QTL MAPPING OF POST - PROCESSING COLOR RETENTION AND OTHER TRAITS IN TWO BLACK BEAN POPULATIONS By Nolan M. Bornowski A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Plant Breeding, Genetics and Biotechnology Crop and Soil Sciences Master of Science 2018 ABSTRACT QTL MAPPING OF POST - PROCESSING COLOR RETENTION AND OTHER TRAITS IN TWO BLACK BEAN POPULATIONS By Nolan M. Bornowski When black beans are processed for consumption, they can lose their dark coloration due to the leaching of water - soluble pigments called anthocyanins. After hydrothermal processing, beans are commonly a faded brown color instead of t he dark black color typical of the dry seed. The aim of this research was to develop mapping populations with different genetic sources of color retention in order to identify regions of the dry bean genome associated with canning quality traits. To this e nd, two half - sibling black bean recombinant inbred line (RIL) populations segregating for post - processing color retention were developed. These RIL populations were phenotyped for canning quality over two years and genotyped using the BARCBean6k_3 BeadChip . A novel phenotyping method using digital image analysis was shown to outperform current methods of quantitative color measurement. QTL for post - processing color retention were detected on six chromosomes, with QTL on Pv03, Pv08, and Pv11 being the most n otable for their co - localization with QTL for quantitative measurements of color. In particular, QTL associated with color retention on Pv11 mapped to a very small physical interval and were consistent across years, populations, and phenotyping methodologi es. C olor retention QTL on Pv08 and Pv11 are good candidates for development of molecular markers that may be used in marker assisted selection (MAS) or early - generation screening to improve post - processing color retention in black beans. iii This thesis is dedicated to Aaron Charles Rodgers . iv ACKNOWLEDGEMENTS First, I want to thank Dr. James Kell y for sharing his knowledge of plant breeding and his passion for dry bean research . I also want to thank my fellow graduate students for providing a great environment in which to learn, live, and love. In addition, thanks go out to my family members, who are incredible advocates of science education. Finally, I want to thank Amber Bassett for her emotional support, scientific discussion, and delicious cooking. v TABLE OF CONTENT S LIST OF TABLES ................................ ................................ ................................ ..................... vi i i LIST OF FIGURE S ................................ ................................ ................................ ..................... i x CHAPTER 1: LITERATURE REVIEW ................................ ................................ .................... 1 INTRODUCTION ................................ ................................ ................................ .............. 1 DRY BEAN PRODUCTION ................................ ................................ ............................ 1 DRY BEAN CONSUMPTION TRENDS ................................ ................................ ........ 2 HEALTH BENEFITS ................................ ................................ ................................ ........ 3 FLAVONOID BIOSYNTHESIS ................................ ................................ ...................... 4 PHASEOLUS MOLECULAR TOOLS ................................ ................................ ............ 8 CANNING QUALITY TRAITS ................................ ................................ ....................... 8 Quantitative Color Measurements ................................ ................................ ....... 9 Breeding for Canning Quality Traits ................................ ................................ . 10 Canning Quality Research ................................ ................................ .................. 11 CONCLUSIONS ................................ ................................ ................................ .............. 13 CHAPTER 2: QTL MAPPING OF POST - PROCESSING COLOR RETENTION AND OTHER TRAITS IN TWO BLACK BEAN POPULATIONS ................................ ...... 1 4 ABSTRACT ................................ ................................ ................................ ...................... 1 4 INTRODUCTION ................................ ................................ ................................ ............ 1 5 MATERIALS AND METHODS ................................ ................................ .................... 1 6 Plant Materials ................................ ................................ ................................ ..... 1 6 Parental Germplasm ................................ ................................ .................. 1 6 Population Development ................................ ................................ ............ 1 7 2016 Field Season ................................ ................................ ...................... 1 7 2017 Field Season ................................ ................................ ...................... 1 8 P henotyping Canning Quality Traits ................................ ................................ . 19 Sample Preparation ................................ ................................ ................... 19 Canning Protocol ................................ ................................ ....................... 19 Reviewer Evaluation ................................ ................................ .................. 2 0 Machine Phenotyping ................................ ................................ ................ 21 Digital Imaging ................................ ................................ .............. 2 1 Other Measurements ................................ ................................ ...... 2 2 Phenotyping Agronomic Traits ................................ ................................ .......... 2 3 Statistical Analysis ................................ ................................ ............................... 2 5 Canning Traits ................................ ................................ ........................... 2 5 Agronomic Traits ................................ ................................ ....................... 2 6 Genotyping ................................ ................................ ................................ ............ 2 7 Linkage Map Construction and QTL Mapping ................................ ................ 2 8 Molecular Marker Analysis ................................ ................................ ................ 29 RESULTS ................................ ................................ ................................ ......................... 3 0 vi Statistical Analysis ................................ ................................ ............................... 3 0 Comparing Methods of Quantitative Color Measurement ........................ 3 0 Agronomic Traits ................................ ................................ ....................... 3 1 QTL M apping of Post - Processing Color Retention ................................ .......... 3 2 C olor Ratings ................................ ................................ ............................. 32 L * ................................ ................................ ................................ ............... 34 a* ................................ ................................ ................................ ................ 37 b* ................................ ................................ ................................ ................ 38 QTL Mapping of Other C anning Quality Traits ................................ .............. 4 0 Appearance Ratings ................................ ................................ ................... 40 Texture ................................ ................................ ................................ ....... 42 Washed and Drained Weight ................................ ................................ ..... 43 QTL M apping of Agronomic Traits ................................ ................................ ... 4 3 S eed Y i eld ................................ ................................ ................................ ... 44 S eed W eight ................................ ................................ ................................ 44 D ays to F lowering ................................ ................................ ...................... 45 D ays to M aturity ................................ ................................ ........................ 46 Canopy Height ................................ ................................ ........................... 47 Desirability Score ................................ ................................ ...................... 48 Ozone Bronzing ................................ ................................ .......................... 49 Molecular Marker Analysis ................................ ................................ ................ 5 0 DISCUSSION ................................ ................................ ................................ ................... 5 1 Comparing Methods of Quantitative Color Measurement .............................. 5 2 Heritability Estimates of Canning Quality Traits ................................ ............ 5 5 QTL Nomenclature ................................ ................................ .............................. 5 6 QTL M apping of Post - Processing Color Retention ................................ .......... 57 Color Ratings ................................ ................................ ............................. 57 L* ................................ ................................ ................................ ............... 63 a* ................................ ................................ ................................ ................ 65 b* ................................ ................................ ................................ ................ 68 QTL M apping of Other Canning Quality Traits ................................ .............. 69 Appearance Ratings ................................ ................................ ................... 70 Texture ................................ ................................ ................................ ....... 73 Washed and Drained Weight ................................ ................................ ..... 74 QTL Mapping of Agronomic Traits ................................ ................................ ... 7 4 Seed Yield ................................ ................................ ................................ ... 75 Seed Weight ................................ ................................ ................................ 75 Days to Flowering ................................ ................................ ...................... 77 Days to Maturity ................................ ................................ ........................ 78 Canopy Height ................................ ................................ ........................... 79 Desirability Score ................................ ................................ ...................... 80 Ozone Bronzing ................................ ................................ .......................... 81 Common Bacterial Blight ................................ ................................ .......... 83 Molecular Marker Analysis ................................ ................................ ................ 8 4 CONCLUSIONS ................................ ................................ ................................ .............. 8 5 vii APPENDIX ................................ ................................ ................................ ................................ ... 87 REFERENCES ................................ ................................ ................................ ........................... 13 1 viii LIST OF TABLES Table 1 . Agronomic and canning quality traits guiding parental selection for black bean RIL populations ................................ ................................ .............. 88 Table 2 . Phenotypic variation in canning quality and agronomic traits for Population 76 (B14311/Zenith) ................................ ................................ ............. 89 Table 3 . Phenotypic variation in canning quality and agronomic traits for Population 86 (B14311/B12724) ................................ ................................ ........... 90 Table 4 . Correlation matrix for canning quality traits in two black bean RIL populations ................................ ................................ ........ 9 1 Table 5 . Correlation matrix for agronomic and selected canning quality traits in two black bean RIL populations ................................ ................................ ........ 9 2 Table 6 . QTL for measurements of post - processing color retention in two black bean RIL populations ................................ ................................ ........ 9 3 Table 7 . QTL for canning quality traits in two black bean RIL populations ..................... 10 0 Table 8 . QTL for agronomic traits in two black bean RIL populations ............................. 10 2 ix LIST OF FIGURES Figure 1 . US production of black beans for the year 2016 ................................ .................. 1 0 5 Figure 2 . US per capita consumption of P. vulgaris dry beans ................................ ........... 1 06 Figure 3 . US per capita consumption of black beans ................................ .......................... 1 07 Figure 4. Canning quality evaluation guidelines ................................ ................................ . 1 08 Figure 5 . Distribution of color ratings for Population 76 (B14311/Zenith) and Population 86 (B14311/B12724) ............... 1 09 Figure 6. Distribution of CIELAB color values in Population 76 (B14311/Zenith) ........... 11 0 Figure 7. Distribution of CIELAB color values in Population 86 (B14311/B12724) ........ 11 1 Figure 8. Distribution of canning quality traits for Population 76 (B14311/Zenith) and Population 86 (B14311/B12724) ............... 11 2 Figure 9. Heritabi lity estimates of canning quality traits in two black bean RIL populations ................................ ................................ ...... 11 3 Figure 10. Distribution of agronomic traits in two black bean RIL populations .................. 1 1 4 Figure 11. Regression of color components and mean visual rating of canned color ........... 1 16 Figure 12. QTL graphs for Population 76 (B14311/Zenith) ................................ ................. 1 17 Figure 13. QTL graphs for Population 86 (B14311/B12724) ................................ ............... 1 24 Figure 14. Screening parents of Populations 76 and 86 with NDSU InDel markers near major color retention QTL on Pv08 and Pv11 ................................ ............. 1 30 1 CHAPTER 1: LITERATURE REVIEW INTRODUCTION Dry edible beans ( Phaseolus vulgaris L.) provide an economical, nutritious food source for millions of people around the world and exhibit a wide diversity of se ed sizes, colors, shapes, and agronomic traits. For US consumers, dry beans are commonly consumed as already - prepared canned products. Dry beans belonging to the black bean market class are unique in that they exhibit significant loss of color during the c anning process, which is attributed to leaching of the anthocyanin pigments that give black beans their characteristic coloration. Color loss during processing may result in a faded brown product that is undesirable to many consumers . Post - processing color retention and other quality traits are important when breeding black bean varieties. This breeding process requires significant time, resource, and labor inputs to develop new lines and evaluate them for quality traits. In order to further improve post - processing color retention of black beans, more information is needed concerning the genetic mechanisms underlying this trait . This genetic mapping study will identify genomic regions associated with black bean color retention that breeders and resear chers can use to meet consumer standards . DRY BEAN PRODUCTION In the United States, dry beans are categorized into commercial classes based on seed characteristics. Commercially - produced beans are generally grown under non - irrigated conditions in the nor th central region of the US, with North Dakota, Michigan, Minnesota, and Nebraska (USDA - NASS, 2016) . US - grown dry beans are mostly purchased by domestic consumers , but approximately 3 0 percent is 2 ex ported annually to countries like Mexico, Canada, the United Kingdom , and Italy, among others (Parr et al., 2018; USDA - NASS, 2018) first dry bean breeding program was founded at Michigan State College, now Michigan State University. Subsequent research efforts generated improved varieties for gro wers within and dry bean production and is the leading producer of small red, cranberry, and black beans (USDA - NASS, 2016) . Black beans are especially important to Michiga n, where approximately half of US black beans are produced annually [ Figure 1 ]. DRY BEAN CONSUMPTION TRENDS US consumption of Phaseolus vulgaris dry edible beans has fluctuated around 6 pounds per capita for several decades. Market class preference among consumers is relatively stable except for navy beans, which have gradually declined in popularity over the years [ Figure 2 ]. Pinto beans, on the other hand, remain the most consumed market class, largely due their presence in both home , institutional , and commercial dishes such as soups, salads, and refried beans (Lucier et al., 2000) . While other market classes have plateaued or declined, black bean consumption has been increasing exponentially since the 1980s [ Figure 3 ], providing an economic justification for quality improvement in processed black beans. Likewise, other grain legumes such as c hickpeas ( Cicer arietinum ), also known as garbanzo beans , have rapidly increased in popularity. The rapid rise in chickpea consumption has prompted growers in north central states to increase planting acreage , mostly by replacing wheat (Bonds, 2017) . 3 Since b oth dry beans and chickpea s undergo similar processing prior to consumption, they share a similar marke t space and dietary function. Of the dry bean market classes, black bean consumption is increasing at a similar rate as chickpea consumption and may play a deciding role in the future of the dry bean industry by competing with other pulse crops for consumer demand. Unlike other dry bean market cl asses, black beans are uniquely susceptible to undesirable color leaching during processing. Genetic improvements in post - processing color retention and other quality traits of processed black beans will provid e an opportunity for black beans to maintain t heir place in future markets. HEALTH B ENEFITS Cooked dry beans are an excellent source of protein, fiber, and other minerals (Hornick and Weiss, 2011) , but black bean pigments are thought to bestow additional health benefits. These pigments belong to a class of flavonoids called anthocyanins that localize to the seed coat of black beans (Takeoka et al., 1997; Beninger and Hosfield, 2003) . Anthocyanins are considered as antioxidants that prevent reactive oxygen species (ROS) from causing damage to cell membranes (Miguel, 2011) . Antioxidant - rich foods are associa ted with lowered risk of cardiovascular disease and cancer (Arts and Hollman, 2005; Wang and Stoner, 2008) . While black beans can vary in amount of total phenolic compounds, the y generally contain more phenolic compounds than other market classes (Luthria and Pas tor - Corrales, 2006; Marles et al., 2010) . Studies by Oomah et al. ( 2005) and Akond et al. ( 2011) found that beans with high anthocyanin s and total phenolic compounds exhibited high antioxidant activity in vitro . Unfortunately , flavonoid antioxidant activity is greatly reduced after processing (Xu and Chang, 2009) . In fact, Lotito and Frei ( 2006) conclude that in vivo antioxidant activity is not due to flavonoids, but urate production attributed 4 to glucose consumption. However, flavonoids represent only a fraction of the nutraceutical potential of dry beans . Compared to cereal grains, d ry beans are an excellent source of resistant starch (Murphy et al., 2008) and fiber (USDA - Agricultural Research Service, 2018) , which result in a low d and metabolized more slowly. Among other claims, foods with low glycemic indices have been shown to help manage diabetes (Brand - Miller et al., 2003) (Goff et al., 2013) (Micha et al., 2010) . In addition, rodents fed milled samples of canned beans reduced their cancer incidence and tumor numbers (Thompson et al., 2009) , though there is scant evidence of anti - cancer properties in human models (Messina, 2014) . Although the health effects of black bean anthocyanins may be overstated, black beans (and dry beans in general) still provide an affordable source of other nutritional benefits to consume rs. In fact, Foyer et al. ( 2016) compr om ised human health, nutritional security an Therefore, i ncreas ing consumption of dry beans and other legumes could have a large positive impact on global public health. FLAVONOID B IOSYNTHESIS Dry bean seed colors and patterns result from various pigments located with in the seed coat. These pigments belong to a class of polyphenolic compounds called flavonoids. Flavonoids are secondary plant metabolites containing two phenyl groups commonly linked by a 3 - carbon oxygenated heterocycle (Bravo, 1998) . Within the flavonoids, several classes can be characterized according to modifications of the 3 - ring flavone backbone : flavanones, flavones, isoflavones, 5 fla van - 4 - ols, dihydroflavonones, and flavonols . Flavonoid biosynthesis [ Supplemental Figure 1 ] is initiated by two Co - enzyme A (CoA) conjugates derived from separate pathways: 4 - Coumaroyl - CoA (also known as p - Coumaroyl - CoA or 4 - Hydroxycinnamoyl - CoA) from the phenylpropanoid pathway and 3 units of malonyl - CoA that have been carboxylated from aceytl - CoA used in the Krebs Cycle. These molecules serve as substrate s for (n aringenin) chalcone synthase ( CHS ) to produce naringenin chalcone (tetrahydroxychalcone) in the first committed step of flavonoid biosynthesis (Martens et al., 2010) . The yellow - colored naringenin chalcone is then con verted into the flavanone naringenin either by a chalcone isomerase (CHI) or spontaneous cyclization at room temperature (Cheng et al., 2018) . Spontaneous cyclization of chalcones to flavanones (e.g. naringenin chalcone to naringenin) occurs at a much slower rate (Bednar and Hadcock, 1988) , yet may produce sufficient quantities of flavanone substrate for use in downstream pigment biosynthesis (Heller an d Forkmann, 1988) . Flavanones are involved in many branches of flavonoid biosynthesis. They can form isoflavones via isoflavone synthase (IFS, synonym 2HIS), flavones via flavone synthases (FSI and FSII), flavan - 4 - ols via dihydroflavonol 4 - reductase (DF R), or flavanonols via flavanone 3 - hydroxylase (F3H, synonym FHT) (Lepiniec et al., 2006). It should be noted that both flavanonols and their flavanone precursors can undergo B - - - resulting in the compounds diverging into different pathways. Flavononols, also referred to as 3 - OH - flavanones or dihydroflavonols, can be catalyzed either by flavonol synthase (FLS) to produce flavonols (3 - hydroxyflavones) or D FR to produce flavan - 3,4 - diols (leucoanthocyanins). These flavan - 3,4 - diols can be catalyzed either by anthocyanidin reductase, ANS (also known as leucocyanidin dioxygenase, LDOX) into 3 - OH - anthocyanidins or by leucoanthocyanin reductase (LAR) into flavan - 3 - ols (also called 6 flavanols). Flavan - 3 - ols like catechin and epicatechin are condensed into oligomers known as condensed tannins, or proanthocyanidins. Proanthocyanidins are colorless, but can be oxidized by polyphenol oxidase (PPO) to become yellow - brow n, or, as their name suggests, can become anthocyanidins via acid hydrolysis. Anthocyanidins are the aglycone form of anthocyanins. Once glycosylated (commonly via 3 - O - glycosylation), the anthocyanin compound increases in polarity , which allows it to be tr ansferred and stored in the vacuole , where it functions as a pigment visible in the seed coat (Corradini et al., 2011) . The se pigments are responsible for the coloration of dry bean seed s . The dark coloring of black beans is attributed to anthocyanins like delphinidin 3 - glucoside, petunidin 3 - glucoside, and malvidin 3 - glucoside (Takeoka et al., 1997; Beni nger and Hosfield, 2003) . In dry beans, some of the genes encoding flavonoid biosynthesis enzymes have been characterized and mapped. According to Feenstra ( 1960 ), Hosfield ( 2001 ), and Bassett ( 2007) , loci P, C, R, J, D, G, B, V, a nd Rk all contribute to seed coat color and often interact epistatically. Efforts by McClean et al. (2002) and Reinprecht et al. (2013) have gener ated markers and genomic positions for these loci and have associated color loci with flavonoid pathway enzymes in dry bean and soybean ( Glycine max ) . These loci controlling flavonoid biosynthesis are fundamental to the characteristic dark coloration of black bean seeds . H owever, it is unknown whether these loci controlling flavonoid production also have a role in post - processing color retention of black bean s . PHASEOLUS M OLECULAR T OOLS Dry bean breeders have been using molecular tool s for many years to indirectly select for disease resistance traits. Using a recombinant inbred population of a cross between BAT93, a Mesoamerican breeding line, and Jalo EEP558, an Andean landrace, a low - density linkage map 7 was created (Nodari et al., 1993a) . This map utilized early generation markers like isozymes, restricted fragment length polymorphisms (RFLPs) , and random amplified polymorphisms (RAPDs) and was integrated with other contemporary linkage maps (Freyre et al., 1998) . RAPD markers were later refined into sequence characterized amplified region (SCAR) markers , which were successfully used in marker assisted selection of major disease res istance loci (Miklas et al., 2000) . In order to improve genome co verage further, microsatellite markers detecting length polymorphisms in simple sequence repeats (SSRs) were developed and aligned to the genomic map (Blair et al., 2003; Grisi et al., 2007) . Despite the ease of using PCR - based markers, low genomic coverage prevented higher - resolution mapping. This obstacle was addressed by single nucleotide polymorphism (SNP) marker discovery resulting from the BeanCAP ( Common B ean C oor dinated A gricultural Pr oject) initiative (Hyten et al., 2010) . In addition to marker development, the BeanCAP also funded nutritional research and student training across public universities and government sectors (http://www.beanca p.org/). Another highly useful BeanCAP outcome was the selection of 5,398 polymorphic SNPs for development of the BARCBean6 k _3 BeadChip (Song et al., 2015) . SNPs selected for the BeadChip are distributed across all chromosomes and can be assigned physical positions according to the dry bean reference genome. Th e SNP chip can detect polymorphisms among and within market classes, aiding various genomic studies (Hoyos - Villegas et al., 2015; Zuiderveen et al., 2016) . Recently, g enomes for G19833, an Andean landrace, and BAT93 have been assembled, revealing indep endent domestication events and extensive synteny with soybean (Schmutz et al., 2014; Vlasova et al., 2016) . Assembly information for these genomes is publicly available online and will be useful in many future research projects. Sequencing costs have fallen since the 8 development of these reference genomes, improv ing access to sequence data for more targeted mapping and cloning experiments. First described by Elshire et al. (2011) , genotyping by sequencing (GBS) can identify thousands of SNPs that may be used for linkage map construction or aligned to a reference genome if desired. Several GBS protocols have recently been published for dry beans with differing restriction enzymes a nd experimental objectives (Zou et al., 2014; Hart and Griffiths, 2015; Ariani et al., 2016; Schröd er et al., 2016) . Further adaptation of bioinformatics and sequencing technologies by the dry bean community will facilitate many genome - scale studies . CANNING Q UALITY T RAITS Dry beans require hydrothermal processing to soften the cotyledons and inacti vate anti - nutritional factors before being consumed (Van Der Poel, 1990; Martínez - Manrique et al., 2011) . In a domestic setting, preparation may include a soaking step followed by heat treatment, either by boiling or pressure - cooking. However, for many consumers the convenience of canned beans is preferable to the long preparation time associated with soaking and cooking dry beans. As part of the industrial canning process, beans are cleaned, soaked in a sal t solution, quickly heated - sterilized under pressure (Matella et al., 2013) . Similar to boiling, canning also causes physico - chemical changes in the beans that influence culinary quality parameters (Wassimi et al., 1990) . Quality parameters like water absorption, texture, color, and appearance after canning vary among and within market classes due to both genetic and genotype by environment effects (Hosfield et al., 1984; Hosfield and Uebersax, 1990) . Washed and drained weight of the canned beans is useful for determining the degree of 9 hydration during the canning process. Texture measure ments of canned beans are based on the amount of force required to compre ss the cooked sample and serve as a proxy for mushiness or firmness of the beans. Color and appearance are typically subjective measurements of processed beans, where a panel of reviewers rate canned samples according to the degree of pigment leaching, seed coat integrity, and other visual characteristics. These measurements , collectively, indicate the quality of the canned product. Quantitative C olor M easurement s Color is commonly measured quantitatively according to CIELAB color space. CIELAB color sp ace is based - (Busse and Bäumer - Schleinkofer, 1996) , where perception of some colors excludes the perception of other colors. For example, humans cannot perceive - developed in 1976 by international collaboration as an improvement on previous color spaces such as CIEXYZ and Hunter LAB (CIE, 2008) and is widely - used across disciplines. In CIELAB color space , a single color is partitioned into three components : L * , a * , and b * [ Supplemental Figure 2 ] . L* measures darkness to lightness from 0 (black) to 100 (white) ; a* measures the level of greenness to redness and ranges from negative values (green) to positiv e values (red); b* measures the level of blueness to yellowness and ranges from negative values (blue) to positive values (yellow). Values of a* and b* near zero are a neutral grey. CIELAB measurements can be compared to other color spaces through transformation if desired (Hunter Laboratories, 1996) . As opposed to the qualitative measurements provided by the reviewer panel, CIELAB values can provide quantitative measurements of canned be an color. 10 Breeding for Canning Q uality T raits The aforementioned quality traits are distinct from agronomic traits and must be selected independently. For example, the color of unprocessed black beans is independent of the color of the canned beans (Ghaderi et al., 1984) . Unfortunately, because canning quality evaluatio n is performed in later generations when germplasm is mostly homozygous, many genotypes selected for favorable agronomic traits may be lacking in these quality traits. Breeding dry beans for canning quality is a time - consuming endeavor, as breeders generally use a pedigree breeding method to develop desirable progeny . The process begins when parents are manually crossed in the greenhouse to produce F 1 progeny. F 1 plants (and all suc ceeding generations) are then allowed to naturally self - pollinate. F 2 plants are grown in the field and harvested as single plant selections until the F 4 generation. Plants are phenotypically selected for highly heritable traits like seed color, size, and shape and agronomic characteristics like growth habit, maturity, and lodging (Kelly and Cichy, 2013) . To accelerate the process, early generations can be grown in greenhouses or warm - climate nurseries, but late - generation selection for more complex traits like yield and canning quality must be evaluated in the target environment. After further selection based on yield, canning quality, agronomic traits, and disease resistance, elite lines are sent to weste rn growers for seed increase in absence of disease pressure. Following several years of data analyses , a team of plant scientists, industry professionals, and administrators then decide if an elite line will be released as a variety (Kelly, 2010) . Because canning quality evaluation is typically performed in the F 6 generation and beyond, selection is limited to later generations when most loci are fixed and sufficient seed is a vailable for replicated field trials. Inbreeding and phenotyping are time - and resource - intensive processes that delay improvement of these traits. P henotyping d ry seed to predict canning quality t raits has been 11 explored with varying results (Mendoza et al., 2014, 2017, 2018) [Supple mental Table 1]. Alternatively, i dentification of linked molecular markers associated with major loci in fluencing processing traits may allow marker assisted selection on early generation material, saving time, resources , and labor. Canning Q uality R esearch Previous research suggests that c anning quality traits like color retention are quantitatively inherited, meaning they are controlled by more than one locus. To locate these genomic regions and their effects on traits of interest, QTL (quantitative trait loci) mapping studies have been de signed to associate genetic markers with phenotypic data (Collard et al., 2005) . Mapping populations can be developed from RIL, F 2 , backcross, or double d haploid populations , provided the populations are segregating for the trait(s) of intere st. Once phenotypic data is collected and populations are genotyped, QTL detection is possible when polymorphic molecular markers are associated with differences in the phenotypic data. Because r eproductive self - compatibility of dry beans allows an individ ual near - homozygous line to be tested over multiple years, RIL mapping populations are commonly used in dry bean genetic mapping studies. Significant genotype by environment interactions have been documented for canning quality traits in pinto (Ghaderi et al., 1984) , navy (Walters et al., 1997) , and black (Hosfield et al., 1984) bean market classes. In the Walters et al. ( 1997) study, navy bean parents with contrasting canned appearance were crossed to generate three RIL populations that were evaluated for canning traits. Small population sizes and use of RAPD markers available at the time limited marker - trait associations, but the authors estimated heritabili (0.59), texture (0.64), and washed - (0.67). 12 Posa - Macalincag et al. (2002) performed a QTL analysis on two kidney bean RIL populations derived from crosses between acceptable and excellent canning genotypes . The authors estimated narrow - sense heritability for canned appearance and degree of splitting to be approximately 0.84 and found high correlation between the two traits across environments (r = 0. 91 to 0 .97). The RAPD markers previously identified by (Walters et al., 1997) were not polymorphic in these populations, but others were associated with canned appearance and splitting, namely OP15.1150 on linkage group (LG) 1 ( anch ored to LG B8, now known as chromosome Pv08) and OG17.1300 on LG 2 (unanchored) . In black beans, Wright and Kelly ( 2011) used a RIL population to map agronomic and quality traits with SSR markers. Seven QTL were identified for post - processing color retention across 5 LGs. A ppearance QTL mapped to Pv05 and Pv08, texture to Pv06 and Pv11, and washed - drained weight to Pv03 and Pv10 , although few QTL were detected across multiple years. More recently, Cichy et al. ( 2014) developed a black bean RIL mapping population derived from crossing parents contrasting as shiny - seeded ( Asp ) and dull - seeded ( asp ). The RIL population was evaluated for canning traits and genotyped with SNP markers generated by DArT and DArTseq platforms (Diversity Arrays Technology, Yarralumla, Australia). SNP genotyping significantly increased the number of mapped markers (n=1449) and improved map resolution compared to previous studies. QTL for canned bean appearance, color retention, texture, and other quality traits were detected across the genome, with some traits exhibiting co - localization . Of note , QTL co - loca lized on Pv05 for anthocyanin content, L * , b * , and color retentio n; for Asp , water uptake , and texture on Pv07 ; and for L * , a * , b * , col or retention, and canned appearance on Pv11 . While Posa - Macalincag et al. (2002) and Wright and Kelly, (2011) both identified a QTL for canned 13 appearance on Pv08, no APP QTL was detected on Pv08 in the RIL population of Cichy et al. (2014) . CONCLUSIONS Black beans have dramatically increased in popularity over the past few y ears. They are healthful, affordable, and very convenient as a ready - to - eat canned product. However, after processing , black beans commonly become faded and lose their dark black coloring due to leaching of water - soluble anthocyanins. This color loss can b e extreme and is undesirable to consumers. Post - processing color retention can be improved through traditional breeding, but despite previous studies, information remains lacking on the genetic mechanisms controlling this trait. This study utilizes two bla ck bean RIL populations for QT L mapping of post - processing color retention. Genotypic data was collected using the BARCBean6 k _3 SNP chip. Several methods of measuring canned bean color retention were evaluated to guide future phenotyping efforts. Identifyi ng the genomic regions influencing color retention and other important traits will provide useful information to dry bean breeders and researchers seeking to create black bean varieties with improved canning quality. Quality gains will allow dry beans , and black beans in particular, to remain competitive in an evolving dietary landscape. 14 CHAPTER 2: QTL MAPPING OF POST - PROCESSING COLOR RET ENTION AND OTHER TRAITS IN TWO BLACK BEAN P OPULATIONS ABSTRACT When black beans are processed for consumption, they can lose their dark coloration due to the leaching of water - soluble pigments called anthocyanins. After hydrothermal processing, beans are commonly a faded brown color instead of the dark black color typical of the dry seed. Genotypes with superior post - pr ocessing color retention have been identified in the Michigan State University breeding program, providing an opportunity to study the genetics underlying this key trait. The aim of this research was to develop mapping populations with different genetic so urces of color retention in order to identify regions of the dry bean genome associated with canning quality traits. To this end, two half - sibling black bean recombinant inbred line (RIL) populations segregating for post - processing color retention were dev eloped. These RIL populations were phenotyped for canning quality over two years and genotyped using the BARCBean6k_3 BeadChip. Quantitative trait loci (QTL) governing color retention, other quality traits, and agronomic traits were identified and compared to previous studies. QTL for post - processing color retention were detected on six chromosomes, with QTL on Pv03, Pv08, and Pv11 being the most consistent across subjective and objective phenotyping methods. Color retention QTL on Pv03 were found at the pr oximal end of the chromosome near 2.2 Mb and explained a modest amount of phenotypic variation. The QTL on Pv08 had high LOD scores and explained a large amount of phenotypic variation, but mapped to a very large physical interval due to low marker coverage . Most encouraging, many QTL for color retention co - localized to a region near 52.5 Mb on Pv11 . This relatively tight physical interval explained a large amount of phenotypic variation ( R 2 15 and had a large effect size on post - processing color retention across population s, years, and methods of measurement . INTRODUCTION Black bean consumption in the US is growing exponentially, with consumers preferring a dark black color in cooked and canned black beans. Black beans typically lose their coloration during soaking and/or thermal processing, which is attributed to leaching of the water - soluble anthocyanin pigments in the seed coat into the soaking or cooking water. Once fully cooked, black beans may have l ost enough pigment ation that they turn a faded brown color that is undesirable to produce r of black beans [ Figure 1 ] , so it is economically important to stay at the forefront of black bean research and improvements. Because of this regional connection to black beans, post - canning color retention is a major breed ing priority for the dry bean breeding program at Michigan State University. Varieties released by the MSU dry bean breeding program need to meet classical agronomic parameters like yield, local adaptation, and disease resistance, but also need to meet qua lity standards. The black bean variety, Zorro, released by the program in 2009, is widely grown in Michigan and set higher standards for black bean color retention (Kelly et al., 2009) . However, the latest MSU varietal release, Zenith, has even better agronomics and color retention than Z orro (Kelly et al., 2015) . This study attempts to identify the genetic mechanis ms contributing to the excellent color retention exhibited by Zenith and breeding line B12724 through biparental QTL mapping. Both genotypes were crossed to a common parent with poor post - processing color retention (breeding line B14311) to develop two hal f - sib ling RIL mapping populations segregating for color retention. These populations were phenotyped for canning quality traits using traditional and novel methodologies and were 16 genotyped with the BARCBean6k_3 Illumina BeadChip (Song et al., 2015) . QTL were detected for color retention, additional quality traits, and agronomic traits. MATERIALS AND METHODS Plant M aterials Parental G ermplasm The various biparental black bean RIL populations developed in this study were derived from crosses between several parental black bean genotypes with contrasting post - processing color retention. MSU advanced breeding lines B14302, B14303, and B14311 demon strated high yields and acceptable agronomic traits when evaluated in yield trials, but showed similar, poor color retention [ Table 1 ] following - scale canning protocol (Hosfield and Uebersax, 1980) . Lines B14302 and B14303 were siblings, both resulting from the cross B09197/B11334, while line B14311 is derived from the cross B11338/B10241 [Supplementa l Figure 3 ] . The superior - canning parents in the RIL populations were the breeding line B12724 and t he commercial variety, Zenith also known as breeding line B10244. These superior - canning parents also performed well in yield trials and maintained their dark black color during canning evaluations [ Table 1 ]. Line B12724 is derive d from the cross B09184/B09135 [Supplemental Figure 4 ] , while Zenith is derived from the cross B04644/Zorro [Supplemental Figure 5 ] . Zorro was released in 200 9 and is currently widely - grown in Michigan . The variety has upright architecture , resistance to anthracnose race 7, common bacterial blight, and avoidance to white 17 mold. (Kelly et al., 2009) . Zenith, released in 2014, has higher yield, additional resistance to anthracnose race 73, and darker - colored canned seed com pared to Zorro (Kelly et al., 2015) . Population D evelopment Parental genotypes were crossed in the greenhouse in the spring of 2015. Eighty - seven crosses were made in total and assigned individual cross numbers 15B261 through 15B347. F 1 seeds were planted at a distance of 20 cm (8 in) at the Saginaw Valley Research and Extension Cen ter (SVREC) near Richville, MI on June 17, 2015. This planting distance was used to improve seed production and facilitate single plant selection at harvest . The soil at SVREC is classified as a Tappan - Londo loam with 0 - 3% slope. Standard agronomic practices were followed throughout the growing season. Individual F 1 plants were harvested separately. F 2 seed from each plant was collected , labeled with a unique accession number and weighed. 150 randomly - selecte d F 2 seeds from accession numbers 15A1005, 15A1011, 15A1031, 15A1045, 15A1076, and 15A1086 were planted in the greenhouse in October 2015. The resulting RIL populations were named according to the last two digits of the accession number (i.e. Population 5 was derived from 15A1005). F 2:3 seeds were harvested from individual plants, stored in coin envelopes, and planted in the greenhouse in February 2016. F 3:4 seeds were harvested from individual plants in May 2016, stored in coin envelopes, and assigned uniq ue numbers for field planting. 2016 F ield S eason F 3:4 seeds were planted at a distance of 20 cm (8 in) in separate rows at SVREC on June 3, 2016 to produce enough seed for canning evaluation and future field trials . Standard agronomic practices were foll owed throughout the growing season to manage weed, insect, and disease 18 pressure. Weather data for the 2016 field season can be found in Supplemental Figure 6 A. No desiccants were applied prior to harvest. At harvest, each row of F 3:4 plants was hand - pulled and threshed separately. F 3:5 seed was collected , labeled , and all samples were opened and stored on an open - air drying rack before being measured for moisture content , weigh ed, and canned. 2017 Field Season Remnant F 3:5 seed from the 2016 fi eld season was used as a seed source for planting at SVREC on June 2 nd 2017. Five RIL populations were planted as separate field experiments (7103 through 7107) with alpha lattice designs and included the parental genotypes, varietal checks Eclipse, Zorro, and Zenith, and selected MSU black bean breeding lines. RILs were planted in 4 - row plots with 50 cm ( 20 in) row spacing and were manually end - trimmed to a length of 4.5 m . The RIL Populations 76 and 86 were selected for genetic study and planted in two re plications, while the other RIL Populations 5, 11, 31, and 45 were entered into breeding program trials. At planting, the field was sporadically littered with straw residue due to the previous wheat crop being disk - ripped in the fall of 2016. Standard agro nomic practices were followed throughout the growing season to manage weed, insect, and disease pressure . Weather data for the 201 7 field season can be found in Supplemental Figure 6 B. No desiccants were applied prior to harvest. At harvest maturity, the c enter two rows of each plot were harvested with a Wintersteiger Classic plot combine (Wintersteiger, Salt Lake City, UT). F 3:6 seed from each RIL plot was collected and labeled with a breeding line designation based on the identification numbers from the previous growing season . The samples were stored on an open - air drying rack before being measured for moisture content , weighed, and canned . Remnant RIL samples from both the greenhouse and field were stored at the MSU Agronomy Farm. 19 Phenotyping C anning Q u ality T raits Canning quality traits were evaluated over two years. F 3:5 RILs grown at SVREC over the 2016 field season were evaluated in winter 2017, while F 3:6 RILs grown at SVREC over the 2017 field season where evaluated in winter 2018. Canning traits evaluated included: canned color rating (COL), canned appearance rating (APP), washed and drained weight (WDW), texture (TXT) , and values of L * , a * , and b * in CIELAB color space . Sample Preparation A subsample of field - harvested seed was hand - cleaned to remove off - types, split seeds, and debris. This subsample was stored and labeled with the RIL identity (unique breeding line number) and a three - digit can code used to identify canning samples. The samples were temporarily stored in a humi difying chamber to raise the moisture content to approximately 12 - 15%. Seeds were measured for % moisture in order to calculate the amount of seeds representing 90 grams of dry weight using the formula : . Weighed canning samples were placed in mesh bags labeled with the - digit can code and fastened with twist ties. Approximately 30 small mesh bags were then placed in a larger mesh bag to faci litate transportation and simultaneous blanching. Canning Protocol After the 2016 and 2017 growing seasons, RILs were canned according to a protocol devised by Bush Brothers & Co. (Knoxville, TN) that differs from the traditional MSU small - scale canning p rotocol developed by Hosfield and Uebersax ( 1980) . The decision to use this new 20 canning protocol was based on anecdotal evidence that beans canned with the Bush Brothers & Co . protocol maintained better seed coat int egrity. Due to time and space limitations, mapping populations from both years were canned on separate days. Large mesh bags containing the canning sample mesh bags were blanched by submerging for 90 seconds in a 0 .03% granular anhydrous calcium chloride s olution heated to 95 ° C in a steam - heated stainless - steel kettle (Groen Mfg. Co, Chicago, IL). After blanching, beans were transferred from the small mesh bags into tin cans labeled with corresponding can code stickers on the bottom of the cans. Cans were then filled to the top with brine heated t o 95 ° C. Brine solution was comprised of 1.5% sugar, 1.25% sodium chloride, and 0 .03% granular anhydrous calcium chloride. Filled cans were placed on a 5.6 m metal - tiled conveyor belt moving at approximately 2.15 cm/s through an exhaust box heated to 95 ° C to facilitate water uptake and removal of air bubbles. Lids were placed on the cans and sealed using a Dixie Double Seamer (Dixie Canner Co, Athens, GA) and placed in a retort (Loveless Mfg. Co, Tulsa, OK). Cans were cooked in the retort at 120 ° C and 15 psi for 30 minutes. After 30 minutes, cold water was pumped into the retort for 15 minutes to cool the cans. Cans were removed from the retort, towel - dried, and stacked in tubs until opened for canning quality evaluation. Reviewer Evaluation Cans were equ ilibrated for at least two weeks before being opened for canning quality evaluation. Over the equilibration period, beans become increasingly hydrated as they equilibrate with the canning liquids. On the day of the evaluation, cans were opened and both bea ns and brine were poured into individual food trays that were arranged on stainless - steel tables in the material handling wing of the MSU Agronomy farm . Reviewers included graduate students, faculty, and 21 members of the dry bean industry. Reviewers were tra ined to evaluate canned beans according to separate 1 - 5 scales for color and appearance, using reference printouts [ Figure 4 ] and varietal checks to assign ratings to each sample. Color of the canned samples was rated on a 1 - 5 attri bute intensity scale, where 1 represented a light brown color and 5 represented a dark black color. Appearance of the canned samples was rated on a 1 - 5 scale largely according to seed coat integrity, but could also account for the amount of extruded starch , amount of clumping, and brine opacity. Therefore, a 1 on the appearance score represented completely split beans, while a 5 represented intact beans with minimal clumping. Both color and appearance ratings were averaged across reviewers within year f or each RIL. Statistical analysis of reviewer consistency was performed in SAS v9.4 (SAS Institute, Cary, NC). Machine Phenotyping Digital Imaging Following evaluation by the panelists, quantitative measurements of cann ed color and texture were obtained . Beans were transferred to plastic colanders and rinsed under cool water to remove brine and extruded starch from the seed coats. Next, beans were evenly distributed on a black plastic plate so that the black plastic was not visible. Beans were then photo graphed in a custom - built photobox containing a mounted digital camera centered between two fluorescent tube lamps, as described by Mendoza et al. ( 2017) . To minim ize effects of internal and external lighting, the box interior was painted matte black and the loading side was covered with a black foam flap. The digital camera was a Canon model EOS Rebel T3i single - lens reflex camera (Canon, Melville, NY). Fluorescent T4 lamps 45 cm long were mounted 35 cm above the imaging stage at a 45 - 22 degree angle and powered on at least 30 minutes before imaging. The camera was connected to a laptop (Latitude E5570 Series, Dell, Round Rock, TX) via USB, using the software EOS Utili ty version 2.1 (Canon) for remote shooting and setting configuration. Settings were: manual exposure, auto focusing, lens aperture of f = 5.6, shutter speed of 1/125, white balanced, and ISO 100. Images were saved according to can code identifiers in both .CR2 (raw) and .JPEG (large size, fine - quality) formats. A grey standard card with 18% reflectance (Kodak, Rochester, NY) and a Munsell color checker card (X - Rite, Grand Rapids, MI) were imaged before and after photographing canned beans. Other Measuremen ts After digital imaging, samples were transferred to plastic boats and weighed to determine washed and drained weights (WDW). Washed and drained weight is a weight measurement of canned beans once they have been rinsed under water to remove brine and the rinse water has drained off. Because an equal amount of bean dry matter was added to each can, this measurement is a proxy of water uptake that has occurred during the canning process. A Hunter Labscan XE spectrophotometer (Hunter Associates Laboratory Inc ., Reston, VA) was used to measure CIE L* , a* , and b* values of canning samples grow n in 2017. L* values measure darkness to lightness from 0 (black) to 100 (white) ; a* values measure the level of greenness to redness and range from negative values (green) to positive values (red); b* values measure the level of blueness to yellowness and range from negative values (blue) to positive values (yellow) [ Supplemental Figure 2 ]. Texture (TXT) was measured as the peak force required to compress a 100 g subsample of canned beans. M easurements on the RILs from the 2016 season were recorded by a Kramer Shear 23 Press (Food Technology Corp., Sterling, VA) and interpreted by visually estimating the peak force reading on a printout. M easurements on the RILs from the 2017 season were recorded by a TA.XTplus100 texture analyzer (Texture Technologies Corp., Hamilton, M A). Raw measurements from the Kramer Shear Press required multiplication by a 1.36 conversion factor ( http://arsftfbean.uprm.edu/bic/wp - content/uploads/2018/04/Bean_Processing.pdf ) to obtain true texture measurements that were then comparable to those from the texture analyzer . Digital images of the canned beans from both years were analyzed with ImageJ software (Schneider et al., 2012) . An experimental macro was developed to m easure CIELAB values from the digital images. As part of the macro, digital images of canned samples were loaded into the software and brightened by a constant gamma correction. Then, reflectance was minimized through noise reduction and images were partit ioned into L * , a * , and b* slices. The mean value of each slice was recorded to obtain L * , a * , and b* values in CIELAB color space. Color values from the Hunter Lab scan spectrophotometer are denoted as L * H , a * H , and b * H , while color values from digital images processed in ImageJ are denoted as L* I , a* I , and b* I . Phenotyping A gronomic T raits Agronomic traits were evaluated in replicated plots during the 2017 growing season at SVREC and included days to flowering, days to maturity, canopy height, lodging, foliar effects of common bacterial blight and ozone damage, and an overall desirability sc ore. After harvest, seed yield (SY) was taken for each RIL as the mean across field replications and seed weight (SW) was recorded as the mass of 100 randomly - selected seeds from each RIL. Days to flowering (DF) were visually estimated as the number of da ys after planting when 50% of a plot was flowering. Days to maturity (DM) were visually estimated as the number of 24 Lodging (LDG) was rated on a per - plot basis using a 1 - 5 scale fallen - trait was not included in QTL analysis. Canopy height (HT) was visually estimated in cm as an average distance from the soil surface to the top of the plants in a plot at harvest . An agronomic desirability score (DS) from 1 - 7 was assigned to each plot based on perceived agronomic potential, where a score of 1 indicated exceptionally poor field performan ce and a 7 indicated excellent field performance. The DS is used by the MSU dry bean breeding and genetics program to guide breeding decisions and is based on a combination of factors that are important to breeding potential: height, architecture, adapta bility, pod load, disease symptoms , and other subtle characteristics. Dry beans affected by ozone stress will sometimes develop bronze or brown patterning on the leaves. In 2017, this foliar bronzing (BRZ) was rated on a 1 - br throughout the plot . Common bacterial blight (CBB) is a disease caused by the fungus Xanthomonas axonopodis pv. phaseoli (Xap). The disease can infect both leaves and pods, but CBB resistance in this study was measured by visually rating each 2017 field plot for foliar symptoms on a 1 - 5 severity scale. A CBB rating of 1 signified no observable disease, while a rating of 5 signified severe foliar lesions widespread throughout the plot . 25 Sta tistical Analysis Canning Traits RILs from both populations segregated for canning quality according to evaluations on samples grown in 2016 and 2017. For Population 76, canned F 3:5 RILs (n=147) grown at SVREC in 2016 were evaluated by 16 reviewers on February 9, 2017, while canned F 3:6 RILs (n=147) grown at SVREC in 2017 were evaluated by 10 reviewers on February 12, 2018. There were eight reviewers in common across both years of evaluation for Population 76 . For Population 86, canned F 3:5 RILs (n=148) were evaluated by 14 reviewers on March 2, 2017, while canned F 3:6 RILs (n=147) were evaluated by 15 reviewers on February 14, 2018. There were 10 reviewers in common across both years of evaluation for Population 86. SAS v9.4 (SAS Institute, Cary, NC) was used for statistical analysis of canning quality traits. Trait data was input as individual reviewer ratings across all RILs and years. Boxplots were generated to visually confirm normality and homogeneity of variance . Several statistical models were tested, and the best was selected according to the lowest Akaike information criterion value (AIC). The final full model was: grand mean, RIL is a f the residual effect of the three - way interaction RIL*Reviewer*Year. Both populations were phenotyped for post - processing color - retention, canning quality, and agronomic traits [ Table 2 ; Table 3 ]. All color retention and canning quality traits were approximately normally distributed in both populations and across both years. For canned color rating, parental lines generally exhibit ed the most extreme phenotypes, though a few transgressive segregants were observed [ Figure 5 ] . Most RILs exhibited average ratings for both color and 26 appearance ratings . CIELAB color components were similarly distributed across both populations [ Figure 6 ; Figure 7 ]. Other canning quality traits like appearance rating, washed and drained weight, and texture were also approximately normal [ Figure 8 ] , although m easurements of washed and drained weights and texture on Population 76 RILs were noticeably lower in 2016 than in 2017 . Pearson correlation coefficients were calculated for all canning quality traits in both populations [ Table 4 ]. Variance components for broad sense heritabilit y estimates of color and appearance ratings were determined stepwise to minimize the confounding effect of Reviewer*Year interaction. First, the model was run with the Year effect excluded to derive least squ are means (LSmeans) incorporating a reviewer effect. LSmeans were then used in the model with the Reviewer effect excluded to derive LSmeans incorporating a year effect. This methodology is justified because reviewers did not give ratings to biological rep licates in any one year and reviewers were not always consistent across years. Broad sense heritabilities for other canning quality traits were determined by using the full model. Variances were estimated using the type 3 sums of squares method and used to calculate broad sense heritabilities (H 2 ) on an entry mean basis according to Fehr (1991) [ Figure 11 ]. Agronomic Traits Although the focus of this research primarily concerned canning quality, RILs were evaluated for various agronomic traits during the 2017 field season [ Table 2 ; Table 3 ]. Agronomic traits were generally normally distributed , although parental lines were similar or identical in several instances [ Figure 1 0 ]. SAS v9.4 (SAS Institute, Cary, NC) was used to determine Pearson correlation coefficients for agronomic traits [ Table 5 ]. 27 Genotypi ng RILs from Populations 76 and 86 were planted in the greenhouse in January 2017 from remnant 2016 F 3:5 seed. Six seeds per RIL were planted in a 20 cm (8 in) clay pot filled with potting soil. Young trifoliates from four plants per pot were pooled togeth er to represent a RIL genotype. Tissue was collected in duplicate; four small trifoliates were pooled in labeled 1.5 ml tubes, while four larger trifoliates were pooled in labeled 15 ml tubes as backup. Tissue samples were submerged in liquid nitrogen imme diately after harvesting and stored at - 80 C. Samples in 1.5 ml tubes were lyophilized using a VirTis Wizard 2.0 lyophilizer (SP Scientific , Stone Ridge, NY ). DNA was isolated with CTAB (hexadecyl - trimethyl - ammonium bromide) in April - May 2017 using a modif ied protocol of the MSU dry bean program, which is itself a modified protocol of Doyle and Doyle (1991) . DNA was suspended in 1.5 ml tubes containing T 10 E 1 buffer and RNase A and stored at - 20 C. Thawed samples were pipet - spectrophotometer (Denville Scientific , Holliston, MA ). A series of dilutions and quantifications - - well plates, with parental genotypes included on each plate. B ecause the population sizes were too large for every RIL to fit on three 96 - well plates, specific RILs were excluded on the basis of creating more normal distribution s of color, appearance, and texture traits from the 2017 canning evaluation on RILs grown in 2016 . After exclu sion , 140 RILs remained in Population 76 and 141 RILs remained in Population 86. Parental genotypes were included on each plate. The three plates were shipped to the USDA Beltsville Agricultural Research Center (BARC) in Beltsville, MD where they were genotyped for 5398 SNP markers from the BARCBean6k_3 BeadChip developed by Song et al. (2015) . 28 Linkage M ap C onstruction and QTL M apping Marker clustering was surveyed in GenomeStudio v2011.1 (Illumina Inc., San Diego, CA) and genotypic calls were exported in AB format. Markers were filtered to remove SNPs monomorphic for the parents and SNP IDs were assigned according to SNP location in the dry bean v2.1 re ference genome (Goodstein et al., 2012; Schmutz et al., 2014) . Locus files were created for each chromosome and imported into JoinMap 4.1 (Van Ooijen, 2011) for linkage map construction. Loci were filtered for segregation distortion and missing data. Questionable markers had chi - square values > 10 or 50% missing genotypic data, but were not always removed to improve marker coverage. The remaining markers were binned into linkage groups representing chromosomes using a LOD threshold. A LOD threshold of 3 was used for all linkage groups with the exception of two chromosomes from RIL Population 76 : Pv01 was grouped at LOD 2 to allow inclusion of markers from the proximal arm , and Pv10 which was grouped at LOD 2 to maximize number of mark er s , though only 5 were ultimately informative . Linkage maps for all chromosomes were generated using the maxim um likelihood mapping algorithm. Map order was optimized using simulated annealing with a chain length of 5000 and a burn - in of 50000. After the burn - in chain, 10 cycles of Monte Carlo Expectation Maximization (MC EM) with chain lengths of 5000 were perfor med to determine map distance over a period of 3 optimization rounds. After initial map construction, markers were examined according to the JoinMap parameters of plausible positions and nearest - neighbor stress and were removed if they had extreme values f or these criteria relative to other markers on the linkage group . Linkage maps were then compared to physical maps with marker order and position based on the v2.1 reference genome. Linkage maps using the fixed marker order from the v2.1 reference genome w ere ultimately selected for QTL mapping. 29 The program Windows QTL Cartographer v2.5 (Wang et al., 2012) was used to analyze marker - trait associations. Preliminary single marker analysis and interval mapping were preformed to identify regions of interest. Composite interval mapping (CIM) with forward and backward regression, background markers (n = 5), a window size of 10 cM, and a LOD threshold of 3 was used to refine regions of interest. These regions were refined again using a window size of 5 cM and a LOD threshold determined on a per - trait basis from 1000 permutations at a significance level of p = 0.05. QTL and marker positions were imported to MapChart (Voorrips, 2002) for graphical visualization. In total, Population 76 (B14311 x Zenith) included 656 SNP markers and 141 RILs, while Population 86 (B14311 x B12724) included 391 SNP markers and 140 RILs for linkage map construction and QTL mapping. However, both populatio ns contained markers that mapp ed to identical genetic positions, which lowered the number of informative markers to 527 in Population 76 and 307 in Population 86. Attempts to integrate linkage maps constructed from the two RIL populations were unsuccessful . Molecular M arker A nalysis Genomic regions significantly associated with QTL for post - processing color retention on Pv08 and Pv11 were further explored with insertion - deletion ( InDel ) markers. The InDel markers were developed by Mogh addam et al. (2014) at North Dakota State University (NDSU) and can be used across and within dry bean market classes. Marker names provide information regarding the university where they were developed (NDSU), type of marker (IND), chromosome number, a nd physical position in the dry bean v1 reference genome (e.g. NDSU_IND_11_47.0739 is located on chromosome Pv11 at 47.0739 Mb in the v1 genome). InDel markers were selected by 30 identifying BARCBean6k_3 SNPs that flanked the InDel s in the v1 protocol of Moghaddam et al. (2014) was followed: 3 minutes at 95 ° C for one cycle, 20 seconds at 95 ° C, 30 seconds at 55 C, an d 1 minute at 72 ° C for 45 cycles, then 10 minutes at 72 ° C for one cycle. PCR products were transferred to a 3% agarose gel containing 3% ethidium bromide that underwent gel electrophoresis in 1x TAE buffer. Bands were visualized under UV light and scored for each genotype. RESULTS Statistical Analysis Comparing M ethods of Q uantitative Color M easurement Both the Hunter Labscan XE spectrophotometer and ImageJ software were used to measure CIELAB color components of canned black beans. The spectrophotomet er was only used on canning samples from 2017 - grown RILs, but the ImageJ macro was applied to saved digital images collected from both years of canning evaluation. L * , a * , and b* color values were different across methods, which was expected because of the post - processing involved in the ImageJ analysis. However, while the methods gave different values, they were mostly consistent relative to each other (e.g. a sample with a high Labscan - derived L * H would typically have a high ImageJ - derived L* I ). Pearson correlation coefficients were generated for both methods to compare consistency across methods [ Table 4 ]. L * , a * , and b* color components measured by different methods were generally strongly positively correlated with each other (r > 0.50). Among the Hunter Lab scan 31 values, b * H - 0.8 2 to - 0.93 ), followed by L* H (r = - 0.73 to - 0.83) and then a * H - 0.7 3 ) . This trend held true across populations. Among the two - year average ImageJ values, b* I - 0.9 5 ), followed by L* I (r - 0.89 ), and then a* I - 0. 81 ). This trend was also consistent across populations. Simple linear regression was used to compare accuracy of color components derived from both methods with actual canned color ratings [ Figure 1 1 ]. Scatterplots of the regressions show that the ImageJ - derived color components ha d less variation and wer e more strongly correlated to visual color ratings. This was ob served across populations. Agronomic Traits In both populations, the most strongly - correlated agronomic traits were desirability score (DS) and canopy height (HT), with correlation coefficients of r = 0.73 and 0.7 8 in Population 76 and Population 86, respectively. Interestingly, DS was also moderately correlated with yield (r = 0.46 to 0.53) and days to maturity (r = 0.32 to 0.40 ). Correlations between these traits make sense, since the MSU breeding program attempts to select high - yielding, more upright plant types that are adapted to the Michigan environment. These results affirm that gains from selection can be Correlations between agronomic t raits and canning quality traits were much weaker. In Population 86, visual appearance ratings of canned beans (APP) were moderately negatively correlated to height (r = - 0.3 5 ), desirability score (r = - 0.3 3 ), yield (r = - 0.3 2 ), and days to maturity (r = - 0.29), which exemplifies the difficulty in breeding high - performing varieties with acceptable canned bean appearance. 32 QTL M apping of P ost - P rocessing C olor R etention A complete list of QTL for post - processing color retention is located in Table 6 . QTL graphs from Population 76 are located in Figure 1 2 , and QTL graphs from Population 86 are located in Figure 1 3 . Color Ratings (COL) QTL for visual color ratings of canned black beans (COL) were found on chromosomes Pv02, Pv03, Pv08, and Pv11 in Population 76 and on Pv03, Pv05, Pv08, Pv10, and Pv11 in Population 86 [ Table 6 ]. The B14311 parent contributed a negative effect on color rating in all instances, except for a single QTL found on Pv10 in Population 86. A QTL for color rating on Pv02 named COL2.1 76 was detected by color ratings for 2017 seed (COL_2017) and the two - ye ar average (COL_2YA) [ Table 6 ] . The QTL was within a 54 - 74 cM interval, with a peak LOD located at near 63 cM . The region associated with 2017 color rating explained 6% of the phenotypic variation and influenced color rating by 0.16. The region associated with the two - year average color rating explained just 5% of the phenotypic variation and influenced color ratings by 0.13. Although below the significance t hreshold, the color ratings from RILs grown in 2016 formed a peak slightly upstream of the COL2.1 76 (data not shown) . QTL were found on Pv03 that were unique to each population. In Population 76, a color rating QTL , COL3.1 76 , was detected over a 26.4 to 32 .8 cM interval using color ratings from 2016 (COL_2016) and the two - year average (COL_ 2YA) . Both COL_2016 and COL_2YA explained about 6% of the total phenotypic variation and affected color ratings by 0.15. In Population 86, two COL QTL were detected on separate chromosome arms. A 2017 color rating QTL , COL3.1 86 , mapped to the top of the chromosome from 0 - 1.5 cM (1.19 - 1.30 Mb), while a 2016 color rating 33 QTL , COL3.2 86 , mapped to the distal end of the chromosome from 102.4 - 121.7 cM (47.28 - 50.39 Mb) [ Table 6 ]. Both of these color QTL remained significant at permutation thresholds below LOD 3, explained approximately 4% of the phenotypic variation, and influenced color ratings by 0.16 and 0.13, respectively. A QTL for color rating on RILs grown in 2017 was found on Pv05 in Population 86 only (COL5.1 86 ) . This peak had an interval of 167.5 - 170.7 cM and was flanked by markers ss715645449 and ss715645459 (38.92 - 38.84 Mb). A peak for the two - year average color rating was just below the sign ificance threshold, but mapped to an adjacent interval from 170.9 - 171.6 cM. QTL for color rating were shared across years and populations despite poor marker coverage on Pv08 [ Table 6 ]. In Population 76, QTL COL8.1 76 was detected across all years of evaluation . Color ratings from 2016 seed (COL_2016) w ere significant across a large map interval from 16.8 - 62.3 cM and contained a peak at 41.55 cM that explained 16% of the phenotypic variation and altered color scores by 0.26. Color ratings from 2017 (COL_2016) and two - year average color ratings (COL_2YA) co - localized within this interval and a shared peak LOD position at 53.42 cM. Both COL_2016 and COL_2YA had peak LOD scores over 7 and affected color ratings by approxim ately 0.24 [ Table 6 ]. In Population 86, QTL COL8.1 86 was detected across all years of evaluation (COL_2016, COL_2017, and COL_2YA) . LOD scores were significant over a large mapping interval from 15. 3 - 60. 4 cM , ranged from 7 to 8, and were all closest to marker ss715647116 located at a physical position of 1.58 Mb. Each individual year explained 12 - 13% of the phenotypic variation and influenced color ratings by approximately 0.26. Although both populations had few markers on Pv10, a QTL for color rating in 2017 was detected in Population 86 (COL10.1 86 ) . The QTL was contained within a 69.6 - 82.1 cM interval 34 and flanked by markers ss715645524 (42.22 Mb) and ss715645501 (43.29 Mb). LOD p eaks for 2016 color rating and two - year average co lor rating were apparent in the same interval, but were below the significance threshold. Similar to Pv08, Pv11 also contained color QTL identified in both populations [ Table 6 ]. In Population 76, peak LOD scores reached 7.4 for 2 016 - grown RILs, 7.01 for the two - year average, and 4.8 for 2017 - grown RILs. The 2016 region (COL_2016) mapped to a 144.3 - 149.6 cM interval, while the other two regions (COL_2017 and COL_2YA) spanned a 149.6 - 150.6 cM interval slightly downstream. The 2016 r egion was most significant, explained 14% of the phenotypic variation, changed color scores by 0.25, and was flanked by markers with physical positions from 52.16 - 52.65 Mb. Since these regions were significant over several years and occupied a similar inte rval, they were considered a single QTL named COL11.1 76 . The yearly color ratings in Population 86 also co - localized and were also considered as one QTL named COL11. 1 86 . T hese regions (COL_2016, COL_2017, and COL_2YA) had higher LODs (>10), R 2 (0.19 - 0.22), and larger additive effects (0.31 - 0.32) than those detected in Population 76. COL11.1 86 mapped to a 26.4 - 28.6 cM interval and was flanked by markers ss715648350 and ss715640405 that have physical positions of 52.47 - 52.84 Mb on Pv11. L * Measu rements of L * describe darkness to lightness and range from 0 (pure black) to 100 (pure white). In this study, L * refers to the color component itself , while L* H and L* I refer to L* values measured by a Hunter Labscan and ImageJ software, respectively. QTL for L* were detected on Pv02 and Pv03 for Population 76, Pv05 and Pv10 for Population 86, and in both populations on Pv08, Pv09, and Pv11. All detectable L* value QTL in Population 86 were derived 35 from the ImageJ macro ( L* I ). All L* QTL but one increas ed L* values when contributed by the B14311 parent. A QTL on Pv02 ( L* 2.1 76 ) was detected in 2017 by the ImageJ macro and had a peak LOD score of 4.66. It covered a 61.6 - 74.1 cM interval bound by flanking markers ss715648552 (13.27 Mb) and ss715651061 (17. 24 Mb). Accounting for 9% of the total phenotypic variation, this QTL affected L* I values by 0.43. The 2017 L* I QTL on Pv03 ( L* 3.1 76 ) was barely above the significance threshold, but co - localized with noticeable peaks from other measurements of L* that fell below the significance threshold. Although the map interval was large (26.4 - 32.8 cM), the physical interval of flanking markers was just 2.02 - 2.43 Mb. Only accounting for 5% of the phenotypic variation, this QTL lightened L* I values by 0.33 when the allele was derived from the B14311 parent. A 2017 L* I QTL ( L* 5.1 86 ) mapped to the interval 121.9 - 126.9 cM on Pv05 and was flanked by markers ss715647683 and ss715639578 (34.33 - 35.96 Mb). It had a LOD score of 3.8, explained 6% of the phenotypic varia tion, and influenced lightness by 0.40. QTL for L* were identified on Pv08 in both populations despite low marker coverage. In Population 76, all measurements of L* were considered as a single QTL, L* 8.1 76 . LOD scores from 2016 L* I values were significant from 0 - 15.3 cM, while 2017 L* values from both Hunter Lab scan and ImageJ methods were significant from 17 - 62 cM and shared p eaks near 52 cM. Th ese shared peak s w ere flanked by markers ss715650193 and ss715648558 with physical positions at 5.86 and 7.15 Mb , respectively . In Population 86, QTL L* 8.1 86 was detected by ImageJ analysis in both years over the map interval 15.8 - 40.1 cM, corresponding to a 1.58 - 6.27 Mb physical region. The 2017 peak ( L* I_2017) explained 17% of phenotypic variation and affected L* I values by 0.67, 36 while the QTL peak in 2016 ( L* I_2016) explained 8% of the phenotypic variation and influenced L* I values by 0.49. Population - dependent L* I QTL were detected on Pv09 [ Table 6 ]. In Population 76, the L* 9.1 76 QTL had a peak LOD score of 6. 8 and spann ed a map interval of 27.0 - 27.2 cM (10.30 - 10.32 Mb based on physical positions of flanking markers). This QTL accounted for 14% of the phenotypic variation and affected L* I values by 0.53. Another QTL detected in Population 76 , L* 9.2 76 , was downstream of L* 9.1 76 . It had a peak LOD of 3.1 near 40 cM and only explained 4.8% of the phenotypic variation. In Population 86, LOD scores for the L* 9.1 86 QTL plateaued at 3.5 across the map interval 2.6 - 5.8 cM , corresponding to a physical interval of 27.58 - 29.1 Mb. QTL L* 9.1 86 accounted for just 5.5% of the phenotypic variation and lowered L* I values by 0.38, the only L* value QTL where the B14311 allele that bestowed a darkening effect. QTL L* 10.1 86 was detected in Population 86 in 2016 on Pv10 . The QTL reached a peak LOD of 3.98, explained 6.8% of the variation, and lightened L* I values by 0.48. Both map and physical intervals were small (65.6 - 67.3 cM and 41.96 - 42.01 Mb, respectively). Pv11 contain ed L* QTL that were shared across population s . P eaks from Hunter Lab scan - derived L* H values were noticeably co - localizing to the same region, but only reached the significance threshold in Population 76 [ Table 6 ]. In Population 76 , the L* 11.1 76 QTL was comprised of measurements fro m the Hunter Labscan ( L* H_2017) an d ImageJ analysis ( L* I_2016, L* I_2017) . Both years of L* I measurements mapped to a 149.6 - 150.6 interval (52.65 - 52.84 Mb) while the Hunter Lab scan - derived L* H QTL was nearby at 150.9 - 154 cM (52.84 - 52.87 Mb). In Population 86 , the L* 11.1 86 QTL mapped to a similar physical position as detected in Population 76. The L* 11.1 86 QTL was comprised of ImageJ measurements from 2016 ( L* I_2016) and 2017 ( L* I_2017). The region associated with L* I_2016 explained 18% of the phenotypic 37 variation, modified L* I values by 0.68, and mapped to 22.9 - 30.8 cM (52.47 - 52.87 Mb). The region associated with L* I_2017 explained 15% of the phenotypic variation, influenced L* I values by 0.62, and mapped to 22.9 - 30.5 cM (52.47 - 52.84 Mb). a* Measurements of a* describe the level of greenness to redness and range from negative values (green) to positive values (red). In this study, a* refers to the color component itself , while a* H and a* I refer to a* values measured by a Hunter Labscan and ImageJ software, respectively. QTL for a* values were found on Pv08 across all populations and included all detection methods and years [ Table 6 ]. A single QTL was found on Pv10 in Population 86 where the B14311 allele had a negative effect on a* . Observable peaks on Pv03 and Pv05 in Population 76 fell below the significance thresholds. On Pv08, LOD scores for a* values were significant over a large mapping interval containing f ew markers. In Population 76, QTL a* 8.1 76 was comprised of a* measurements from all years and methods. ImageJ - derived a* I values in 2017 ( a* I_2017) had a peak LOD of 9.8 spanning 17.5 - 49.7 cM, and explain ing 27% of the phenotypic variation , while the Hunter Labscan - derived a* H value s ( a* H_2017) explained 13% of the phenotypic variation in that same interval. Local LOD peaks for were detected downstream of the peak LOD. These l ocalized peaks were from 49.7 - 50.4 cM (5.86 - 6.00 Mb). Two regions for 2016 a* I values ( a* I_2016) mapped outside of the larger interval at intervals of 0 - 17.54 cM (0.484 - 1.54 Mb) and 55.3 - 83.9 cM (7.16 - 18.75 Mb) , though only the upstream one was visualized in MapChart [ Figure 1 2 H ] . Like Population 76, Population 86 a* value s co - localiz ed to a very large mapping interval from 15.3 - 60.4 cM (1.57 - 53.68 Mb) that was considered a single QTL named a* 8.1 86 . Peaks for all 38 measurements of Population 86 a* values explained approximately 20% of the phenotypic variation and influenced a* value s by approximately 0.48. A QTL for a* I from RILs grown in 2017 was detected on Pv10 in Population 86. This QTL was named a* 10.1 86 and spanned the map interval 69.7 - 82.1 cM , with flanking markers ss715645524 and ss715645501 located at physical positions 42.22 - 43.29 Mb. This QTL explained 8% of the phenotypic variation and was the only QTL where the B14311 allele decreased the a* I value by 0.31. LOD peaks for other a* values from 2016 ImageJ and 2017 Hunter Labscan measurements were apparent near this QTL, but were below the significance threshold. b* The b* value measures the level of blueness to yellowness and ranges from negative values (blue) to positive values (yellow). In this study, b* refers to the color component itself, while b* H and b* I refer to b* values measured by a Hunter Labscan and ImageJ software, respectively. Population 76 had QTL for b* on Pv02, Pv03, Pv08, and Pv11, while Population 86 had QTL on Pv03, Pv 08, and Pv11 [ Table 6 ]. All alleles from the B14311 parent increased b* values, signifying a more yellow coloration. A b * I QTL on Pv02 ( b* 2.1 76 ) was detected by ImageJ software in 2017. It plateaued at 4.38 LOD, explained 9% of the phenotypic variation, and influenced the b * I value by 0.41. Flanking markers occupied an interval from 54 - 74 cM (11.0 - 17.2 Mb). Two b* value QTL on Pv03 were detected in separate populations by the ImageJ softwar e : b* 3.1 76 in 2016, and b* 3.1 86 i n 2017 . In Population 76, the b* 3.1 76 QTL mapped to an interval of 3.4 - 14.7 cM, and a physical interval of 1.00 - 1.19 Mb. This QTL explained just 7% of the phenotypic variation and affected b* I values by 0.41. In Population 86, t he b* 3.1 86 QTL barely 39 cleared the LOD permutation threshold of 2.7 and only explained 3.9% of the phenotypic variation. This QTL mapped to a larger mapping interval of 0 - 1.51 cM, but had a similar physical interval of 1.19 - 1.30 Mb. A QTL for b* on Pv 08 was detected across populations using both methods and years . In Population 76, b* 8.1 76 mapped to the interval 16.8 - 62.3 cM, (1.53 - 7.25 Mb). Individual measurements b* I _2016, b* I_2017, and b* H_2017 explained a range of phenotypic variation from 7 - 16% and modified the b* value by a range of 0.39 to 0.47. In Population 86, all measurements of b* co - localized to a large map interval 15.3 - 40.1 cM in length and were designated as QTL b* 8.1 86 . Markers flan king this QTL had physical positions of 1.57 - 6.27 Mb, which is within the physical interval identified in Population 76. The three measurements ( b* I_2016, b* I_2017, and b* H_2017) explained a range of phenotypic variation from 9 - 12.6% . Pv11 also contained b* QTL that were shared across populations and detected over methods and years . In Population 76, QTL b* 11. 76 was comprised of individual measurements of b* with very high peak LOD scores: 8.8 for b* I_2016 , 6.9 for b* I_2017 , and 5.3 for b* H_2017 . These regions explained 11 - 21% of the phenotypic variation and affected the b * value by 0.38 to 0.45. All three peaks co - localized to 149.3 cM and were within a physical interval of 52.12 - 52.84 Mb. In Population 86, QTL b* 11.1 86 was comprised of b* measurements with even higher LOD scores (13.3 - 14.6), larger amount of phenotypic variation explained (26 - 27%), and greater influence on b* values ( 0.62 to 0.81 ) . No tably, the b* 11. 86 QTL from Population 86 co - localized to an interval from 22.9 - 30.5 cM (52.47 - 52.84 Mb), which is within the physical interval of b* 11.1 76 from Population 76. 40 QTL M apping of Ot her C anning Qu ality Traits A complete list of QTL for appearance, washed and drained weight, and texture is located in Ta ble 7 . QTL graphs from Population 76 are located in Figure 1 2 , and QTL graphs from Population 86 are located in Figure 1 3 . Appearance Ratings (APP) QTL for canned bean appearance (APP) were unique to populations; Population 76 had QTL on Pv02, Pv05, and Pv08, while Population 86 had QTL on Pv02, Pv03, Pv04, Pv06, and Pv10 [ Table 7 ]. QTL on Pv02 were similar in their effects on appearance rating s , but were considered unique between populations because of large differences in their physical positions. The APP2.1 76 QTL in Population 76 mapped to the interval 142.8 - 164.4 cM corresponding to the physical interval 37.81 - 44.97 Mb, while the APP2.1 86 QTL in Population 86 mapped to the interval 0 - 1.83 cM corresponding to the physical interval 3.90 - 4.48 Mb. The QTL from Population 76 was detected in 2017 and explained 9% of the ph enotypic variation with the B14311 allel e lowering appearance ratings by 0.12. The QTL from Population 86 was detected in 2016 and explained 8.7% of the phenotypic variation with the B14311 allele lowering appearance ratings by 0.16. In Population 86, Pv0 3 contained the APP 3.1 86 QTL that included regions significant for 2016 (APP_2016) and two - year average appearance (APP_2YA) ratings . Both regions had minor effects , explaining just 9.7 and 6.7% of the total phenotypic variation, and decreasing appearance scores by 0.16 and 0.11 from the B14311 allele , respectively. Th e APP3.1 86 QTL mapped to a 1.5 cM interval at the top of the chromosome, which corresponded to a physical interval of 1.2 - 1.3 Mb. 41 Population 86 also had an APP QTL on Pv04 , APP4.1 86 . The QTL had an R 2 of 7% and the B14311 allele improved appearance rating by 0.13. This QTL mapped to 51.1 - 52.1 cM and was flanked by markers ss715646227 and ss715646218 (2.75 - 2.89 Mb) [ Table 7 ]. The APP 5.1 76 QTL on Pv05 was detected in 2016 in Population 76. This QTL mapped to 40.6 - 57.1 cM, corresponding to a physical interval of just 4.47 - 4.75 Mb. It explained 8% of the total phenotypic variation and was responsible for lowering appearance rating by 0.12 whe n the allele was derived from B14311. Population 86 also detected a barely - significant QTL on Pv06 in spite of low marker density. The APP6.1 86 QTL explained 6% of the phenotypic variation and improved appearance scores by 0.13 when the allele was donated by the B14311 parent. The QTL mapped to a small interval and was flanked by markers located at 20.9 - 21.7 cM (28.97 - 29.04 Mb). One QTL from each year was detected on Pv08 in Population 7 6 , where both QTL explained 7% of the phenotypic variation . The APP 8. 1 76 QTL detected in 2016 was located at 145.5 - 147.0 cM (60.97 - 61.30 Mb) and the B14311 allele improved APP by 0.12 . T he APP8.1 76 QTL was detected in 2017 and located at 49.7 - 50.4 cM (5.86 - 6.00 Mb) and the B12724 allele improved APP by 0.10. In Population 86, the APP10.1 86 QTL was detected at the end of the chromosome Pv10 spanning a 69.65 - 90.23 cM interval (42.22 - 44.22 Mb). Two local peaks occurred within this interval, but they were considered as part of one QTL. Both local peaks had similar LOD scores, e xplained about 9% of the phenotypic variation and improved appearance by 0.13 when the allele was donated by the B14311 parent. 42 Texture (TXT) QTL for texture (TXT) were identified on Pv02, Pv05, Pv09, and Pv10. Separate, minor effect texture QTL for 2016 were detected in each population. A TXT QTL (TXT2.1 76 ) was detected on Pv02 in Population 76 and was located in the mapping interval from 81.8 - 88.9 cM (17.31 - 20.44 Mb). This QTL explained 8% of the phenotypic variation and increased texture by 2.64 kg when B14311 contributed the allele. In Population 86, a TXT QTL on Pv02 was located from 0 - 1.83 cM (3.90 - 4.48 Mb). It explained 6% of the variation and decreased texture by 2.25 kg when B14311 contributed the allele. A QTL for texture on Pv05 was detected acr oss years within Population 76 and named TXT5.1 76 . Measurements from both 2016 (TXT_2016) and 2017 (TXT_2017) mapped near 138 cM and were very significant, having peak LOD scores of 7.0 in 2016 and 11.6 in 2017. The 2016 peak explained 14% of the phenotypi c variation and influenced texture by 2.4 kg, while the 2017 peak explained 25% of the phenotypic variation and influenced texture by 2.5 kg. In both instances, the B14311 allele increased texture measurements. Both peaks were flanked by markers ss71564953 9 and ss715646996 which mapped to a 128.2 - 141.5 cM interval (27.70 - 36.79 Mb). A TXT QTL (TXT9.1 76 ) was identified on Pv09 in Population 76 in 2017. This QTL contained two localized peaks at 27 and 33.5 cM with similar LOD scores and was considered as a sin gle QTL spanning the interval 16.3 - 39.6 cM (7.87 - 13.55 Mb). This QTL explained 7.6% of the phenotypic variation and softened texture by 1.38 kg when the allele was donated by B14311. The TXT10.1 86 QTL at the distal end of Pv10 in Population 86 had a peak LOD greater than 11 and mapped the interval 82.06 - 90.23 cM (43.29 - 44.20 Mb) in 2016. This QTL explained over 26% of the phenotypic variation for that year and the B14311 allele decreased texture by 4.45 43 kg. A noticeable peak for texture measurements from 2 017 co - localized to the exact same region, but its peak LOD of 3.17 was below its permutation threshold of 3.42. Washed and D rained W eight (WDW) QTL for the washed and drained weight trait (WDW) of canned beans were detected only in 2016 on chromosomes Pv02 and Pv08 [ Table 7 ]. A peak from 2016 with a LOD of 3.15 was detected on Pv04 in Population 86, but it was below the 3 .7 LOD permutation threshold. Although both populations shared a distinct peak for 2017 WDW near 46.2 Mb on Pv01, it was below the significance threshold. The top of Pv02 contained WDW 2.1 76 detected in 2016 in Population 86. It explained 11% of the phenot ypic variation and increased WDW by 1.6 g when the allele was contributed by the B14311 parent. This QTL was positioned from 0 - 25.4 cM, which corresponded to a large physical interval of 3.90 - 30.15 Mb. The only WDW QTL found in Population 76 was located a t the end of Pv08 from 158.4 - 160.1 cM (62.27 - 62.75 Mb). This QTL was named WDW 8.1 76 and had a sharp peak at 158.4 cM with a LOD of 4.4. The R 2 was 12%, and it decreased the washed and drained weight by 2.43 g when the allele was donated by the B14311 paren t. QTL M apping of A gronomic T raits Agronomic traits were not the primary focus of this research, but agronomic data was useful for identifying RILs that could potentially contribute to the MSU breeding program. Many of these traits are polygenic in nature and are thus affected by many loci, each with a small contribution to a trait phenotype. A complete list of QTL for agronomic traits is located in Table 44 8 . QTL graphs from Population 76 are located in Figure 1 2 , and QTL graphs from Population 86 are located in Figure 1 3 . Seed Y ield (SY) A single QTL for seed yield (SY) was identified on Pv08 for both populations; these QTL were considered distinct due to large difference s in the physical positions of the flanking markers (60.07 - 60.56 Mb in Population 76, compared to 0.374 - 1.41 Mb in Population 86) [ Table 8 ]. In Population 76, the SY 8.1 7 6 QTL had a peak LOD of 3.5 at 137.7 cM , explained 9% of the phenotypic variation, and lowered yield by 110 kg/ha when the B14311 allele was present. In Population 86, the SY8.1 86 QTL had a peak LOD of 6.85 at 0.01 cM, explained 14.8% of the phenotypic variation, and lowered yield by 130 kg/ha when the B14311 allele was present. S eed W eight (SW) QTL for seed weight (SW) were detected on Pv03 and Pv04 in both populations, Pv05 in Population 86, and Pv07 and Pv08 in Population 76. A SW QTL was identified in both populations within a similar physical interval on Pv03 [ Table 8 ] . In Population 76, th e SW3.1 76 QTL had a peak LOD of 12.4 located within the interval 119.3 - 122.8 cM (11.47 - 11.82 Mb). It explained 23% of the phenotypic variation and the B14311 allele increase d seed weight by 0.79 g. In Population 86, the SW3.1 86 QTL had a peak LOD of 4.5 located within the interval 43.6 - 60.2 cM (3.82 - 12.30 Mb). This QTL explained 12% of the phenotypic variation and the B14311 allele increased seed weight by 0.42 g . The SW 4.1 76 QTL was detected in Population 76 on Pv04 with an R 2 of 13% and an additive effect of 0.60 [ Table 8 ]. Its peak LOD of 8.1 was located within the interval 10.3 - 20.6 cM 45 (0.16 - 1.90 Mb). In Population 86, two QTL were found, SW4.1 8 6 at the proximal end of Pv04 (.73 - 1.2 cM) and SW4.2 86 within the interval 51.1 - 63.9 cM. The SW4.1 86 QTL near the top of Pv04 was between 2.2 - 2.33 Mb, had an R 2 of 8% and decreased seed weight by 0.34 g when the B14311 allele was present. The SW4.2 86 QTL c ontained two local peaks, but was considered as a single QTL because the local peaks did not have a sufficiently large drop in LOD between them. This QTL spanned the physical interval 2.75 - 3.59 Mb, explained about 9.6% of the phenotypic variation and decre ased seed weight by 0.38 g when the allele was contributed by the B14311 parent. A SW QTL (SW5.1 86 ) was found on Pv05 in Population 86 that explained 9% of the phenotypic variation and increased seed weight by 0.36 when the B14311 allele was present. This QTL had a peak LOD of 4.1 that mapped between 126.9 - 133.3 cM (35.96 - 36.79 Mb). A minor effect SW QTL (SW7.1 76 ) on Pv07 in Population 76 explained just 6% of the total phenotypic variation and affected seed weight by 0.5 g. It was located over a large map distance from 83.1 - 109.5 cM, but a relatively small physical distance from 4.25 - 4.39 Mb. Another minor effect S W QTL from Population 76 was found on Pv08 , named SW8.1 76 . This QTL had a relatively high LOD score of 5.6 and mapped near the end of the chromosome from 151.1 - 158.4 cM (62.06 - 62.27 Mb). This QTL explained 8% of the phenotypic variation and decreased seed weight by 0.45 g when donated by the B14311 parent. Days to F lowering (DF) Despite a lack of phenotypic variation in days to flowering amongst the parents of both populations [ Table 2 ; Table 3 ], days to flowering QTL were found in Population 76 on Pv07, Pv08, and Pv11 [ Table 8 ]. 46 A QTL on Pv07 (DF7.1 76 ) contained two local peaks and reached the significance threshold with a LOD score of 3 . Markers flanking the region had physical positions of 27.48 - 30.85 Mb. This QTL only explained 5% of the phenotypic variation and affected flowering by 0.25 days. The DF 8.1 76 QTL Pv08 had a highly - significant peak with a LOD score of 10. The QTL mapped to a tight mapping interval of 145.5 - 147.0 cM, corresponding to a physical interval of 60.97 - 61.30 Mb. This QTL explained 20% of the phenotypic variation and the B14311 allele delayed flowering by approximately half of a day. Like the DF7.1 76 on Pv07, the DF11.1 76 QTL peak on Pv11 was barely signi ficant with a LOD score of 2.99 from 143.7 - 144.0 cM. Th e DF11.1 76 QTL influenced days to flowering by 0.25, explained only 5% of the phenotypic variation, and mapped to the tight physical region from 51.95 - 51.96 Mb. Days to M aturity (DM) Days to maturity QTL were found on Pv02 in Population 76 and on Pv04 and Pv11 in Population 86 [ Table 8 ]. The DM 2.1 76 QTL on Pv02 was found in Population 76 from 110.1 - 113.2 cM (31.67 - 33.65 Mb). It had a peak LOD of 4.69, an R 2 of 11%, and the B1 4311 allele hastened maturity by 0.43 days. A DM QTL with a tight ly peak ed LOD score of 4.29 was detected near the top of Pv04 from 0.73 - 1.2 cM (2.2 - 2.41 Mb). This QTL (DM4.1 86 ) explained 11.6% of the phenotypic variation and hastened maturity by 0.39 days when the allele was contributed by B14311. 47 On Pv11, the DM 11.1 86 QTL explained 8% of the phenotypic variation and delayed maturity by 0.35 days when the B14311 allele was pres ent. It was located over a mapping interval of 0 - 10.8 cM, which corresponded to a modest physical interval of 49.59 - 51.12 Mb [ Table 8 ]. Canopy H eight (HT) Several minor - effect QTL for canopy height were found above the significan ce threshold, but all were population - dependent. Population 76 had QTL on Pv02, Pv03, Pv07, and Pv11, with the B14311 parent generally increasing height, whereas Population 86 had QTL on Pv01, Pv04, and Pv08, with the B4311 parent decreasing height in all instances. The Pv01 QTL (HT 1.1 86 ) was located over a large interval from 43.6 - 63.7 cM (1.29 - 2.85 Mb). It had an R 2 of 6.7% and affected height by 0.34 cm. The HT 2.1 76 QTL on Pv02 explained 10% of the phenotypic variation and was the only HT QTL in Popula tion 76 where the B14311 allele decreased height by 0.45 cm. It was positioned over the large interval 95.3 - 110.1 cM, corresponding to 25.39 - 31.67 Mb. QTL HT3.1 76 on Pv03 contained a peak that barely reached the significance threshold with a LOD score of 3.08. It explained just 6% of the phenotypic variation and was located from over the interval 92.4 - 108.8 cM (10.69 - 11.25 Mb). The HT4.1 86 QTL on Pv04 had a LOD peak o f 10.62 located at 50.86 cM. It explained 17% of the phenotypic variation and influenced height by approximately 0.5 cm. It mapped to an interval from 44.49 - 52.13 cM and the flanking markers had physical positions from 2.55 - 2.89 Mb. On Pv07, the HT7.1 76 Q TL had a peak LOD score of 3.07 that barely reached the significance threshold. It was located over 5.85 - 9.43 cM (0.65 - 0.91 Mb), only explained 5% of the phenotypic variation, and changed height by 0.34 cm. 48 The HT8.1 86 QTL on Pv08 had the highest peak LOD score of any HT QTL detected in this study at 12.79. This QTL explained 22% of the phenotypic variation and the B14311 parent lowered height by nearly 0.6 cm. It was located at the top of the chromosome from 0 - 12.88 cM, which corresponded to a physical in terval of 0.374 - 1.50 Mb. The HT 11.1 76 QTL on Pv11 had a peak LOD of 4.6 at 144.34 cM. This peak lay within the interval 144.33 - 149.57 cM (52.16 - 52.65 Mb). It was responsible for 9% of the phenotypic variation and impacted height by 0.42 cm. Desirability S core (DS) Desirability score QTL were population - dependent; Population 76 had QTL on Pv04 and Pv08, while Population 86 had QTL on Pv02 and Pv09. In all cases, the B14311 parent contributed a negative additive effect. A QTL on Pv02 named DS2.1 76 barely m et its permutation threshold with a LOD score of 3.0. It was located from 50.6 - 51.1 cM (5.86 - 7.1 Mb), explained 7% of the phenotypic variation, and affected DS by 0.14. Similarly, the DS 4.1 86 QTL on Pv04 explained just 7% of the phenotypic variation and influenced the desirability by 0.13. This QTL was located on the interval 51.1 - 52.1 cM (2.75 - 2.89 Mb). The strongest QTL for desirability score (DS 8.1 86 ) was detected at the proximal end of Pv 08 from 0 - 12.9 cM, which corresponded to a physical interval of 0.37 - 1.5 Mb. This QTL had a peak LOD of 6.7, explained over 14% of the phenotypic variation, and affected the score by 0.19. 49 On Pv09, the DS9.1 76 QTL had a peak located at 14.9 cM with a LOD score of 4.92. It fell within the interval 14.9 - 15.8 cM, corresponding to a small 7.70 - 7.79 Mb physical region. This QTL explained 11% of the phenotypic variation and impacted desirability score by 0.18. Ozone Bronzing (BRZ) In Population 7 6, QTL for foli ar bronzing were discovered on Pv06 and Pv08 , while in Population 8 6, QTL for foliar bronzing were discovered on Pv05, Pv07, and Pv09 . A noticeable, yet insignificant peak was also identified in Population 76 near the same physical region as the Pv07 BRZ Q TL in Population 86. The BRZ 5.1 86 QTL on Pv05 plateaued at a LOD score of 3.2 over the map interval 174.7 - 188.5 cM, corresponding to a small physical interval of 39.24 - 39.34 Mb. This QTL explained about 7% of the phenotypic variation, and the B14311 allel e reduced bronzing rating s by 0.19. The BRZ 6.1 76 QTL in Pv06 plateaued over a large map distance containing few markers (0.51 - 22.9 cM), although the physical positions of the flanking markers were located at a smaller interval from 12.21 - 13.75 Mb. The QTL explained just 6% of the phenotypic variation and bronzing ratings were reduced by 0.25 when the B14311 allele was present. The BRZ 7.1 86 QTL on Pv07 was the most significant QTL detected for this trait. It had a peak LOD of 6.6 at 46.1 cM, explained 17% o f the phenotypic variation, and the B14311 allele reduced bronzing rating s by 0.3. This QTL was located within the map interval 46.1 - 49.8 cM and was flanked by markers ss715649276 and ss715646465 (3.99 - 4.17 Mb). The BRZ8.1 76 QTL was detected near the end of Pv08 over the interval 144.8 - 158.4 cM, corresponding to a physical interval of 60.71 - 62.27 Mb. This QTL explained 10% of the phenotypic variation and the B14311 allele increased bronzing ratings by 0.31. 50 The BRZ9.1 86 QTL was located on Pv09 and was sig nificant from 12.2 - 19.2 cM. This small map interval corresponded to a small physical interval of 31.40 - 33.35 Mb that was flanked by markers ss715646279 and ss715645629. The QTL explained about 7% of the phenotypic variation and the B14311 allele increased bronzing ratings by 0.18. Molecular M arker A nalysis Parental lines were screened with seven InDel markers with v2.1 physical positions near COL QTL located on Pv08 [ Figure 1 4 A ]. This region had low marker coverage, but high LOD scores. The markers spanned an approximate physical interval of 5.43 to 7.16 Mb in the v2.1 genome. InDel markers screened were: NDSU_IND_8_5.4417 (between 5.43 - 5.50 Mb), NDSU_IND_8_6.0169 (between 5.95 - 6.27 Mb), NDSU_IND_8_6.2923 (between 6.27 - 6.44 Mb), NDSU_IND_8_6.6519 (near 6.71 Mb), NDSU_IND_8_6.6880 (between 6.71 - 6.99 Mb), NDSU_IND_8_6.7497 (between 6.99 - 7.04 Mb), NDSU_IND_8_7.0078 (near 7.16 Mb). None of the markers showed a polymorphism between B1 4311 and both of the other parents. The B12724 product was polymorphic to products from B14311 and Zenith for marker NDSU_IND_8_6.2923. The product from Zenith was polymorphic to products from B14311 and B12724 for marker NDSU_IND_8_7.0078. Markers NDSU_IN D_8_6.6880 and NDSU_IND_8_6.7497 were heterozygous amongst all parental lines. Parental lines were also screened with six InDel markers with v2.1 physical positions near COL QTL located on the distal end of Pv11 [ Figure 1 4 B ]. The m arkers spanned an approximate physical interval of 50.68 to 53.17 Mb in the v2.1 genome. InDel markers screened were: NDSU_IND_11_47.0739 (between 50.68 - 50.75 Mb), NDSU_IND_11_47.7708 (between 51.50 - 51.53 Mb), NDSU_IND_11_47.9412 (between 51.72 - 51.75 Mb), NDSU_IND_11_48.4937 51 (between 52.23 - 52.48 Mb), NDSU_IND_11_48.7818 (between 52.53 - 52.54 Mb), and NDSU_IND_11_49.5223 (between 52.96 - 53.17 Mb). The B14311 product was polymorphic to - 50.75 Mb) , NDSU_IND_11_47.7708 (51.50 - 51.53 Mb), and NDSU_IND_11_49.5223 (52.96 - 53.17 Mb). Based on the physical positions of the flanking BARCBean6k_3 SNPs, these three InDel markers lie outside the most - significant COL QTL interval from 52.16 - 52.84 Mb. DISCUSSIO N Post - processing color retention is an important quality trait in black beans. To identify genomic regions controlling this trait, black bean RIL populations were developed. Two half - sib ling populations were selected for genetic mapping, Population 76 and Population 86. These populations shared a common female parent, B14311, that had poor post - processing color retention. Populations derived from this MSU breeding line were purposefully selec ted over the other similarly poor - canning breeding lines because B14311 had greater seed coat integrity when canned, reflected in higher appearance ratings [ Table 1 ]. Having acceptable canned appearance (e.g. no splits) was import ant to the study design in order to minimize the effect of pigment leaching due to mechanical breakdown or splitting of the seed coat. This also mitigates potential bias amongst reviewers who may unintentionally confound appearance and color instead of tre ating them as distinct characteristics. Both B12724 and Zenith exhibit a similar level of superior post - processing color retention, yet possess different genetic backgrounds that may uniquely contribute to color retention [Supplemental Figure s 4 and 5 ]. Re sults from this study reveal both population - specific QTL and those shared across populations, meaning that the superior - canning parents contain both unique and shared QTL for post - processing color retention. 52 Comparing M ethods of Q uantitative C olor M easurement The MSU dry bean breeding program has traditionally measured canned bean color objectively with a Hunter Lab scan XE spectrophotometer to obtain L * , a * , and b * values. Of these measurements, L * was thought to be the closest representation of bla ck bean darkness, but has been shown in this study to be less accurate at describing perceived darkness than previously thought. The Hunter Labscan XE is easy - to - use, quick, and has been used for many years, but comes with some disadvantages. For one, this instrument can only measure a small portion of a canned sample at a time ( 4.4 cm diameter) , which may not be representative of the entire sample. Furthermore, this instrument was designed to measure color according to human perception, so glare and glossi ness are included in measurements. While rinsing the canned samples before imaging is important to remove brine, excess water on the seed coat surface creates glare that introduces varying amounts of reflection on each sample. To address these drawbacks as sociated with the current methodology, a new protocol was developed to measure C IE L* , a* , and b* values from digital images of canned beans using ImageJ software (Schneider et a l., 2012) . CIE L* , a* , and b* values were different between measurement methods, which was to be expected due to the difference in sample size and confounding effect of glare . Using the same CIELAB color sp ace for color components derived from both Hunter Labscan and ImageJ analysis enabled comparison between the two methods to determine which was mo re effective at measuring canned black bean color. The Hunter Labscan XE spectrophotometer and ImageJ software measured color components with varying deg rees of labor input and accuracy. When canned bean samples underwent machine phenotyping after reviewer evaluation, capturing digital images took longer than measuring CIELAB values from the Hunter Lab scan spectrophotometer. This was mostly due 53 to the for photographing . After rinsing, each canned bean sample was placed on a tray where it was manually distributed to a uniform depth that completely covered the bottom of the tray. The plated sample then had to be precisely positioned beneath the digital camera to fill the viewing area . Before taking the photograph, the can code for each sample had to be manually entered so that each image file was accurately recorded. In practice, the digital imaging process took about 30 - 40 seconds per sample, compared to approximately 20 seconds per sample from the Hunter Labscan. Once images were collected in a f older , the ImageJ macro took 5 - 10 seconds to process a single image before automatically moving on to the next o ne. A major advantage of digital images is that they hold a large amount of spatial information that can be used in a myriad of downstream analyses (e.g. creation of custom color spaces, segmentation of beans, and measurement of distances and angles). Qua ntitative color components of L* , a* , and b* were similarly correlated to visual color ratings in both population s [ Table 4 ] . Color values derived from the ImageJ macro were more accurate and precise compared to those from the Hun ter Labscan across years and populations [ Table 4 ; Figure 1 1 ]. Regardless of measurement method, the b* component was most strongly correlated with visual color ratings, though b * I values from the ImageJ macro were more strongly correlated with visual color ratings than the b * H values from the Hunter Labscan. Considering only measurements from 2017 , the L* H values from the Hunter Labscan were strongly correlated with visual color ratings ( r = - 0.73 to - 0.83), but were less descriptive than the L* I values from the ImageJ analysis (r = - 0. 87 to - 0. 91 ) . Using two years of ImageJ - derived L* I values further improves the correlation with visual color ratings (r = - 0.87 to - 0.91), though the two - year averages of L* I did not reach the strength of the two - year averages of b* I (r = - 0.93 to - 0.96). Th e Hunter 54 Labscan - derived L * H value has typically been treated as the standard for quantitative color measurement of black beans, a practice that needs to be re - visited in light of these findings. I nstead of relying on L * values, b* values should be used to measure perceived seed coat darkness of canned black beans . Considering only measurements from 2017 , the Hunter Labscan b * H values had a very strong correlation (r = - 0.82 to - 0.93) with visual color ratings, but did not exceed the ImageJ b * I correlation coefficients of - 0. 91 to - 0 .93. When two years of ImageJ b* I values were used , the re was a nearly perfect negative correlation with visual color ratings (r = - 0.93 to - 0.96 ). In reality, t he b* I value may be a more accurate measurement of canned bean color than the consensus reviewe r rating . Although reviewers were trained and provided physical copies of the 1 - 5 visual color scale , their ratings of canned beans may be influenced by the rating experience of the reviewer s , brine on the surface of the seed coats, or fatigue and loss of focus from the large number of samples . All of t hese aforementioned issues can be mitigated through machine - derived measurements of color. Interestingly, L* , a* , and b* were moderately correlated with v isual appearance ratings across populations and measurement methods [ Table 4 ]. This could be caused by reviewers confounding color ratings with appearance ratings or by the machines measuring the exposed cotyledon tissue of split beans , thereby increas ing the mean brightness of the digital image . Regardless of the potential cause, L* values were more strongly correlated to visual appearance ratings than a* and b * values. T he ImageJ L* I values were similarly correlated to visual appearance ratings (r = - 0.46 to - 0.52) than the Hunter Labscan L* H values (r = - 0.42 to - 0.49) . While simple measurements from machine vision can provide unbiased and repeatable measurements of canned color, further improvements are required to more accurately describe canned appearance. Research by Mendoza et al. (2017) utilized digital images of canned black 55 beans and brine to develop p artial least square regression (PLSR) models for prediction of color and appearance ratings . In the study, d igital images of canned beans underwent an additional segmentation process that separated the canned beans from the background. Information from the segmented beans was combined with several other image features into PLSR models for color and appearan ce. The model s retu rned correlation coefficients of 0.873 to 0.937 for color ratings and 0.806 to 0.871 for appearance ratings . However, i mplementation of the methodology used by Mendoza et al. (2017) in breeding programs may be hindered by the lack of a comprehensive, easy - to - use phenotyping pipeline . Overall, research from the present study has shown that color retention of canned black beans can be accurately measured by CIE L* , a* , and b* values generated after minimal post - proce ssing of digital images. U sing ImageJ to measure color components from digital images i s a more reliable method than the traditional Hunter Labscan, and b* values are more strongly correlated with visual color ratings than L* values. Heritability E stimates of C anning Q uality T raits Heritability estimates of canned bean color retention have not been previously reported. Broad - sense heritabilities of reviewer color ratings were high in both Population 76 (0.87) and Population 86 (0.91) [ Figure 11 ]. This suggest s that a consensus approach can be effecti ve at phenotyping this trait. However, CIELAB color components as measured by ImageJ analysis were also highly heritable in these populations, with b* I and L* I having heritabilities near 0.90 in both populations. T hese results validate anecdotal evidence suggesting that color retention is moderately to highly heritable in the MSU dry bean breeding program (J. Kelly, pers. comm.) . 56 Furthermore, color retention heritabilities estimated by objective measurements can meet or exceed those provided by subjective reviewer ratings. Populations 76 and 86 both had appearance heritabilities near 0.58 [ Figure 11 ], which is nearly identical to the 0.58 estimated by Walters et al. ( 1997) bean s. H owever, Posa - Macalincag et al. (2002) estimated the narrow - sense heritability of canned kidney bean appearance as approximately 0.84. Estimates of broad - sense heritabilities for canned bean texture ranged from 0.46 to 0.67 in Populations 86 and 76, respectively. These were comparable to the estimate of 0.64 determined by Walters et al. (1997) . Broad - sense heritabilities of washed and drained weight were very low [ Figure 11 ]. In Population 76, the heritability was just 0.06 and in Population 86, the herita bility was 0.30. Contrastingly, Walters et al. (1997) determined a moderate heritability of 0.67 for washed and drained mass, an equivalent measurement of washed and drained weight. QTL Nomenclature QTL are named according to the guidelines proposed by Miklas and Porch (2010) . Briefly, each trait is assigned a two - to three - letter abbreviation. Common abbreviations are listed in a separate document (unpublished), but canning q uality QTL are uncommon traits so new trait abbreviations were created for this study. After the abbreviation, the linkage group or chromosome number is listed. This study mapped QTL on each of the 11 Phaseolus vulgaris chromosomes so all numbers 1 through 11 are used. Specific QTL within a linkage group are noted successively by QTL are tagged according to the population in which they were discovered. This stud 57 (B14311/B12724). These population descriptors are useful for consistency within the context of (B14311/Zen ith) (B14311/B12724) in future publications to maintain compliance with nomenclature guidelines . QTL M apping of P ost - P rocessing C olor R etention A complete list of QTL for post - processing color retention is located in Table 6 . QTL graphs from Population 76 are located in Figure 12 , and QTL graphs from Population 86 are located in Figure 13 . Color Ratings (COL) QTL for visual color ratings were detected on five chromosomes: Pv02, Pv03, Pv05, Pv08, Pv10, and Pv11. As expected, QTL contributed by the B14311 parent decreased color ratings, except in a single instance. Many significant, yet small - effect QTL were dete cted across and within the two RIL populations. Previous studies by Wright and Kelly (2011) and Cichy et al. (2014) also measured canned black bean color, but used methods that differed from this study. Wright and Kelly (2011) used HunterLAB color space L value as a proxy for color, while Cichy et al. (2014) used a 1 - preference. The COL2.1 76 QTL on Pv02 was detected by color ratings from 201 7 and the two - year average ( COL _2017 and COL _2YA, respectively ) and was specific to Population 76. Th is QTL influenced color ratings only slightly, but co - localized with QTL for quantitative color measureme nts like L* 2.1 7 6 and b* 2.1 76 w hich were both detected by ImageJ software in 2017. 58 LOD peaks for the Hunter Labscan L* H and b* H values also co - localized to this region, but were below the significance threshold. The 2017 ratings mapped to a smaller interval than the two - year average ratings , but both shared a terminal flanking marker, ss715651061 located at 17.24 Mb . Neither Wright a nd Kelly (2011) , nor Cichy et al. (2014) detected COL QTL on Pv02, but co - localizatio n with quantitative color measurements supports the detection of a population - specific COL QTL in this region. COL QTL on Pv03 were detected in both populations, but in different years for each. In Population 76, the COL 3.1 76 QTL was detect ed from 2016 and the two - year average color ratings ( COL _2016 and COL _2YA, respectively ) . Both measurements were found within the same, small physical interval of 2.02 - 2.4 3 Mb and had small effects on color retention, explaining just 6% of the phenotypic variation and affecting color ratings by approximately 0 .15. This COL 3.1 76 QTL co - localized with QTL for ImageJ color measurements . A QTL for L* I from 201 7 RILs ( L* 3.1 76 ) mapped to the same region, while a QTL for b* I from 2016 ( b* 31. 76 ) mapped slightly upstream. In Population 86, a 2017 COL QTL was detected near the top of Pv03 from 1.1 9 - 1.30 Mb, named COL3.1 86 . This QTL had an extremely small effect on color retention, but co - localized with the APP 3.1 86 QTL detected from 2016 and two - year average and the b* 3.1 86 QTL from ImageJ software in RILs grown in 201 7 . Another small effect COL QTL (COL3.2 86 ) , this one from 2016 , was detected near the distal end of the chromosome from 47.28 - 50.39 Mb. Both of these COL QTL in Population 86 mapped to regions of sparse marker coverage, and overall Population 86 had fewer markers mapping to unique positions on Pv03 (n=22) than Population 76 (n=50). For example, COL3.2 86 at the distal end of Pv03 had a left - flanking marker that mapped to 102.4 cM (47.28 Mb) and the next closest u pstream marker was located at 75.4 cM (37.2 Mb). 59 The COL 5.1 86 QTL on Pv05 mapped to a small physical interval from 38.84 - 38.92 Mb and contributed very little to color retention. The QTL mapped within the interval for the two - year average color rating, though the two - year average color ratings fell just short of the permutation LOD threshold. Interestingly, COL5.1 86 was the only COL QTL that mapped independently of quantitative color measu rements, which somewhat weaken s its validity. A highly - significant COL QTL was detected across populations on Pv08 . In Population 76, the COL 8.1 76 QTL was detected by color ratings from 2016 , 2017 , and the two - year average and mapped to a tight region near 53 cM, although the 2016 ratings extended further upstream than the others. Each of these individual measurements contributed a moderate amount toward color retention by explaining 12 - 16% of phenotypic variation and influencin g color ratings by approximately 0 .25 on the 1 - 5 rating scale. Although this region spanned a large map distance of nearly 50 cM, the physical interval was 1.53 - 7.25 Mb. Only four markers were located within that interval , and there was a large 4 Mb gap fr om 1.53 - 5.86 Mb without any markers. This means that this region may actually contain several QTL if mapping resolution was increased. Nevertheless, post - processing color retention because QTL for all quantitative color measurements like L * , a * , and b * co - localized in or near this interval. Likewise, the COL 8.1 86 QTL from Population 86 was detected by ratings from 2016 , 2017 , and the two - year average (COL _2016 , COL _2017 , and COL8 _2YA , respectively) and mapped to the interval 15.32 - 60.37 cM, which corresponded to a physical region of 1.57 - 53.68 Mb. Th is COL QTL explained a similar amount of phenotypic variation as those identified in Population 76, with nearly identical effect sizes. However, COL8.1 86 spanned a much larger interval overall. The region encompassed by COL_2017 in Population 86 had the smallest physical interval from 1.57 - 6.27 Mb, which fits within the interval of the Population 76 COL 8.1 76 QTL (1.53 - 7.25 Mb). As 60 mentioned before, this re gion of Pv08 contained major gaps in marker coverage. In Population 86, t here were no markers from 1.58 - 6.27 Mb or from 6.27 - 53.59 Mb, which contributed to the large QTL interval. Like Population 76 , quantitative color measurements co - localized with COL QT L in this population, as well . Overall, the region from 1.5 - 7.25 Mb on Pv08 was found to be a key determinant of post - processing color retention in both populations. The Co - 4 locus conditioning resistance to anthracnose ( Colletotrichum lindemuthianum ) resides within this interval at approximately 2.8 Mb (Oblessuc et al., 2015) , and the complex C locus [C R Prp] also maps in this region (McClean et al., 2002) . Interestingly, a ll loci within the complex C loc us are involved in pigmentation: C determines seed coat patterning (Prakken, 1974) ; R determines red seed coat coloration (Prakken, 1974) ; and Prp determines pod pigmentation (Bassett, 1994) . While the complex C locus is an important determinant of pigmentation of dry beans, it is unknown if it also plays a role in seed coat color retention of canned beans. In any case, this region of Pv08 is crucial to dry bean pigmentat ion and canned color retention , but additional markers are needed to determine the actual physical location of the COL QTL identified in this study. Genes within t he 1.5 - 7.3 Mb QTL interval w ere examined using PhytoMine ( DOE - JGI and USDA - NIFA, http://phytozome.jgi.doe.gov/ phytomine/ begin.do ) to generate a list of 599 genes (data not shown) . Genes encoded for unknown proteins, leucine - rich repeat proteins, and transferases involved in flavonoid biosynthesis , among many others . Gene ontology (GO) terms were generated [ Supplemental Figure 7 ] . H owever , deducing the biological mechanism(s) behind post - processing color retention was beyond the scope of this study . A single QTL on Pv10 (COL10.1 86 ) was found in Population 86 that had a very minor effect on color retention. It was noteworthy because it was the only COL QTL found in this study 61 where the B14311 parent contributed a beneficial allele for color retention. COL10.1 86 co - localize d with QTL for a * , APP, and TXT over a 12 cM map interval corresponding to a smaller physical interval of 42.22 - 43.29. A QTL for L* ( L * 10.1 86 ) mapped just upstream. COL QTL on Pv11 were major determinants of post - processing color retention and were identified across populations and instruments . In Population 76, the COL 11.1 76 QTL detected by ratings from 2016, 2017, and the two - year average mapped to a region near the distal end of the chromosome from 52.16 - 52.84 Mb. These individual regions explaine d a range of phenotypic variation from 8.4 - 14.3% and influenced color ratings by 0 .19 - 0 .25. QTL for quantitative color measurements of L * ( L* 11.1 76 ) and b * ( b* 11.1 76 ) co - localized to this region as well, showing that the COL QTL at this location can be detected consistently by machine phenotyping. In Population 86, the COL11.1 86 QTL mapped to the end of Pv11 just like the COL11.1 76 QTL from Population 76. Th is COL 11.1 86 QTL was detected across all years (2016, 2017, and the two - year average) , and had associated LOD scores greater than any other color rating s in the study , at 10.55, 12.74, and 12.96, respectively. Individually, t hey explained 19 - 22% of the phenotypic variation and influenced color ratings by nearly a third of a score. These regions ass ociated with yearly color ratings all mapped to the exact same physical interval from 52.47 - 52.84 Mb , which lies within the interval of the COL 11. 1 76 QTL detected in Population 76. In Population 86, COL11.1 86 co - localized with all measurements of L * ( L* 11.1 86 ) and b * ( b* 11.1 86 ) , with the exception of L * H measured by the HunterLab spectrophotometer o n 201 7 RILs . Taken together, the Pv11 COL QTL represent the most influential source of color retention detected in this study. Previous work by Cichy et al. (2014) also revealed QTL associated with color on Pv11. In that study, significan t QTL for hedonic color ratings in 2010 and 2011 mapped near the top of the chromosome and co - localized with QTL for putative L * H , a * H , and b * H values measured by a 62 Hunter Lab scan spectrophotometer . T h eir QTL cluster was reported as a 0 - 13 cM interval at the top of chromosome 11 . H oweve r , t he mapping positions must have been mistakenly inverted , since physical positions of the flanking SNPs are actually located around 52 - 53 Mb in the v2.1 dry bean genome . The SNP marker M27933 was closest to QTL for color retention in both years , L * H in 2011, and b * H in both years. No sequence data was provided for M27933, but a BLAST query of adjacent markers D05338 and D30369 placed them at 52.61 and 53.47 Mb, respectively. Disregarding the published genetic positio ns in favor of the latest physical positions, the QTL identified by Cichy et al. (20 14) are in the same physical region as the co - localizing QTL for color retention identified in the present study. This region of COL QTL co - localization on the distal end of Pv11 is nearby several loci conferring disease resistance in dry beans. Slightl y upstream of the QTL, the 51 - 52.2 Mb interval contains many genes encode leucine - rich repeat proteins, sulfotransferases, and albumins (data not shown) . Furthermore, the anthracnose resistance locus Co - 2 (Geffroy et al., 1998) maps to the distal end of chromosome Pv11, along with loci involved in rust resistance, Ur - 3, Ur - 6, Ur - 11, and Ur - Dorado (Miklas et al., 2006) . Of these loci, physical positions have been determined for Ur - 11 near 51.93 Mb (McClean , unpublished) and Ur - 3 from 46.97 - 47.01 Mb (Hurtado - Gonzales et al., 2017) . The region from 5 2 .41 - 5 2.85 Mb was screened for candidate genes using PhytoMine ( DOE - JGI and USDA - NIFA, http://phytozome.jgi.doe.gov/ phytomine/ begin.do ) . This region contained genes encoding unk nown proteins, aspartyl proteases , and annexin s [ Supplemental Table 2 ] . The biological role of a spartyl proteases is not completely understood, but the ir ability to hydrolyze proteins ma y contribute to protein storage or disease resistance (Simões and Faro, 2004) . An Arabidopsis thaliana homolog , CDR1 , was found to confer resistance to Pseudomonas 63 syringae (Xia et al., 2004) . According to PhytoMine results, p utative aspartyl protease genes Phvul.011G208100 and Phvul.011G208900 had perfectly correlated expression levels with paralogs related to pectin breakdown : Phvul.007G202000 , Phvul.007G271650 , and Phvul.007G271600 . Annexins are thought to have several cellular function s, ran ging from C a 2+ - dependent membrane binding (Gerke et al., 2005) to polar growth and stress response (Konopka - Post upolska et al., 2011) . However , the focus of the present study was toward genetic mapping of post - processing color retention ; the biological mechanism behind the phenotype was not explored . A dditional experiments like microscopy, NIL development, metabolomic analyses, RNA - Seq, and comparative mapping may be useful in deducing a biological mechanis m behind post - processing color retention. L * L * values describe the luminosity of a sample on a 0 - 100 scale where 0 is black and 100 is white. L* values from a Hunter Labscan XE spectrophotometer (Hunter Associates Laboratory Inc., Reston, VA) have traditionally been used by the MSU breeding program as a quantitative method of black bean post - processing color and are referred to in this study as L* H . While these L * H values are able to detect phenotypic variation in canned black beans, the measurement process can create artifacts ( F. Mendoza, pers. comm.). In an attempt to improve color measurements, an experimental macro was developed for ImageJ tha t measured L * , a * , and b* values from processed digital images of canned beans. QTL for L * values from the digital imaging software ( L* I ) were more detectable and co - localized more frequently with QTL for color retention than QTL for L * H values from the Hunter LabScan XE spectrophotometer . Hunter Labscan - derived L * H QTL from this study can be compared to previous studies by and Wright and Kelly ( 2011) and Cichy et al. 64 (2014) where both authors us ed a Hunter Labscan XE to measure QTL for L* respectively. The L* 2.1 76 QTL was detected by ImageJ analysis and located on Pv02 in Population 76. This QTL had a small, yet detectable effect on L * I and co - localized with QTL detected by visual color rating ( COL2.1 76 ) and b * I ( b* 2.1 76 ) . Overlapping this QTL, LOD scores were elevated for L * H and b * H values measured by the Hunter Labscan, but were below the significance threshold. Another lone L* QTL ( L * 3.1 76 ) mapped to a 2.02 - 2.4 3 Mb physical interval on Pv03. T his QTL co - localized with COL 3.1 76 and b* 3. 1 76 . The L* 3.1 76 QTL was significant, but did not have a major effect on L* value due to a low R 2 of 6.2% and low additive effect of 0 .40. Wright and Kelly ( 2011) detected a putative L * H QTL on Pv03 in 2005 that accounted for 21% of the phenotypic variation, but the nearest marker (F9R1.150) was not given a physical position. QTL L * 5.1 86 mapped to a physical region of 34.33 - 35.96 Mb on Pv05 . The peak LOD was located toward the right side of interval above th e significance threshold. While this QTL appear ed to be near the region associated with the two - year average color ratings (COL_2YA) , they are actually separated by 40 cM (4 Mb). Wright and Kelly ( 2011) identified a putative L * H QTL on Pv05 that was found in years 2005 and 2006. The 2005 QTL was nearest marker IAC96 and explained 10% of the phenotypic variation, while the 2006 QTL was nearest marker F22R1.400 and explained 13% of the phenotypic variation. A BLAST query of the SSR marker IAC96 against the v2.1 dry bean genome gave a top hit at the physical position near 3. 2 Mb, which is very distant from the physical interval of the L * 5.1 86 QTL detected in the present study. In the Cichy et al. (2014) study, putative L * H QTL ( L*10 and L*11 ) also mapped to Pv05 where they explained ap proximately 10% of the phenotypic variation and co - localized with QTL for b * and color ratings. The DArT marker sequences from the Cichy et al. (2014) study were provided by K. Cichy and 65 used in BLAST queries against the v2.1 dry bean gen ome. Based on the nearest marker , D33359 , t he L*10 QTL on Pv05 map s to 8.42 Mb , and the closest flanking markers are located several Mb away at 5.4 and 18.0 Mb. The nearest marker to the L*11 QTL on Pv05 was D23441 at 106.28 cM (18 Mb) ; however, this same genetic position was shared with markers that have physical positions as far away as 16.76 and 27.6 Mb. The physical intervals from the Cichy et al. ( 2014) study did not match the physical interval of the L* 5.1 86 QTL detected in the present study . Both studies have large gaps on Pv05 that require additional markers to more accurately determine QTL locations. L * QTL were found on Pv08 in both populations . As mentioned in the COL QTL section, L * QTL co - segregated with COL QTL around a 1.53 - 7.25 Mb region in both populations. Surprisingly , in Population 76, the region associated with the Hunter Labscan - derived values ( L* H_2017) was more tightly co - localized with COL _2017 and COL_2YA, while the reg ion associated with ImageJ - derived values ( L* I_2016 and L* I_2017) co - localized with color ratings from 2016 (COL _2016 ). In Population 86, ImageJ - derived measurements from both years ( L* I_2016 and L* I_2017 ) tightly co - localized with the COL8.1 86 QTL, while the Hunter Labscan - derived value s were unable to detect the QTL . T hree separate L * QTL w ere detected by ImageJ software that mapped to Pv09. In Population 76, L * 9.1 76 and L * 9.2 76 were detected in the absence of QTL for COL or other color measurements. While both QTL occupied tight intervals, L* 9.1 76 had a tight LOD peak from 10.30 - 10.32 Mb, and L * 9.2 76 had a more gradual peak from 13.55 - 13.71 Mb. The L * 9 .1 76 QTL had a higher LOD score, more explained phenotypic variation, and larger additive effect. In Population 86, L * 9.1 86 had a very small effect on the L* value and was noteworthy as the only L * QTL where the B14311 allele had an additive effect that darkened L* values. 66 Pv10 contained the L * 10.1 86 QTL that mapped to a very small physical interval from 41.96 - 42.01 Mb. However, its LOD scores had an extremely gradual ascent over a 50 cM region where no markers were present, which cast a level of uncertainty on its actual position. It was located slightly upstream of QTL for color rating (COL10.1 86 ) and a* value ( a * 10.1 86 ). L* QTL were found on Pv11 in both populations . In Population 76, the L* 11.1 76 QTL was comprised of regions from individual years ( L* I_2016, L* I_2017, and L* H_2017) that all explained less than 10% of the phenotypic variation, but neatly co - localized with the COL 11.1 76 QTL near 52.47 - 52.87 Mb. In Population 86, L * 11.1 86 was comprised of ImageJ - derived L* I_2016 and L* I_2017 regions that mapped to a 400 kb interval containing COL 11.1 86 and b * 11.1 86 QTL within its 52.47 - 52.87 Mb region. Wright and Kelly (2011) detected a putative L* H QTL C on Pv11 in 2007 near marker F5R10.475 that explained 9% of the phenotypic variation . Cichy et al. (2014) detected putative L * H QTL ( L*10 and L*11 ) near the proximal end of Pv11 that both explained approximately 25% of the phenotypic variation . However, r e - locating the Cichy et al. (2014) L * QTL to their correct physical positions at the distal end of Pv11 places them near 5 2.6 - 53.4 Mb in the v2.1 genome, which is similar to the physical region of the COL, L* , and b* QTL detected in the present study. a* QTL for a * were almost exclusively located on Pv08 where they co - localized with QTL for color ratings, L* values, and b* values across populations. Other instances on Pv03 and Pv05 show regions of elevated, but insignificant LOD scores that co - localiz ed with L* QTL. In Population 76, the a * 8.1 76 QTL was detected by all measurements of a* that mapped within a region from 0 .48 - 7.25 Mb. Strangely, the a* measurements from 2016 ImageJ ( a * I_2016 ) 67 mapped to two peaks on either end of the major color - determining region, but the first peak was more in - line with other color measurements and was subsequently the one selected for vis ualization in MapChart. Aside from th e 2016 measurements , a* I_2017 and a* H_2017 explained a surprisingly large amount of phenotypic variation for the trait ; values of R 2 for a* I_2017 and a* H_2017 wer e 27.0 % and 2 1.4 %, respectively. In Population 86 , the a* 8.1 86 QTL was also detected across all a* measurements, but mapped more closely with other color values than a* 8.1 76 from Population 76 [ Figure 1 3 H ] . Again, the range of explained phenotypic variation for the underlying regions associated with a* 8.1 86 was moderately large, from 17 - 21.4%. Interestingly, a* H measurements from the Hunter Labscan explained this trait as well as a* I measurements from the ImageJ software, which was not the case for QTL derived from the other color component measurements L* and b * . However, the a* value does not seem to be as strongly correlated with visual color ratings as L* and b* . [ Table 4 ]. A small - effect a* QTL ( a * 10.1 86 ) was detected by ImageJ in 201 7 RILs on Pv10 . The QTL mapped to a small physical interval of 42.22 - 42.29 Mb that was in between an L* QTL from 2016 ( L * 10.1 86 ) and a COL QTL from 2017 (COL10.1 86 ). Despite having minor effects, both the a* QTL and the COL QTL were the only instances of the B14311 parent contributing a beneficial additive effect on color retention. The B14311 a * 10.1 86 allele decreased the a* I value by 0 .31, while the COL10.1 86 allele i mproved color ratings by 0 .15. Cichy et al. (2014) mapped putative a* H QTL ( on Pv07 and Pv11 to regions containing clusters of QTL involved in color ratings. Interestingly, their a* H QTL consistently mapped adjacent to, not inside, these clusters, a pattern that was also observed for the a* H QTL detected in this study on Pv08. 68 b* QTL for b* were detected on Pv02 in Population 76 and Pv03, Pv08, and Pv11 in both populations [ Table 6 ] . The b* value measures color on a scale of blue to yellow, where smaller values are more blue and larger values are more yellow. On Pv02, a single small - effect b* QTL from 2017 ( b * 2.1 76 ) co - localized with QTL for color ratings (COL 2.1 76 ) and L* ( L * 2.1 76 ). The Hunter Labscan measurement of b* H was unable to detect a significant QTL, though elevated LOD scores were found over the same interval, similar to what happened with L* H from the Hunter Labscan [ Figure 1 2 B ] . This is evidence that ImageJ software is generally more accurate at detecting color retention QTL than the spectrophotometer. A b * QTL mapped near the top of chromosome Pv03 in both populations. Across the two popul ations, both b* QTL shared the flanking marker ss715646396 at 1.19 Mb. In Population 76, b * 3.1 76 mapped to an interval of 1.00 - 1.19 Mb slightly upstream of QTL for COL and L * . In Population 86, b * 3.1 86 tightly co - localized with QTL for COL and APP over the interval 1.19 - 1.30 Mb. The b* 3.1 76 QTL from Population 76 explained nearly double the phenotypic variation compared to the b* 3.1 86 QTL from Population 86, but both had very small effects on this trait. Although Cichy et al. (2014) detected a putative b* H QTL ( b*11 ) on Pv05 that co - localized with QTL for color, L * , and anthocyanin content of canned beans, no b* QTL were found on Pv05 in the present study. Significant LOD scores for all measurements of b* were detected on Pv08 , resulting in b* QTL for both populations . Both t he b* 8.1 76 QTL from Population 76 and the b* 8.1 86 QTL from Population 86 co - localized with QTL for color ratings, a* values, and L* values. In Population 76, the b* QTL mapped more closely to the COL QTL than the L* QTL, while in Population 86, both b* and L* QTL mapped to the exact same intervals. Due to the absence of markers in the regions 69 1.58 - 6.27 Mb and 6.27 - 53.59 Mb, it remains unclear if the b* QTL on Pv08 are identical or if there are multiple QTL contributing to b* in this interval. Pv11 contained b * QTL that were detected by all methods and years , similar to Pv08 . These QTL , b * 11.1 76 in Population 76 and b * 11.1 86 in Population 86 , had component regions explain ing a range of phenotypic variation from 11.4% to over 27%. The individual measurements ass ociated with b* in this region mapped to a similar or smaller interval than COL or L* QTL. Even b * H values from the HunterLab spectrophotometer were strongly significant, though they explained a lesser amount of phenotypic variation compared to the b* I QTL from the ImageJ analysis . QTL for b* were mapped in both populations and co - localized with color ratings to a small physical interval on the distal end of Pv11 . Cichy et al. (2014) also identified putative b * H QTL on Pv11 ( b*10 and b*11 ) that co - localized with QTL for color ratings ( color10 and color11 ) and putative L * H values ( L*10 and L*11 ). In that study, the putative b * H QTL explained a larger amount of phenotypic variation (33 - 44%) than determined in the present study. As mentioned previously, these QTL were all placed near the top of Pv11, but should be re - positioned to the distal end of the chromosome. QTL M apping of O ther C anning Q uality T raits A complete list of QTL for appearance, washed and drained weight, and texture is located in Table 7 . QTL graphs from Population 76 are located in Figure 12 , and QTL graphs from Population 86 are located in Figure 13 . 70 Appearance Ratings (APP) Recently, n ew methods of canning evaluation have been implemented by the MSU breeding program that differ from previous canning quality stud ies . Walters et al. ( 1997 ) and Posa - Macalincag et al. ( 2002) used a 1 - kidney beans, respectively. In Wright and Kelly (2011) , canned bean appearance ( visual appearance ) was rated by reviewers on a 1 - also factored in the perceived darkness of the seed coat color. Similarly, reviewers from Cichy et al. (2014) rated canned bean appearance ( overall appearance ) on a 1 - included seed coat color. In the present study, appearance a nd color were rated as distinct characteristics on 1 - 5 scales based on seed coat integrity and perceived darkness of seed coat color , respectively [ Figure 4 ]. In most instances, APP QTL typically mapped independently of COL QTL and were detected on many chromosomes. Furthermore, the percentage of phenotypic variation explained by any one APP QTL was always less than 10%. Interestingly, the poor - canning parent B1 4311 was responsible for improved appearance ratings in four out of nine detected APP QTL. APP QTL on Pv02 differed across years, populations, and physical positions, which weakens their reliability. For example, APP2.1 76 was detected in 2017 and mapped in isolation near 40 Mb, while APP2.1 86 was detected in 2016 and co - localized with QTL for 2016 texture (TXT2.1 86 ) and 2016 washed and drained weight (WDW2.1 86 ) near 4 Mb. The co - localization of these QTL in Population 86 lends validity to this region becaus e the QTL involved could realistically be explaining a shared physiological mechanism. When the allele was contributed by B14311, APP2.1 86 decreased appearance scores (more split beans), TXT2.1 86 decreased texture 71 scores (mushier beans), and WDW2.1 86 increased washed and drained weights (more hydrated beans). Appearance ratings for 2016 and the two - year average were combined into the APP3.1 86 QTL that co - localized with QTL for COL and b* on Pv03. These APP QTL mapped to a relatively tight physical in terval of 1.2 - 1.3 Mb, but explained very little phenotypic variation. An APP QTL on Pv04 (APP4.1 86 ) mapped to a small physical interval from 2.75 - 2.89 Mb. Unexpectedly, it co - localized with QTL for several agronomic traits: height, desirability score and seed weight. A QTL for 2016 washed and drained weight s did not meet the permutation threshold, but showed a region of elevated LOD scores mapping slightly upstream of the APP4.1 86 QTL. Pv05 contained an APP QTL (APP5.1 76 ) over a 300 kb interval from 4.47 - 4.7 5 Mb. This APP QTL mapped in complete isolation of other QTL. Wright and Kelly (2011) also detected an appearance QTL on Pv05 that shared its nearest marker IAC96 with a putative L * H QTL ( reported C genome placed it near 3. 2 Mb, which is moderately cl ose to the physical position of the APP5.1 76 QTL detected in this study. The APP6.1 86 QTL on Pv06 mapped to the small physical interval of 28.97 - 29.04 Mb. The LOD scores for this QTL fluctuated around 2.4 LOD over the entire chromosome outside of the peak . Only 8 markers mapped to unique genetic positions on Pv08, so this QTL is somewhat questionable . In the Cichy et al. (2014) study, the authors found an isolated, yet highly significant APP QTL on Pv06 that expla ined nearly 20% of the phenotypic variation. A BLAST query of the closest DArT marker D17956 against the v2.1 dry bean genome returned a top hit near 18.74 Mb, which is approximately 10 Mb up stream of the APP QTL identified in the present study. 72 Two indep endent APP QTL (APP8.1 76 and APP8.2 76 ) mapped to Pv08 in Population 76 . APP8.1 76 co - localized with the washed and drained weight QTL WDW8.1 76 and also with QTL for agronomic traits BRZ, DF, SY, and SW . The co - localization of APP QTL and agronomic QTL was a lso observed on Pv04. These QTL may be located in gene - rich areas or exhibit pleiotropic effects. APP8.2 76 co - localized to the top of Pv08 along with many QTL related to color retention: color ratings, L* values, a* values, and b* values. Color ratings and appearance ratings we re weakly positively correlated (r = 0 .24) [ Table 4 ] and reviewers may inadvertently rate severely split beans with poor color ratings. Conversely, if the canned color is dark black, cotyledon staining Wright and Kelly (2011) detected a visual appearance QTL on Pv08 that was nearest marker TE1/6.340. This marker was also the closest marker for a putative L* H QTL ( Color ) and a seed weight QTL ( seed size ) . While both studies were somewhat limited by low marker number, it is very interesting to observe that canned appearance has co - segregated with seed weight in three different instances . The APP10.1 86 QTL mapped to Pv10 and co - localized with QTL for textu re (TXT10.1 86 ) and color (COL10.1 86 ) near the distal end of the chromosome. Although the map intervals between markers in this region were large, APP10.1 86 mapped to a relatively small region from 42.22 - 44.22 Mb. This was just the second instance of APP QT L co - localizing with COL QTL, which supports the practice of rating both traits separately. Cichy et al. (2014) identified an APP QTL on Pv11 ( app11 ) that mapped near putative QTL for COL, L * H , a * H , and b * H at the proximal end of the linkage group . In this case, co - localization of visual appearance and color ratings may have been confounded by rating color darkness of the seed coat. 73 Texture (TXT) Six TXT QTL were identified across four chromosomes. Two minor - e ffect TXT QTL on Pv02 mapped to opposite ends of the chromosome. When donated by B14311, TXT2.1 76 increased texture measurements by 2.64 kg, while the TXT2.1 86 decreased texture measurements by 2.25 kg. Both were detected in just one population - year and ex plained very little phenotypic variation, casting doubt on their validity. The TXT5.1 76 QTL on Pv05 was comprised of texture measurements from 2016 and 2017 that mapped to the exact same physical positions from 27.70 - 36.79 Mb. Both TXT_2016 and TXT_2017 h ad strong LOD scores, and the reproducibility across years lends credence to the QTL. Unfortunately, there was a large gap in marker coverage from 27.72 - 35.96 Mb, which would need to be addressed in future mapping projects. Wright and Kelly (2011) detected a TXT QTL on Pv06 near marker F8R2.350, which was not given a physical position. Cichy et al. (2014) detected a consistent TXT QTL over two years that co - localized with other canning quality measurements at the A sp locus on Pv07. TXT9.1 76 mapped to a large region on Pv09 from 16.26 - 39.61 cM (7.87 - 13.55 Mb). This interval overlapped with QTL for agronomic desirability score s and L* I values measured by ImageJ software. It was not surprising for this TXT QTL to overlap with unrelated QTL because of its large interval. The TXT10.1 86 QTL on Pv10 was detected in 2016 and mapped to a small interval at the distal end of the chromosome from 43.29 - 44.22 Mb. TXT10.1 86 co - localized with QTL for APP ( APP10.1 86 ), COL ( COL1 0.1 86 ), and also 2017 texture measurements that showed an elevated LOD score , but did not reach the significance threshold. Taking into account the high LOD score 74 (11.3), large R 2 , small physical interval, and potential for replication across years, the re gion near TXT10.1 86 may be useful in future research on the trait. Washed and D rained W eight (WDW) Only two WDW QTL were detected in this study, which was partly expected due to minimal variation for this trait in the parental lines. WDW2.1 86 mapped to t he top arm of Pv02 and spanned a very large physical interval from 3.90 - 30.15 Mb. As mentioned in the discussion section for APP QTL, WDW2.1 86 co - localized with QTL for APP and TXT. On Pv08, a WDW QTL (WDW8.1 76 ) also mapped near an APP QTL (APP8.1 76 ). Agai n, this co - localization is discussed in the previous section pertaining to APP QTL. Finer mapping and more contrasting phenotypes would be needed to refine the WDW QTL identified in this study. Wright and Kelly (2011) detected WDW QTL on Pv03 and Pv10, but did not provide physical positions. Cichy et al. (2014) detected QTL for a trait called washed drained coefficient ( WDC ). This trait was calculated by dividing a canned sample before canning. Two WDC QTL ( wdc10 and wdc11 ) mapped to the Asp locus on Pv07 along with many other QTL involved in water uptake. Parental lines in the present study both contained the recessive asp allele (matte seed coat) and the Asp locus was not a factor in this study . QTL M apping of A gronomic T raits This study was mostly concerned with canning quality traits so agronomic traits were only measured to identify RILs that might be useful for breeding goals. Parents were not greatly different for agronomic traits, and many of these traits are under complex genetic control, which limited detection of significant QTL. Furthermore, agronomic traits were only measured in one 75 environment (SVREC) and year ( 201 7 ) with significant, but mostly negligible effects. Considering these major limitations, small - effect QTL detected for agronomic traits may not be compelling enough for further validation. A complete list of QTL for agronomic traits is located in Table 8 . QTL graphs from Population 76 are located in Figure 1 2 , and QTL graphs from Population 86 are located in Figure 1 3 . Seed Y ield (SY) Although the parents differed slightly in seed yield, it was not a mapping priority in this study. Two SY QTL were detected on Pv08, each with high LOD scores , but a small effect on the trait. SY8.1 76 mapped to the distal end of the chromosome in Population 76, while SY8.1 86 mapped to the proximal end of the chromosome in Population 86. Interestingly, SY8.1 86 co - localized to the exact same interval as QTL for DS ( DS8.1 86 ) and HT (HT8.1 86 ) . In Population 86, SY8.1 76 mapped to a tight window from 60.07 - 60.56 Mb that may be of interest if genomic selection methods gain popularity in dry bean breeding. While both QTL are currently impractical as breeding targets, they may be useful to document for future agronomic studies . S eed W eight (SW) Measurements of seed weight can vary drastically for a genotype, even within year s or replication s . The parents in this study varied only slightly for seed weight in 2016 , but were by B12724, and then B14311 [ Table 2 ; Table 3 ]. A major SW QTL on Pv03 was found in both p opulations. SW3.1 76 and SW3.1 86 explained a large amount of phenotypic variation (R 2 = 22.7 and 12.2, respectively) and mapped to a similar 76 physical interval. SW3.1 76 had a physical region of 11.47 - 11.82 Mb , and SW3.1 86 had a physical region of 3.82 - 12.30 Mb that included the region of the SW QTL in Population 76 . These may be the same QTL detected across populations, although there were large gaps in marker coverage over the SW3.1 86 interval, which limited resolution . S W QTL on Pv04 were also detected in both populations. SW4.1 76 , SW4.1 86 , and SW4.2 86 had smaller effects than the SW QTL on Pv03, but still mapped closely together. Although the two SW QTL from Population 86 appear separate based on a LOD graph, these may a ctually be the same QTL. They were separated by a single marker ss715646243, that upon further inspection was separated from neighboring markers by a 20 cM gap despite differing in physical positions of less than 100 kb. This is evidence that the marker di d not fit well at its mapped location. When the marker was temporarily excluded, the two peaks were joined, but the right - most peak from SW4.1 86 was still the most prominent. A SW QTL on Pv05 (SW5.1 86 ) mapped to a relatively tight peak from 35.96 - 36.79 Mb. It had a minor effect on the trait and mapped to a similar location as L * 5.1 86 . The QTL SW7.1 76 was detected in the interval from 4.25 - 4.39 Mb on Pv07 . It spanned a large map distance that was slightly downstream of a QTL for canopy height, HT7.1 76 . A SW QTL on Pv08 (SW8.1 76 ) was found over a 200 kb region near 62 Mb. This QTL was part of a cluster that contained agronomic QTL (SY, DF, BRZ) and canning quality QTL (APP, WDW), all detected from RILs grown in 2017. Wright and Kelly (2011) found putati v e SW ( seed size ) QTL on Pv05, Pv06, Pv08, and Pv11 that were mostly year - dependent. Cichy et al. (2014) found three putative SW ( seedwt ) QTL on Pv08 over two years that mapped in isolation from each other. D Ar T m arkers near the Cichy et al. (2014) SW QTL on Pv08 were BLASTed against the v2.1 dry bean genome and returned top hits from 61 - 62 Mb, which is in the same physical 77 region as SW8.1 76 detected in the present study. As exemplified in other studies, SW QTL are strongly year - dependent. SW QTL from this study were detected in just one year and require further validation. Days to F lowering (DF) QTL were detected for DF on three chromosomes, only in Population 76. These QTL should be regarded as somewhat questionable, given that in 2017 the parents differed in flowering date by just one day. Both DF7.1 76 and DF11.1 76 were barely significant, with LOD scores of 3.04 and 2.99, respectively. DF7.1 76 mapped in isolation, while DF11.1 76 mapped just upstream of a large cluster that contained QTL for HT and many color retention - related measurements. Wright and Kelly (2011) w ere able to detect a DF QTL on Pv11 in all three years that consistently mapped with marker F17R8.420, but comparison of between the studies is impossible due to lack of physical positions. The DF QTL DF8.1 76 was the most interesting (and unex pected), given its high percentage of explained phenotypic variation (R 2 =20%). As mentioned previously, this QTL mapped from 60.97 - 61.30 Mb along with QTL for SY, SW, BRZ, APP, and WDW. This QTL was not detected in the black bean RIL population of Wright and Kelly (2011) , nor in the Middle American GWAS performed by Moghaddam et al. (2016) . However, Kamfwa et al. ( 2015) detected a significant SNP for DF, ss715646088, reported at a physical position of 57.73 Mb. This physical position is in the vicinity of the 60.97 - 61.30 Mb region for DF8.1 76 , but the v2.1 dry bean genome places it even closer at 61.16 Mb. Althou gh SNP ss715646088 was not included in linkage mapping, its updated physical position is directly between the flanking markers for DF8.1 76 , strongly validating the QTL. 78 Days to M aturity (DM) There was very little phenotypic variation for days to maturity among the parents (and RILs) in this study. [ Table 2 ; Table 3 ] . Nevertheless, a few QTL were detected that should be treated with a n appropr iate amount of skepticism . QTL DM2.1 76 co - localized with the height QTL HT2.1 76 . Both of these traits are involved in agronomic adaptation and may suggest a pleiotropic effect between them. This QTL was not found in other studies. A DM QTL on Pv04 (DM4.1 86 ) mapped in isolation to a small physical region from 2.2 - 2.4 Mb and had the largest effect for a DM QTL by explaining 11.6% of the phenotypic variation. Moghaddam et al. (2016) found significant SNPs for DM near the top of Pv04. In their supplementary material, a DM SNP on Pv04 mapped to 1 Mb, but other significant SNPs near the proximal end of Pv04 were apparent in Supplementary Figure S1f showing Ma nhattan plots of days to maturity among beans from the Mesoamerica n subpopulation . A DM QTL on Pv11 (DM11.1 86 ) was significant over a broad map ping interval, but small physical interval from 49.59 - 51.12. It is important to note that t h e Population 86 Pv11 linkage group was not representative of the entire chromosome; the first marker was located at 49.59 Mb and the last marker was located at 52.88 Mb. Moghaddam et al. (2016) detected GWAS peaks and candidate genes associate d with DM on Pv11 at 4. 3 Mb and 41 Mb. According to the authors, the peak at 4 Mb was detected in Nebraska and Michigan, while the peak at 41 Mb was detected in North Dakota. Additional mapping with a more complete linkage map is required to validate DM11.1 86 . 79 Canopy H eight (HT) Many small - effect QTL for canopy height were detected for the 2017 field season. Normally, this would be expected since height is widely considered to be under polygenic control. However, in 2017 the parental lines only differed in height by 0 .5 cm, as estimated on a per - plot basis. Lack of phenotypic variability calls into question the validity of these QTL. Most HT QTL in this study had an R 2 below 10%; QTL explaining more than 10% phenotypic variation will be otherwise noted. The QTL, HT1.1 86 mapp ed in isolation to an interval of 1.30 - 2.85 Mb on Pv01 . A few somewhat - significant peaks were observed in this region according to the first Manhattan plot in the S upplementa ry Figure S1r of Moghaddam et al. (2016) . Other studies did not detect significant markers on Pv01. On Pv02, HT2.1 76 mapped to a broad 6 Mb interval from 25 - 31 Mb that co - localized within a QTL for days to maturity, DM2.1 76 . This QTL has not been supported by other s tudies. Pv03 contained a HT QTL (HT3.1 76 ) from 10.69 - 11.25 Mb that mapped upstream of the SW3.1 76 QTL at 11.47 - 11.82 Mb. A significant peak was also detected near the top of the Manhattan plot for the Michigan location in the third plot of the Moghaddam et al. (2016) S upplementary Figure S1r. On Pv04, HT4.1 86 explained over 17% of the phenotypic variation for height in Population 86. Interestingly, this QTL co - localized with other 2017 QTL for seed weight, agronomic desirability score, and canned app earance ratings. It mapped to the physical interval 2.55 - 2.89 Mb, which is extremely close to 2.9 Mb where Moghaddam et al. (2016) found a GWAS peak by excluding shorter statured beans with type 1 (determinate) architecture. This height QTL is the most likely to be validated in other studies. 80 A HT QTL on Pv07 (HT7.1 76 ) was barely significant in this study and mapped to the proximal end of Pv07. This location does not match the position of the strong peak detected at 46 Mb by Moghaddam et al. (201 6) . The HT8.1 86 QTL was detected within the physical interval 0 .37 - 1.5 Mb on Pv08 that explained over 21% of the phenotypic variation. According to the third Manhattan plot of S upplementary Figure S1r in Moghaddam et al. (2016) , a significant peak for height was detected near the top of Pv08. Interestingly, this peak was only observed with Mesoamerican genotypes grown in Michigan, suggesting a QTL x environment interaction. On Pv11, HT11.1 76 mapped t o a 52.16 - 52.65 Mb region at the end of the chromosome along with QTL for DF and nearly every measurement of canned bean color. Moghaddam et al. (2016) describe d a height QTL at 43 Mb that contained several SNPs within a gene having pleiotropic effect on architectural traits. Wright and Kelly ( 2011) also detected a height QTL on Pv11 in 2004 linked that was linked to marker F17R8.420. Neither of these previously documented QTL give strong support to HT11.1 76 given its position. Desirability S core (DS) The agronomic desirability score is an in - house measurement used to guide breeding decisions. The score is visually assigned on a per - plot basis and factors in a variety of agronomic desirability scores and few QTL were detected for this trait in the mapping populations. The B14311 contributed a negative additive effect in all cases. Most DS QTL mapped to relatively tight physical positions. 81 On Pv02, DS2.1 76 mapped near 51 cM (5.86 - 7.10 Mb), which was upstream of QTL for color measurements that had left - most flanking markers near 54 cM (11.03 Mb). B14311 was responsible for negative effects toward both DS and COL, and the proximity of the QTL suggests a degree of genetic linkage. DS4.1 86 on Pv04 co - localized wi th QTL for height, seed weight, and canned appearance ratings to a 2.75 - 2.89 Mb interval. The B14311 allele lowered values for all co - localizing traits except canned appearance ratings. This region appears to be important for both agronomic and quality tra its, but values of R 2 were generally low , which may prohibit effective molecular markers. DS8.1 86 mapped to a physical interval from 0 .37 - 1.50 Mb and neatly co - localized with QTL for yield (SY8.1 76 ) and height (HT8.1 86 ). This was the strongest DS QTL detected according to its LOD score of 6.7 with an R 2 value of nearly 20%. This region is clearly important for key agronomic traits and suggests the desirability score can be a useful tool when rated with a skilled A DS QTL on Pv09 (DS9.1 76 ) had a tight LOD peak at 7.70 - 7.79 Mb. This QTL overlapped with TXT9.1 76 and was also near QTL for L * values, although these traits would seem to be unrelated. Ozone Bronzing (BRZ) Foliar bronzing attributed to ozone damage was observed in the 2017 field trials, however bronzing did not appear to be evenly distributed across the field. Since bronzing ratings were taken on each plot, they were included in QTL analysis in the spirit of curiosity. Several population - specific bronzing QTL from the 2017 growing season were detected. T 82 knowledge , this is the first time this trait has been mapped in dry beans , though v alidation in different years and populations is required. Ratings for foliar bronzing may have been confounded by CBB lesions, which may explain the proximity of bronzing QTL with known CBB QTL. QTL BRZ 5.1 86 mapped to a very tight physical interval on Pv05 from 39.24 - 39.34 Mb. This QTL co - localized within a QTL for canned bean color (COL5.1 86 ) and explained very little phenotypic variation. In Population 76, BRZ6.1 76 was distributed across 0.5 - 22.9 cM . This gene tic distance covered over half of the poorly - covered linkage group representing Pv06 (40.4 cM in length ). This QTL explained minimal phenotypic variation (R 2 = 6.3%). T he most - significant BRZ QTL ( BRZ 7.1 86 ) resided on Pv07 . With a peak LOD of 6.6 and a high R 2 of 30%, this QTL mapped to the tight physical interval from 3.9 9 - 4.17 Mb . This location is relatively close to the Phs locus that encodes for phaseolin, the main seed storage protein in dry beans (Ma and Bliss, 1978) . The Phs CDS from black beans identified by Diniz et al. ( 2014) were BLASTed against the v2.1 dry bean genome . Both queries returned top hits at approximately 5.026 Mb near a n annotated P. vulgaris gene, Phvul.007G059700 .1 . This gene encodes a Cupin family seed storage protein that also has an Arabidopsis thaliana homolog , At3g22640 . Several CBB QTL have been mapped near the Phs locus in three independent studies: (Nodari et al., 1993b; Miklas et al., 1996; Jung et al., 1996) . Interestingly, this bronzing QTL on Pv07 is extremely close to a CBB resistance locus near 4 Mb that is currently being fine - mapped ( P. Miklas, pers. comm.). This may suggest that bronzing ratings in the present study were confounded by presence of CBB or that there is a simi lar physiological response to these stresses. In Population 76, BRZ8.1 76 co - localized with QTL for APP, SY, SW, and DF at the proximal end of Pv08 near 61 Mb , however it did not explain much phenotypic variation . A major 83 CBB resistance QTL, SU91 is also located on Pv08 . Shi et al. (2012) found a soybean predicted protein homolog (UDP - glycosyltransferase) associated with a tepary bean EST near the SU 91 locus. A BLAST p query of the predicted protein sequence against the v2.1 dry bean genome returned t op hit s for genes Phvul.008G290200.1 and Phvul.008G290 3 00.1 located at 62.81 Mb and Phvul.008G26200.1 and Phvul.008G262100.1 located at 60.89 Mb . Recently, Lobaton et al. (2018) ha ve developed a KASP marker for the SU91 locus that has a physical position near 6 2.95 Mb in the v2.1 genome (B. Raatz, pers. c omm.). Although the BRZ8.1 76 QTL is in the same general region as the SU91 QTL, they are most likely different QTL since t he B14311 allele for BRZ8.1 76 was associated with increased foliar bronzing, whereas the B14311 allele for SU91 - CG11 was associated wi th a reduction in foliar symptoms of CBB. A BRZ QTL on Pv09 was also detected ( BRZ 9.1 86 ) that mapped between 31.40 - 33.38 Mb. Although no published bronzing QTL were found in this region, t wo CBB QTL have been mapped to this chromosome from mapping populations BAT93/Jalo EEP558 (Freyre et al., 1998; Gepts, 1999; Geffroy et al., 2000) and Belnab - RR - 1/A55 population (Ariyarathne et al., 1999; Jung et al., 2003; Fourie et al., 2004) . The se CBB QTL were not given physical positions to compare with BRZ9.1 86 . Common B acterial B light R esistance (CBB) Parental lines used in this study exhibited slight , but significant phenotypic variation in CBB resistanc e in 2017 [ Table 2 ] . N o CBB QTL were detected, though several BRZ QTL were detected near previously - identified CBB QTL . Parents were genotyped with the codominant SCAR marker SU91 - CG 11 developed by Shi et al. (2012) from the SCAR marker SU91 first published by Pedraza Garc í a et al. (1997) . Thi s marker is tightly linked to a major locus governing 84 CBB resis tance that was derived from tepary bean ( Phaseolus acutifolius ) PI 319433 via common bean XAN 159 (Miklas et al., 2003) . Parental lines B12724 and B14311 were shown to have the product size corresponding to CBB resistance , while Zenith did not [Supplemental Figure 8 ]. Although SU91 has been mapped to Pv08, no CBB QTL were detected by SNP markers on Pv08 in Population 76 (B14311/Zenith) that should have segregated at the SU91 locus. Absence of CBB QTL might be explained by mild or uneven disease pressure in the field , confounding effects of ozone - induced foliar bronzing, and the paucity of markers on the Pv08 linkage map (n=20). Molecular M arker A nalysis Parental lines were genotyped with InDel markers developed by Moghaddam et al. (2014) that were located near COL QTL on Pv08 [ Figure 1 4 A ]. This region contained QTL for post - processing color retention, L* , a* , and b* , but all spanned a large physical interval from approximately 1.5 - 7 Mb. The region from 5.43 - 7.13 Mb was selected for exploratory genotyping because several highly - significant SNPs mapped to the region. Seven markers polymorphic across navy, black, and li ght red kidney market classes were selected based on their proximity to the physical positions of SNPs included on the BARCBean6k_3 BeadChip. InDel markers from non - black market classes were included because there were not many in the region, with only fiv e InDel markers spanning the 6.00 to 7.00 Mb interval. The B14311 parent was not polymorphic to both parents for any marker ; however , B12724 and Zenith were polymorphic to the other genotypes for markers NDSU_IND_8_6.2923. and NDSU_IND_8_7.0078. Although n one of the tested markers were able to discriminate B14311 against the other parents, fine - mapping the 1.5 - 7 Mb interval could refine detected QTL or uncover multiple QTL. 85 Parental lines were also genotyped with InDel markers developed by Moghaddam et al. (2014) that were located near COL QTL at the distal end of Pv11 [ Figure 1 4 B ]. Six markers polymorphic in the black bean market class were selected based on their proximity to the physical positions of SNPs included on the BARCBean6k_3 BeadChip. The B14311 parent was polymorphic to the other parents for three of these markers, tho ugh these three markers were located outside of the physical region from 52.16 - 52.84 Mb. Within the region of interest, B14311 had the same product size as Zenith and B12724, except for marker NDSU_IND_11.487818 where B14311 and Zenith had a larger product size than B12724. Additional InDel markers developed by Moghaddam et al. (2014) are polymorphic in other market classes and are located near this region, but their usefulness in genotyping black beans for color retention is unteste d. CONCLUSIONS Black beans are an increasingly popular dietary option for US consumers. In order to meet consumer demand for a bean that remains dark black after processing, it is necessary to explore improvements on both phenotypic and gen et ic aspects of this trait. In this study, a novel and comprehensive method of phenotyping canned bean color via digital image analysis was developed . Extracting CIELAB color values from canned bean photographs eliminate d many confounding factors associated with traditional phenotyping such as high reflectance, small sample size, and time - consuming ( and often highly - variable) reviewer ratings . On the genetic side , t he RIL mapping populations that were developed through this research were used to i dentify regions in the dry bean genome associated with color retention. Many small - effect QTL were detected for black bean post - processing color retention, supporting previous research that this trait is under polygenic control. Several of these QTL co - loc alized to the same genomic regions on Pv03, Pv08, 86 and Pv11 across years, phenotyping methods and populations, while other QTL were population - or year - dependent and require additional validation. QTL for other canning quality traits and agronomic traits we re also detected. In most cases, marker development for these QTL is impractical for two reasons: most QTL explained only a small amount of the total phenotypic variation and lack of markers limited the resolution of mapping intervals. That said, if marker s were to be developed for color retention QTL, they may be useful to screen early generation material for canning quality potential. Those r egions on Pv03, Pv08, and Pv11 where visual color ratings co - localize with quantitative color values would be the best areas to target for molecular marker development. The region from 52.5 - 52.9 Mb on Pv11 shows potential for molecular marker development due to a high R 2 and small physical interval. Alternatively, genomic selection for canning quality traits would be an interesting continuation of this research, as many small - effect loci are involved that may not always be detectable from year to year. This study gives dry bean breeders and scientists a better understanding of the genetics controlling color retention s o that they can generate darker - colored processed black beans to meet the growing consumer demand. 87 APPENDIX 88 Table 1. Agronomic and canning quality t raits guiding parental selection for black bean RIL populations. Parental lines are bolded and shaded according to color of their canned seed. *Zenith and B12724 have excellent color retention, while a black bean variety widely - grown in Michigan. Data : MSU 2015 Standard Black Bean Yield Trial. Abbreviations: SY: seed yield, SW: 100 - seed weight, DF : days to flowering, DM: days to maturity, H T: canopy height, DS: desirability score, CBB: common bacterial blight resistance, ANT 73 : resistance to anthracn ose race 73 , COL: canned color rating, APP: canned appearance rating, WDW: washed and drained weight; TXT: texture ; BRZ: ozone bronzing (not measured in 2015) Agronomic Traits Canning Traits Parent Pedigree SY SW DF DM HT DS CBB ANT 73 COL APP WDW TXT (kg/ha) (g) (days) (days) (cm) (1 - 7) (1 - 5) (R/S) (1 - 5) (1 - 5) (g) (kg) B14303 B09197/B11334 3571 18.9 45 96 51 5.8 1 R 1.7 2.8 256.8 40 B14302 B09197/B11334 3386 18.2 45 97 51.3 5.8 1 R 2.2 2.5 256.3 36 B14311 B11338/B10241 2907 18.7 45 96 48.3 5 1 S 1.7 255.4 34 Zenith B04644/ZORRO 2803 22.4 44 96 50.5 4.8 4.3 R 5* 4.2 255.8 29 B12724 B09184/B09135 2638 21.2 45 101 49.3 3.5 1.8 R 4.8* 3.5 257.4 35 (ZORRO) B00103*/X00822 2211 19.4 45 97 50.3 4.3 4.3 S 3.5 3.3 262.2 36 Mean (n=30) 2856 20.3 44.9 95.8 48.5 4.1 3 LSD (.05) 424 1.1 0.7 1.7 1.5 0.6 0.7 CV (%) 12.6 4.5 0.9 1.5 2.7 13.4 18.8 89 Table 2 . Phenotypic variation in canning quality and agronomic traits for Popula tion 76 (B14311/Zenith) . Population 76 B14311 Zenith RILs Trait Mean Mean p value Mean ± SD Range p value Color (1 - 5) 1 1.37 4.91 <.0001 3.13 ± 0.56 1.54 - 4.55 <.0001 L* I 2 17.07 10.49 <.0001 12.72 ± 1.27 9.6 - 16.41 <.0001 a* I 7.93 3.49 <.0001 5.34 ± 0.91 3.16 - 7.72 <.0001 b* I 9.18 1.90 <.0001 5.85 ± 1.22 2.39 - 8.71 <.0001 L* H 19.08 12.64 N/A 16.32 ± 1.34 11.89 - 20.11 N/A a* H 8.50 3.29 N/A 5.73 ± 0.81 4.13 - 8.46 N/A b* H 6.62 1.42 N/A 3.65 ± 0.93 1.54 - 6.24 N/A Appearance (1 - 5) 3 2.81 3.91 0.0012 3.12 ± 0.33 2.28 - 3.88 <.0001 Washed and drained weight (g) 4 255.2 261.3 0.2783 258.6 ± 4.2 247.9 - 274.8 0.3268 Texture (kg) 5 54.9 43.8 0.0078 53.2 ± 4.4 41.5 - 62.1 <.0001 Seed yield (kg/ha) 3328 2977 0.761 3073 ± 351 1326 - 3776 0.0013 Seed weight (g/100 seeds) 20.7 25.1 <.0001 21.9 ± 1.5 18.2 - 25.6 <.0001 Days to flowering 46 47 0.0769 46.8 ± 1 45 - 50 <.0001 Days to maturity 93 95 0.0491 93.2 ± 1.2 91 - 98 <.0001 Lodging (1 - 5) 1 1 1 1 ± 0 1 - 1 1 Height (cm) 46.5 46.5 1 46.1 ± 1.3 43 - 49 <.0001 Desirability score (1 - 5) 4.5 4.5 1 4.4 ± 0.5 3 - 6 0.0141 Bronzing (1 - 5) 2 2 0.392 2.3 ± 0.9 1 - 5 <.0001 CBB (1 - 5) 1 3 0.0026 2.2 ± 0.7 1 - 4.5 <.0001 Means of canning traits are listed as two - year averages, except for Hunter Labscan - derived L* H, a* H, and b* H values which were only measured on 2017 samples. Means of agronomic traits are listed as plot averages from 2017. 1 Color ratings of canned beans were averaged across reviewers on 2016 and 2017 samples. 2 L* , a* , b* : CIELAB color values, where L* measures lightness, a* measures greenness/redness, b* measures blueness/yellowness. 'I' indicates values measured by ImageJ analysis and 'H' indicates values measured by a Hunter Labscan. 3 Appearance ratings of canned beans were averaged across reviewers on 2016 and 2017 samples. 4 Washed and drained weights were measured after briefly rinsing canned beans under cool water. 5 Texture was measured as the peak force (kg) required to compress a 100 g sample of canned beans. 90 Ta ble 3 . Phenotypic variation in canning quality and agronomic traits for Population 8 6 (B14311/B12724) . Population 86 B14311 B12724 RILs Trait Mean Mean p value Mean ± SD Range p value Color (1 - 5) 1 1.58 4.93 <.0001 3.25 ± 0.64 1.72 - 4.96 <.0001 L* I 2 15.97 10.23 <.0001 12.11 ± 1.4 8.4 - 15.95 <.0001 a* I 6.60 2.90 <.0001 4.63 ± 0.86 2.2 - 6.68 <.0001 b* I 8.42 1.53 <.0001 5.3 ± 1.42 1.73 - 8.54 <.0001 L* H 16.44 9.95 N/A 11.92 ± 1.5 8.56 - 16.45 N/A a* H 6.87 3.51 N/A 4.69 ± 0.97 1.88 - 7.05 N/A b* H 8.52 1.92 N/A 5.26 ± 1.46 1.51 - 8.77 N/A Appearance (1 - 5) 3 3.15 3.72 0.06 3.22 ± 0.4 2.19 - 4.01 <.0001 Washed and drained weight (g) 4 250.2 255.6 0.7409 255.7 ± 4.5 244.6 - 266.8 0.7922 Texture (kg) 5 61.8 63.0 0.8382 59.5 ± 5.7 44.8 - 72.3 0.0001 Seed yield (kg/ha) 3346 3690 0.5704 3315 ± 314 2285 - 3959 0.0442 Seed weight (g/100 seeds) 21.8 23.1 0.1257 22.3 ± 1.1 19.7 - 26.2 <.0001 Days to flowering 45 47 0.1098 46 ± 0.9 44 - 49 0.001 Days to maturity 93 93 1 93.3 ± 1.1 91 - 96 <.0001 Lodging (1 - 5) 1 1 1 1 ± 0 1 - 1.5 0.5 Height (cm) 47 47 1 46.7 ± 1.2 44 - 49.5 <.0001 Desirability score (1 - 5) 5.5 5 0.3744 4.6 ± 0.5 3.5 - 6 0.0233 Bronzing (1 - 5) 3 1 0.0039 1.6 ± 0.6 1 - 3 0.0001 CBB (1 - 5) N/A N/A N/A N/A N/A N/A Means of canning traits are listed as two - year averages, except for Hunter Labscan - derived L* H, a* H, and b* H values which were only measured on 2017 samples. Means of agronomic traits are listed as plot averages from 2017. 1 Color ratings of canned beans w ere averaged across reviewers on 2016 and 2017 samples. 2 L* , a* , b* : CIELAB color values, where L* measures lightness, a* measures greenness/redness, b* measures blueness/yellowness. 'I' indicates values were measured by ImageJ analysis and 'H' indicates values were measured by a Hunter Labscan. 3 Appearance ratings of canned beans were averaged across reviewers on 2016 and 2017 samples. 4 Washed and drained weights were measured after briefly rinsing canned beans under cool water. 5 Texture was measured as the peak force (kg) required to compress a 100 g sample of canned beans. 91 Table 4. Correlation matrix for canning quality traits in two black bean RIL populations . Correlations for Population 76 (B14311/Zenith) are on the left axi s, while correlations for Population 86 (B14311/B12724) are on the top axis. Abbreviations for traits are given in Table 1. 'I' indicates values measured by ImageJ analysis and 'H' indicates values measured by a Hunter Labscan . Pearson correlation coeffici ents and p - values are given for each comparison. 92 Table 5. Correlation matrix for agronomic and selected canning quality t raits in two black bean RIL populations . Correlations for Population 76 (B14311/Zenith) are on the left axis, while correlations for Population 86 (B14311/B12724) are on the top axis. Abbreviations for traits are given in Table 1. 'I' indicates values measured by ImageJ analysis and 'H' indicates values measured by a Hunter Labscan . Pearson correlation coefficients and p - values are given for each comparison. 93 Table 6. QTL for measurements of post - processing color retention in two black bean RIL populations . QTL Chr 1 Year Peak Pos (cM) 2 Peak LOD 3 R 2 (%) 4 a 5 Map interval (cM) Physical interval (Mb) Left - flanking SNP Right - flanking SNP Color rating COL2.1 76 Flanking region 3.2 - 3.7 4.9 - 6.2 - (0.14 - 0.16) 54.0 - 74.1 11.03 - 17.24 ss715649088 ss715651061 Shared region 61.1 - 74.1 12.78 - 17.24 ss715649961 ss715651061 COL_2017 2 2017 63.1 3.7 6.2 - 0.16 54.0 - 74.1 11.03 - 17.24 ss715649088 ss715651061 COL_2YA 2 2YA 64.1 3.2 4.9 - 0.14 61.1 - 74.1 12.78 - 17.24 ss715649961 ss715651061 COL3.1 76 Shared region 3.5 - 4.0 6.3 - (0.15 - 0.16) 26.4 - 32.8 2.02 - 2.43 ss715646879 ss715647570 COL_2016 3 2016 26.4 3.5 6.3 - 0.16 26.4 - 32.8 2.02 - 2.43 ss715646879 ss715647570 COL_2YA 3 2YA 26.4 4.0 6.3 - 0.15 26.4 - 32.8 2.02 - 2.43 ss715646879 ss715647570 COL3.1 86 3 2017 0.0 2.8 3.6 - 0.13 0 - 1.51 1.19 - 1.30 ss715646396 ss715646392 COL3.2 86 3 2016 102.4 2.9 4.4 - 0.17 102.4 - 121.7 47.28 - 50.39 ss715650580 ss715647338 COL5.1 86 5 2017 170.5 3.1 4.3 - 0.14 167.5 - 170.7 38.84 - 38.92 ss715645449 ss715645459 COL8.1 76 Flanking region 5.8 - 7.4 11.8 - 16.2 - (0.23 - 0.26) 16.8 - 62.3 1.53 - 7.25 ss715647112 ss715648559 Shared region 17.5 - 62.3 1.54 - 7.25 ss715647113 ss715648559 COL_2016 8 2016 41.6 5.8 16.2 - 0.26 16.8 - 62.3 1.53 - 7.254 ss715647112 ss715648559 COL_2017 8 2017 53.4 7.4 13.0 - 0.25 17.5 - 62.3 1.54 - 7.254 ss715647113 ss715648559 COL_2YA 8 2YA 53.4 7.0 11.8 - 0.23 17.5 - 62.3 1.54 - 7.254 ss715647113 ss715648559 94 Table 6. (cont d) QTL Chr 1 Year Peak Pos (cM) 2 Peak LOD 3 R 2 (%) 4 a 5 Map interval (cM) Physical interval (Mb) Left - flanking SNP Right - flanking SNP COL8.1 86 Flanking region 6.7 - 8.0 12.3 - 13.7 - (0.26 - 0.27) 15.3 - 60.4 1.57 - 53.68 ss715647115 ss715648232 Shared region 15.3 - 40.1 1.57 - 6.27 ss715647115 ss715647905 COL_2016 8 2016 39.8 6.7 12.3 - 0.27 15.3 - 60.4 1.57 - 53.68 ss715647115 ss715648232 COL_2017 8 2017 35.8 7.0 13.7 - 0.27 15.3 - 40.1 1.57 - 6.27 ss715647115 ss715647905 COL_2YA 8 2YA 39.8 8.0 13.4 - 0.26 15.3 - 60.4 1.57 - 53.68 ss715647115 ss715648232 COL10.1 86 10 2017 75.7 2.9 4.5 0.15 69.7 - 82.1 42.22 - 43.29 ss715645524 ss715645501 COL11.1 76 Shared region 4.8 - 7.4 8.4 - 14.3 - (0.19 - 0.25) 144.3 - 150.9 52.16 - 52.84 ss715649459 ss715640405 COL_2016 11 2016 149.3 7.4 14.3 - 0.25 144.3 - 149.6 52.16 - 52.65 ss715649459 ss715650816 COL_2017 11 2017 149.6 4.8 8.4 - 0.19 149.6 - 150.6 52.65 - 52.84 ss715650816 ss715649382 COL_2YA 11 2YA 149.6 7.0 11.9 - 0.21 149.6 - 150.9 52.65 - 52.84 ss715650816 ss715640405 COL11.1 86 Shared region 10.6 - 13.0 18.7 - 22.0 - (0.31 - 0.32) 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 COL_2016 11 2016 28.4 10.6 18.7 - 0.31 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 COL_2017 11 2017 28.4 12.7 22.0 - 0.32 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 COL_2YA 11 2YA 28.4 13.0 21.2 - 0.31 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 L* value L* 2.1 76 2 2017 61.6 4.7 8.8 0.43 61.1 - 74.1 12.78 - 17.24 ss715649961 ss715651061 L* 3.1 76 3 2017 26.4 3.1 5.2 0.33 26.4 - 32.8 2.02 - 2.43 ss715646879 ss715647570 L* 5.1 86 5 2017 125.9 3.8 6.2 0.40 121.9 - 126.9 34.33 - 35.96 ss715647683 ss715639578 95 Table 6. (cont d) QTL Chr 1 Year Peak Pos (cM) 2 Peak LOD 3 R 2 (%) 4 a 5 Map interval (cM) Physical interval (Mb) Left - flanking SNP Right - flanking SNP L* 8.1 76 Shared region 4.1 - 7.7 8.4 - 14.4 0.43 - 0.60 0 - 62.3 0.48 - 7.25 ss715646686 ss715648559 L* I_2016 8 2016 15.0 4.1 8.4 0.43 0 - 15.2 0.484 - 1.41 ss715646686 ss715647114 L* I_2017 8 2017 52.4 7.7 14.4 0.60 16.8 - 62.3 1.53 - 7.25 ss715647112 ss715648559 L* H_2017 8 2017 51.4 4.2 10.5 0.47 17.54 - 62.3 1.54 - 7.25 ss715647113 ss715648559 L* 8.1 86 Flanking region 4.4 - 7.8 8.4 - 16.7 0.49 - 0.67 15.3 - 40.1 1.57 - 6.27 ss715647115 ss715647905 Shared region 15.8 - 40.1 1.58 - 6.27 ss715647116 ss715647905 L* I_2016 8 2016 39.83 4.4 8.4 0.49 15.8 - 40.1 1.58 - 6.27 ss715647116 ss715647905 L* I_2017 8 2017 38.83 7.8 16.7 0.67 15.3 - 40.1 1.57 - 6.27 ss715647115 ss715647905 L* 9.1 76 9 2016 27.0 6.8 13.9 0.53 27.0 - 27.1 10.30 - 10.32 ss715646178 ss715646179 L* 9.2 76 9 2017 39.6 3.1 4.8 0.31 39.6 - 40.1 13.55 - 13.71 ss715647980 ss715647985 L* 9.1 86 9 2016 2.7 3.5 5.5 - 0.38 2.6 - 5.8 27.58 - 29.10 ss715647620 ss715649156 L* 10.1 86 10 2016 66.6 3.8 6.8 0.48 65.6 - 67.3 41.96 - 42.01 ss715645508 ss715645510 L* 11.1 76 Shared region 3.0 - 4.5 7.3 - 8.8 0.41 - 0.43 149.6 - 154.0 52.65 - 52.87 ss715650816 ss715650160 L* I_2016 11 2016 149.6 4.5 8.8 0.43 149.6 - 150.6 52.65 - 52.84 ss715650816 ss715649382 L* I_2017 11 2017 149.6 4.4 8.1 0.41 149.6 - 150.6 52.65 - 52.84 ss715650816 ss715649382 L* H_2017 11 2017 150.9 3.0 7.3 0.41 150.9 - 154.0 52.84 - 52.87 ss715640405 ss715650160 96 Table 6. (cont d) QTL Chr 1 Year Peak Pos (cM) 2 Peak LOD 3 R 2 (%) 4 a 5 Map interval (cM) Physical interval (Mb) Left - flanking SNP Right - flanking SNP L* 11.1 86 Flanking region 8.1 - 8.9 15.0 - 18.0 0.62 - 0.68 22.9 - 30.8 52.47 - 52.87 ss715648350 ss715650160 Shared region 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 L* I_2016 11 2016 30.5 8.9 18.0 0.68 22.9 - 30.8 52.47 - 52.87 ss715648350 ss715650160 L* I_2017 11 2017 28.4 8.1 15.0 0.62 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 a* value a* 8.1 76 Flanking region 3.9 - 9.8 8.6 - 27.0 0.34 - 0.57 0 - 62.3 0.48 - 7.25 ss715646686 ss715648559 Shared region 15.2 - 17.5 1.41 - 1.54 ss715647114 ss715647113 a* I_2016 8 2016 15.0 3.9 8.6 0.34 0 - 17.5 0.48 - 1.54 ss715646686 ss715647113 a* I_2017 8 2017 23.6 9.8 27.0 0.57 15.2 - 62.3 1.41 - 7.25 ss715647114 ss715648559 a* H_2017 8 2017 36.6 7.2 21.4 0.40 16.8 - 62.3 1.53 - 7.25 ss715647112 ss715648559 a* 8.1 86 Flanking region 6.6 - 8.7 17.3 - 20.7 0.46 - 0.55 15.3 - 60.4 1.53 - 53.68 ss715647115 ss715648232 Shared region 15.3 - 40.1 1.57 - 6.27 ss715647115 ss715647905 a* I_2016 8 2016 39.8 8.7 20.7 0.49 15.3 - 60.4 1.57 - 53.68 ss715647115 ss715648232 a* I_2017 8 2017 35.8 6.6 17.3 0.46 15.3 - 40.1 1.57 - 6.27 ss715647115 ss715647905 a* H_2017 8 2017 37.8 7.7 20.6 0.55 15.3 - 40.1 1.57 - 6.27 ss715647115 ss715647905 a* 10.1 86 10 2017 69.7 3.8 8.3 - 0.31 69.7 - 82.1 42.22 - 43.29 ss715645524 ss715645501 b* value b* 2.1 76 2 2017 64.1 4.4 8.6 0.41 54.0 - 74.1 11.03 - 17.24 ss715649088 ss715651061 b* 3.1 76 3 2016 14.2 3.5 7.2 0.41 3.4 - 14.7 1.01 - 1.19 ss715650435 ss715646396 b* 3.1 86 3 2017 0.0 2.7 3.9 0.31 0 - 1.5 1.19 - 1.30 ss715646396 ss715646392 97 Table 6. (cont d) QTL Chr 1 Year Peak Pos (cM) 2 Peak LOD 3 R 2 (%) 4 a 5 Map interval (cM) Physical interval (Mb) Left - flanking SNP Right - flanking SNP b* 8.1 76 Flanking region 3.5 - 7.8 6.7 - 16.4 0.39 - 0.47 16.8 - 62.3 1.53 - 7.25 ss715647112 ss715648559 Shared region 17.5 - 54.1 1.54 - 7.04 ss715647113 ss715640331 b* I_2016 8 2016 51.4 4.9 10.8 0.47 17.5 - 50.4 1.54 - 6.00 ss715647113 ss715648337 b* I_2017 8 2017 53.4 3.5 6.7 0.39 50.4 - 54.1 6.00 - 7.04 ss715648337 ss715640331 b* H_2017 8 2017 52.4 7.8 16.4 0.41 16.8 - 62.3 1.53 - 7.25 ss715647112 ss715648559 b* 8.1 86 Flanking region 4.3 - 7.0 9.1 - 12.6 0.45 - 0.55 15.3 - 60.4 1.57 - 53.68 ss715647115 ss715648232 Shared region 15.8 - 40.1 1.58 - 6.27 ss715647116 ss715647905 b* I_2016 8 2016 38.8 7.0 11.7 0.55 15.3 - 60.37 1.57 - 53.68 ss715647115 ss715648232 b* I_2017 8 2017 36.8 4.3 9.1 0.49 15.8 - 40.1 1.58 - 6.27 ss715647116 ss715647905 b* H_2017 8 2017 39.8 6.8 12.6 0.45 15.3 - 40.1 1.57 - 6.27 ss715647115 ss715647905 b* 11.1 76 Flanking region 5.3 - 8.8 11.4 - 20.7 0.31 - 0.66 123.3 - 150.9 51.12 - 52.84 ss715649251 ss715640405 Shared region 144.3 - 149.6 52.16 - 52.65 ss715649459 ss715650816 b* I_2016 11 2016 149.3 8.8 20.7 0.66 144.3 - 149.6 52.16 - 52.65 ss715649459 ss715650816 b* I_2017 11 2017 149.3 6.9 13.9 0.51 123.3 - 149.6 51.12 - 52.65 ss715649251 ss715650816 b* H_2017 11 2017 149.3 5.3 11.4 0.34 144.3 - 150.9 52.16 - 52.84 ss715649459 ss715640405 b* 11.1 86 Shared region 13.3 - 14.6 25.9 - 27.3 0.62 - 0.81 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 b* I_2016 11 2016 28.4 14.6 26.3 0.79 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 b* I_2017 11 2017 28.4 13.3 27.3 0.81 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 b* H_2017 11 2017 28.4 13.5 25.9 0.62 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 98 Table 6. (cont d) QTL Chr 1 Year Peak Pos (cM) 2 Peak LOD 3 R 2 (%) 4 a 5 Map interval (cM) Physical interval (Mb) Left - flanking SNP Right - flanking SNP L* 11.1 86 Flanking region 8.1 - 8.9 15.0 - 18.0 0.62 - 0.68 22.9 - 30.8 52.47 - 52.87 ss715648350 ss715650160 Shared region 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 L* I_2016 11 2016 30.5 8.9 18.0 0.68 22.9 - 30.8 52.47 - 52.87 ss715648350 ss715650160 L* I_2017 11 2017 28.4 8.1 15.0 0.62 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 a* value a* 8.1 76 Flanking region 3.9 - 9.8 8.6 - 27.0 0.34 - 0.57 0 - 62.3 0.48 - 7.25 ss715646686 ss715648559 Shared region 15.2 - 17.5 1.41 - 1.54 ss715647114 ss715647113 a* I_2016 8 2016 15.0 3.9 8.6 0.34 0 - 17.5 0.48 - 1.54 ss715646686 ss715647113 a* I_2017 8 2017 23.6 9.8 27.0 0.57 15.2 - 62.3 1.41 - 7.25 ss715647114 ss715648559 a* H_2017 8 2017 36.6 7.2 21.4 0.40 16.8 - 62.3 1.53 - 7.25 ss715647112 ss715648559 a* 8.1 86 Flanking region 6.6 - 8.7 17.3 - 20.7 0.46 - 0.55 15.3 - 60.4 1.53 - 53.68 ss715647115 ss715648232 Shared region 15.3 - 40.1 1.57 - 6.27 ss715647115 ss715647905 a* I_2016 8 2016 39.8 8.7 20.7 0.49 15.3 - 60.4 1.57 - 53.68 ss715647115 ss715648232 a* I_2017 8 2017 35.8 6.6 17.3 0.46 15.3 - 40.1 1.57 - 6.27 ss715647115 ss715647905 a* H_2017 8 2017 37.8 7.7 20.6 0.55 15.3 - 40.1 1.57 - 6.27 ss715647115 ss715647905 a* 10.1 86 10 2017 69.7 3.8 8.3 - 0.31 69.7 - 82.1 42.22 - 43.29 ss715645524 ss715645501 b* value b* 2.1 76 2 2017 64.1 4.4 8.6 0.41 54.0 - 74.1 11.03 - 17.24 ss715649088 ss715651061 b* 3.1 76 3 2016 14.2 3.5 7.2 0.41 3.4 - 14.7 1.01 - 1.19 ss715650435 ss715646396 b* 3.1 86 3 2017 0.0 2.7 3.9 0.31 0 - 1.5 1.19 - 1.30 ss715646396 ss715646392 99 Table 6. (cont d) QTL Chr 1 Year Peak Pos (cM) 2 Peak LOD 3 R 2 (%) 4 a 5 Map interval (cM) Physical interval (Mb) Left - flanking SNP Right - flanking SNP b* 8.1 76 Flanking region 3.5 - 7.8 6.7 - 16.4 0.39 - 0.47 16.8 - 62.3 1.53 - 7.25 ss715647112 ss715648559 Shared region 17.5 - 54.1 1.54 - 7.04 ss715647113 ss715640331 b* I_2016 8 2016 51.4 4.9 10.8 0.47 17.5 - 50.4 1.54 - 6.00 ss715647113 ss715648337 b* I_2017 8 2017 53.4 3.5 6.7 0.39 50.4 - 54.1 6.00 - 7.04 ss715648337 ss715640331 b* H_2017 8 2017 52.4 7.8 16.4 0.41 16.8 - 62.3 1.53 - 7.25 ss715647112 ss715648559 b* 8.1 86 Flanking region 4.3 - 7.0 9.1 - 12.6 0.45 - 0.55 15.3 - 60.4 1.57 - 53.68 ss715647115 ss715648232 Shared region 15.8 - 40.1 1.58 - 6.27 ss715647116 ss715647905 b* I_2016 8 2016 38.8 7.0 11.7 0.55 15.3 - 60.37 1.57 - 53.68 ss715647115 ss715648232 b* I_2017 8 2017 36.8 4.3 9.1 0.49 15.8 - 40.1 1.58 - 6.27 ss715647116 ss715647905 b* H_2017 8 2017 39.8 6.8 12.6 0.45 15.3 - 40.1 1.57 - 6.27 ss715647115 ss715647905 b* 11.1 76 Flanking region 5.3 - 8.8 11.4 - 20.7 0.31 - 0.66 123.3 - 150.9 51.12 - 52.84 ss715649251 ss715640405 Shared region 144.3 - 149.6 52.16 - 52.65 ss715649459 ss715650816 b* I_2016 11 2016 149.3 8.8 20.7 0.66 144.3 - 149.6 52.16 - 52.65 ss715649459 ss715650816 b* I_2017 11 2017 149.3 6.9 13.9 0.51 123.3 - 149.6 51.12 - 52.65 ss715649251 ss715650816 b* H_2017 11 2017 149.3 5.3 11.4 0.34 144.3 - 150.9 52.16 - 52.84 ss715649459 ss715640405 b* 11.1 86 Shared region 13.3 - 14.6 25.9 - 27.3 0.62 - 0.81 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 b* I_2016 11 2016 28.4 14.6 26.3 0.79 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 b* I_2017 11 2017 28.4 13.3 27.3 0.81 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 b* H_2017 11 2017 28.4 13.5 25.9 0.62 22.9 - 30.5 52.47 - 52.84 ss715648350 ss715640405 QTL names (bolded) are assigned according to the dry bean QTL nomenclature established by Miklas and Porch (2010) . QTL names consist of an abbreviated trait name, the chromosome of detection, unique number identifier on the chromosome, and the popula tion of detection in superscript. QTL detected by multiple years or methods are listed below the putative QTL. Population 76 was derived from B14311/Zenith, and Population 86 was derived from B14311/B12724. 'I' indicates values measured by ImageJ analysis and 'H' indicates values measured by a Hunter Labscan . 1 Chr: chromosome number ; 2 Peak Pos: genetic position of the peak LOD ; 3 LOD: logarithm of odds ; 4 R 2 : percentage of phenotypic variation explained ; 5 a: additive effect of the allele donated by breeding line B14311 100 Table 7 . QTL for canning quality traits in two black bean RIL populations . QTL Chr 1 Year Peak Pos (cM) 2 Peak LOD 3 R 2 (%) 4 a 5 Map interval (cM) Physical interval (Mb) Left - flanking SNP Right - flanking SNP Appearance rating APP2.1 76 2 2017 164.0 4.2 8.9 - 0.12 142.8 - 164.4 37.81 - 44.97 ss715648834 ss715647236 APP2.1 86 2 2016 1.0 3.8 8.8 - 0.16 0 - 1.8 3.90 - 4.48 ss715647803 ss715639861 APP3.1 86 Shared region 3.3 - 4.5 6.7 - 9.7 - (0.11 - 0.16) 0 - 1.5 1.19 - 1.30 ss715646396 ss715646392 APP_2016 3 2016 0.0 4.5 9.7 - 0.16 0 - 1.5 1.19 - 1.30 ss715646396 ss715646392 APP_2YA 3 2YA 0.0 3.3 6.7 - 0.11 0 - 1.5 1.19 - 1.30 ss715646396 ss715646392 APP4.1 86 4 2016 52.1 3.3 7.1 0.13 51.1 - 52.1 2.75 - 2.89 ss715646227 ss715646218 APP5.1 76 5 2016 49.2 3.1 7.5 - 0.12 40.6 - 57.1 4.47 - 4.75 ss715648066 ss715649111 APP6.1 86 6 2016 21.0 3.2 6.0 0.13 20.9 - 21.7 28.97 - 29.04 ss715645203 ss715645202 APP8.1 76 8 2017 49.7 3.2 6.6 - 0.10 49.7 - 50.4 5.86 - 6.00 ss715650193 ss715648337 APP8.2 76 8 2016 146.5 3.0 7.0 0.12 145.5 - 147.0 60.97 - 61.30 ss715646515 ss715646092 APP10.1 86 10 2017 89.2 3.6 9.5 0.14 69.7 - 90.2 42.22 - 44.22 ss715645524 ss715649823 101 Table 7 . (cont d) QTL Chr 1 Year Peak Pos (cM) 2 Peak LOD 3 R 2 (%) 4 a 5 Map interval (cM) Physical interval (Mb) Left - flanking SNP Right - flanking SNP Texture TXT2.1 76 2 2016 85.7 4.1 7.7 2.64 81.8 - 88.9 17.31 - 20.44 ss715650059 ss715651192 TXT2.1 86 2 2016 0.0 3.1 5.9 - 2.25 0 - 1.8 3.90 - 4.48 ss715647803 ss715639861 TXT5.1 76 Shared region 7.0 - 11.6 13.7 - 25.2 2.4 - 2.5 128.2 - 141.5 27.70 - 36.79 ss715649539 ss715646996 TXT_2016 5 2016 138.8 7.0 13.7 2.38 128.2 - 141.5 27.70 - 36.79 ss715649539 ss715646996 TXT_2017 5 2017 137.1 11.6 25.2 2.49 128.2 - 141.5 27.70 - 36.79 ss715649539 ss715646996 TXT9.1 76 9 2017 33.5 3.9 7.6 - 1.38 16.3 - 39.6 7.87 - 13.55 ss715645741 ss715647980 TXT10.1 86 10 2016 89.2 11.3 26.4 - 4.45 82.1 - 90.2 43.29 - 44.22 ss715645501 ss715649823 Washed drained weight WDW2.1 86 2 2016 0.0 4.5 10.2 1.64 0 - 25.4 3.90 - 30.15 ss715647803 ss715647526 WDW8.1 76 8 2016 158.4 4.4 11.6 - 2.43 158.4 - 160.1 62.27 - 62.75 ss715646764 ss715647397 QTL names (bolded) are assigned according to the dry bean QTL nomenclature established by Miklas and Porch (2010) . QTL names consist of an abbreviated trait name, the chromosome of detection, unique number identifier on the chromosome, and the popula tion of detection in superscript. QTL detected by multiple years or methods are listed below the putative QTL . Population 76 was derived from B14311/Zenith, and Population 86 was derived from B14311/B12724 . 1 Chr: chromosome number ; 2 Peak Pos: genetic position of the peak LOD ; 3 LOD: logarithm of odds ; 4 R 2 : percentage of phenotypic variation explained ; 5 a: additive effect of the allele donated by breeding line B14311 102 Table 8. QTL for agronomic traits in two black bean RIL populations . QTL Chr 1 Year Peak Pos (cM) 2 Peak LOD 3 R 2 (%) 4 a 5 Map interval (cM) Physical interval (Mb) Left - flanking SNP Right - flanking SNP Seed yield SY8.1 76 8 2017 137.7 3.5 8.9 - 110 137.7 - 144.2 60.07 - 60.56 ss715646529 ss715646508 SY8.1 86 8 2017 0.0 6.8 14. 8 - 1 30 0 - 12.9 0.37 - 1.50 ss715646680 ss715647128 Seed weight SW3.1 76 3 2017 119.3 12.4 22.7 0.79 119.3 - 122.7 11.47 - 11.82 ss715646286 ss715646290 SW3.1 86 3 2017 59.8 4.5 12. 2 0.42 43.6 - 60.2 3.82 - 12.30 ss715649325 ss715649868 SW4.1 76 4 2017 12.8 8.1 13.2 - 0.60 10.3 - 20.6 0.16 - 1.90 ss715648682 ss715646916 SW4.1 86 4 2017 0.7 3.4 7. 7 - 0.33 0.7 - 1.2 2.20 - 2.41 ss715647817 ss715646249 SW4.2 86 4 2017 54.7 4.3 9. 6 - 0.38 51.1 - 63.9 2.75 - 3.59 ss715646227 ss715650365 SW5.1 86 5 2017 132.9 4.1 9.3 0.36 126.9 - 133.3 35.96 - 36.79 ss715639578 ss715646996 SW7.1 76 7 2017 90.1 3.3 6.2 0.50 83.1 - 109.5 4.25 - 4.39 ss715646463 ss715646455 SW8.1 76 8 2017 155.7 5.6 8.2 - 0.45 151.1 - 158.4 62.06 - 62.27 ss715646750 ss715646764 Days to flowering DF7.1 76 7 2017 211.3 3.0 5.0 - 0.25 203.9 - 215.0 27.48 - 30.85 ss715650972 ss715639231 DF8.1 76 8 2017 145.5 10.0 20.5 - 0.47 145.5 - 147.0 60.97 - 61.30 ss715646515 ss715646092 DF11.1 76 11 2017 143.7 3.0 4.8 0.23 143.7 - 144.0 51.95 - 51.96 ss715649909 ss715640836 103 Table 8. (con d) QTL Chr 1 Year Peak Pos (cM) 2 Peak LOD 3 R 2 (%) 4 a 5 Map interval (cM) Physical interval (Mb) Left - flanking SNP Right - flanking SNP Days to maturity DM2.1 76 2 2017 113.1 4.7 11.0 - 0.43 110.1 - 113.2 31.67 - 33.65 ss715647744 ss715647098 DM4.1 86 4 2017 0.7 4.3 11.6 - 0.39 0.7 - 1.2 2.20 - 2.41 ss715647817 ss715646249 DM11.1 86 11 2017 4.0 3.7 8.3 0.35 0 - 10.8 49.59 - 51.12 ss715649023 ss715649251 Canopy height HT1.1 86 1 2017 43.6 4.5 6.7 - 0.34 43.6 - 63.7 1.29 - 2.85 ss715646260 ss715647676 HT2.1 76 2 2017 110.1 4.9 9.6 - 0.45 95.3 - 110.1 25.39 - 31.67 ss715648526 ss715647744 HT3.1 76 3 2017 108.3 3.1 5.8 0.35 92.4 - 108.8 10.69 - 11.25 ss715647445 ss715649740 HT4.1 86 4 2017 50.9 10.6 17.2 - 0.49 44.5 - 52.1 2.55 - 2.89 ss715646239 ss715646218 HT7.1 76 7 2017 8.9 3.1 5.5 0.34 5.8 - 9.43 0.65 - 0.91 ss715645687 ss715645692 HT8.1 86 8 2017 0.0 12.8 21.5 - 0.59 0 - 12.9 0.37 - 1.50 ss715646680 ss715647128 HT11.1 76 11 2017 144.3 4.6 9.0 0.42 144.3 - 149.6 52.16 - 52.65 ss715649459 ss715650816 104 Table 8. (cont d) QTL Chr 1 Year Peak Pos (cM) 2 Peak LOD 3 R 2 (%) 4 a 5 Map interval (cM) Physical interval (Mb) Left - flanking SNP Right - flanking SNP Desirability score DS2.1 76 2 2017 50.6 3.0 6.8 - 0.14 50.6 - 51.0 5.86 - 7.10 ss715650567 ss715649765 DS4.1 86 4 2017 52.1 3.6 7.3 - 0.12 51.1 - 52.1 2.75 - 2.89 ss715646227 ss715646218 DS8.1 86 8 2017 0.0 6.7 14.5 - 0.19 0 - 12.9 0.37 - 1.50 ss715646680 ss715647128 DS9.1 76 9 2017 14.9 4.9 11.1 - 0.18 14.8 - 15.8 7.70 - 7.79 ss715645748 ss715645745 Bronzing BRZ5.1 86 5 2017 178.5 3.2 7.4 - 0.19 174.7 - 188.5 39.24 - 39.34 ss715645318 ss715645331 BRZ6.1 76 6 2017 9.6 3.4 6.3 - 0.25 0.5 - 22.9 12.21 - 13.75 ss715649979 ss715647424 BRZ7.1 86 7 2017 46.1 6.6 16.9 - 0.30 46.1 - 49.8 3.99 - 4.17 ss715649276 ss715646465 BRZ8.1 76 8 2017 151.1 5.0 10.3 0.31 144.8 - 158.4 60.71 - 62.27 ss715646503 ss715646764 BRZ9.1 86 9 2017 19.0 3.5 7.3 0.18 12.1 - 19.2 31.40 - 33.38 ss715646279 ss715645631 QTL names (bolded) are assigned according to the dry bean QTL nomenclature established by Miklas and Porch (2010) . QTL names consist of an abbreviated trait name, the chromosome of detection, unique number identifier on the chromosome, and the popula tion of detection in superscript. QTL detected by multiple years or methods are listed below the putative QTL . Population 76 was derived from B14311/Zenith, and Population 86 was derived from B14311/B12724. 1 Chr: chromosome number ; 2 Peak Pos: genetic position of the peak L OD ; 3 LOD: logarithm of odds ; 4 R 2 : percentage of phenotypic variation explained ; 5 a: additive effect of the allele donated by breeding line B14311 105 Figure 1. US production of black beans for the year 2016. Michigan led the nation with 113,624 US tons produced (47% of total production). Data: 2017 USDA - NASS Crop Production Summary. Michigan North Dakota Minnesota 2016 Black bean production (1000 cwt) Michigan North Dakota Minnesota Nebraska Washington Idaho 106 Figure 2. US per capita consumption of P. vulgaris dry beans. edible beans after 1979. Garbanzo beans ( Cicer aritinum ) are also included for comparison. Data: Oct. 2017 Vegetable and Pulse Yearbook Table 5. USDA - ERS. 0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 Consumption (pounds/person) US per capita dry bean consumption Pinto Navy Great Northern Red Kidney Small Red Black Pink Other Total Garbanzo 107 Figure 3. US per capita consumption of black beans. Consumption of black beans ( Phaseolus vulgaris ) is compared with consumption of garbanzo beans ( Cicer arietinum ). Data: Oct. 2017 Vegetable and Pulse Yearbook Table 5. USDA - ERS. 0.00 0.20 0.40 0.60 0.80 1.00 1.20 1.40 1.60 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 Consumption (pounds/person) US per capita black bean consumption Black Garbanzo 108 Figure 4 . Canning quality evaluation guidelines. Evaluation charts were developed by Mendoza et al. (2017) and provided to reviewers during evaluations. Canned bean color and canned bean appearance are rated as independent traits. A: Color ratings are based on the perceived color of the seed coat darkness. B: Appearance ratings are mostly based on see d coat integrity, but may take into account brine consistency, clumping, extruded starch, and seed size. A B 109 Figure 5 . Distribution of color ratings for Population 76 (B14311/Zenith) and Population 86 (B14311/B12724) . Color ratings for each genotype were averaged across reviewers within years . 0 5 10 15 20 25 30 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3 3.25 3.5 3.75 4 4.25 4.5 4.75 5 Frequency Pop. 76 Color rating (1 - 5) 2016 2017 0 5 10 15 20 25 30 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3 3.25 3.5 3.75 4 4.25 4.5 4.75 5 Frequency Pop. 86 Color rating (1 - 5) 2016 2017 Key: B14311 Zenith B12724 110 Figure 6 . Distribution of CIELAB color values in Population 76 (B14311/Zenith) . 'I' indicates values measured by ImageJ analysis and 'H' indicates values measured by a Hunter Labscan. 0 10 20 30 40 50 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Frequency L* value Pop. 76 L* L*I 2016 L*I 2017 L*H 2017 0 20 40 60 80 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Frequency a* value Pop. 76 a* a*I 2016 a*I 2017 a*H 2017 0 10 20 30 40 50 60 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Frequency b* value Pop. 76 b* b*I 2016 b*I 2017 b*H 2017 L* I 2016 L* I 2017 L* H 2017 a * I 2016 a * I 2017 a * H 2017 b * I 2016 b * I 2017 b * H 2017 B14311 Zenith B14311 Zenith B14311 Zenith 111 Figure 7 . Distribution of CIELAB color values in Population 86 (B14311/B12724) . 'I' indicates values measured by ImageJ analysis and 'H' indicates values measured by a Hunter Labscan. 0 10 20 30 40 50 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Frequency L* value Pop. 86 L* L*I 2016 L*I 2017 L*H 2017 0 20 40 60 80 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Frequency a* value Pop. 86 a* a*I 2016 a*I 2017 a*H 2017 0 10 20 30 40 50 60 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Frequency b* value Pop. 86 b* b*I 2016 b*I 2017 b*H 2017 L* I 2016 L* I 2017 L* H 2017 a * I 2016 a * I 2017 a * H 2017 b * I 2016 b * I 2017 b * H 2017 B14311 B12724 B14311 B12724 B14311 B12724 112 Figure 8 . Distribution of canning quality trait s for Population 76 (B14311/Zenith) and Population 86 (B14311/B12724) . Appearance ratings for each genotype were averaged across reviewers within years. Washed and drained weights and textures were not replicated within years. 0 10 20 30 40 50 1 1.5 2 2.5 3 3.5 4 4.5 5 Frequency Pop. 76 Appearance rating (1 - 5) 2016 2017 0 10 20 30 40 50 1 1.5 2 2.5 3 3.5 4 4.5 5 Frequency Pop. 86 Appearance rating (1 - 5) 2016 2017 0 10 20 30 40 50 60 30 35 40 45 50 55 60 65 70 75 80 85 Frequency Pop. 76 Texture (kg/100 g) 2016 2017 0 10 20 30 40 50 60 30 35 40 45 50 55 60 65 70 75 80 85 Frequency Pop. 86 Texture (kg/100 g) 2016 2017 0 10 20 30 240 244 248 252 256 260 264 268 272 276 280 Frequency Pop. 76 Washed and drained weight (g) 2016 2017 0 10 20 30 240 244 248 252 256 260 264 268 272 276 280 Frequency Pop. 86 Washed and drained weight (g) 2016 2017 Key: B14311 Zenith B12724 113 Figure 9 . Heritabilit y estimates of canning quality traits in two black bean RIL populations . A: Population 76 is derived from B14311/Zenith. B: Population 86 is derived from B14311/B12724. COL L*I a*I b*I APP TXT WDW 90% Upper CI 0.90 0.90 0.82 0.91 0.68 0.67 0.29 Heritability 0.87 0.86 0.77 0.88 0.57 0.57 0.06 90% Lower CI 0.83 0.82 0.69 0.85 0.44 0.43 -0.23 -0.40 -0.20 0.00 0.20 0.40 0.60 0.80 1.00 Heritability Population 76 Canning Trait Heritabilities COL L*I a*I b*I APP TXT WDW 90% Upper CI 0.93 0.89 0.79 0.95 0.76 0.59 0.47 Heritability 0.91 0.85 0.72 0.93 0.69 0.46 0.30 90% Lower CI 0.88 0.81 0.63 0.91 0.59 0.29 0.07 0.00 0.20 0.40 0.60 0.80 1.00 Heritability Population 86 Canning Trait Heritabilities A B 114 Figure 1 0 . Distribution of ag ronomic traits in two black bean RIL populations . 0 10 20 30 40 50 Frequency Pop. 76 Seed yield (kg/ha) 0 10 20 30 40 50 Frequency Pop. 86 Seed yield (kg/ha) 0 10 20 30 40 50 19 20 21 22 23 24 25 26 27 Frequency Pop. 76 Seed weight (g/100 seed) 0 10 20 30 40 50 19 20 21 22 23 24 25 26 27 Frequency Pop. 86 Seed weight (g/100 seed) 0 10 20 30 40 50 60 70 44 45 46 47 48 49 50 Frequency Pop. 76 Days to flowering 0 10 20 30 40 50 60 70 44 45 46 47 48 49 50 Frequency Pop. 86 Days to flowering 0 20 40 60 91 92 93 94 95 96 97 98 Frequency Pop. 76 Days to maturity 0 20 40 60 91 92 93 94 95 96 97 98 Frequency Pop. 86 Days to maturity 115 Figure 1 0 . (con d) Measurements for each genotype were averaged across 2017 field replications. 0 10 20 30 40 50 43 44 45 46 47 48 49 50 51 Frequency Pop. 76 Canopy height (cm) 0 10 20 30 40 50 43 44 45 46 47 48 49 50 51 Frequency Pop. 86 Canopy height (cm) 0 20 40 60 3 3.5 4 4.5 5 5.5 6 Frequency Pop. 76 Desirability score (1 - 7) 0 20 40 60 3 3.5 4 4.5 5 5.5 6 Frequency Pop. 86 Desirability score (1 - 7) 0 20 40 60 80 1 1.5 2 2.5 3 3.5 4 4.5 5 Frequency Pop. 76 Bronzing rating (1 - 5) 0 20 40 60 80 1 1.5 2 2.5 3 3.5 4 4.5 5 Frequency Pop. 86 Bronzing rating (1 - 5) 0 10 20 30 40 50 1 1.5 2 2.5 3 3.5 4 4.5 5 Frequency Pop. 76 CBB rating (1 - 5) Key: B14311 Zenith B12724 116 Figure 1 1 . Regression of color components and mean visual rating of canned color. ImageJ gives more precise measurements, and L* and b* are good descriptors of perceived color. R² = 0.5179 R² = 0.8481 5 10 15 20 1 2 3 4 5 L* value Reviewer Color Rating 2017 Pop. 76 Color ( L* ) Hunter Labscan ImageJ R² = 0.2109 R² = 0.7851 5 10 15 20 1 2 3 4 5 L* value Reviewer Color Rating 2017 Pop. 86 Color ( L* ) Hunter Labscan ImageJ R² = 0.5594 R² = 0.544 2 3 4 5 6 7 8 9 1 2 3 4 5 a* value Reviewer Color Rating 2017 Pop. 76 Color ( a* ) Hunter Labscan ImageJ R² = 0.3767 R² = 0.5396 2 3 4 5 6 7 8 9 1 2 3 4 5 a* value Reviewer Color Rating 2017 Pop. 86 Color ( a* ) Hunter Labscan ImageJ R² = 0.6988 R² = 0.826 0 2 4 6 8 10 1 2 3 4 5 b* value Reviewer Color Rating 2017 Pop. 76 Color ( b* ) Hunter Labscan ImageJ R² = 0.7376 R² = 0.8561 0 2 4 6 8 10 1 2 3 4 5 b* value Reviewer Color Rating 2017 Pop. 86 Color ( b* ) Hunter Labscan ImageJ 117 Figure 1 2 . QTL graphs for Population 76 ( B14311/Zenith ) . A: Pop. 76 Pv01 (no QTL detected) B: Pop. 76 Pv0 2 118 Figure 1 2 . (cont d) C: Pop. 76 Pv0 3 D: Pop. 76 Pv0 4 119 Figure 1 2 . (cont d) E: Pop. 76 Pv0 5 F: Pop. 76 Pv0 6 120 Figure 1 2 . (cont d) G: Pop. 76 Pv0 7 121 Figure 1 2 . (cont d ) H: Pop. 76 Pv0 8 122 Figure 1 2 . ( cont d ) I: Pop. 76 Pv0 9 J: Pop. 76 Pv 10 (no QTL detected) 123 Figure 1 2 . ( cont d ) SNP markers and their genetic positions in centimorgans (cM) are shown on linkage maps. Abbreviations: COL: canned color rating, APP: canned appearance rating, TXT: texture ; WDW: washed and drained weight; SY: seed yield, SW: 100 - seed weight, DF: days to flowering, DM: days to maturity, HT: canopy height, DS: desirability score, BRZ: ozone bronzing, CBB: common bacterial blight resistance K: Pop. 76 Pv 11 124 Figure 1 3 . QTL graphs for Population 86 (B14311/B12724). A: Pop. 86 Pv 01 B: Pop. 86 Pv 02 125 Figure 1 3 . ( cont d ). C: Pop. 86 Pv 03 D: Pop. 86 Pv 04 126 Figure 1 3 . ( cont d ). E: Pop. 86 Pv 05 F: Pop. 86 Pv 06 127 Figure 1 3 . ( cont d ). G: Pop. 86 Pv 07 H: Pop. 86 Pv 08 128 Figure 1 3 . ( cont d ). I: Pop. 86 Pv 09 J: Pop. 86 P v10 129 Figure 1 3 . ( cont d ). SNP markers and their genetic positions in centimorgans (cM) are shown on linkage maps . Abbreviations: COL: canned color rating, APP: canned appearance rating, TXT: texture ; WDW: washed and drained weight; SY: seed yield, SW: 100 - seed weight, DF: days to flowering, DM: days to maturity, HT: canopy height, DS: desirability score, BRZ: ozone bronzing, CBB: common bacterial blight resistance K: Pop. 86 P v11 130 Figure 1 4 . Screening parents of Populations 76 and 86 with NDSU InDel markers near major color retention QTL on Pv08 and Pv11 . Markers were run on 3% agarose gels according to Moghaddam et al. (2014) . InDel marker names correspond to their physical positions in the v1 dry bean genome. Flanking BARCBean6k_3 SNPs and their v2.1 physical positions are given. SNPs that were mapped in at least one mapping population are bolded. A: InDel markers near COL8.1 76 and COL8.1 86 . COL QTL on Pv08 were significant over a 1.53 - 7.25 Mb interval estimated by the brown bar. B: InD el markers near COL11.1 76 and COL11.1 86 . COL QTL on Pv11 were most significant over a 52.16 - 52.84 Mb interval estimated by the brown bar. A B 131 REFERENCES 132 REFERENCES Akond, G.M.A., L. Khandaker, J. Berthold, L. Gates, K. Peters, H. Delong, and K. Hossain. 2011. Anthocyanin, total polyphenols and antioxidant activity of common bean. Am. J. Food Technol. 6(5): 385 394. Ariani, A., J.C. Berny Mier y Teran, and P. Gepts. 2016. Genome - wid e identification of SNPs and copy number variation in common bean ( Phaseolus vulgaris L.) using genotyping - by - sequencing (GBS). Mol. Breed. 36(7): 1 11. Ariyarathne, H.M., D.P. Coyne, G. Jung, P.W. Skroch, A.K. Vidaver, J.R. Steadman, P.N. Miklas, and M.J. Bassett. 1999. Molecular mapping of disease resistance genes for halo blight, common bacterial blight, and bean common mosaic virus in a segregating population of common bean. J. Am. Soc. Hortic. Sci. 124(6): 654 662 . Arts, I.C.W., and P.C.H. Hollman. 200 5. Polyphenols and disease risk in epidemiologic studies. Am. J. Clin. Nutr. 81(January): 317S 325S. Bassett, M.J. 1994. Tight linkage of purple pod character and the complex C locus in common bean. J. Hered. 85(4): 288 290 . Bassett, M.J. 2007. Genetics of seed coat color and pattern in common bean. p. 239 315. In Janick, J. (ed.), Plant Breeding Reviews. John Wiley & Sons, Inc. Bednar, R.A., and J.R. Hadcock. 1988. Purification and characterization of chalcone isomerase from soybeans. J. Biol. Chem. 263(20 ): 9582 9588. Beninger, C.W., and G.L. Hosfield. 2003. Antioxidant a ctivity of e xtracts, c ondensed t annin f ractions, and p ure f lavonoids from Phaseolus vulgaris L. s eed c oat c olor g enotypes. J. Agric. Food Chem. 51(27): 7879 7883. Blair, M.W., F. Pedraza, H.F. Buendía, E. Gaitán - Solís, S.E. Beebe, P. Gepts, and J. Tohme. 2003. Development of a genome - wide anchored microsatellite map for common bean ( Phaseolus vulgaris L.). Theor. Appl. Genet. 107(8): 1362 1374. Bonds, J.K. 2017. Pulses production expanding as consumers cultivate a taste for U.S. lentils and chickpeas . United States Department of Agriculture - Economic Research Service . Brand - Miller, J., S. Hayne, P. Petocz, and S. Colagiuri. 2003. Low glycemic index diets in the management of diabetes. Diabet es Care 26(8): 2261 LP - 2267 . Bravo, L. 1998. Polyphenols: chemistry, dietary sources, metabolism, and nutritional significance. Nutr. Rev. 56(11): 317 333 . Busse, M., and Ä. Bäumer - Schleinkofer. 1996. Ewald Hering und die Gegenfarbtheorie. NTM Int. J. Hist. Ethics Nat. Sci. Technol. Med. 4(1): 159 172 . 133 Cheng, A. - X., X. Zhang, X. - J. Han, Y. - Y. Zhang, S. Gao, C. - J. Liu, and H. - X. Lou. 2018. Identification of chalcone isomerase in the basal land plants reveals an ancient evolution of enzymatic cyclization activity for synthesis of flavonoids. New Phytol. 217(2): 909 924 . Cichy, K.A., A.C. Fernandez, A. Kilian, J.D. Kelly, C.H. Galeano, S. Shaw, M.A. Brick, D. Hodkinson, and E. Troxtell. 2014. QTL analysis of canning quality and color retention in black bean s ( Phaseolus vulgaris L.). Mol. Breed. 33(1): 139 154 . CIE. 2008. ISO 1164 - 4. Part 4: CIE 1976 L*a*b* Colour Space. Geneva, Switzerland. Collard, B.C.Y., M.Z.Z. Jahufer, J.B. Brouwer, and E.C.K. Pang. 2005. An introduction to markers, quantitative trait lo ci (QTL) mapping and marker - assisted selection for crop improvement: The basic concepts. Euphytica 142(1 2): 169 196. Corradini, E., P. Foglia, P. Giansanti, R. Gubbiotti, R. Samperi, and A. Lagana. 2011. Flavonoids: chemical properties and analytical meth odologies of identification and quantitation in foods and plants. Nat. Prod. Res. 25(5): 469 495. Diniz, A.L., M.I. Zucchi, L. Santini, L.L. Benchimol - Reis, M.H.P. Fungaro, and M.L.C. Vieira. 2014. Nucleotide diversity based on phaseolin and iron reductase genes in common bean accessions of different geographical origins. Genome 57(2): 69 77 . Doyle, J.., and J.. Doyle. 1991. DNA Isolation from Small Amount of Plant Tissue. Phytochem. Bull. 57: 13 15. Elshire, R.J., J.C. Glaubitz, Q. Sun, J.A. Poland, K. Kaw amoto, E.S. Buckler, and S.E. Mitchell. 2011. A robust, simple genotyping - by - sequencing (GBS) approach for high diversity species. PLoS One 6(5): 1 10. Feenstra, W.J. 1960. Biochemical aspects of seedcoat colour inheritance in Phaseolus vulgaris L. Fehr, W .R. 1991. Principles of Cul t ivar Development Vol I: Theory and Technique. 2nd ed. Iowa State University, Ames, IA. Fourie, D., P.N. Miklas, and H.M. Ariyarathne. 2004. Genes conditioning halo blight resistance to races 1, 7, and 9 occur in a tight cluster. Annu. Rep. Bean Improv. Coop. 47: 103 104. Foyer, C.H., H. - M. Lam, H.T. Nguyen, K.H.M. Siddique, R.K. Varshney, T.D. Colmer, W. Cowling, H. Bramley, T.A. Mori, J.M. Hodgson, J.W. Cooper, A.J. Miller, K. Kunert, J. Vorster, C. Cullis, J.A. Ozga, M.L. Wahlq vist, Y. Liang, H. Shou, K. Shi, J. Yu, N. Fodor, B.N. Kaiser, F. - L. Wong, B. Valliyodan, and M.J. Considine. 2016. Neglecting legumes has compromised human health and sustainable food production. Nat. Plants 2(8): 16112 . Freyre, R., P.W. Skroch, V. Geffro y, A. - F. Adam - Blondon, A. Shirmohamadali, W.C. Johnson, V. Llaca, R.O. Nodari, P.A. Pereira, S.M. Tsai, J. Tohme, M. Dron, J. Nienhuis, C.E. Vallejos, and P. Gepts. 1998. Toward an integrated linkage map of common bean. 4. Development of a core linkage map and alignment of RFLP maps. Theor. Appl. Genet. 97: 847 856 . 134 Geffroy, V., F. Creusot, J. Falquet, M. Sévignac, A. - F. Adam - Blondon, H. Bannerot, P. Gepts, and M. Dron. 1998. A family of LRR sequences in the vicinity of the Co - 2 locus for anthracnose resist ance in Phaseolus vulgaris and its potential use in marker - assisted selection. Theor. Appl. Genet. 96(3 4): 494 502 . Geffroy, V., M. Sévignac, J.C.F. De Oliveira, G. Fouilloux, P. Skroch, P. Thoquet, P. Gepts, T. Langin, and M. Dron. 2000. Inheritance of p artial resistance against Colletotrichum lindemuthianum in Phaseolus vulgaris and co - localization of quantitative trait loci with genes involved in specific resistance. Mol. Plant - Microbe Interact. 13(3): 287 296 . Gepts, P. 1999. Development of an integrat ed linkage map. p. 53 91. In Singh, S.P. (ed.), Common Bean Improvement in the Twenty - First Century. 1st ed. Developments in Plant Breeding. Springer Netherlands, Dordrecht. Gerke, V., C.E. Creutz, and S.E. Moss. 2005. Annexins: linking Ca2+ signalling to membrane dynamics. Nat. Rev. Mol. Cell Biol. 6(6): 449 461 . Ghaderi, A., G.L. Hosfield, M.W. Adams, and M.A. Uebersax. 1984. Variability in culinary quality, component interrelationships, and breeding implications in navy and pinto beans. J. Amer. Soc. Hor t. Sci. 109(1): 85 90. Goff, L.M., D.E. Cowland, L. Hooper, and G.S. Frost. 2013. Low glycaemic index diets and blood lipids: A systematic review and meta - analysis of randomised controlled trials. Nutr. Metab. Cardiovasc. Dis. 23(1): 1 10 . Goodstein, D.M., S. Shu, R. Howson, R. Neupane, R.D. Hayes, J. Fazo, T. Mitros, W. Dirks, U. Hellsten, N. Putnam, and D.S. Rokhsar. 2012. Phytozome: a comparative platform for green plant genomics. Nucleic Acids Res. 40 (Database issue): D1178 - 86 . Grisi, M.C.M., M.W. Blai r, P. Gepts, C. Brondani, P.A.A. Pereira, and R.P. V Brondani. 2007. Genetic mapping of a new set of microsatellite markers in a reference common bean ( Phaseolus vulgaris ) population BAT93 x Jalo EEP558. Genet. Mol. Res. 6(3): 691 706. Hart, J.P., and P.D. Griffiths. 2015. Genotyping - by - s equencing e nabled m apping and m arker d evelopment for the By - 2 p otyvirus r esistance a llele in c ommon b ean. Plant Genome 8(1): 1 14 . Heller, W., and G. Forkmann. 1988. Biosynthesis. p. 400 425. In Harbone, J. (ed.), The Flavo noids: Advances in r esearch s ince 1980. Chapman and Hall, London. Hornick, B.A., and L. Weiss. 2011. Comparative nutrient analysis of commonly consumed vegetables. Nutr. Today 46(3): 130 137. Hosfield, G.L. 2001. Seed c oat c olor in Phaseolus vulgaris L.: Its c hemistry and as sociated h ealth r elated b enefits. Annu. Rep. Bean Improv. Coop. Hosfield, G.L., and M.A. Uebersax. 1980. Variability in physico - chemical properties and nutritional components of tropical and domestic dry bean germplasm. J. Amer. So c. Hort Sci 105(2): 246 252. 135 Hosfield, G.L., and M.A. Uebersax. 1990. Culinary quality in dry bean - Can it be improved? Annu. Rep. Bean Improv. Coop. 33: 17 18. Hosfield, G.L., M.A. Uebersax, and T.G. Isleib. 1984. Seasonal and genotypic effects on yield a nd physico - chemical seed characteristics related to food quality in dry, edible beans. J. Amer. Soc. Hort. Sci. 109(2): 182 189. Hoyos - Villegas, V., W. Mkwaila, P.B. Cregan, and J.D. Kelly. 2015. Quantitative t rait l oci a nalysis of w hite m old a voidance in p into b ean. Crop Sci. 55(5): 2116 . Hunter Laboratories. 1996. Insight on Color: Hunter Lab Color Scale. Reston, VA. Hurtado - Gonzales, O.P., G. Valentini, T.A.S. Gilio, A.M. Martins, Q. Song, and M.A. Pastor - Corrales. 2017. Fine mapping of Ur - 3, a historica lly important rust resistance locus in common bean. G3 7(2): 557 569 . Hyten, D.L., Q. Song, E.W. Fickus, C. V Quigley, J. - S. Lim, I. - Y. Choi, E. - Y. Hwang, M.A. Pastor - Corrales, and P.B. Cregan. 2010. High - throughput SNP discovery and assay development in c ommon bean. BMC Genomics 11(1): 475 . Jung, G., H.M. Ariyarathne, D.P. Coyne, and J. Nienhuis. 2003. Mapping QTL for bacterial brown spot resistance under natural infection in field and seedling stem inoculation in growth chamber in common bean. Crop Sci. 4 3(1): 350 357 . Jung, G., D.P. Coyne, P.W. Skroch, J. Nienhuis, E. Arnaud - Santana, J. Bokosi, H.M. Ariyarathne, J.R. Steadman, J.S. Beaver, and S.M. Kaeppler. 1996. Molecular markers associated with plant architecture and resistance to common blight, web bl ight, and rust in common beans. J. Am. Soc. Hortic. Sci. 121(5): 794 803 . Kamfwa, K., K.A. Cichy, and J.D. Kelly. 2015. Genome - w ide a ssociation s tudy of a gronomic t raits in c ommon b ean. Plant Genome 8(2): 1 12 . Kelly, J.D. 2010. The Story of Bean Breeding. East Lansing , MI . Kelly, J.D., and K.A. Cichy. 2013. Dry Bean Breeding and Production Technologies. p. 23 54. In Siddiq, M., Uebersax, M.A. (eds.), Dry Beans and Pulses Production, Processing and Nutrition. First. John Wiley & Sons, Inc., Oxford, UK. Kell Bean. J. Plant Regist. 9(1): 15 20 . J. Plant Regist. 3(3): 226 . Konopka - Postupolska, D., G. Clark, and A. Hofmann. 2011. Structure, function and membrane interactions of plant annexins: An update. Plant Sci. 181(3): 230 241 . 136 Lobaton, J.D., T. Miller, J. Gil, D. Ariza, J.F. de la Hoz, A. Soler, S. Beebe, J. Duitama, P. Gepts, and B. Raatz. 2018. Resequencing of common bean identifies regions of inter gene pool introgression and provides comprehensive resources for molecular breeding. Plant Genome 11 ( 2 ): 1 - 21 . Lotito, S.B., and B. Frei. 2006. Consumption of flavonoid - rich foods and increased plasma antioxidant capacity in humans: Cause, consequence, or epiphenomenon? Free Radic. Biol. Med. 41(12): 1727 1746. Lucier, G., L. Biing - Hwan, J. Allshouse, and L.S. Kantor. 2000. Factors a ffecting d ry b ean c onsumption in the United States. Econ. Res. Serv. (April): 26 34 . Luthria, D.L., and M.A. Pastor - Corrales. 2006. Phenolic acids content of fifteen dry edible bean ( Phaseolus vulgaris L.) varieties. J. Food Compos. Anal. 19(2 3): 205 211. Ma, Y., and F.A. Bliss. 1978. Seed proteins of common bean. Crop Sci. 18: 431 437. Marles, M.A.S., P. Balasubramanian, and K.E. Bett. 2010. Differential accumulation of polyphenolics in black bean genotypes grown in four environments. J. Agric. Food Chem. 58(11): 7001 7006. Martens, S., A. Preuß, a nd U. Matern. 2010. Multifunctional flavonoid dioxygenases: Flavonol and anthocyanin biosynthesis in Arabidopsis thaliana L. Phytochemistry 71(10): 1040 1049 . Martínez - Manrique, E., C. Jacinto - Hernández, R. Garza - García, A. Campos, E. Moreno, and I. Bernal - Lugo. 2011. Enzymatic changes in pectic polysaccharides related to the beneficial effect of soaking on bean cooking time. J. Sci. Food Agric. 91(13): 2394 2398. Matella, N.J., D.K. Mishra, and K.D. Dolan. 2013. Hydration, Blanching and Thermal Processing of Dry Beans. p. 129 154. In Siddiq, M., Uebersax, M.A. (eds.), Dry Beans and Pulses Production, Processing and Nutrition. Blackwell Publishing Ltd., Oxford, UK. McClean, P.E., R.K. Lee, C. Otto, P. Gepts, and M.J. Bassett. 2002. Molecular and phenotypic m apping of genes controlling seed coat pattern and color in common bean ( Phaseolus vulgaris L.). J. Hered. 93(2): 148 52 . Mendoza, F.A., K.A. Cichy, R. Lu, and J.D. Kelly. 2014. Evaluation of c anning q uality t raits in b lack b eans ( Phaseolus vulgaris L.) by Visible/Near - Infrared Spectroscopy. Food Bioprocess Technol. 7(9): 2666 2678. Mendoza, F.A., K.A. Cichy, C. Sprague, A. Goffnett, R. Lu, and J.D. Kelly. 2018. Prediction of canned black bean texture ( Phaseolus vulgaris L.) from intact dry seeds using visible/near infrared spectroscopy and hyperspectral imaging data. J. Sci. Food Agric. 98(1): 283 290. Mendoza, F.A., J.D. Kelly, and K.A. Cichy. 2017. Automated prediction of sensory scores for color and appearance in cann ed black beans ( Phaseolus vulgaris L.) using machine vision. Int. J. Food Prop. 20(1): 83 99. 137 Messina, V. 2014. Nutritional and health benefits of dried beans. Discovery Service for Endeavour College of Natural Health Library. Am. J. Clin. Nutr. 100(1): 43 7 . Micha, R., P.J. Rogers, and M. Nelson. 2010. The glycaemic potency of breakfast and cognitive function in school children. Eur. J. Clin. Nutr. 64(9): 948 957 . Miguel, M.G. 2011. Anthocyanins: Antioxidant and/or anti - inflammatory activities. J. Appl. Pha rm. Sci. 1(6): 7 15. Miklas, P.N., D.P. Coyne, K.F. Grafton, N. Mutlu, J. Reiser, D.T. Lindgren, and S.P. Singh. 2003. A major QTL for common bacterial blight resistance derives from the common bean great northern landrace cultivar Montana No.5. Euphytica 131(1): 137 146 . Miklas, P.N., E. Johnson, V. Stone, J.S. Beaver, C. Montoya, and M. Zapata. 1996. Selective mapping of QTL conditioning disease resistance in common bean. Crop Sci. 36(5): 1344 . Miklas, P.N., J.D. Kelly, S.E. Beebe, and M.W. Blair. 2006. C ommon bean breeding for resistance against biotic and abiotic stresses: From classical to MAS breeding. Euphytica 147(1 2): 105 131. Miklas, P.N., and T.G. Porch. 2010. Guidelines for common bean QTL nomenclature. Annu. Rep. Bean Improv. Coop. 53: 202 204. Miklas, P.N., R. Riley, K.F. Grafton, S.P. Singh, G. Jung, and D.P. Coyne. 2000. Marker - a ssisted b reeding for p yramided r esistance t o c ommon b acterial b light in c ommon b ean. Annu. Rep. Bean Improv. Coop. 43(l): 49 40. Moghaddam, S.M., S. Mamidi, J.M. Osorno, R. Lee, M.A. Brick, J.D. Kelly, P. Miklas, C. Urrea, Q. Song, P.B. Cregan, J. Grimwood, J. Schmutz, and P.E. McClean. 2016. Genome - wide association study identifies candidate loci underlying agronomic traits in a M i ddle American diversity panel of common bean. Plant Genome 9(3): 1 21 . Moghaddam, S.M., Q. Song, S. Mamidi, J. Schmutz, R. Lee, P.B. Cregan, J.M. Osorno, and P.E. McClean. 2014. Developing market class specific InDel markers from next generation sequence d ata in Phaseolus vulgaris L. Front. Plant Sci. 5(May): 185 . Murphy, M.M., J.S. Douglass, and A. Birkett. 2008. Resistant starch intakes in the United States. J. Am. Diet. Assoc. 108(1): 67 78 . Nodari, R.O., S.M. Tsai, R.L. Gilbertson, and P. Gepts. 1993a. Toward an integrated linkage map of common bean. II. Development of an RFLP - based linkage map. Theor. Appl. Genet. 85(5): 513 520. Nodari, R.O., S.M. Tsai, P. Guzmán, R.L. Gilbertson, and P. Gepts. 1993b. Toward an Integrated Linkage Map of common bean. II I. Mapping genetic factors controlling host - bacteria interactions. Genetics 134(1): 341 50 . 138 Oblessuc, P.R., C. Francisco, and M. Melotto. 2015. The Co - 4 locus on chromosome Pv08 contains a unique cluster of 18 COK - 4 genes and is regulated by immune respons e in common bean. Theor. Appl. Genet. 128(6): 1193 1208. Van Ooijen, J.W. 2011. Multipoint maximum likelihood mapping in a full - sib family of an outbreeding species. Genet. Res. (Camb). 93(5): 343 349 . Oomah, B.D., A. Cardador - Martínez, and G. Loarca - Piña. 2005. Phenolics and antioxidative activities in common beans ( Phaseolus vulgaris L). J. Sci. Food Agric. 85(6): 935 942 . Parr, B., J.K. Bond, and T. Minor. 2018. Vegetables and Pulses Outlook. Pedraza García, F., G.J. Gallego, S.E. Beebe, and J.M. Tohme. 1997. Marcadores SCAR y RAPD para la resistancia a la Bacteriosis Común (CBB) del frijol. p. 130 134. In Singh, S.P., Voysesy, O. (eds.), Taller de mejoramiento de frijol para el Siglo XXI: Bases para una estrategia para América Latina. CIAT, Cali, Colombi a. Van Der Poel, A.F.B. 1990. Effect of processing on antinutritional factors and protein nutritional value of dry beans ( Phaseolus vulgaris L.): A review. Anim. Feed Sci. Technol. 29(3 4): 179 208 . Posa - Macalincag, M.C.T., G.L. Hosfield, K.F. Grafton, M.A . Uebersax, and J.D. Kelly. 2002. Quantitative trait loci (QTL) analysis of canning quality traits in kidney bean ( Phaseolus vulgaris L.). J. Amer. Soc. Hort. Sci. 127(4): 608 615 . Prakken, R. 1974. Inheritance of colours in Phaseolus vulgaris L.: IV recom bination within the 36 . Reinprecht, Y., Z. Yadegari, G.E. Perry, M. Siddiqua, L.C. Wright, P.E. McClean, and K.P. Pauls. 2013. In silico comparison of genomic regions containing genes coding for enzymes and transcription factors for the phenylpropanoid pathway in Phaseolus vulgaris L. and Glycine max L. Merr. Front. Plant Sci. 4(September): 317 . Schmutz, J., P.E. McClean, S. Mamidi, G.A. Wu, S.B. Cannon, J. Grimwood, J. Jenkins, S. Shu, Q. Song, C. Chavarro, M. Torres - Torres, V. Geffroy, S.M. Moghaddam, D. Gao, B. Abernathy, K. Barry, M. Blair, M.A. Brick, M. Chovatia, P. Gepts, D.M. Goodstein, M. Gonzales, U. Hellsten, D.L. Hyten, G. Jia, J.D. Kelly, D. Kudrna, R. Lee, M.M.S. Richard, P.N. Miklas, J.M. Osorno , J. Rodrigues, V. Thareau, C.A. Urrea, M. Wang, Y. Yu, M. Zhang, R.A. Wing, P.B. Cregan, D.S. Rokhsar, and S.A. Jackson. 2014. A reference genome for common bean and genome - wide analysis of dual domestications. Nat. Genet. 46(7): 707 713 . Schneider, C.A., W.S. Rasband, and K.W. Eliceiri. 2012. NIH Image to ImageJ: 25 years of image analysis. Nat. Methods 9(7): 671 5 . Schröder, S., S. Mamidi, R. Lee, M.R. McKain, P.E. McClean, and J.M. Osorno. 2016. Optimization of genotyping by sequencing (GBS) data in com mon bean ( Phaseolus vulgaris L.). Mol. Breed. 36(1): 1 9. 139 Shi, C., K. Yu, W. Xie, G.E. Perry, A. Navabi, K.P. Pauls, P.N. Miklas, and D. Fourie. 2012. Development of candidate gene markers associated to common bacterial blight resistance in common bean. Th eor. Appl. Genet. 125(7): 1525 1537. Simões, I., and C. Faro. 2004. Structure and function of plant aspartic proteinases. Eur. J. Biochem. 271(11): 2067 2075 . Song, Q., G. Jia, D.L. Hyten, J. Jenkins, E. - Y. Hwang, S.G. Schroeder, J.M. Osorno, J. Schmutz, S.A. Jackson, P.E. McClean, and P.B. Cregan. 2015. SNP assay development for linkage map construction, anchoring whole genome sequence and other genetic and genomic applications in common bean. G3 Genes|Genomes|Genetics 5(11): 2285 2290 . Takeoka, G.R., L.T . Dao, G.H. Full, R.Y. Wong, L.A. Harden, R.H. Edwards, and J.D.J. Berrios. 1997. Characterization of black bean ( Phaseolus vulgaris L.) anthocyanins. J. Agric. Food Chem. 45(97): 3395 3400 . Thompson, M.D., M.A. Brick, J.N. McGinley, and H.J. Thompson. 200 9. Chemical composition and mammary cancer inhibitory activity of dry bean. Crop Sci. 49(1): 179 186 . USDA - Agricultural Research Service. 2018. Nutrient search. Natl. Nutr. Database Stand. Ref. Leg. Release . Available at https://ndb.nal.usda.gov/ndb/nutri ents/index. USDA - NASS. 2016. Crop Production 2015 Summary. USDA - NASS. 2018. Crop Production 2017 Summary. Vlasova, A., S. Capella - Gutiérrez, M. Rendón - Anaya, M. Hernández - Oñate, A.E. Minoche, I. Erb, F. Câmara, P. Prieto - Barja, A. Corvelo, W. Sanseverino , G. Westergaard, J.C. Dohm, G.J. Pappas, S. Saburido - Alvarez, D. Kedra, I. Gonzalez, L. Cozzuto, J. Gómez - Garrido, M.A. Aguilar - Morón, N. Andreu, O.M. Aguilar, J. Garcia - Mas, M. Zehnsdorf, M.P. Vázquez, A. Delgado - Salinas, L. Delaye, E. Lowy, A. Mentaberr y, R.P. Vianello - Brondani, J.L. García, T. Alioto, F. Sánchez, H. Himmelbauer, M. Santalla, C. Notredame, T. Gabaldón, A. Herrera - Estrella, R. Guigó, P. Graham, C. Vance, J. Kami, V. Velásquez, D. Debouck, P. Gepts, M. Kwak, E. Bitocchi, L. Nanni, E. Bellu cci, M. Rossi, A. Giardini, P. Zeuli, S. Mamidi, S. Moghaddam, D. Annam, R. Lee, R. Papa, J. Schmutz, P. McClean, G. Wu, S.B. Cannon, J. Grimwood, S. Moghaddam, S. Chacón, B. Pickersgill, A. Delgado - Salinas, R. Bibler, M. Lavin, D. Rau, M. Rodriguez, E. Bi agetti, J.A. Acosta - Gallegos, M.W. Blair, A. Soler, A.J. Cortés, B. Gaut, X. Huang, Q. Feng, Q. Qian, Q. Zhao, L. Wang, A. Wang, K. Arumuganathan, E. Earle, M. Bennett, J. Smith, G. Parra, K. Bradnam, I. Korf, G. Arimura, R. Ozawa, S. Kugimiya, J. Takabaya shi, J. Bohlmann, J.D. Kelly, P. Miklas, D.P. Coyne, V. Geffroy, M. Sévignac, J. Oliveira, G. Fouilloux, P. Skroch, P. Thoquet, U. Ørom, T. Derrien, M. Beringer, K. Gumireddy, A. Gardini, G. Bussotti, J. Huerta - Cepas, L. Pryszcz, M. Marcet - Houben, T. Gabal dón, I. Denisov, D. Kormes, Y. Jiao, N. Wickett, S. Ayyampalayam, A. Chanderbali, L. Landherr, P. Ralph, Y. Yang, M. Moore, S. Brockington, D. Soltis, G. Wong, L. Sterck, S. Rombauts, S. Sato, F. Cheung, J. Gouzy, J. Schlueter, J. Ma, T. Mitros, W. Nelson, A. Ferrarini, A. Benazzo, S. Woody, Y. - T. Bolon, B. Joseph, B. Diers, A.D. Farmer, L. Mao, J. Hemert, S. Dash, J. Dickerson, J. Huerta - cepas, J. Dopazo, M. Huyne n, T. Padawer, R. Leighty, D. Wang, M. 140 McConnell, S. Chikara, M. Ramírez, M. Graham, L. Blanco - lo, S. Silvente, A. Medrano - soto, M. Melotto, C. Monteiro - Vitorello, A. Bruschi, L. Camargo, F. Belzile, J. Tian, P. Venkatachalam, H. Liao, X. Yan, K. Raghotham a, V. Kalavacharla, Z. Liu, B. Meyers, J. Thimmapuram, K. Melmaiee, B. Le, J. Wagmaister, T. Kawashima, A. Bui, J. Harada, R. Goldberg, V. Singh, R. Garg, M. Jain, J. Verdier, I. Torres - Jerez, M. Wang, A. Andriankaja, S. Allen, J. He, R. Johnson, A. Tanzer , S. Djebali, H. Tilgner, J. Sun, Y. Lin, J. Wu, C. Kutter, S. Watt, K. Stefflova, M. Wilson, A. Goncalves, C. Ponting, M. Mutwil, B. Usadel, M. Schütte, A. Loraine, O. Ebenhöh, S. Persson, K. Aoki, Y. Ogata, D. Shibata, S. Ma, S. Shah, H. Bohnert, M. Snyd er, S. Dinesh - Kumar, M. Lynch, A. Force, V. Prince, F. Pickett, S. Renny - Byfield, J. Gallagher, C. Grover, E. Szadkowski, J. Page, J. Udall, J. Duarte, L. Cui, P. Wall, Q. Zhang, X. Zhang, J. Leebens - Mack, R. Ferreira, P. Ramos, E. Franco, C. Ricardo, A. T eixeira, D. Sicard, S. Cohen, S. Chisholm, G. Coaker, B. Day, B. Staskawicz, D. Cook, T. Lee, X. Guo, S. Melito, K. Wang, A. Bayless, M.C.M. Grisi, C. Brondani, P. Pereira, R. Brondani, T. Flutre, E. Duprat, C. Feuillet, H. Quesneville, D. Ellinghaus, S. K urtz, U. Willhoeft, S. Steinbiss, G. Gremme, J. Jurka, V. Kapitonov, A. Pavlicek, P. Klonowski, O. Kohany, J. Walichiewicz, S. EMBL, P. Agarwal, J. Abril, T. Wiehe, J. Fickett, R. Guigo, M. Stanke, O. Keller, I. Gunduz, A. Hayes, S. Waack, B. Morgenstern, W. Majoros, M. Pertea, S. Salzberg, B. Haas, W. Zhu, J. Allen, J. Orvis, S. Hunter, P. Jones, A. Mitchell, R. Apweiler, T. Attwood, A. Bateman, M. Kanehisa, S. Goto, Y. Sato, M. Furumichi, M. Tanabe, D. Croft, A. Mundo, R. Haw, M. Milacic, J. Weiser, G. Wu , T. Petersen, S. Brunak, G. Heijne, H. Nielsen, S. Götz, J. García - Gómez, J. Terol, T. Williams, S. Nagaraj, Frusciante, S. Altschul, W. Gish, W. Miller, E. Myers, D. Lipman , G. Slater, E. Birney, C. Notredame, D. Higgins, J. Heringa, C. Trapnell, B. Williams, G. Pertea, A. Mortazavi, G. Kwan, M. Baren, E. Nawrocki, D. Kolbe, S.R. Eddy, K. McCue, L. Schaeffer, B. Wold, M. Robinson, D. McCarthy, G. Smyth, J. Friedman, T. Hasti e, R. Tibshirani, T. Fruchterman, M. Reingold, A. Clauset, M. Newman, C. Moore, R. Edgar, K. Katoh, H. Toh, T. Lassmann, O. Frings, E. - Martínez, O. Gascuel, S. Guindon, J. Dufayard, V. Lefort, M. Anisimova, W . Hordijk, D.M. Goodstein, S. Shu, R. Howson, R. Neupane, R. Hayes, J. Fazo, H. Li, and R. Durbin. 2016. Genome and transcriptome analysis of the Mesoamerican common bean and the role of gene duplications in establishing tissue and temporal specialization of genes. Genome Biol. 17(1): 32 . Voorrips, R.E. 2002. MapChart: software for the graphical presentation of linkage maps and QTLs. J. Hered. 93(1): 77 8 . Walters, K.J., G.L. Hosfield, M.A. Uebersax, and J.D. Kelly. 1997. Navy bean canning quality: Correlat ions, heritability estimates, and randomly amplified polymorphic DNA markers associated with component traits. J. Amer. Soc. Hort. Sci. 122(3): 338 343 . Wang, S., C.J. Basten, and Z. - B. Zeng. 2012. Windows QTL Cartographer. Wang, L. - S., and G.D. Stoner. 20 08. Anthocyanins and their role in cancer prevention. Cancer Lett. 269(2): 281 290. Wassimi, N.N., G.L. Hosfield, and M.A. Uebersax. 1990. Inheritance of physico - chemical seed characters related to culinary quality in dry bean. J. Amer. Soc. Hort Sci 115(3 ): 492 499. 141 Wright, E.M., and J.D. Kelly. 2011. Mapping QTL for seed yield and canning quality following processing of black bean ( Phaseolus vulgaris L.). Euphytica 179(3): 471 484. Xia, Y., H. Suzuki, J. Borevitz, J. Blount, Z. Guo, K. Patel, R.A. Dixon, and C. Lamb. 2004. An extracellular aspartic protease functions in Arabidopsis disease resistance signaling. EMBO J. 23(4): 980 988 . Xu, B., and S.K.C. Chang. 2009. Total phenolic, phenolic acid, anthocyanin, flavan - 3 - ol, and flavonol profiles and antioxid ant properties of pinto and black beans ( Phaseolus vulgaris L.) as affected by thermal processing. J. Agric. Food Chem. 57(11): 4754 4764. Zou, X., C. Shi, R.S. Austin, D. Merico, S. Munholland, F. Marsolais, A. Navabi, W.L. Crosby, K.P. Pauls, K. Yu, and Y. Cui. 2014. Genome - wide single nucleotide polymorphism and Insertion - Deletion discovery through next - generation sequencing of reduced representation libraries in common bean. Mol. Breed. 33(4): 769 778. Zuiderveen, G.H., B.A. Padder, K. Kamfwa, Q. Song, and J.D. Kelly. 2016. Genome - Wide association study of anthracnose resistance in A ndean beans ( Phaseolus vulgaris ). PLoS One 11(6): 1 17 .