3-4). While darker color has previously been associated with longer cooking times (Cichy et al., 2015b), this correlation was also expected considering PI527538 is both darker and slower-cooking than Ervilha. If changes in color due to differences in retort time impact consumer preference, then color should be considered when evaluating canning quality in fast-cooking genotypes. For black beans, color is an important trait for consumer acceptance, and breeding programs use a separate scale for color to indicate degree of blackness (Cichy et al., 2014; Mendoza et al., 2017). CIELAB color measurements were recently found to be strongly correlated with the color scores from trained panelists in black beans indicating that CIELAB values are sufficient for evaluating color (Bornowski, 2018). In order for new dry bean varieties to be successful, they must meet industry standards for canning quality (Kelly and Cichy, 2012). Canned products address the consumer need for convenience and as such are a common method of dry bean consumption in the USA. For these reasons, canning quality has been prioritized in breeding programs to the detriment of cooking time. While cooking time varies across all market classes, some market classes including yellow beans are more consistently fast-cooking, which makes them well-suited for the development of traits related to convenience. Currently, yellow beans have a low, but increasing prevalence in the USA, and it will be critical that they meet industry standards for canning quality to be successful as a market class. By prioritizing fast cooking time as an aspect of canning quality in future yellow bean varieties, they will be marketable to consumers prioritizing convenience and canning companies prioritizing low energy costs. Other market classes could also benefit from improved cooking times, but seed mixing at grain elevators could impact homogeneity of canned products if 206 the cooking times are not similar across genotypes. It could be worthwhile to isolate genotypes by cooking time to allow canning companies and consumers to benefit from fast-cooking times where possible, as well as allow growers to have a wider selection of varieties to choose from that meet canning quality standards when processed appropriately for their cooking time. CONCLUSIONS This study identified that fast-cooking genotypes benefit from shorter retort processing times, allowing for evaluation of their optimal canning quality. By reducing retort processing time, the bias toward slow-cooking beans can be mitigated, allowing future variety releases to maintain both fast-cooking time and high canning quality. This would benefit the canning industry, which may value the reduced time and energy required to prepare fast-cooking beans, as well as growers, who would have more options when selecting varieties to grow while still accounting for canning quality standards. The improved understanding gained from this research will allow dry bean breeders to better meet the needs of both consumers and the canning industry through the development of varieties that are convenient and cost efficient to prepare both in the kitchen and in the can. ACKNOWLEDGEMENTS This work was supported in part by funding from the U.S. Department of Agriculture, National Institute of Food and Agriculture (2017 67013 26212) and the Agricultural Research Service Projects 5050-21430-01000D (K.C.). We also thank Dr. Mohammad Siddiq and Dr. Jeff Swada for technical assistance with the retort and the Michigan State University dry bean breeding program for assistance with canning and evaluation. 207 APPENDICES 208 APPENDIX A: CHAPTER 4 TABLES AND FIGURES Table 4.1 Average values of seed weight, soak water uptake, cooking time, and total water uptake for Ervilha, PI527538, and RILs. Trait Parents Ervilha PI527538 Seed weight (g/100 seeds) 52.71a ± 2.3 49.72a ± 0.9 RILs Mean Range 54.06 45.3 - 64.5 P value <0.0001 Soak water uptake (%) 114.05a ± 4.8 100.62b ±3.1 Cooking time (min) 18.84b ± 0.3 26.25a ± 1.1 103.25 98.4 - 116.4 0.0017 23.10 18.0 - 27.7 <0.0001 Total water uptake (%) 148.89a ± 11.7 159.19a ± 15.6 132.57 120.4 - 141.4 NS P-values indicate the significance of the genotype effect determined via ANOVA. Mean separation between parents is indicated by superscript letter. Table 4.2 ANOVA results indicating the significance of the fixed effects genotype, retort time, and genotype by retort time for intactness, washed-drained weight, texture, and CIELAB color. Trait Genotype Retort Time Genotype x Retort Time Intactness <0.0001 <0.0001 <0.0001 Washed-Drained Wt. <0.0001 <0.0001 0.0331 Texture <0.0001 <0.0001 <0.0001 L* a* b* <0.0001 <0.0001 NS <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 209 Table 4.3 Means and ranges of intactness (1–5 scale), washed-drained weight (W-D Wt.) (g), texture measurements (kg), and CIELAB values for Ervilha, PI527538, and the RILs at the five retort processing times. Parents Ervilha PI527538 P value 10 min 2.5a ± 0.1 3.8a ± 0.2 <0.0001 15 min 2.8a ± 0.2 3.4ab ± 0.2 0.0071 RILs Mean 2.9a ± 0.1 2.7b ± 0.1 Range P value 1.3 - 3.6 <0.0001 1.2 - 3.4 <0.0001 Intactness† 20 min 2.2a ± 0.3 3.1b ± 0.3 0.0031 30 min 2.3a ± 0.2 3.2b ± 0.2 0.0002 2.5bc ± 0.1 1.1 - 3.5 2.5c ± 0.1 1.2 - 3.3 45 min 2.8a ± 0.3 3.0b ± 0.2 0.4635 2.4c ± 0.2 0.8 - 3.6 <0.0001 <0.0001 <0.0001 10 min 271.1a ± 2.8 246.0a ± 6.4 <0.0001 15 min 280.0a ± 1.7 247.8a ± 1.5 <0.0001 W-D Wt. 20 min 273.2a ± 8.5 252.2a ± 0.2 <0.0001 30 min 280.0a ± 4.0 255.4a ± 2.1 <0.0001 260.0c ± 1.4 239.6 - 278.6 <0.0001 260.2bc ± 1.2 246.3 - 281.7 <0.0001 263.5ab ± 1.3 252.0 - 281.6 0.0007 264.9a ± 1.0 253.3 - 283.9 <0.0001 45 min 280.7a ± 3.1 256.3a ± 0.6 <0.0001 266.4a ± 1.2 244.4 - 283.8 0.0033 10 min 56.8a ± 1.0 84.6a ± 1.6 <0.0001 15 min 40.9b ± 1.2 68.7b ± 1.1 <0.0001 Texture 20 min 32.9c ± 1.0 51.9c ± 0.6 <0.0001 30 min 35.8c ± 0.2 55.4c ± 0.3 <0.0001 69.0a ± 1.5 58.3b ± 1.1 44.5 - 96.4 <0.0001 37.4 - 76.4 <0.0001 43.8c ± 0.9 28.5 - 59.1 <0.0001 44.3c ± 0.8 31.3 - 56.8 <0.0001 45 min 25.9d ± 1.3 38.9d ± 1.4 <0.0001 32.1d ± 0.5 21.3 - 41.7 <0.0001 10 min 66.7a ± 0.7 48.8a ± 0.6 <0.0001 15 min 66.2a ± 0.4 48.0a ± 0.2 <0.0001 L* 20 min 64.7a ± 0.5 47.6ab ± 0.3 <0.0001 30 min 65.3a ± 0.5 47.4ab ± 0.8 <0.0001 57.8a ± 1.1 57.2a ± 1.2 48.0 - 68.7 <0.0001 47.4 - 68 <0.0001 56.6a ± 1.1 46.8 - 67.3 <0.0001 55.4a ± 1.2 44.8 - 66.7 <0.0001 45 min 62.4a ± 1.3 44.3b ± 0.5 <0.0001 52.8b ± 1.2 42.8 - 65.0 <0.0001 10 min 5.5b ± 0.1 8.4d ± 0.0 <0.0001 15 min 5.3b ± 0.1 9.1cd ± 0.2 <0.0001 a* 20 min 6.0b ± 0.0 9.8c ± 0.2 <0.0001 30 min 6.3ab ± 0.1 11.5b ± 0.1 <0.0001 6.8d ± 0.2 7.3d ± 0.2 8.0c ± 0.3 9.2b ± 0.3 5.0 - 8.8 <0.0001 4.7 - 9.9 <0.0001 5.3 - 10.3 <0.0001 5.9 - 12.6 <0.0001 45 min 7.3a ± 0.4 14.6a ± 0.0 <0.0001 9.9a ± 0.4 6.2 - 13.0 <0.0001 10 min 23.6b ± 0.3 18.3c ± 0.0 <0.0001 15 min 23.9b ± 0.0 19.4bc ± 0.4 <0.0001 b* 20 min 24.1b ± 0.3 20.5b ± 0.0 <0.0001 30 min 24.9ab ± 0.3 22.3a ± 0.4 <0.0001 20.2d ± 0.3 21.0d ± 0.3 16.6 - 24.4 <0.0001 18.1 - 24.4 <0.0001 22.2c ± 0.2 20.1 - 26.2 <0.0001 23.0b ± 0.2 20.4 -26.8 <0.0001 45 min 25.9a ± 0.4 22.9a ± 0.1 <0.0001 23.7a ± 0.2 21.1 - 26.6 <0.0001 † Means for Intactness are least squares estimates. P-values indicate the least squares differences between the parents and the significance of the genotype effect determined via ANOVA. Mean separation across retort times is indicated by superscript letter. 210 Table 4.4 ANOVA results indicating the significance of the fixed effects cooking group, retort time, and cooking group by retort time for intactness, washed-drained weight, texture, and CIELAB color. Trait Cooking Group Retort Time Cooking Group x Retort Time Intactness <0.0001 Washed-Drained Wt. <0.0001 <0.0001 <0.0001 NS NS Texture L* a* b* <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 NS <0.0001 <0.0001 <0.0001 0.0138 Table 4.5 Means and ranges of intactness (1–5 scale), washed-drained weight, texture measurements, and CIELAB values for the fast-, medium-, and slow-cooking groups across all retort times. Trait Fast Mean Medium Slow Range Mean Range Mean Range Intactness† 2.4b ± 0.1 1.6 - 3.5 2.5b ± 0.1 0.8 - 3.6 2.9a ± 0.1 1.9 - 3.8 W-D Wt. 269.6a ± 1.0 253.6 - 283.9 259.8b ± 0.8 242.4 - 277.6 258.0b ± 0.9 239.6 - 270.5 Texture 40.5c ± 1.0 21.3 - 73.9 51.7b ± 1.2 25.5 - 96.4 55.6a ± 1.5 31.0 - 90.5 L* a* b* 65.0a ± 0.3 58.3 - 68.7 54.0b ± 0.7 42.8 - 67.9 49.6c ± 0.4 43.8 - 57.3 6.1c ± 0.1 4.7 - 8.7 8.5b ± 0.2 5.0 - 12.5 9.9a ± 0.2 6.8 - 13.0 23.4a ± 0.2 20.5 - 26.4 21.7b ± 0.3 16.6 - 26.8 21.0c ± 0.2 18.2 - 24.2 † Means for Intactness are least squares estimates. Mean separation across cooking groups is indicated by superscript letter. 211 Table 4.6 Means and ranges of intactness (1–5 scale), washed-drained weight, texture measurements, and CIELAB values for the fast-, medium-, and slow-cooking groups at the five retort times. Fast Mean Medium Slow Range Mean Range Mean Range 10 min 2.7a ± 0.2 1.9 - 3.5 15 min 2.6a ± 0.2 1.9 - 3.3 Intactness† 20 min 2.4ab ± 0.2 1.8 - 3.0 30 min 2.2b ± 0.2 1.6 - 3.2 2.8a ± 0.2 2.6ab ± 0.3 1.3 - 3.6 1.2 - 3.4 2.3bc ± 0.3 1.1 - 3.2 2.4abc ± 0.3 1.2 - 3.2 3.3a ± 0.2 2.9b ± 0.2 2.5 - 3.8 1.9 - 3.4 2.8b ± 0.2 2.4 - 3.5 2.8b ± 0.2 2.1 - 3.3 45 min 2.3ab ± 0.2 1.8 - 2.8 2.1c ± 0.3 0.8 - 3.6 2.7b ± 0.2 2.2 - 3.2 10 min 265.4a ± 2.1 253.6 - 278.6 253.0c ± 1.4 242.4 - 262.5 250.5b ± 1.8 239.6 - 262.5 15 min 266.3a ± 2.2 256.8 - 281.7 20 min 271.9a ± 2.3 257.0 - 281.6 30 min 271.8a ± 1.7 263.7 - 283.9 257.8bc ± 1.7 246.3 - 273.7 261.0a ± 1.5 254.2 - 277.1 262.1a ± 1.1 253.9 - 271.0 258.4a ± 1.4 252.0 - 265.7 257.3a ± 1.4 246.3 - 263.6 261.7a ± 2.0 253.3 - 270.5 W-D Wt. 45 min 272.4a ± 2.0 258.6 - 283.8 265.1a ± 1.5 255.2 - 277.6 262.2a ± 2.1 244.4 - 270.5 10 min 55.5a ± 1.5 44.6 - 73.9 15 min 48.0b ± 1.5 37.4 - 64.8 70.3a ± 2.4 61.9b ± 1.7 49.4 - 96.4 41.9 - 76.4 80.9a ± 1.0 70.0 - 90.5 63.9b ± 1.0 55.0 - 71.3 Texture 20 min 35.3c ± 1.1 28.5 - 46.7 30 min 36.9c ± 0.7 31.3 - 44.0 46.6c ± 1.2 31.2 - 59.1 46.2c ± 1.0 35.5 - 56.8 48.6c ± 0.6 44.0 - 54.2 49.2c ± 0.9 42.9 - 55.7 45 min 27.0d ± 0.5 21.3 - 30.9 33.4d ± 0.7 25.5 - 41.7 35.4d ± 0.6 31.0 - 41.4 10 min 66.6a ± 0.5 63.3 - 68.7 15 min 66.2a ± 0.5 62.5 - 68.0 55.8a ± 1.4 55.2a ± 1.4 48.7 - 67.9 47.8 - 67.3 51.5a ± 0.8 48.0 - 56.7 50.9a ± 1.0 47.4 - 57.3 L* 20 min 65.2a ± 0.3 63.1 - 66.5 30 min 64.7ab ± 0.5 61.0 - 66.3 55.0a ± 1.4 47.2 - 67.3 53.5a ± 1.6 45.9 - 66.7 49.9a ± 0.8 46.8 - 54.6 50.1ab ± 1.2 46.3 - 61.2 45 min 62.4b ± 0.6 58.3 - 64.8 50.5a ± 1.5 42.8 - 65.0 46.4b ± 0.6 43.8 - 51.6 10 min 5.3c ± 0.1 5.0 - 5.7 15 min 5.5bc ± 0.1 4.7 - 6.2 a* 20 min 6.0b ± 0.1 5.6 - 6.6 30 min 6.7a ± 0.2 6.2 - 7.9 6.9c ± 0.2 7.6c ± 0.3 5.2 - 8.7 5.0 - 9.3 8.2bc ± 0.3 5.3 - 10.2 9.6ab ± 0.4 5.9 - 11.8 8.2d ± 0.2 8.6d ± 0.2 6.8 - 8.8 7.3 - 9.9 9.6c ± 0.2 8.5 - 10.3 10.8b ± 0.3 7.7 - 11.7 45 min 7.3a ± 0.2 6.2 - 8.7 10.3a ± 0.4 6.3 - 12.5 12.0a ± 0.3 9.8 - 13.0 10 min 22.5c ± 0.3 20.5 - 24.0 15 min 22.7bc ± 0.3 20.7 - 24.4 19.6c ± 0.6 20.7bc ± 0.4 18.1 - 24.4 16.6 - 24.4 b* 20 min 23.2bc ± 0.3 21.6 - 25.0 30 min 23.9ab ± 0.4 22.0 - 26.0 22.1ab ± 0.4 20.1 - 26.2 22.8a ± 0.5 20.4 - 26.8 18.9d ± 0.2 18.2 - 20.2 19.9c ± 0.2 18.9 - 21.3 21.2b ± 0.2 20.4 - 23.0 22.6a ± 0.4 21.5 - 26.4 45 min 24.9a ± 0.2 24.0 - 26.4 23.4a ± 0.4 21.1 - 26.6 23.1a ± 0.2 22.1 - 24.2 † Means for Intactness are least squares estimates. Mean separation across retort times is indicated by superscript letter. 212 Figure 4.1 Histogram of the cooking times of Ervilha, PI527538, and the 242 RILs, determined using the Mattson cooker method following a 12 h soak. Seeds were grown at the Montcalm Research Farm in Michigan, USA in 2016. The nine fastest (in blue) and slowest cooking lines (in red) were selected for this study. 213 Figure 4.2 Pearson correlation matrix of seed weight, soak water uptake, cooking time, and total water uptake. Correlation coefficients are indicated in the lower left and represented by colored, directional ellipses in the upper right. *P < 0.05, **P < 0.01, ***P < 0.001. 214 Figure 4.3 Boxplots of seed weights, soak water uptake, cooking times, and total water uptake for Ervilha, PI527538, and selected RILs separated into fast-, medium-, and slow-cooking groups. Lines indicate Ervilha (yellow) and PI527538 (brown). Mean separation is indicated by letters above each boxplot. 215 Figure 4.4 Pearson correlation matrix of retort time, cooking time, washed-drained weight, texture, intactness, and CIELAB color values across all genotypes and retort times. Correlation coefficients are indicated in the lower left and represented by colored, directional ellipses in the upper right. *P < 0.05, **P < 0.01, ***P < 0.001. 216 Figure 4.5 Scatterplots showing the relationship between cooking time and washed-drained weight, texture, intactness, and CIELAB color values separated by retort time. The five retort time series are indicated by colors and symbols as specified. 217 Figure 4.6 Boxplots of washed-drained weights, texture and intactness values across all retort times for Ervilha, PI527538, and selected RILs separated into fast-, medium-, and slow-cooking groups. Lines indicate Ervilha (yellow) and PI527538 (brown). Mean separation within each retort time is indicated by letters above each boxplot. 218 Figure 4.7 Boxplots of CIELAB color values across all retort times for Ervilha, PI527538, and selected RILs separated into fast-, medium-, and slow-cooking groups. Lines indicate Ervilha (yellow) and PI527538 (brown). Mean separation within each retort time is indicated by letters above each boxplot. 219 APPENDIX B: CHAPTER 4 SUPPLEMENTAL TABLES AND FIGURES Table S4.1 Pearson correlation coefficients and P-values for correlations between cooking time and washed-drained weight, texture, intactness, and CIELAB color values at the five retort times. 10 min 15 min 20 min 30 min 45 min r Trait Intactness Washed-Drained Wt Texture L* a* b* 0.39 P-value NS r -0.71 P-value 0.0004 r 0.87 P-value <0.0001 r -0.90 P-value <0.0001 r 0.93 P-value <0.0001 r -0.70 0.25 NS -0.55 0.39 NS -0.72 0.0114 0.71 0.0005 -0.90 <0.0001 0.90 <0.0001 -0.70 0.0003 0.78 <0.0001 -0.92 <0.0001 0.93 <0.0001 -0.58 0.46 0.0408 -0.69 0.0007 0.83 <0.0001 -0.91 <0.0001 0.93 <0.0001 -0.47 0.39 NS -0.59 0.0064 0.81 <0.0001 -0.90 <0.0001 0.91 <0.0001 -0.62 P-value 0.0006 0.0007 0.0071 0.0378 0.0033 220 Figure S4.1 Images of the raw seed of Ervilha, PI527538, and the RILs selected for this study separated into fast-, medium-, and slow-cooking groups. 221 Figure S4.2 Images of the washed-drained canned samples for Ervilha and PI527538 after retort processing for 10, 15, 20, 30, and 45 minutes. 222 Figure S4.3 Boxplots of the CIELAB values for the raw seed of Ervilha, PI527538, and the RILs selected for this study separated into fast, medium, and slow cooking groups. Lines indicate Ervilha (yellow) and PI527538 (brown). Mean separation is indicated by the letters above each boxplot. 223 Figure S4.4 Pearson correlation matrix of retort time, cooking time, washed-drained weight, texture, intactness, and CIELAB color values (canned and raw) across all genotypes and retort times. Correlation coefficients are indicated in the lower left and represented by colored, directional ellipses in the upper right. P-values are indicated by asterisks, where *, **, and *** represent <0.05, <0.01, and <0.001 respectively. 224 REFERENCES 225 REFERENCES Bassett, A., and Cichy, K. (2020). QTL analysis of a yellow Phaseolus vulgaris recombinant inbred line population for a fast-cooking, flavorful, and flourishing future of dry beans. Annu. Rep. Bean Improv. Coop. 63, 95–96. Bornowski, N. (2018). QTL mapping of post-processing color retention and other traits in two black bean populations. Thesis, Michigan State University, East Lansing, MI (2018). Cichy, K. A., Fernandez, A., Kilian, A., Kelly, J. D., Galeano, C. H., Shaw, S., et al. (2014). QTL analysis of canning quality and color retention in black beans (Phaseolus vulgaris L.). Mol. Breed. 33, 139–154. doi:10.1007/s11032-013-9940-y. Cichy, K. A., Porch, T. G., Beaver, J. S., Cregan, P., Fourie, D., Glahn, R. P., et al. (2015a). A Phaseolus vulgaris diversity panel for Andean bean improvement. Crop Sci. 55, 2149–2160. doi:10.2135/cropsci2014.09.0653. Cichy, K. A., Wiesinger, J. A., and Mendoza, F. A. (2015b). Genetic diversity and genome-wide association analysis of cooking time in dry bean (Phaseolus vulgaris L.). Theor. Appl. Genet. 128, 1555–1567. doi:10.1007/s00122-015-2531-z. Coelho, C. M. M., de Mattos Bellato, C., Santos, J. C. P., Ortega, E. M. M., and Tsai, S. M. (2007). Effect of phytate and storage conditions on the development of the ‘hard-to-cook’ phenomenon in common beans. J. Sci. Food Agric. 87, 1237–1243. doi:10.1002/jsfa.2822. Davis, D. R., Twogood, M. L., and Black, K. R. (1980). Effect of blanch treatment on quality attributes of canned dry pinto and small and large lima beans. J. Food Sci. 45, 817–820. doi:0022- 1147/80/0004-0817$02.25/0. Deshpande, S. S., Sathe, S. K., Salunkhe, D. K., and Rackis, J. J. (1984). Dry beans of Phaseolus: A review. Part 3. C R C Crit. Rev. Food Sci. Nutr. 21, 137–195. doi:10.1080/10408398409527399. Dhurandhar, N. V., and Chang, K. C. (1990). Effect of cooking on firmness, trypsin inhibitors, lectins and cystine/cysteine content of navy and red kidney beans (Phaseolus vulgaris). J. Food Sci. 55, 470–474. doi:10.1111/j.1365-2621.1990.tb06789.x. Featherstone, S. (2015a). “Energy supply and requirements,” in A Complete Course in Canning and Related Processes Volume 1: Fundamental Information on Canning, 14th edn, ed. by Featherstone, S. Woodhead Publishing, Oxford, 129–148. doi:10.1016/B978-0-85709-677- 7.00007-4. Featherstone, S. (2015b). Optimising retort operations for canned goods. A Complete Course in Canning and Related Processes Volume 2: Microbiology, Packaging, HACCP and Ingredients, 226 - Food Sci. 50, LWT Technol. observations. 14th edn, ed. by Featherstone, S. Woodhead Publishing, Oxford, 59–71. doi:10.1016/b978-0- 85709-678-4.00004-x. Ghasemlou, M., Gharibzahedi, S. M. T., and Emam-Djomeh, Z. (2013). Relating consumer preferences to textural attributes of cooked beans: Development of an industrial protocol and microstructural 88–98. doi:10.1016/j.lwt.2012.06.018. Hernandez-Unzon, H. Y., and Ortega-Delgado, M. L. (1989). Phytic acid in stored common bean seeds (Phaseolus vulgaris L.). Plant Foods Hum. Nutr., 209–221. Hornick, B. A., and Weiss, L. (2011). Comparative nutrient analysis of commonly consumed vegetables. Nutr. Today 46, 130–137. Hosfield, G. L. (1991). Genetic control of production and food quality factors in dry bean. Food Technol. 45, 99–103. Hosfield, G. L., Ghaderi, A., and Uebersax, M. A. (1984). A factor analysis of yield and sensory and physico-chemical data from tests used to measure culinary quality in dry edible beans. Can. J. Plant Sci. 64, 285–294. Hosfield, G. L., Lansing, E., and Kelly, J. D. (1995). Eight small-red dry bean germplasm lines with upright architecture, narrow profile, and short vine growth habit. HortScience 30, 1479–1482. Hosfield, G. L., and Uebersax, M. A. (1980). Variability in physico-chemical properties and nutritional components of tropical and domestic dry bean germplasm. J. Am. Soc. Hortic. Sci. 105, 246–252. Available at: https://www.cabdirect.org/cabdirect/abstract/19801688890. Junek, J. J., Sistrunk, W. A., and Neely, M. B. (1980). Influence of processing methodology on quality of canned dry beans. J. Food Sci. 45, 821–824. Katuuramu, D. N., Luyima, G. B., Nkalubo, S. T., Wiesinger, J. A., Kelly, J. D., and Cichy, K. A. (2020). On-farm multi-location evaluation of genotype by environment interactions for seed yield and cooking time in common bean. Sci. Rep. 10, 3628. doi:10.1038/s41598-020-60087-2. Kelly, J. D., and Cichy, K. A. (2012). Dry bean breeding and production technologies. Dry Beans Pulses Prod. Process. Nutr., 23–54. doi:10.1002/9781118448298.ch2. Lajolo, F. M., and Genovese, M. I. (2002). Nutritional significance of lectins and enzyme inhibitors from legumes. J. Agric. Food Chem. 50, 6592–6598. doi:10.1021/jf020191k. Liener, I. (1979). Significance for humans of biologically active factors in soybeans and other food legumes. J. Am. Oil Chem. Soc. 56, 121–129. 227 Liu, K., Bourne, D. M. C., and Bourne, M. C. (1995). Cellular, biological, and physicochemical basis for the hard‐to‐cook defect in legume seeds. Crit. Rev. Food Sci. Nutr. 35, 263–298. doi:10.1080/10408399509527702. Long, Y., Bassett, A., Cichy, K., Thompson, A., and Morris, D. (2019). “Bean split ratio for dry bean canning quality and variety analysis,” in The IEEE Conference on Computer Vision and Pattern Recognition (CVPR) Workshops. Matella, N. J., Mishra, D. K., and Dolan, K. D. (2013). “Hydration, blanching and thermal processing of dry beans,” in Dry Beans and Pulses Production, Processing and Nutrition, eds. M. Siddiq and M. A. Uebersax (Oxford, UK: Blackwell Publishing Ltd.), 129–154. doi:10.1002/9781118448298. Mendoza, F. A., Kelly, J. D., and Cichy, K. A. (2017). Automated prediction of sensory scores for color and appearance in canned black beans (Phaseolus vulgaris L.) using machine vision. Int. J. Food Prop. 20, 83–99. doi:10.1080/10942912.2015.1136939. Nciri, N., and Cho, N. (2018). New research highlights: Impact of chronic ingestion of white kidney beans (Phaseolus vulgaris L. var. Beldia) on small-intestinal disaccharidase activity in Wistar rats. Toxicol. Reports 5, 46–55. doi:10.1016/j.toxrep.2017.12.016. Nordstrom, C. L., and Sistrunk, W. A. (1977). Effect of type of bean, soak time, canning media and storage time on quality attributes and nutritional value of canned dry beans. J. Food Sci. 42, 795–798. Nordstrom, L., and Sistrunk, W. A. (1979). Effect of type of bean, moisture level, blanch treatment and storage time on quality attributes and nutrient content of canned dry beans. J. Food Sci. 44, 392–395. Povey, R., Conner, M., Sparks, P., James, R., and Shepherd, R. (1998). Interpretation of healthy and unhealthy eating and implication for dietary change. Health Educ. Res. 13, 171–83. doi:10.1016/S1361-3111(98)80053-8. Santoro, N., Cantu, S. L., Tornqvist, C. E., Falbel, T. G., Bolivar, J. L., Patterson, S. E., et al. (2010). A high-throughput platform for screening milligram quantities of plant biomass for lignocellulose digestibility. Bioenergy Res. 3, 93–102. doi:10.1007/s12155-009-9074-6. Thompson, H. J. (2019). Improving human dietary choices through understanding of the tolerance and constituents. Curr. Opin. Food Sci. 30, 93–97. doi:10.1016/j.cofs.2019.01.001. Wang, N., and Daun, J. K. (2005). Determination of cooking times of pulses using an automated Mattson cooker apparatus. J. Sci. Food Agric. 85, 1631–1635. doi:10.1002/jsfa.2134. toxicity of pulse crop 228 Winham, D., Tisue, M., Palmer, S., Cichy, K., and Shelley, M. (2019). Dry bean preferences and attitudes among Midwest Hispanic and non-Hispanic White women. Nutrients 11, 178. doi:10.3390/nu11010178. 229 SUMMARY AND CONCLUSIONS Cooking time, flavor, and texture are important consumer-valued traits that contribute to consumer purchasing decisions. Incorporating these traits into breeding programs will expand appeal of beans to consumers that are deterred by the long cooking times and undesirable flavor and texture present in dry beans as well as contribute to success of bean products in new markets. These studies explored genetic variability and the mechanism of cooking time as it relates to the seed coat and cell wall, identified genomic loci relevant for cooking time and sensory attributes using quantitative genetics approaches, and determined the relevance of cooking time to the canning industry. Chapter 1 evaluated cooking time, pre-soaking time, physical traits, and cell wall and seed coat compositional traits across four seed types of dry beans. The relationships among cooking time and these attributes suggest that cooking time of unsoaked and pre-soaked beans are controlled by different mechanisms. Cooking time of pre-soaked beans was associated with seed weight, cotyledon/seed coat percent, cotyledon cell wall thickness, insoluble cell wall isolate, and total and insoluble whole seed dietary fiber. Previous studies have associated these traits with cell separation, water uptake, and water transport during cooking. Cooking time of unsoaked beans was associated with thicknesses of seed coat layers. These traits also affect water uptake and transport, but at an earlier stage in the hydration process. Both seed coat and cotyledon cell wall traits have been previously associated with cooking time, but only in the context of hardshell and the hard-to-cook phenomenon. This work revealed that genetic variability for these traits contributes to genetic variability for cooking time outside the context of textural defects. Understanding the factors associated with genetic variability for cooking time in unsoaked and 230 pre-soaked beans can help direct progress in breeding fast-cooking beans as well as reveal potential consequences of faster-cooking germplasm, including trade-offs like reduced fiber or seed coat integrity. Chapter 2 lays a foundation for incorporating sensory attributes into dry bean breeding programs and contributes to the limited genetic resources available for breeding fast cooking beans. Broad ranges of sensory attribute intensities and cooking times were observed both across and within seed types, revealing a lack of uniformity within seed type, but also a wealth of genetic variability for sensory quality and cooking time. This genetic variability can be harnessed to improve cooking time in new varieties and target specific sensory profiles to be defined according to consumer preference for each seed type. Limited correlations were observed among sensory attributes and cooking time, indicating that they can combine in multiple ways with limited effort required to break undesirable linkages. The modified QDA approach used to screen materials and the significant genetic SNPs identified for flavor, texture, and cooking time could allow breeders to improve agronomic traits without sacrificing desirable sensory quality and cooking time. The set of genotypes exhibiting extreme sensory attribute intensities identified during this study can be used for panel training as well as future work exploring sensory attributes and consumer preference. Improving flavor, texture, and cooking time in dry beans can ensure they are appreciated as a delicious and tasteful component of a healthful diet in all the versatile ways consumers choose to eat them. Chapter 3 further adds to the currently limited pool of resources available for dry bean breeders to target fast cooking time, flavor, and texture in their breeding programs. This chapter also highlights the potential for yellow beans, particularly the Manteca seed type, to deliver desirable traits to consumers in an easily identifiable package. The QTL identified can be used to 231 develop molecular markers for the incorporation of fast cooking time and desired sensory attribute intensities into new bean varieties. Yellow beans may appeal to USA consumers, who are seeking bean products with improved culinary characteristics and unique appearance. With the recent increased interest in plant-based proteins, now is an opportune time to address consumer preference in dry beans to remain competitive with other pulses, and yellow beans might be an ideal vehicle to a fast-cooking, flavorful, and flourishing future of dry beans. Chapter 4 identified the relationship between cooking time and canning quality. Fast- cooking genotypes were found to benefit from shorter retort processing times, allowing for evaluation of their optimal canning quality. Current processing times used to evaluate germplasm are biased toward slow-cooking genotypes, as they overcook fast-cooking genotypes, causing reduced canning quality. By reducing retort processing time, the bias toward slow-cooking beans can be mitigated, allowing future variety releases to maintain both fast-cooking time and high canning quality. Reduced processing times could benefit the canning industry, which may value the reduced time and energy required to prepare fast-cooking beans, as energy is a major expanse for canning facilities. This may also benefit growers, as they would have more options when selecting varieties to grow while still accounting for canning quality standards. Understanding the relationship between cooking time and canning quality will allow dry bean breeders to better meet the needs of both consumers and the canning industry through the development of varieties that are convenient and cost efficient to prepare both in the kitchen and in the can. Ultimately, this work aims to support breeders and researchers in their goals to increase bean consumption and ensure this nutritious crop is accessible to a growing global population. The genetic resources provided by this work will be useful to breeders as they develop new varieties with a focus on cooking time and sensory quality. 232