Agronomy Publications Agronomy

3-2019 Stability Analysis of Kernel Quality Traits in Exotic- Derived Doubled Haploid Maize Lines Adam Vanous Iowa State University, [email protected]

Candice Gardner United States Department of Agriculture, [email protected]

Michael Blanco United States Department of Agriculture

Adam Martin-Schwarze Iowa State University

Jinyu Wang Iowa State University, [email protected]

SeFoe nelloxtw pa thige fors aaddndition addal aitutionhorsal works at: https://lib.dr.iastate.edu/agron_pubs Part of the Agriculture Commons, Agronomy and Crop Sciences Commons, and the Plant Breeding and Genetics Commons The ompc lete bibliographic information for this item can be found at https://lib.dr.iastate.edu/ agron_pubs/536. For information on how to cite this item, please visit http://lib.dr.iastate.edu/ howtocite.html.

This Article is brought to you for free and open access by the Agronomy at Iowa State University Digital Repository. It has been accepted for inclusion in Agronomy Publications by an authorized administrator of Iowa State University Digital Repository. For more information, please contact [email protected]. Stability Analysis of Kernel Quality Traits in Exotic-Derived Doubled Haploid Maize Lines

Abstract Variation in kernel composition across maize (Zea mays L.) germplasm is affected by a combination of the plant’s genotype, the environment in which it is grown, and the interaction between these two elements. Adapting exotic germplasm to the US Corn Belt is highly dependent on the plant’s genotype, the environment where it is grown, and the interaction between these components. Phenotypic plasticity is ill-defined when specific exotic germplasm is moved over large latitudinal distances and for the adapted variants being created. Reduced plasticity (or stability) is desired for the adapted variants, as it allows for a more rapid implementation into breeding programs throughout the Corn Belt. Here, doubled haploid lines derived from exotic maize and adapted through backcrossing exotic germplasm to elite adapted lines were used in conjunction with genome-wide association studies to explore stability in four kernel composition traits. Genotypes demonstrated a response to environments that paralleled the mean response of all genotypes used across all traits, with protein content and kernel density exhibiting the highest levels of Type II stability. Genes such as opaque10, empty pericarp 16, and floury 1 were identified as potential candidates within quantitative trait locus regions. The findings within this study aid in validating previously identified genomic regions and identified novel genomic regions affecting kernel quality traits.

Disciplines Agriculture | Agronomy and Crop Sciences | Plant Breeding and Genetics

Comments This article is published as Vanous, Adam, Candice Gardner, Michael Blanco, Adam Martin-Schwarze, Jinyu Wang, Xianran Li, Alexander E. Lipka et al. "Stability Analysis of Kernel Quality Traits in Exotic-Derived Doubled Haploid Maize Lines." The Plant Genome 12 (2018): 1-14. doi: 10.3835/plantgenome2017.12.0114.

Rights Works produced by employees of the U.S. Government as part of their official duties are not copyrighted within the U.S. The onc tent of this document is not copyrighted.

Authors Adam Vanous, Candice Gardner, Michael Blanco, Adam Martin-Schwarze, Jinyu Wang, Xianran Li, Alexander E. Lipka, Sherry Flint-Garcia, Martin Bohn, Jode Edwards, and Thomas Lubberstedt

This article is available at Iowa State University Digital Repository: https://lib.dr.iastate.edu/agron_pubs/536 Published online November 21, 2018

The Plant Genome ori ginal research

Stability Analysis of Kernel Quality Traits in Exotic-Derived Doubled Haploid Maize Lines

Adam Vanous, Candice Gardner,* Michael Blanco, Adam Martin-Schwarze, Jinyu Wang, Xianran Li, Alexander E. Lipka, Sherry Flint-Garcia, Martin Bohn, Jode Edwards, and Thomas Lübberstedt

A. Vanous, J. Wang, X. Li, T. Lübberstedt, Dep. of Agronomy, Iowa State Univ., Ames, IA 50011; M. Blanco, J. Edwards, C. Gardner, USDA- ARS and Dep. of Agronomy, Iowa State Univ., Ames, IA 50011; A. Martin-Schwarze, Dep. of Statistics, Iowa State Univ., Ames, IA 50011; S. Flint-Garcia, Division of Plant Sciences, Univ. of Missouri, Columbia, MO 65211; M. Bohn, Dep. of Crop Sciences and the Illinois Plant Breeding Center, Univ. of Illinois at Urbana-Champaign, Urbana, IL 61801; A. Lipka, Dep. of Crop Sciences, Univ. of Illinois at Urbana- Champaign, Urbana, IL 61801.

Abstract Variation in kernel composition across maize (Zea core ideas mays L.) germplasm is affected by a combination of the plant’s genotype, the environment in which it is grown, and the interaction • Exotic germplasm may be useful for altering between these two elements. Adapting exotic germplasm to the temperate kernel characteristics. US Corn Belt is highly dependent on the plant’s genotype, the • Single nucleotide polymorphism markers were environment where it is grown, and the interaction between these associated with composition traits. components. Phenotypic plasticity is ill-defined when specific • The stability and usefulness of exotic germplasm was exotic germplasm is moved over large latitudinal distances and demonstrated. for the adapted variants being created. Reduced plasticity (or stability) is desired for the adapted variants, as it allows for a more rapid implementation into breeding programs throughout the Corn Belt. Here, doubled haploid lines derived from exotic he overall fitness of a plant is highly dependent on maize and adapted through backcrossing exotic germplasm to Tthe plant’s ability to adapt to changing environmental elite adapted lines were used in conjunction with genome-wide factors through changes made in physiology, metabolism, association studies to explore stability in four kernel composition growth, and intermediate development (Lee et al., 2002; traits. Genotypes demonstrated a response to environments that Marais et al., 2013). These heritable changes are known paralleled the mean response of all genotypes used across as phenotypic plasticity, which can be selected for or all traits, with protein content and kernel density exhibiting the against with traditional plant breeding techniques (Gage highest levels of Type II stability. Genes such as opaque10, empty et al., 2017; Pigliucci, 2005). However, even with maize pericarp 16, and floury 1 were identified as potential candidates showing large amounts of phenotypic plasticity, relatively within quantitative trait locus regions. The findings within this little emphasis has been placed on introducing tropical study aid in validating previously identified genomic regions and germplasm to widen the U.S. germplasm pool (Pollak, identified novel genomic regions affecting kernel quality traits. 2003). Long daylength Corn Belt environments cause tropical maize to exhibit delayed flowering (Allison and Abbreviations: BGEM, inbred line from Iowa State University in the Daynard, 1979; Warrington and Kanemasu, 1983), which Germplasm Enhancement of Maize project; BLUP, best linear unbiased prediction; DH, doubled haploid; GWAS, genome-wide association study;

LD, linkage disequilibrium; Meff, the effective number of independent tests; Citation: Wang, J., S. Li, A.E. Lipka, S. Flint-Garcia, M. Bohn, J. MLM, mixed linear model; NIRT, near-infrared transmission spectroscopy; Edwards, and T. Lübberstedt. 2019. Stability analysis of kernel quality QTL, quantitative trait locus; SNP, single nucleotide polymorphism. traits in exotic-derived double haploid maize lines. Plant Genome 12:170114. doi: 10.3835/plantgenome2017.12.0114

Received 15 Dec. 2017. Accepted 15 July 2018. *Corresponding author ([email protected]).

This is an open access article distributed under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Copyright © Crop Science Society of America 5585 Guilford Rd., Madison, WI 53711 USA the plant genome  vol. 12, no. 1  march 2019 1 of 14 makes adapting the germplasm to its new environment Vasal, 1992; Worral et al., 2015). Additional studies were precarious at best. Hallauer and Carena (2014) theorized, completed with the aim of finding additional protein however, that for some traits, the addition of tropical quality QTLs without yield reduction (Cook et al., 2012; germplasm could confer better characteristics than those Dudley et al., 2007, 2004; Laurie et al., 2004); however, of the 100% temperate germplasm currently being used yield reductions were still noted. in the Corn Belt. This is especially important, since maize Starch composition in maize kernels is well under- is an important food staple for many people around the stood (Wang et al., 2015). Shrunken1, shrunken2, brittle2, world, where it is directly consumed by humans or indi- waxy1, sugary2, dull1, amylose extender1, and sugary1 rectly through livestock consumption (Cook et al., 2012). (Hennen‐Bierwagen and Myers, 2013; Huang et al., Oil, starch, and protein content directly affect the 2014; Klösgen et al., 1986) are a few of the known genes quality of food and feed (Baker et al., 1969; Lewis et al., involved in starch biosynthesis. However, little is known 1982; Worral et al., 2015). Maize is also a popular fuel about the connections of these genes in the regula- source, as its starch content has become important in the tion of starch biosynthesis and starch accumulation in production of ethanol, potentially creating bottlenecks in maize (Wang et al., 2015). Many QTL studies have been the stable supply of sufficient maize for food, feed, and fuel conducted to further explain the connections of known (Ray et al., 2013; Wang et al., 2015). Balancing maize kernel maize starch genes to their actual roles in starch biosyn- composition for both food and fuel use is an even greater thesis in the maize kernel (Cook et al., 2012; Dudley et challenge, as altering kernel composition greatly impacts al., 2007, 2004; Goldman et al., 1993; Guo et al., 2013) yield (Bjarnason and Vasal, 1992; Worral et al., 2015). with few causative genetic factors having been found Zhang et al. (2016) reported that increasing starch content (Wang et al., 2015). Wang et al. (2015) found large-effect has little impact on yield, but the grain is less nutritious for QTLs in a biparental mapping population. They suggest food and feed (Yang et al., 2016). Increasing oil and pro- that marker-assisted selection be implemented for these tein content is associated with substantially reduced grain few known large-effect QTLs, as progress with many yields (Bjarnason and Vasal, 1992, Worral et al., 2015). known small-effect QTLs is not advantageous. Maize Finding the optimal balance of starch, oil, and protein is hybrids that are high in starch have lower protein and a struggle that modern plant breeders must undertake to oil, although yield reductions are not noted, as starch supply the demand of maize grain in the future. imposes lower energy demands on the plant than pro- Specific breeding populations have been developed ducing oil and protein (Zhang et al., 2016). for high-oil maize (Li et al., 2013), such as the Illinois Many of the previous studies concentrated on lines high-oil population (Dudley and Lambert, 2010), where that were developed from the Illinois long-term selection selection has been going on for over a century. Others are program or closely related materials. Here, a diverse set the Alexho single-kernel synthetic population (Lambert of exotic-derived doubled haploid (DH) lines that repre- et al., 2010) and the Beijing high-oil population (Song sent 52 maize races are used to investigate the following and Chen, 2004). These populations have been of great maize kernel composition traits: (i) oil content, (ii) protein importance in elucidating the underlying pathways in content, (iii) starch content, and (iv) kernel density. The oil biosynthesis in maize (Laurie et al., 2004; Yang et al., overall goals of the study were to: (i) characterize kernel 2010). The nested association mapping population and composition traits in this diverse panel by collecting the panel of 282 diverse inbred lines (Flint-Garcia et al., open-pollinated ears and subjecting them to near-infrared 2005) were used to find 22 quantitative trait loci (QTLs) transmission spectroscopy (NIRT), (ii) examine adapta- affecting oil concentration, explaining nearly 70% of the tion of DH lines through use of a modified Finlay–Wilkin- phenotypic variation (Cook et al., 2012). Li et al. (2013) son joint-regression analysis (Finlay and Wilkinson, found 26 QTLs that explained 83% of the phenotypic 1963) to determine genotype × environment effects, and variation in a diverse panel of inbred lines that consisted (iii) conduct genome-wide association studies (GWAS) of normal and high-oil inbred lines. Although large to identify the causative genomic regions responsible for amounts of phenotypic variation were explained in these adaptation-related effects on kernel composition. studies, no large-effect QTLs were found. Efforts to alter protein content have been success- Materials and Methods ful in many selection experiments. The Illinois high-oil population successfully increased protein content levels Germplasm and Data Collection to 27% (Moose et al., 2004). Worral et al. (2015) recently A total of 252 DH lines were developed from crosses released several quality protein maize lines with average made between exotic maize accessions and the expired protein levels but increased concentrations of lysine and Plant Variety Protection lines ‘PHB47’ and ‘PHZ51’, tryptophan, which are important in overall protein qual- which were backcrossed to PHB47 or PHZ51, respec- ity. The success of these lines was attributed to the use of tively, prior to undergoing the DH process to produce opaque2 modifier genes (Paez et al., 1969; Prasanna et al., isogenic doubled haploid lines. This process is described 2001; Vasal et al., 1980; Villegas et al., 1992; Worral et al., in greater detail by Brenner et al. (2012) and Vanous et 2015; Yang et al., 2016). However, yield reductions have al. (2018). The DH lines were developed and released in been noted when opaque2 was present (Bjarnason and a joint collaboration between the Iowa State University

2 of 14 the plant genome  vol. 12, no. 1  march 2019 Doubled Haploid Facility and the USDA-ARS Germ- be fixed effects. Random effects included the BGEM lines plasm Enhancement of Maize project. Released lines and incomplete blocks. Outliers were then identified and are known as BGEM lines, B indicating Iowa State Uni- removed. Outliers were identified via the studentized con- versity inbred line and GEM indicating the Germplasm ditional residual and by determining the 95th quantile, Enhancement of Maize project. Supplemental Table S1 with Bonferroni correction, based on a t-distribution. shows the BGEM line, race, accession number, country of Data falling outside the defined regions were removed origin, and elevation for the 252 BGEM lines. from further analysis. Data were assembled to include The BGEM lines were planted in the field in 2013 in an α all locations once the outliers were removed. These data incomplete block design across six environments. Rows were were fitted to a MLM containing locations, replications, 9 m long and 15 seeds were planted per row. The six environ- incomplete blocks, and BGEM lines as random effects. ments were: (i) Crop Sciences Research and Education Center, Replications were nested within locations and incom- Champaign, IL, University of Illinois: two replications; (ii) plete blocks were nested within replications. Data files for North Central Region Plant Introduction Station, Ames, IA, the estimates of random and fixed effects were obtained Iowa State University: three replications; (iii) Bradford Farm, from SAS version 9.2 (SAS Inst. Inc., Cary, NC) and were Columbia, MO, University of Missouri: two replications; (iv) used in R (R Core Team, 2014) to calculate the best linear Genetics Farm, Columbia, MO, University of Missouri: one unbiased predictions (BLUPs) via a custom script. The replication; (v) North Central Region Plant Introduction Sta- custom script added the fixed effect intercept and random tion, Ames, IA, Iowa State University: two replications; and location, genotype, and genotype × environment BLUP (vi) Burkey Farm, Ames, IA, Iowa State University: three estimates for each BGEM line within each location and replications. At each location, four inbred lines, ‘B73’ (Rus- across locations. Phenotypic trait correlations were cal- sell, 1972), ‘Mo17’ (Zuber, 1973), PHB47, and PHZ51, were culated via Pearson’s product–moment correlation coef- included as checks. Each check was repeated a minimum of ficient r( ) in the package Hmisc (Harrell, 2016) in R (R six times and a maximum of nine times per replication. Core Team, 2014) with cross-location BLUP values. In the fall, three to five open-pollinated ears were A modified Finlay–Wilkinson (Finlay and Wilkinson, harvested from the center of each row across all loca- 1963) approach, which was similar to that of Eberhart and tions. Ears were dried and bulk shelled. Near-infrared Russell (1966), was used to assess stability of the BGEM transmission spectroscopy was used to analyze oil, lines. Single-location BLUP values from each BGEM line starch, and protein content, as well as the density of were used as input data for the stability equation: each bulked sample. This NIRT analysis was conducted y=+ u g ++( 1 bh) + e with a Foss Infratec 1241 Grain Analyzer (Serial num- ij i i j ij , [1] ber 12412188, Foss North America, Eden Prairie, MN), th with five subsamples per sample. This device performs where yij is the BLUP value phenotype of the i BGEM th th NIRT via near-infrared transmittance. Calibrations were line in the j location, gi is the main effect of the i th obtained from the Grain Quality Laboratory, Iowa State BGEM line, hj is the main effect of the j location, and eij University. The process used for calibration by the Grain is the residual error term. Following Kusmec et al. (2017),

Quality Laboratory is described in Rippke et al. (1995) the variance of eij, or δ, for each BGEM line was used as and calibrations were based on Buchmann et al. (2001). a measure of nonlinearity in the BGEM lines’ response The calibration ranges for protein were 3.78 to 17.37%. to location. All stability analyses were performed in R (R The ranges for oil and starch were 1.63 to 13.87 and 39.12 Core Team, 2014) via a custom script. to 69.98%, respectively. The density ranges were 1.12 to Broad-sense heritability (h2), on an entry-mean basis, 1.38 g cm–3. The SE of predictions were 0.48, 0.45, 0.76, was calculated from the equation: and 0.02 for protein, oil, starch, and density, respectively. sˆ 2 h2 = g , [2] The percentage of oil, starch, and protein content are sˆ 2 sˆ 2 s+ˆ 2 gxe + e reported on a dry matter percentage weight basis (%wt). g e re Density values are reported at a 15% moisture level. 2 2 where sˆ g is the genotypic variance estimate, sˆ gxe is the 2 Statistical Analysis genotype × environment interaction, sˆ e is the error vari- Phenotypic analyses were conducted using a mixed linear ance estimation, r is the number of replications, and e is model (MLM) in SAS version 9.2 (SAS Inst. Inc., Cary, the number of environments (Hallauer et al., 2010). Stan- NC) as described by (Wolfinger et al., 1997). The analysis dard errors were estimated via Dickerson’s approximation described by (Wolfinger et al., 1997) allowed for the recov- (Dickerson, 1969), as explained in Hallauer et al. (2010). ery of incomplete blocking and inter-variety information when the blocking and varieties are random effects. Model Marker Data parameters were estimated via Henderson’s mixed model Genotyping of the exotic-derived BGEM lines was equations and the variances components were estimated performed via genotyping-by-sequencing (Elshire et via the restricted maximum likelihood method. al., 2011) by the Cornell University Genomic Diversity First, a MLM was fitted to the data from each loca- Facility, resulting in 955,690 single nucleotide polymor- tion. Replications and check inbreds were considered to phisms (SNPs). The Buckler Lab for Maize Genetics and vanous et al. 3 of 14 Diversity performed the assembly process, as described LD among SNPs to capture the correlation and derives Meff by Glaubitz et al. (2014). Marker imputation was per- from the number of principal components that contribute formed in TASSEL version 5.0 (Bradbury et al., 2007) to 99.5% of variation (Gao et al., 2010). The number of inde- via the FILLIN option (Swarts et al., 2014). Markers with pendent tests was found to be 15,729. Bonferroni correction greater than 25% missing data and minor allele frequen- was then conducted via the Meff estimate [Eq. 3]: cies less than 2.5% were removed. The remaining 247,775 0.5 P = , [3] SNPs were filtered further by removing SNPs that were n in the same position according to their genetic map posi- tion. Map positions were based on the intermated B73 where n is the number of markers or the Meff estimate. The × Mo17 (Lee et al., 2002) genetic map, described by Wei resulting significance threshold was set at P = 3.17 × 10−6. et al. (2007). Within these regions, a single random SNP Quantile–quantile plots were constructed for each analy- was chosen, resulting in the retention of 62,077 SNPs. sis. Comparisons of quantile–quantile plots were used to With these markers, the proportion of recurrent par- determine the most appropriate analysis for each measure ents found within the BGEM lines was higher than the from the Finlay–Wilkinson analysis for each trait. expected 75% and the average number of recombination Significant SNPs identified via GWAS were com- events estimated was far greater than expected (Sanchez pared with previously identified QTL regions. First, et al., 2018; Vanous et al., 2018). Hence, SNP marker data genetic map bins that contained the significant GWAS were corrected for monomorphic markers through use of a SNPs were identified. Linkage disequilibrium was com- Bayes theorem correction. Briefly, markers that comprised puted for the significant SNPs and all SNPs that were short distances of recurrent parent contributions and located with the genetic map bin from the marker dataset that were flanked by donor parent genotypes were more used in this study. Linkage disequilibrium was calculated likely to be from the identity of marker alleles between the for these SNPs in TASSEL version 5.0 (Bradbury et al., recurrent and donor parents than a double crossover event 2007) via a sliding window approach. Mean LD values (Sanchez et al., 2018). Regions containing short recurrent were calculated and a threshold of r2 = 0.2 was used to parent regions were tested with the null hypothesis that a determine whether the SNP identified via GWAS over- double crossover event took place, and the correction was lapped with the previous QTL regions. For significant based on the P-values obtained from the test (Rice et al., SNPs found to be concurrent with previously identi- 2018). After correction, donor parent contributions were fied QTLs, 0.5-Mb regions centered on the SNPs were closer to the expected 25% and the number of recombina- scanned for known kernel composition genes, and LD tion events was greatly reduced (Sanchez et al., 2018). between these candidate genes and the associated SNPs were tested in the same manner as described for QTLs. Genome-wide Association Studies Linkage disequilibrium (LD) between SNP markers was Gene Ontology Term Enrichment conducted via a sliding window in TASSEL version 5.0 Candidate gene regions for gene ontology term enrichment (Bradbury et al., 2007). R (R Core Team, 2014) was used to tests were defined by 100kb windows that centered on each fit a smooth line to the data via local regression. Genome- significant SNP from the MLM and FarmCPU GWAS wide association studies were conducted with the intercept, results. Candidate genes were declared if a gene, according slope, and δ estimates obtained from the stability analysis to the B73 RefGen_v2 (Schnable et al., 2009) coordinates, for 232 BGEM lines. Twenty lines were discarded because fell within the predefined windows. The complete maize of segregation being noted in field trials, missing genotypic gene list was obtained from MaizeGDB (Andorf et al., data, or heterozygosity being noted in the genotypic data 2016). Candidate genes within significant regions were (Vanous et al., 2018). Three analytical software packages tested for gene ontology term enrichment. First, candidate were used to conduct GWAS. First, TASSEL version 5.0 genes were connected with the respected gene ontology (Bradbury et al., 2007) was used to conduct a general linear terms obtained from Gramene (ftp://ftp.gramene.org/pub/ model approach. Next, GAPIT (Lipka et al., 2012) was used gramene/archives/PAST_RELEASES/release34b/data/ to perform a MLM approach that included an additive ontology/go/go_ensembl_zea_mays.txt, accessed 13 May genetic relatedness matrix (VanRaden, 2008) to estimate 2018), and gene ontology annotations were retrieved from the variance–covariance between individuals. The model the Gene Ontology Consortium (http://purl.obolibrary.org/ selection option was also used to determine the optimal obo/go.obo, accessed 23 Aug. 2018). A proportion test was number of principal components to retain for the analysis. conducted as described by Li et al. (2012). The optimal number was found to be zero principal com- ponents. Finally, FarmCPU (Liu et al., 2016) was used. In Results each of the three models, a family-wise error rate obtained from simpleM (Gao et al., 2010) was obtained with R (R Phenotypic Analysis Core Team, 2014).The simpleM package calculates the Supplemental Table S2 shows the variance estimates from effective number of independent tests, Meff (Cheverud, the phenotypic trait data analyses. After outlier removal, 2001; Gao et al., 2008; Li and Ji, 2005; Moskvina and 0.10, 0.00, 24.67, and 2.45% of the BGEM samples fell Schmidt, 2008; Nyholt, 2004, 2005), by using composite outside the calibration ranges for protein, oil, and starch

4 of 14 the plant genome  vol. 12, no. 1  march 2019 Table 1. Summary statistics of kernel composition traits for PHZ51-derived, PHB47-derived, and combined PHZ51- and PHB47-derived inbred lines from Iowa State University and the Germplasm Enhancement of Maize project (BGEM lines).

RP† BC1F1 Group Trait Mean‡ SD‡ Mean‡ Min.‡ Max.‡ Range‡ SD‡ h2§ (SE)

RP: PHZ51 BC1F1: Protein (%wt) 11.2 0.86 12.2 10.0 14.1 4.1 0.76 – PHZ51-derived BGEM lines Oil (%wt) 4.0 0.12 4.0 3.6 4.6 1.0 0.19 – Starch (%wt) 70.6 0.88 69.4 67.3 71.5 4.1 0.76 – Density (g cm–3) 1.3 0.01 1.3 1.3 1.4 0.1 0.02 –

RP: PHB47 BC1F1: Protein (%wt) 12.4 0.84 12.9 11.0 15.3 4.3 0.91 – PHB47-derived BGEM lines Oil (%wt) 4.2 0.09 4.2 3.8 4.9 1.0 0.22 – Starch (%wt) 69.1 0.62 68.5 65.9 70.7 4.8 0.99 – Density (g cm–3) 1.3 0.01 1.3 1.3 1.4 0.1 0.02 – Combined Protein (%wt) – – 12.6 10.0 15.2 5.2 0.93 0.85 (0.09) Oil (%wt) – – 4.2 3.6 4.9 1.3 0.23 0.86 (0.09) Starch (%wt) – – 68.9 65.9 71.5 5.6 1.00 0.88 (0.09) Density (g cm–3) – – 1.3 1.3 1.4 0.1 0.02 0.88 (0.09) † RP, recurrent parent. ‡ Values are estimated from trait best linear unbiased predictions of n lines within each group. n = 98, PHZ51; n = 134, PHB47. § h2, broad-sense heritability. content, and kernel density, respectively. All of the BGEM Table 2. Pearson’s product-moment correlation coefficient r( ) between four kernel composition traits for mean and two stability measures. line samples that were outside the calibration ranges for starch were over the 69.86%wt maximum, with 73.56%wt JRA† Trait Protein Oil Starch being the highest recorded sample. PHZ51 and PHB47 Intercept Protein – – – has ranges of 11.2 to 12.4, 4.0 to 4.2, and 69.1 to 70.6%wt Oil 0.43** – – for protein, oil, and starch content, respectively (Table 1). Starch -0.95** -0.64** – PHB47 had higher protein and oil weight percentages, but Density 0.19* 0.04 -0.09 had a lower weight percentage for starch. The BGEM lines Slope Protein – – – had ranges of 10.0 to 15.2, 3.6 to 4.9, and 65.9 to 71.5%wt Oil 0.06 – – for protein, oil, and starch, respectively. Both recurrent Starch 0.85 -0.19 – parents had a density of 1.3 g cm–3, whereas the BGEM Density 0.15 0.06 0.07 lines ranged from 1.3 to 1.4 g cm–3. Supplemental Table S3 δ Protein – – – shows the trait BLUP values of the 232 BGEM lines used Oil 0.22** – – in the association analyses. Broad-sense heritability esti- Starch 0.88** 0.38** – mates were high for all kernel composition traits, ranging Density 0.28** 0.18* 0.26** from 0.85 to 0.88 (Table 1). Oil and protein content had a * Significant atP < 0.01. significant positive correlation r( = 0.43; Table 2, Supple- ** Significant atP < 0.001. mental Fig. S1), whereas oil and protein content both had † JRA, joint regression analysis. significant negative correlations with starch r( = -0.64 and r = -0.95, respectively). Density significantly correlated (Sanchez et al., 2018; Vanous et al., 2018). However, LD with protein (r = 0.19). Correlations among oil, protein, within the BGEM population is complex. The BGEM and starch were consistent with previous reports, such as lines derived from either of the recurrent parent back- those by (Cook et al., 2012). grounds are 75% related, which results in high levels of Reaction norms and fitted response curves (Fig. 1) linkage within lines sharing a common recurrent parent were used to examine stability of the four kernel com- and causing the large decay distances. position traits. Protein, oil, and density fitted response Quantile–quantile plots (Fig. 2, Supplemental Fig. S3 curves (Fig. 1b, d, f, h) showed a linear increase across to Supplemental Fig. S11) were obtained from the three environments. However, the fitted response curves for GWAS analyses with the estimates of intercept, slope, starch showed a more general flat line response (Fig. 1f). and δ response of four kernel composition traits. All Intercept and slope values followed a bell-shaped normal general linear model analyses showed the overestima- distribution (Supplemental Fig. S2; Fig. 3), with intercept tion of observed values compared with expected values values centering on the population mean and the slope (Supplemental Fig. S3, Supplemental Fig. S4, Supplemen- values centering on one. Delta values were found to be tal Fig. S9), whereas MLM tended to slightly underesti- small, with most values near zero (Fig. 4). mate the observed values (Fig. 1, Supplemental Fig. S4, Supplemental Fig. S7, Supplemental Fig. S10). Analyses Genome-wide Association Studies with FarmCPU software generally showed the best per- Linkage disequilibrium decayed (r2 = 0.2 threshold) formance out of the three approaches (Fig. 1; Supplemen- over a distance greater than 500 kb for all chromosomes tal Figure S5; S8; S11), but only minor differences were vanous et al. 5 of 14 Fig. 1. Kernel composition trait stability based on categorical order and joint regression fitted values. (a) Reaction norm based on the categori- cal order of population means for six environments. (b) Fitted response curves of 232 inbred lines from Iowa State University in the Germplasm Enhancement of Maize project (BGEM lines) across six environments. Red lines indicate PHZ51-derived BGEM lines; blue lines indicate PHB47- derived BGEM lines. noted when the FarmCPU results were compared with were found with the slope rather than the intercept. Pro- MLM quantile–quantile plots. For this reason, FarmCPU tein and starch were the only traits that had significant results were chosen for further analysis for 10 of the 12 associations with a SNP when δ values were used in the GWASs performed. The MLM results were chosen for oil GWAS analyses. The genetic effects of significant associa- δ and kernel density δ analysis. tions were identified via the mean and two measures of Significant marker–trait associations P( = 3.17 × 10−6 stability are listed in Supplemental Table S4. threshold) were found in four kernel composition traits Significant associations were compared with previ- with two measures of stability (slope and δ) and a single ously published QTLs (Berke and Rocheford, 1995; Cook measure of the mean (intercept) (Supplemental Fig. S12 et al., 2012; Goldman et al., 1993; Gustafson and de Leon, to Supplemental Fig. S20). Use of the intercept values 2010; Lübberstedt et al., 1997; Melchinger et al., 1998) resulted in finding two, seven, two, and five SNPs for and many associations were found to be located in the protein, oil, starch, and density, respectively. When slope same genetic map bin. Most of these QTLs were found to estimates were used as the phenotype, significant mark- be concurrent with previously identified QTLs (Table 3). ers were found for each trait; however, different SNPs For example, protein associations S1_264529186 (r2 =

6 of 14 the plant genome  vol. 12, no. 1  march 2019 Fig. 2. Quantile–quantile plots of selected analysis for four kernel composition traits with the phenotypic mean and two measures of stability. (a) Protein intercept, (b) protein slope, (c) protein ∆, (d) oil intercept, (e) oil slope, (f) oil δ, (g) starch intercept, (h) starch slope, (i) starch δ, (j) density intercept, (k) density slope, and (l) density δ. vanous et al. 7 of 14 Fig. 3. Histogram of slope values from a modified Finlay–Wilkinson joint regression analysis.

0.24), S2_18205663 (r2 = 0.29), S3_189984261 (r2 = 0.24), composition genes identified eight genes. These genes and S8_150046222 (r2 = 0.22) overlapped with QTL include opaque10 (GRMZM2G346263), empty pericarp 16 regions identified in three different studies (Berke and (GRMZM2G060516), and floury 1 (GRMZM2G094532). A Rocheford, 1995; Cook et al., 2012; Melchinger et al., complete list of key kernel quality genes identified in this 1998); however, three protein associations were found to study is given in Supplemental Table S6. be different from previously identified regions (Table 3). Enrichment tests showed significant enrichment Oil and starch associations were also found to be concur- for catalytic activity, protein serine or threonine kinase rent with previously identified QTLs. Density was the activity, protein phosphorylation, protein kinase activ- only trait that did not have a significant association con- ity, and plasma membrane (Table 4). These results were current with a previously identified QTLs (Table 3). found for protein, oil, and starch, but no significant enrichment was noted for kernel density. Candidate Genes and Gene Ontology Candidate genes are listed in Supplemental Table S5. The Discussion number of genes per region varied from 8 to 33, with a Genotype × environment interactions were studied within mean of 18.2 genes per region. The intercept had a total of four kernel composition traits to determine the level of 37 genes within the defined regions, whereas slope and δ phenotypic plasticity displayed within a diverse set of had 133 and 22 genes, respectively. Protein, oil, and starch exotic derived DH lines. The BGEM lines have previously had 63, 69, and 68 genes within the candidate regions, been shown to exhibit general adaptation in terms of flow- respectively. Comparing these gene lists with known kernel ering and stature (Vanous et al., 2018) but the effects of

8 of 14 the plant genome  vol. 12, no. 1  march 2019 Fig. 4. Histogram of δ values from a modified Finlay–Wilkinson joint regression analysis. the exotic contribution within these lines is unknown for of these increases, one of the remaining two components other traits. Kernel composition traits, obtained via NIRT, must decrease (Cook et al., 2012; Zhang et al., 2016). In were used to explore the contribution of the exotic compo- this study, significant SNPs identified through GWAS were nent of these lines to these traits within varying environ- found to overlap with previously identified QTL regions, ments via a modified Finlay–Wilkinson regression. Trait suggesting that these findings are in agreement with those correlations between the two stability measures and the pleiotropic relationships, and also suggests that many can- mean were found to be moderate to weak (Supplemental didate genes play a role in the genetic architecture control- Table S1). This differs from the results reported by Kusmec ling one or more kernel composition traits. et al. (2017), who reported moderate to strong correla- Stability was assessed via two different measures tions. The two studies differed in the approaches used for obtained from a joint regression analysis. Four types of joint regression and in the traits studied, which explain stability were noted by Bernardo (2002) for assessing the different correlations. Correlations with trait BLUP stability with slope estimates: (i) slope = 0, indicating a values were found to be similar to those reported by Cook constant performance across environments; (ii) slope = et al. (2012). This is in agreement with the premise of plei- 1, indicating that the response to different environments otropy having been suggested for the underlying genetic is the same as the mean response; (iii), slope > 1, indicat- mechanisms controlling maize kernel composition (Cook ing a better than average response in favorable environ- et al., 2012; Goldman et al., 1993; Zhang et al., 2016). Oil, ments and a less than average response to unfavorable starch, and protein content make up the major dry mat- environments; and (iv), slope < 1, indicating a better than ter components in the maize kernel, indicating that if one average response in unfavorable environments and a less vanous et al. 9 of 14 Table 3. Linkage disequilibrium (LD) between significant single nucleotide polymorphisms (SNPs) and previously identified quantitative trait loci (QTLs). Previously identified region Trait JRA† Chr. QTL Bin Start End SNP LD (r2) Reference Protein Intercept 1 qproc24 1.09 250,093,155 267,860,952 S1_264529186 0.24 Melchinger et al. (1998) 10 qproc42 10.03 13,554,330 88,334,564 S10_34900374 0.14 Berke and Rocheford (1995) Slope 2 qproc11 2.03 15,005,045 28,142,812 S2_18205663 0.29 Goldman et al. (1993) qproc40 Berke and Rocheford (1995) m209 Cook et al. (2012) 3 qproc28 3.06 168,366,251 191,054,985 S3_189984261 0.24 Melchinger et al. (1998) qproc29 Melchinger et al. (1998) 8 qproc38 8.03 21,148,080 109,164,471 S8_88275245 0.10 Melchinger et al. (1998) m872 Cook et al. (2012) m912 8.06 146,891,660 165,718,074 S8_150046222 0.22 Cook et al. (2012) m930 Cook et al. (2012) δ 4 qproc30 4.01 694,004 5,041,255 S4_2435702 0.16 Melchinger et al. (1998) Oil Intercept 1 m170 1.11 283,188,047 297,960,525 S1_287722726 0.32 Cook et al. (2012) 2 m221 2.04 28,142,812 71,112,564 S2_41022257 0.23 Cook et al. (2012) qoilc2 Goldman et al. (1993) Slope 4 m538 4.09 204,960,848 236,702,403 S4_228746780 0.19 Cook et al. (2012) 5 qoilc9 5.07 204,605,587 211,726,962 S5_205228239 0.40 Goldman et al. (1993) 6 m739 6.05 120,850,613 153,762,257 S6_128483729 0.30 Cook et al. (2012) qoilc11 Goldman et al. (1993) Starch Slope 3 m353 3.05 126,236,346 168,366,251 S3_166390887 0.30 Cook et al. (2012) 8 m922 8.06 146,891,660 165,718,074 S8_159888337 0.26 Cook et al. (2012) qstc17 Lübberstedt et al. (1997) qsty11 Lübberstedt et al. (1997) δ 7 m820 7.04 156,011,630 168,294,916 S7_164995787 0.23 Cook et al. (2012) qsty7 Lübberstedt et al. (1997) Density Intercept 10 – 10.01 2,647,442 5,005,413 S10_4911495 0.18 Gustafson and de Leon (2010) † JRA, joint regression analysis. than average response in favorable environments. The and functionally distinct groups in controlling the phe- BGEM lines exhibited Type II stability for protein, oil, notypic trait value. starch, and kernel density. This response is expected, as Multiple candidate genes were found within defined the major contributor to the genetic background of these regions surrounding significant SNPs. The candidate gene lines is adapted, temperate lines that both exhibited Type opaque10 was identified and is known to affect zein assem- II stability across the Corn Belt environments. A second bly and protein body morphology (Yao et al., 2016). It is measure of stability, δ, is the variance among residu- also known as a cereal-specific protein body protein (Yao als for each BGEM line. According to Bernardo (2002), et al., 2016). Interestingly, floury 1 was also noted as a pos- a low value of δ indicates that the slope accounts for a sible candidate gene in this study. Floury 1 has been shown large portion of the variation from a genotype’s response to encode a novel endoplasmic reticulum protein that is across environments. Delta estimates for the four kernel also involved in zein protein body formation (Holding et composition traits were all found to be near zero, another al., 2007). Yao et al. (2016) noted that the cotransfection indication of stable performance of the BGEM lines of floury 1 and opaque10 could produce strong luciferase across Corn Belt environments. activity, showing the interaction between these two genes. In the current study, GWAS was conducted with the Finally, empty pericarp 16 was identified as a possible intercept, slope, and δ estimates for four kernel compo- candidate gene. Xiu et al. (2016) suggested that this gene sition traits. Thirty-nine significant SNPs were found is required for mitochondrial nad2 Intron 4 cis‐splicing through these analyses; however, no overlap among and is essential for Complex I assembly and embryogen- intercept, slope, and δ for significant SNPs were noted. esis and the development of endosperm in maize. These Other studies of phenotypic plasticity also noted little to three candidate genes are the best known candidates from no overlap of the SNPs identified through GWAS (Kus- our results. It should also be noted that known candidate mec et al., 2017). Kusmec et al. (2017) noted that these genes were only found within QTL regions for protein and findings are concurrent with the premise that different oil. Many of the starch QTLs tested for LD with significant genes are in involved in controlling the mean and the SNPs were initially discovered from starch concentration plasticity of a trait and that these genes form structurally and starch yield in forage maize; however, forage yields are largely impacted by grain yield (Lorenz et al., 2010) and

10 of 14 the plant genome  vol. 12, no. 1  march 2019 Table 4. Gene ontology enrichment test of four kernel composition traits. Trait JRA† Annotation Group Candidate Genome SD Test statistic P-value ——––— % ——––— Protein Intercept Catalytic activity molecular_function 0.04 0.01 0.01 2.77 P < 0.01 ATP‡ binding molecular_function 0.04 0.03 0.02 0.73 0.47 Oil Intercept Plasma membrane cellular_component 0.04 0.02 0.01 1.73 0.08 Nucleus cellular_component 0.04 0.02 0.01 0.92 0.36 δ ATP binding molecular_function 0.05 0.03 0.02 1.62 0.11 Protein serine or threonine kinase activity molecular_function 0.04 0.01 0.01 2.36 P < 0.05 Protein phosphorylation biological_process 0.04 0.01 0.01 2.27 P < 0.05 Protein kinase activity molecular_function 0.04 0.01 0.01 2.49 P < 0.05 Starch Intercept Plasma membrane cellular_component 0.05 0.02 0.01 2.37 P < 0.05 Density Intercept ATP binding molecular_function 0.05 0.03 0.01 1.60 0.11 Plasma membrane cellular_component 0.03 0.02 0.01 0.94 0.35 Membrane cellular_component 0.05 0.03 0.01 1.59 0.11 δ Protein binding molecular_function 0.03 0.02 0.01 0.90 0.37 ATP binding molecular_function 0.03 0.03 0.01 0.21 0.83 Plasma membrane cellular_component 0.04 0.02 0.01 1.58 0.11 † JRA, joint regression analysis. ‡ ATP, adenosine 5’-triphosphate. increased starch content accompanies higher grain yields including oil, protein, starch, and density, and these asso- (Zhang et al., 2016), leading to justification for including ciations were compared with previously conducted QTL these QTLs for comparison. In addition, only a limited and GWAS studies, validating these previously identified amount of research has been conducted for kernel density, regions. Several significant SNP associations were found thus limiting the number of QTLs for comparison. that correspond with gene models supported by previous In this study, 232 BGEM lines were used to screen findings; however, many could not be validated from pre- kernel composition traits. These lines represented 52 exotic vious studies. The novel regions represent an opportunity maize races but only for 25% of the BGEM lines’ genome. to be further explored and indicate the need for addi- The 25% exotic genome contribution made by these acces- tional studies to be performed to understand and explain sions support the basis for discovery of novel QTL regions. maize kernel composition. Additionally, the use of the However, the 75% contribution of the adapted temper- diverse panel of BGEM lines is a valuable tool to aid in ate lines explains why overlap with previously identified identifying novel regions and genes that control impor- regions was also found. Many previous studies (Dudley tant agronomic traits. Additional BGEM lines will be et al., 2007; Goldman et al., 1993; Gustafson and de Leon, needed to increase the power to detect rare alleles within 2010; Li et al., 2013; Song and Chen, 2004) used germ- the panel, and data on the donor populations could aid plasm derived from a limited pool, comprised mostly of in identifying important lines or populations for specific popular temperate lines (Lee et al., 2002) and derivatives trait improvements. Additional understanding of these of the Illinois high-oil population (Dudley et al., 2004), the trait pathways and how they interact would also serve as Alexho single-kernel synthetic population (Lambert et al., a means for plant breeders to develop maize hybrids that 2010), and the Beijing high-oil population (Song and Chen, are more specifically targeted for food and/or fuel uses. 2004). In summary, diverse germplasm needs to be used in future studies to aid in detecting kernel composition Supplemental Information variants that can be used to supply the world’s demands Supplemental Table S1. BGEM code, exotic donor race for maize and the material used in the current study is a and accession number, donor country of origin, and eleva- starting point. The BGEM population is over-represented tion for 252 BGEM lines. by Corn Belt germplasm; however, with additional lines Supplemental Table S2. Variance component estimates that would lead to the increased representation of exotic from the combined analysis across six Corn Belt environ- germplasm’s contribution, as well as the improvement in ments for four kernel composition traits. the approaches used to identify causal regions for traits Supplemental Table S3. Trait BLUPs of kernel composi- of interest, the BGEM collection of lines could serve as a tion traits for 232 BGEM lines and two temperate recurrent valuable contributor in identifying unique regions effect- parents. ing important agronomic traits. Supplemental Table S4. List of significant markers found through association mapping using two stability Conclusions measures and the mean. This study has supplied new information on the under- Supplemental Table S5. List of candidate genes within lying genetic mechanisms of kernel composition traits. 0.5 Mb of significant SNPs identified through GWAS and Significant associations were found for all four traits, found to be in agreeance with previously identified QTLs. vanous et al. 11 of 14 Supplemental Table S6. Linkage disequilibrium Supplemental Fig. S15. Manhattan plots for the slope between candidate genes with known effects on kernel stability measure of four kernel composition traits analyzed composition and SNP markers identified with association via GLM GWAS. Protein content (% wt), oil content (% wt), mapping. starch content (% wt), and density (g cm–3). Supplemental Fig. S1. Correlation between the pheno- Supplemental Fig. S16. Manhattan plots for the slope typic mean (intercept) and two stability measures across stability measure of four kernel composition traits analyzed four kernel composition traits. via MLM GWAS. Protein content (% wt), oil content (% wt), Supplemental Fig. S2. Histogram of intercept values starch content (% wt), and density (g cm–3). from a modified Finlay–Wilkinson joint regression analysis. Supplemental Fig. S17. Manhattan plots for the slope Supplemental Fig. S3. Quantile–quantile plots for the stability measure of four kernel composition traits analyzed phenotypic mean of four kernel composition traits analyzed via FarmCPU GWAS. Protein content (% wt), oil content (% via GLM GWAS. Protein content (% wt), oil content (% wt), wt), starch content (% wt), and density (g cm–3). starch content (% wt), and density (g cm3). Supplemental Fig. S18. Manhattan plots for the δ stabil- Supplemental Fig. S4. Quantile–quantile plots for the ity measure of four kernel composition traits analyzed via phenotypic mean of four kernel composition traits analyzed GLM GWAS. Protein content (% wt), oil content (% wt), via MLM GWAS. Protein content (% wt), oil content (% wt), starch content (% wt), and density (g cm–3). starch content (% wt), and density (g cm–3). Supplemental Fig. S19. Manhattan plots for the δ stabil- Supplemental Fig. S5. Quantile–quantile plots for the ity measure of four kernel composition traits analyzed via phenotypic mean of four kernel composition traits analyzed MLM GWAS. Protein content (% wt), oil content (% wt), via FarmCPU GWAS. Protein content (% wt), oil content (% starch content (% wt), and density (g cm–3). wt), starch content (% wt), and density (g cm–3). Supplemental Fig. S20. Manhattan plots for the δ stabil- Supplemental Fig. S6. Quantile–quantile plots for the ity measure of four kernel composition traits analyzed via slope stability measure of four kernel composition traits FarmCPU GWAS. Protein content (% wt), oil content (% analyzed via GLM GWAS. Protein content (% wt), oil con- wt), starch content (% wt), and density (g cm–3). tent (% wt), starch content (% wt), and density (g cm–3). Supplemental Fig. S7. Quantile–quantile plots for the Conflict of Interest Disclosure slope stability measure of four kernel composition traits The authors declare that there is no conflict of interest. analyzed via MLM GWAS. Protein content (% wt), oil con- Acknowledgments –3 tent (% wt), starch content (% wt), and density (g cm ). The authors thank the staff of the Iowa State Doubled Haploid Facility and Supplemental Fig. S8. Quantile–quantile plots for the the USDA-ARS Germplasm Enhancement of Maize project. Without the slope stability measure of four kernel composition traits aid of these two groups, completing this work would not have been possible. analyzed via FarmCPU GWAS. Protein content (% wt), oil This work was supported by the Raymond F. Baker Center for Plant Breed- –3 ing, the Iowa State University, the Iowa State University Doubled Haploid content (% wt), starch content (% wt), and density (g cm ). Facility, the USDA-ARS Germplasm Enhancement of Maize Project, and Supplemental Fig. S9. Quantile–quantile plots for the δ the Hatch Multi-State Project NC-007. The authors also thank the two stability measure of four kernel composition traits analyzed anonymous reviewers for their helpful suggestions for the manuscript. using GLM GWAS. Protein content (% wt), oil content (% wt), starch content (% wt), and density (g cm–3). References Allison, J.C.S., and T.B. Daynard. 1979. Effect of change in time of flowering, Supplemental Fig. S10. Quantile–quantile plots for the δ induced by altering photoperiod or temperature, on attributes related stability measure of four kernel composition traits analyzed to yield in maize. Crop Sci. 19:1–4. doi:10.2135/cropsci1979.0011183X0 via MLM GWAS. Protein content (% wt), oil content (% wt), 01900010001x starch content (% wt), and density (g cm–3). Andorf, C.M., E.K. Cannon, J.L. Portwood, 2nd, J.M. Gardiner, L.C. Harper, Supplemental Fig. S11. Quantile–quantile plots for the δ M.L. Schaeffer, et al. 2016. MaizeGDB update: New tools, data and interface for the maize model organism database. Nucleic Acids Res. stability measure of four kernel composition traits analyzed 44:D1195–D1201. doi:10.1093/nar/gkv1007 via FarmCPU GWAS. Protein content (% wt), oil content (% Baker, D., D. Becker, H. Norton, A. Jensen, and B. Harmon. 1969. Lysine wt), starch content (% wt), and density (g cm–3). imbalance of corn protein in the growing pig. J. Anim. Sci. 28:23–26. Supplemental Fig. S12. Manhattan plots for the pheno- doi:10.2527/jas1969.28123x typic mean of four kernel composition traits analyzed via Berke, T.G., and T.R. Rocheford. 1995. Quantitative trait loci for flowering, plant and ear height, and kernel traits in maize. Crop Sci. 35:1542–1549. GLM GWAS. Protein content (% wt), oil content (% wt), doi:10.2135/cropsci1995.0011183X003500060004x –3 starch content (% wt), and density (g cm ). Bernardo, R. 2002. Breeding for quantitative traits in plants. Stemma Press, Supplemental Fig. S13. Manhattan plots for the pheno- Woodbury, MN. typic mean of four kernel composition traits analyzed via Bjarnason, M., and S. Vasal. 1992. Breeding of quality protein maize (QPM). MLM GWAS. Protein content (% wt), oil content (% wt), Plant Breed. Rev. 9:181–216. 3 Bradbury, P.J., Z. Zhang, D.E. Kroon, T.M. Casstevens, Y. Ramdoss, and E.S. starch content (% wt), and density (g cm ). Buckler. 2007. TASSEL: Software for association mapping of complex Supplemental Fig. S14. Manhattan plots for the pheno- traits in diverse samples. Bioinformatics 23:2633–2635. doi:10.1093/ typic mean of four kernel composition traits analyzed via bioinformatics/btm308 FarmCPU GWAS. Protein content (% wt), oil content (% Brenner, E.A., M. Blanco, C. Gardner, and T. Lübberstedt. 2012. Genotypic wt), starch content (% wt), and density (g cm–3). and phenotypic characterization of isogenic doubled haploid exotic

12 of 14 the plant genome  vol. 12, no. 1  march 2019 introgression lines in maize. Mol. Breed. 30:1001–1016. doi:10.1007/ Hallauer, A.R., M.J. Carena, and J.D. Miranda Filho. 2010. Quantitative s11032-011-9684-5 genetics in maize breeding. Springer Science & Business Media, New Buchmann, N., H. Josefsson, and I. Cowe. 2001. Performance of European York. artificial neural network (ANN) calibrations for moisture and protein Harrell F.E., Jr. 2016. Hmisc: Harrell miscellaneous. R package version 3.17-4. in cereals using the Danish near-infrared transmission (NIT) network. R Foundation for Statistical Computing. http://CRAN.R-project.org/ Cereal Chem. 78:572–577. doi:10.1094/CCHEM.2001.78.5.572 package=Hmisc (accessed 23 Aug. 2018). Cheverud, J.M. 2001. A simple correction for multiple comparisons in Hennen‐Bierwagen, T.A. and A.M. Myers. 2013. Genomic specification of interval mapping genome scans. Heredity 87:52–58. doi:10.1046/j.1365- starch biosynthesis in maize endosperm. In: P.W. Becraft, editor, Seed 2540.2001.00901.x genomics. John Wiley & Sons, Inc., Hoboken, NJ. p. 123–137. Cook, J.P., M.D. McMullen, J.B. Holland, F. Tian, P. Bradbury, J. Ross-Ibarra, Holding, D.R., M.S. Otegui, B. Li, R.B. Meeley, T. Dam, B.G. Hunter, et al. et al. 2012. Genetic architecture of maize kernel composition in the 2007. The maize floury1 gene encodes a novel endoplasmic reticulum nested association mapping and inbred association panels. Plant protein involved in zein protein body formation. Plant Cell 19:2569– Physiol. 158:824–834. doi:10.1104/pp.111.185033 2582. doi:10.1105/tpc.107.053538 Dickerson, G. 1969. Techniques for research in quantitative animal genetics Huang, B., T.A. Hennen-Bierwagen, and A.M. Myers. 2014. Functions of In: A.B. Chapman, editor, Techniques and procedures in animal science multiple genes encoding ADP-glucose pyrophosphorylase subunits research. American Society of Animal Science, Albany, NY. in maize endosperm, embryo, and leaf. Plant Physiol. 164:596–611. Dudley, J., and R. Lambert. 2010. 100 generations of selection for oil and pro- doi:10.1104/pp.113.231605 tein in corn. Plant Breed. Rev. 24:79–110. doi:10.1002/9780470650240. Klösgen, R.B., A. Gierl, Z. Schwarz-Sommer, and H. Saedler. 1986. Molecular ch5 analysis of the waxy locus of Zea mays. Mol. Gen. Genet. 203:237–244. Dudley, J.W., D. Clark, T.R. Rocheford, and J.R. LeDeaux. 2007. Genetic anal- doi:10.1007/BF00333960 ysis of corn kernel chemical composition in the random mated 7 gen- Kusmec, A., S. Srinivasan, D. Nettleton and P.S. Schnable. 2017. Distinct eration of the cross of generations 70 of IHP × ILP. Crop Sci. 47:45–57. genetic architectures for phenotype means and plasticities in Zea mays. doi:10.2135/cropsci2006.03.0207 Nature Plants 3:715–723. doi:10.1038/s41477-017-0007-7 Dudley, J.W., A. Dijkhuizen, C. Paul, S.T. Coates, and T.R. Rocheford. 2004. Lambert, R., D. Alexander, and I. Mejaya. 2010. Single kernel selection for Effects of random mating on marker–QTL associations in the cross of increased grain oil in maize synthetics and high‐oil hybrid develop- the Illinois High Protein × Illinois Low Protein maize strains. Crop Sci. ment. Plant Breed. Rev. 24:153–175. 44:1419–1428. doi:10.2135/cropsci2004.1419 Laurie, C.C., S.D. Chasalow, J.R. LeDeaux, R. McCarroll, D. Bush, B. Hauge, Eberhart, S. A., and W. Russell. 1966. Stability parameters for comparing et al. 2004. The genetic architecture of response to long-term artificial varieties. Crop Sci. 6:36–40. doi:10.2135/cropsci1966.0011183X000600 selection for oil concentration in the maize kernel. Genetics 168:2141– 010011x 2155. doi:10.1534/genetics.104.029686 Elshire, R.J., J.C. Glaubitz, Q. Sun, J.A. Poland, K. Kawamoto, E.S. Buckler, et Lee, M., N. Sharopova, W.D. Beavis, D. Grant, M. Katt, D. Blair, et al. al. 2011. A robust, simple genotyping-by-sequencing (GBS) approach 2002. Expanding the genetic map of maize with the intermated for high diversity species. PLoS One 6:E19379. doi:10.1371/journal. B73 x Mo17 (IBM) population. Plant Mol. Biol. 48:453–461. pone.0019379 doi:10.1023/A:1014893521186 Finlay, K., and G. Wilkinson. 1963. The analysis of adaptation in a plant- Lewis, A.J., M.B. Barnes, D.A. Grosbach, and E.R. Peo, Jr. 1982. Sequence in breeding programme. Aust. J. Agric. Res. 14:742–754. doi:10.1071/ which the amino acids of corn (Zea mays) become limiting for growing AR9630742 rats. J. Nutr. 112:782–788. doi:10.1093/jn/112.4.782 Flint-Garcia, S.A., A.C. Thuillet, J.M. Yu, G. Pressoir, S.M. Romero, S.E. Li, H., Z.Y. Peng, X.H. Yang, W.D. Wang, J.J. Fu, J.H. Wang, et al. 2013. Mitchell, et al. 2005. Maize association population: A high-resolution Genome-wide association study dissects the genetic architecture of oil platform for quantitative trait locus dissection. Plant J. 44:1054–1064. biosynthesis in maize kernels. Nat. Genet. 45:43–50. 10.1038/ng.2484 doi:10.1111/j.1365-313X.2005.02591.x Li, J., and L. Ji. 2005. Adjusting multiple testing in multilocus analyses Gage, J.L., D. Jarquin, C. Romay, A. Lorenz, E.S. Buckler, S. Kaeppler, et al. using the eigenvalues of a correlation matrix. Heredity 95:221–227. 2017. The effect of artificial selection on phenotypic plasticity in maize. doi:10.1038/sj.hdy.6800717 Nat. Commun. 8:1348. doi:10.1038/s41467-017-01450-2 Li, X., C. Zhu, C.-T. Yeh, W. Wu, E. Takacs, K. Petsch, et al. 2012. Genic and Gao, X., L.C. Becker, D.M. Becker, J.D. Starmer, and M.A. Province. 2010. non-genic contributions to natural variation of quantitative traits in Avoiding the high Bonferroni penalty in genome-wide association stud- maize. Genome Res. 22:2436–2444. doi:10.1101/gr.140277.112 ies. Genet. Epidemiol. 34:100–105. doi:10.1002/gepi.20430 Lipka, A.E., F. Tian, Q.S. Wang, J. Peiffer, M. Li, P.J. Bradbury, et al. 2012. Gao, X.Y., J. Stamier, and E.R. Martin. 2008. A multiple testing correc- GAPIT: Genome association and prediction integrated tool. Bioinfor- tion method for genetic association studies using correlated single matics 28:2397–2399. doi:10.1093/bioinformatics/bts444 nucleotide polymorphisms. Genet. Epidemiol. 32:361–369. doi:10.1002/ Liu, X., M. Huang, B. Fan, E.S. Buckler, and Z. Zhang. 2016. Iterative usage gepi.20310 of fixed and random effect models for powerful and efficient genome- Glaubitz, J.C., T.M. Casstevens, F. Lu, J. Harriman, R.J. Elshire, Q. Sun, et al. wide association studies. PLoS Genet. 12:E1005767. doi:10.1371/journal. 2014. TASSEL-GBS: A high capacity genotyping by sequencing analysis pgen.1005767 pipeline. PLoS One 9:E90346. doi:10.1371/journal.pone.0090346 Lorenz, A., T. Gustafson, J. Coors, and N. de Leon. 2010. Breeding maize Goldman, I., T. Rocheford, and J. Dudley. 1993. Quantitative trait loci for a bioeconomy: A literature survey examining harvest index and influencing protein and starch concentration in the Illinois long term stover yield and their relationship to grain yield. Crop Sci. 50:1–12. selection maize strains. Theor. Appl. Genet. 87:217–224. doi:10.1007/ doi:10.2135/cropsci2009.02.0086 BF00223767 Lübberstedt, T., A.E. Melchinger, C.C. Schön, H.F. Utz, and D. Klein. 1997. Guo, Y., X. Yang, S. Chander, J. Yan, J. Zhang, T. Song, et al. 2013. Identifica- QTL mapping in testcrosses of European flint lines of maize: I. Com- tion of unconditional and conditional QTL for oil, protein and starch parison of different testers for forage yield traits. Crop Sci. 37:921–931. content in maize. The Crop Journal 1:34–42. doi:10.1016/j.cj.2013.07.010 doi:10.2135/cropsci1997.0011183X003700030037x Gustafson, T., and N. de Leon. 2010. Genetic analysis of maize (Zea mays L.) Marais, D.L.D., K.M. Hernandez, and T.E. Juenger. 2013. Genotype-by- endosperm vitreousness and related hardness traits in the intermated environment interaction and plasticity: Exploring genomic responses of B73 × Mo17 recombinant inbred line population. Crop Sci. 50:2318– plants to the abiotic environment. Annu. Rev. Ecol. Evol. Syst. 44:5–29. 2327. doi:10.2135/cropsci2009.12.0726 doi:10.1146/annurev-ecolsys-110512-135806 Hallauer, A.R., and M.J. Carena. 2014. Adaptation of tropical maize germ- Melchinger, A.E., H.F. Utz, and C.C. Schön. 1998. Quantitative trait locus plasm to temperate environments. Euphytica 196:1–11. doi:10.1007/ (QTL) mapping using different testers and independent population s10681-013-1017-9 samples in maize reveals low power of QTL detection and large bias in estimates of QTL effects. Genetics 149:383–403. vanous et al. 13 of 14 Moose, S.P., J.W. Dudley, and T.R. Rocheford. 2004. Maize selection passes VanRaden, P.M. 2008. Efficient methods to compute genomic predictions. J. the century mark: A unique resource for 21st century genomics. Trends Dairy Sci. 91:4414–4423. doi:10.3168/jds.2007-0980 Plant Sci. 9:358–364. doi:10.1016/j.tplants.2004.05.005 Vasal, S.K., E. Villegas, M. Bjarnason, B. Gelaw, and P. Goertz. 1980. Genetic Moskvina, V., and K.M. Schmidt. 2008. On multiple-testing correction modifiers and breeding strategies in developing hard endosperm in genome-wide association studies. Genet. Epidemiol. 32:567–573. opaque-2 materials. In: W.G. Pollmers and R.H. Phillips, editors, Qual- doi:10.1002/gepi.20331 ity traits of maize for grain and silage use. Martinus Nijhoff, London. Nyholt, D.R. 2004. A simple correction for multiple testing for single-nucleo- p. 37–73. tide polymorphisms in linkage disequilibrium with each other. Am. J. Villegas, E., S. Vasal, M. Bjarnason, and E. Mertz. 1992. Quality protein Hum. Genet. 74:765–769. doi:10.1086/383251 maize—What is it and how was it developed? In: E.T. Mertz, editor. Nyholt, D.R. 2005. Evaluation of Nyholt’s procedure for multiple testing cor- Quality protein maize. American Association of Cereal Chemists, St. rection–author’s reply. Hum. Hered. 60:61–62. doi:10.1159/000087919 Paul, MN. p. 27–48.. Paez, A.V., J.L. Helm, and M.S. Zuber. 1969. Lysine content of opaque-2 maize Wang, T.T., M. Wang, S.T. Hu, Y.N. Xiao, H. Tong, Q.C. Pan, et al. 2015. kernels having different phenotypes. Crop Sci. 9:251–252. doi:10.2135/ Genetic basis of maize kernel starch content revealed by high-density cropsci1969.0011183X000900020045x single nucleotide polymorphism markers in a recombinant inbred line Pigliucci, M. 2005. Evolution of phenotypic plasticity: Where are we going population. BMC Plant Biol. 15: 288. doi:10.1186/s12870-015-0675-2 now? Trends Ecol. Evol. 20:481–486. doi:10.1016/j.tree.2005.06.001 Warrington, I.J., and E.T. Kanemasu. 1983. Corn growth-response to tem- Pollak, L.M. 2003. The history and success of the public–private project on perature and photoperiod.1. Seedling emergence, tassel initiation, and germplasm enhancement of maize (GEM). Adv. Agron. 78:45–87. anthesis. Agron. J. 75:749–754. doi:10.2134/agronj1983.0002196200750 doi:10.1016/S0065-2113(02)78002-4 0050008x Prasanna, B., S. Vasal, B. Kassahun and N. Singh. 2001. Quality protein Wei, F., E. Coe, W. Nelson, A.K. Bharti, F. Engler, E. Butler, et al. 2007. maize. Curr. Sci. India 81:1308–1319. Physical and genetic structure of the maize genome reflects its com- R Core Team. 2014. R: A language and environment for statistical computing. plex evolutionary history. PLoS Genet. 3:E123. doi:10.1371/journal. R Foundation for Statistical Computing, Vienna, Austria. pgen.0030123 Ray, D.K., N.D. Mueller, P.C. West and J.A. Foley. 2013. Yield trends are insuf- Wolfinger, R.D., W.T. Federer, and O. Cordero-Brana. 1997. Recovering ficient to double global crop production by 2050. PLoS One 8: E66428. information in augmented designs, using SAS PROC GLM and PROC doi:10.1371/journal.pone.0066428. MIXED. Agron. J. 89:856–859. doi:10.2134/agronj1997.0002196200890 Rice, B.R., D.L. Sanchez, D. Nettleton, T. Lubberstedt, and A.E. Lipka. 2018. A 0060002x description of monomorphic marker correction. In preparation. Worral, H.M., M.P. Scott, and A.R. Hallauer. 2015. Registration of temper- Rippke, G.R., C.L. Hardy, C.R. Hurburgh, Jr., and T.J. Brumm. 1995. Calibra- ate quality protein maize (QPM) lines BQPM9, BQPM10, BQPM11, tion and field standardization of Tecator Infratec analyzers for corn BQPM12, BQPM13, BQPM14, BQPM15, BQPM16, and BQPM17. J. and soybeans. In: A.M.C. Davies and P. Williams, editors, Proceedings Plant Reg. 9:371–375. doi:10.3198/jpr2014.11.0080crg of the 7th International Conference on Near Infrared Spectroscopy, Xiu, Z., F. Sun, Y. Shen, X. Zhang, R. Jiang, G. Bonnard, et al. 2016. EMPTY Montreal, Canada. 6–11 Aug. 1995. p. 122–131. NIR Publications, PERICARP16 is required for mitochondrial nad2 intron 4 cis‐splicing, Chichester, UK. complex I assembly and seed development in maize. Plant J. 85:507–519. Russell, W.A. 1972. Registration of B70 and B73 parental lines of maize. Crop Yang, J., C. Ji, and Y.R. Wu. 2016. Divergent transactivation of maize storage Sci. 12:721. doi:10.2135/cropsci1972.0011183X001200050085x protein zein genes by the transcription factors Opaque2 and OHPs. Sanchez, D.L., S. Liu, R. Ibrahim, M. Blanco, and T. Lubberstedt. 2018. Genetics 204:581–591. doi:10.1534/genetics.116.192385 Genome-wide association studies of doubled haploid exotic introgres- Yang, X.H., Y.Q. Guo, J.B. Yan, J. Zhang, T.M. Song, T. Rocheford, et al. 2010. sion lines for root system architecture traits in maize (Zea mays L.). Major and minor QTL and epistasis contribute to fatty acid composi- Plant Sci. 268:30–38. doi:10.1016/j.plantsci.2017.12.004 tions and oil concentration in high-oil maize. Theor. Appl. Genet. Schnable, P.S., D. Ware, R.S. Fulton, J.C. Stein, F. Wei, S. Pasternak, et al. 120:665–678. doi:10.1007/s00122-009-1184-1 2009. The B73 maize genome: Complexity, diversity, and dynamics. Yao, D., W. Qi, X. Li, Q. Yang, S. Yan, H. Ling, et al. 2016. Maize opaque10 Science 326:1112–1115. doi:10.1126/science.1178534 encodes a cereal-specific protein that is essential for the proper distribu- Song, T.M., and S.J. Chen. 2004. Long term selection for oil concentration in tion of zeins in endosperm protein bodies. PLoS Genet. 12:E1006270. five maize populations. Maydica 49:9–14. doi:10.1371/journal.pgen.1006270 Swarts, K., H. Li, J.A. Romero Navarro, D. An, M.C. Romay, S. Hearne, et al. Zhang, Z.Y., X.X. Zheng, J. Yang, J. Messing, and Y.R. Wu. 2016. Maize 2014. Novel methods to optimize genotypic imputation for low-cover- endosperm-specific transcription factors O2 and PBF network the age, next-generation sequence data in crop plants. Plant Genome 7(3). regulation of protein and starch synthesis. Proc. Natl. Acad. Sci. USA doi:10.3835/plantgenome2014.05.0023 113:10842–10847. doi:10.1073/pnas.1613721113 Vanous, A., C. Gardner, M. Blanco, A. Martin-Schwarze, A.E. Lipka, S. Flint- Zuber, M.S. 1973. Registration of 20 maize parental lines. Crop Sci. 13:779– Garcia et al. 2018. Association mapping of flowering and height traits in 780. doi:10.2135/cropsci1973.0011183X001300060085x germplasm enhancement of maize doubled haploid (GEM-DH) lines. Plant Genome 1:170083. doi:10.3835/plantgenome2017.09.0083

14 of 14 the plant genome  vol. 12, no. 1  march 2019 Supplemental Table S1. BGEM code, exotic donor race and accession number, donor country of origin and elevation for 252 BGEM lines.

BGEM Race Accession Country Elevation (m) BGEM-0001-N Altiplano PI 485364 Bolivia 2980 BGEM-0002-N Altiplano PI 485364 Bolivia 2980 BGEM-0003-N Altiplano PI 485364 Bolivia 2980 BGEM-0004-N Altiplano PI 485364 Bolivia 2980 BGEM-0005-N Altiplano PI 485364 Bolivia 2980 BGEM-0006-S Altiplano PI 485364 Bolivia 2980 BGEM-0007-S Altiplano PI 485364 Bolivia 2980 BGEM-0008-S Altiplano PI 485364 Bolivia 2980 BGEM-0009-S* Altiplano PI 485364 Bolivia 2980 BGEM-0010-S Altiplano PI 485364 Bolivia 2980 BGEM-0011-S Altiplano PI 485364 Bolivia 2980 BGEM-0012-S Altiplano PI 485364 Bolivia 2980 BGEM-0013-S Altiplano PI 485364 Bolivia 2980 BGEM-0014-S* Altiplano PI 485364 Bolivia 2980 BGEM-0015-S* Ancashino PI 514763 Peru 2700-3100 BGEM-0016-S* Ancashino PI 514763 Peru 2700-3100 BGEM-0017-S Andaqui PI 444284 Colombia 610 BGEM-0018-S Andaqui PI 444284 Colombia 610 BGEM-0019-S Andaqui PI 444284 Colombia 610 BGEM-0020-S* Andaqui PI 444284 Colombia 610 BGEM-0021-S Andaqui PI 444284 Colombia 610 BGEM-0022-S Andaqui PI 444284 Colombia 610 BGEM-0023-S Andaqui PI 444284 Colombia 610 BGEM-0024-S Andaqui PI 444284 Colombia 610 BGEM-0025-S Andaqui PI 444284 Colombia 610 BGEM-0026-S Araguito NSL 283561 Venezuela 170 BGEM-0027-S Araguito NSL 283561 Venezuela 170 BGEM-0028-S Araguito NSL 283561 Venezuela 170 BGEM-0029-S Araguito NSL 283561 Venezuela 170 BGEM-0030-S Araguito NSL 283561 Venezuela 170 BGEM-0031-S Araguito NSL 283561 Venezuela 170 BGEM-0032-S Araguito NSL 283561 Venezuela 170 BGEM-0033-S Araguito NSL 283561 Venezuela 170 BGEM-0034-S Araguito NSL 283561 Venezuela 170 BGEM-0035-S Araguito NSL 283561 Venezuela 170 BGEM-0036-S Araguito NSL 283561 Venezuela 170 BGEM-0037-S Araguito NSL 283561 Venezuela 170 BGEM-0038-N Arequipeno Ames 28878 Peru 1000 BGEM-0039-N Arizona PI 485359 Peru 1500 BGEM-0040-N Arizona PI 485359 Peru 1500 BGEM-0041-S Arizona PI 485359 Peru 1500 BGEM-0042-S Avati Moroti Guapi PI 485458 Paraguay 600 BGEM-0043-S Avati Moroti Guapi PI 485458 Paraguay 600 BGEM-0044-S Blanco Blandito PI 488113 Ecuador 2660 BGEM-0045-N Bofo Ames 28481 Mexico 2458 BGEM-0046-N Bofo Ames 28481 Mexico 2458 BGEM-0047-S Bofo Ames 28481 Mexico 2458 BGEM-0048-S Bofo Ames 28481 Mexico 2458 BGEM-0049-S Bofo Ames 28481 Mexico 2458 BGEM-0050-S Bofo Ames 28481 Mexico 2458 BGEM-0051-S Bofo Ames 28481 Mexico 2458 BGEM-0052-S Bofo Ames 28481 Mexico 2458 BGEM-0053-S Cabuya PI 445323 Colombia 2380 BGEM-0054-S Camelia NSL 42755 Chile 510 BGEM-0055-S Camelia NSL 42755 Chile 510 BGEM-0056-S* Camelia NSL 42755 Chile 510 BGEM-0057-S* Camelia NSL 42755 Chile 510 BGEM-0058-N Candela NSL 287040 Ecuador 10-300 BGEM-0059-S Candela NSL 287040 Ecuador 10-300 BGEM-0060-S Candela NSL 287040 Ecuador 10-300 BGEM-0061-N Capio rosado Ames 28794 Argentina 2400 BGEM-0062-N* Caraja Ames 28919 Brazil BGEM-0063-N Chandelle Ames 28574 Cuba Low BGEM-0064-N Chandelle Ames 28574 Cuba Low BGEM-0065-N Confite Puntiagudo Ames 28653 Peru 2500-3500 BGEM-0066-N Confite Puntiagudo Ames 28653 Peru 2500-3500 BGEM-0067-S Confite Puntiagudo Ames 28653 Peru 2500-3500 BGEM-0068-S Confite Puntiagudo Ames 28653 Peru 2500-3500 BGEM-0069-S Conico Norteno PI 515577 Mexico 1600-2100 BGEM-0070-S Conico Norteno PI 515577 Mexico 1600-2100 BGEM-0071-S Conico Norteno PI 515577 Mexico 1600-2100 BGEM-0072-S Conico Norteno PI 515577 Mexico 1600-2100 BGEM-0073-S Conico Norteno PI 515577 Mexico 1600-2100 BGEM-0074-S Conico Norteno PI 515577 Mexico 1600-2100 BGEM-0075-S Conico Norteno PI 515577 Mexico 1600-2100 BGEM-0076-S Conico Norteno PI 515577 Mexico 1600-2100 BGEM-0077-S Cristalino amarallo PI 516163 Argentina 345 BGEM-0078-S Cristalino amarallo PI 516163 Argentina 345 BGEM-0079-S Cristalino amarallo PI 516163 Argentina 345 BGEM-0080-S Cristalino amarallo PI 516163 Argentina 345 BGEM-0081-S Cristalino amarallo PI 516163 Argentina 345 BGEM-0082-S Cristalino amarallo PI 516163 Argentina 345 BGEM-0083-S* Cubano dentado PI 485383 Bolivia 440 BGEM-0084-S Curagua Grande PI 485412 Chile 490 BGEM-0085-N Cuzco PI 485274 Peru 2400-3300 BGEM-0086-N Cuzco PI 485274 Peru 2400-3300 BGEM-0087-N Dulcillo del Noroeste PI 490973 Mexico 1500 BGEM-0088-N Dulcillo del Noroeste PI 490973 Mexico 1500 BGEM-0089-N Dulcillo del Noroeste PI 490973 Mexico 1500 BGEM-0090-N Dulcillo del Noroeste PI 490973 Mexico 1500 BGEM-0091-N Dulcillo del Noroeste PI 490973 Mexico 1500 BGEM-0092-S Dulcillo del Noroeste PI 490973 Mexico 1500 BGEM-0093-S Dulcillo del Noroeste PI 490973 Mexico 1500 BGEM-0094-S Dulcillo del Noroeste PI 490973 Mexico 1500 BGEM-0095-S Dulcillo del Noroeste PI 490973 Mexico 1500 BGEM-0096-N Early Caribbean Ames 28579 Martinique Low BGEM-0097-S Early Caribbean Ames 28579 Martinique Low BGEM-0098-S Early Caribbean Ames 28579 Martinique Low BGEM-0099-S Early Caribbean Ames 28579 Martinique Low BGEM-0100-S Early Caribbean Ames 28579 Martinique Low BGEM-0101-S Early Caribbean Ames 28579 Martinique Low BGEM-0102-N Elotes Occidentales PI 484828 Mexico 0-1500 BGEM-0103-N* Elotes Occidentales PI 484828 Mexico 0-1500 BGEM-0104-N* Elotes Occidentales PI 484828 Mexico 0-1500 BGEM-0105-N Elotes Occidentales PI 484828 Mexico 0-1500 BGEM-0106-N Elotes Occidentales PI 484828 Mexico 0-1500 BGEM-0107-N Elotes Occidentales PI 484828 Mexico 0-1500 BGEM-0108-S Elotes Occidentales PI 484828 Mexico 0-1500 BGEM-0109-N Elotes Occidentales PI 628414 Mexico 1200 BGEM-0110-N Elotes Occidentales PI 628414 Mexico 1200 BGEM-0111-S Gordo PI 484406 Mexico 2912 BGEM-0112-S Gordo PI 484406 Mexico 2912 BGEM-0113-S Gordo PI 484406 Mexico 2912 BGEM-0114-S Gordo PI 484406 Mexico 2912 BGEM-0115-S Gordo PI 484406 Mexico 2912 BGEM-0116-S Gordo PI 484406 Mexico 2912 BGEM-0117-S Gordo PI 484406 Mexico 2912 BGEM-0118-S Gordo PI 484406 Mexico 2912 BGEM-0119-S Gordo PI 484406 Mexico 2912 BGEM-0120-N Jora PI 571477 Peru 400 BGEM-0121-N Jora PI 571477 Peru 400 BGEM-0122-N Jora PI 571477 Peru 400 BGEM-0123-N* Karapampa NSL 286824 Bolivia 2120 BGEM-0124-N Karapampa NSL 286824 Bolivia 2120 BGEM-0125-N Karapampa NSL 286824 Bolivia 2120 BGEM-0126-N Karapampa NSL 286824 Bolivia 2120 BGEM-0127-N Karapampa NSL 286824 Bolivia 2120 BGEM-0128-N Karapampa NSL 286824 Bolivia 2120 BGEM-0129-N Karapampa NSL 286824 Bolivia 2120 BGEM-0130-N Karapampa NSL 286824 Bolivia 2120 BGEM-0131-N Karapampa NSL 286824 Bolivia 2120 BGEM-0132-N Karapampa NSL 286824 Bolivia 2120 BGEM-0133-S Karapampa NSL 286824 Bolivia 2120 BGEM-0134-S Karapampa NSL 286824 Bolivia 2120 BGEM-0135-S Karapampa NSL 286824 Bolivia 2120 BGEM-0136-S Karapampa NSL 286824 Bolivia 2120 BGEM-0137-S Karapampa NSL 286824 Bolivia 2120 BGEM-0138-S Karapampa NSL 286824 Bolivia 2120 BGEM-0139-N Kcello Ecuatoriano Ames 28740 Bolivia 3560 BGEM-0140-N Kcello Ecuatoriano Ames 28740 Bolivia 3560 BGEM-0141-N Kcello Ecuatoriano Ames 28740 Bolivia 3560 BGEM-0142-N* Kcello Ecuatoriano Ames 28740 Bolivia 3560 BGEM-0143-N Kcello Ecuatoriano Ames 28740 Bolivia 3560 BGEM-0144-N Kcello Ecuatoriano Ames 28740 Bolivia 3560 BGEM-0145-N Kcello Ecuatoriano Ames 28740 Bolivia 3560 BGEM-0146-N Kcello Ecuatoriano Ames 28740 Bolivia 3560 BGEM-0147-S Kcello Ecuatoriano Ames 28740 Bolivia 3560 BGEM-0148-S Kcello Ecuatoriano Ames 28740 Bolivia 3560 BGEM-0149-S Kcello Ecuatoriano Ames 28740 Bolivia 3560 BGEM-0150-S Kcello Ecuatoriano Ames 28740 Bolivia 3560 BGEM-0151-N Mishca PI 488016 Ecuador 2620 BGEM-0152-N Mishca PI 488016 Ecuador 2620 BGEM-0153-S Mishca PI 488016 Ecuador 2620 BGEM-0154-S Mishca PI 488016 Ecuador 2620 BGEM-0155-S Mishca PI 488016 Ecuador 2620 BGEM-0156-S* Mishca PI 488016 Ecuador 2620 BGEM-0157-N Mixed Creole PI 489361 Cuba Low BGEM-0158-S Montana PI 445252 Colombia 2105 BGEM-0159-S Montana PI 445252 Colombia 2105 BGEM-0160-S Montana PI 445252 Colombia 2105 BGEM-0161-S Montana PI 445252 Colombia 2105 BGEM-0162-S Morado PI 485373 Bolivia 1590 BGEM-0163-S Morado PI 485373 Bolivia 1590 BGEM-0164-S Morado PI 485373 Bolivia 1590 BGEM-0165-S Morado PI 485373 Bolivia 1590 BGEM-0166-S Morado PI 485373 Bolivia 1590 BGEM-0167-S Morado PI 485373 Bolivia 1590 BGEM-0168-S Morado PI 485373 Bolivia 1590 BGEM-0169-S Morado PI 485373 Bolivia 1590 BGEM-0170-S Morado PI 485373 Bolivia 1590 BGEM-0171-S Morado Canteno PI 515026 Peru 1900 BGEM-0172-S Morado Canteno PI 515026 Peru 1900 BGEM-0173-S Morado Canteno PI 515026 Peru 1900 BGEM-0174-N Morocho PI 571413 Peru 2700 BGEM-0175-S Morocho PI 503511 Peru 2700 BGEM-0176-S Morocho PI 571413 Peru 2700 BGEM-0177-S Morocho PI 571413 Peru 2700 BGEM-0178-S* Oloton Ames 28539 Guatemala 1200-2500 BGEM-0179-S* Oloton Ames 28539 Guatemala 1200-2500 BGEM-0180-S* Oloton Ames 28539 Guatemala 1200-2500 BGEM-0181-N Onaveno PI 484880 Mexico 0-500 BGEM-0182-N Onaveno PI 484880 Mexico 0-500 BGEM-0183-N Onaveno PI 484880 Mexico 0-500 BGEM-0184-N Onaveno PI 484880 Mexico 0-500 BGEM-0185-N Onaveno PI 484880 Mexico 0-500 BGEM-0186-S Onaveno PI 484880 Mexico 0-500 BGEM-0187-S Onaveno PI 484880 Mexico 0-500 BGEM-0188-S Onaveno PI 484880 Mexico 0-500 BGEM-0189-S Onaveno PI 484880 Mexico 0-500 BGEM-0190-N Patillo PI 488039 Ecuador 2600 BGEM-0191-N Patillo PI 488039 Bolivia 3280 BGEM-0192-N Patillo PI 488039 Bolivia 3280 BGEM-0193-N Patillo PI 488039 Bolivia 3280 BGEM-0194-N Patillo Grande Ames 28748 Bolivia 2320 BGEM-0195-N* Patillo Grande Ames 28748 Bolivia 2320 BGEM-0196-N Patillo Grande Ames 28748 Bolivia 2320 BGEM-0197-S* Patillo Grande Ames 28748 Bolivia 2320 BGEM-0198-S Patillo Grande Ames 28748 Bolivia 2320 BGEM-0199-S Patillo Grande Ames 28748 Bolivia 2320 BGEM-0200-S Perla PI 571479 Peru 10-900 BGEM-0201-N Pira PI 445528 Colombia 1100 BGEM-0202-N Pira PI 445528 Colombia 1100 BGEM-0203-N Pisankalla NSL 286568 Bolivia 2240 BGEM-0204-N Pisankalla NSL 286568 Bolivia 2240 BGEM-0205-N Pisankalla NSL 286568 Bolivia 2240 BGEM-0206-N Pisankalla NSL 286568 Bolivia 2240 BGEM-0207-N Pisankalla NSL 286568 Bolivia 2240 BGEM-0208-N Pisankalla NSL 286568 Bolivia 2240 BGEM-0209-N Pisankalla NSL 286568 Bolivia 2240 BGEM-0210-N Pisankalla NSL 286568 Bolivia 2240 BGEM-0211-N Pisankalla NSL 286568 Bolivia 2240 BGEM-0212-N* Pisankalla NSL 286568 Bolivia 2240 BGEM-0213-S Pisankalla NSL 286568 Bolivia 2240 BGEM-0214-S Pisankalla NSL 286568 Bolivia 2240 BGEM-0215-N Pojoso Chico NSL 286760 Bolivia 920 BGEM-0216-N Pojoso Chico NSL 286760 Bolivia 920 BGEM-0217-S Pojoso Chico NSL 286760 Bolivia 920 BGEM-0218-S Pojoso Chico NSL 286760 Bolivia 920 BGEM-0219-S Pojoso Chico NSL 286760 Bolivia 920 BGEM-0220-S Pojoso Chico NSL 286760 Bolivia 920 BGEM-0221-S Pojoso Chico NSL 286760 Bolivia 920 BGEM-0222-S Pojoso Chico NSL 286760 Bolivia 920 BGEM-0223-N San Geronimo Huancavelicano PI 503711 Peru 2500-3500 BGEM-0224-N San Geronimo Huancavelicano PI 503711 Peru 2500-3500 BGEM-0225-N San Geronimo Huancavelicano PI 503711 Peru 2500-3500 BGEM-0226-S Semi dentado paulista PI 449576 Paraguay 500 BGEM-0227-N Suwan Ames 26251 Brazil Low BGEM-0228-N Suwan Ames 26251 Brazil Low BGEM-0229-N Suwan Ames 26251 Brazil Low BGEM-0230-N Suwan Ames 26251 Brazil Low BGEM-0231-N Suwan Ames 26251 Brazil Low BGEM-0232-N Tehua Ames 29075 Mexico 800 BGEM-0233-S Tehua Ames 29075 Mexico 800 BGEM-0234-N Tuxpeno Ames 26252 Brazil Low BGEM-0235-N Tuxpeño Ames 28567 Guatemala 400 BGEM-0236-S Vandeño Ames 28466 Mexico 500 BGEM-0237-N Yucatan PI 445514 Colombia 585 BGEM-0238-N Yucatan PI 445514 Colombia 585 BGEM-0239-N Yucatan PI 445514 Colombia 585 BGEM-0240-N Yucatan PI 445514 Colombia 585 BGEM-0241-N Yucatan PI 445514 Colombia 585 BGEM-0242-N Yucatan PI 445514 Colombia 585 BGEM-0243-S Yucatan PI 445514 Colombia 585 BGEM-0244-S Yucatan PI 445514 Colombia 585 BGEM-0245-S Yucatan PI 445514 Colombia 585 BGEM-0246-N Yungueno NSL 286578 Bolivia 1020 BGEM-0247-N Yungueno NSL 286578 Bolivia 1020 BGEM-0248-N Yungueno NSL 286578 Bolivia 1020 BGEM-0249-N Yungueno NSL 286578 Bolivia 1020 BGEM-0250-S Yungueno NSL 286578 Bolivia 1020 BGEM-0251-S Yunquillano forma andaqui PI 485436 Ecuador 450 BGEM-0252-S Yunquillano forma andaqui PI 485436 Ecuador 450 *lines remove from association analysis Supplemental Table S2. Variance component estimates from the combined analysis across six Corn-Belt environments for four kernel composition traits.

Variance Variance Percent Trait Z Value P- value Component Estimate of Total L 5.28E-03 1.45 0.0729 3.95 R(L) 5.73E-04 1.36 0.0873 0.43 B(L*R) 7.53E-20 - - - Oil G 0.05 9.44 <1E-04 37.97 GxE 0.02 11.07 <1E-04 17.57 error 0.05 31.01 <1E-04 40.07 L 0.18 1.38 0.0841 6.54 R(L) 0.05 1.75 0.0398 2.01 B(L*R) - - - - Protein G 0.91 9.73 <1E-04 33.57 GxE 0.39 8.79 <1E-04 14.44 error 1.18 30.27 <1E-04 43.44 L 0.05 1.14 0.1272 1.96 R(L) 0.04 1.76 0.0396 1.53 B(L*R) - - - - Starch G 1.03 9.63 <1E-04 40.39 GxE 0.4 9.66 <1E-04 15.56 error 1.03 30.28 <1E-04 40.56 L 9.70E-05 1.54 0.0622 11.23 R(L) 4.46E-06 1.48 0.0688 0.51 B(L*R) - - - - Density G 3.33E-04 9.66 <1E-04 38.68 GxE 1.45E-04 11.66 <1E-04 16.79 error 2.83E-04 30.51 <1E-04 32.77 L, Location; R(L), Replication nested within Location; B(L*R), Incomplete Block nest within Replication; G, genotype; GxE, genotype by location interaction. Supplemental Table S3. Trait BLUPs of kernel composition traits for 232 BGEM lines and two temperate recurrent parents.

BGEM Protein Oil Starch Density PHB47 12.4 4.2 69.1 1.3 PHZ51 11.2 4 70.6 1.3 BGEM-0001-N 11.11 3.7 70.77 1.3 BGEM-0002-N 12.32 3.83 69.58 1.31 BGEM-0003-N 12.2 3.9 69.67 1.31 BGEM-0004-N 11.86 4.1 69.47 1.32 BGEM-0005-N 12.09 4.14 69.28 1.32 BGEM-0006-S 14.09 4.27 67.41 1.33 BGEM-0007-S 11.56 4.32 69.6 1.31 BGEM-0008-S 13.41 4.24 68.1 1.34 BGEM-0010-S 13.6 4.13 68.13 1.32 BGEM-0011-S 11.81 4.04 69.68 1.32 BGEM-0012-S 13.14 4.29 68.26 1.34 BGEM-0013-S 12.47 4.01 69.07 1.34 BGEM-0017-S 12.48 4.17 69.02 1.32 BGEM-0018-S 13.39 4.24 68.06 1.33 BGEM-0019-S 12.65 4.1 68.97 1.33 BGEM-0021-S 12.65 3.93 68.93 1.33 BGEM-0022-S 11.77 4.18 69.75 1.33 BGEM-0023-S 12.68 3.87 69.2 1.31 BGEM-0024-S 11.51 4.04 70.04 1.33 BGEM-0025-S 11.72 4.18 69.92 1.33 BGEM-0026-S 13.69 4.01 68.14 1.35 BGEM-0027-S 12.44 4.21 68.9 1.33 BGEM-0028-S 13.98 4.2 67.53 1.35 BGEM-0029-S 13.65 4.23 67.71 1.34 BGEM-0030-S 12.79 4.44 68.39 1.34 BGEM-0031-S 12.52 4.36 68.78 1.33 BGEM-0032-S 12.97 3.91 68.73 1.34 BGEM-0033-S 13.15 4.36 67.95 1.36 BGEM-0034-S 13.44 4.35 67.62 1.34 BGEM-0035-S 12.95 3.83 68.92 1.33 BGEM-0036-S 13.06 4.11 68.44 1.36 BGEM-0037-S 12.63 4.1 68.97 1.35 BGEM-0038-N 11.39 4.03 70.14 1.32 BGEM-0039-N 12.58 3.92 69.21 1.33 BGEM-0040-N 12.22 3.88 69.63 1.34 BGEM-0041-S 12.53 4.04 69.27 1.33 BGEM-0042-S 12.45 4.23 68.94 1.35 BGEM-0043-S 12.26 3.91 69.58 1.3 BGEM-0044-S 11.04 4.31 70.22 1.33 BGEM-0045-N 11.13 4.25 70.33 1.34 BGEM-0046-N 12.17 3.96 69.31 1.29 BGEM-0047-S 12.29 4.24 69.19 1.33 BGEM-0048-S 14.55 4.16 67.21 1.37 BGEM-0049-S 14.49 4.25 67.08 1.34 BGEM-0050-S 12.9 4.02 68.81 1.32 BGEM-0051-S 14.68 4.25 67.01 1.35 BGEM-0052-S 13.06 4.06 68.8 1.35 BGEM-0053-S 13.08 4.47 68.12 1.32 BGEM-0054-S 13.12 3.97 68.71 1.33 BGEM-0055-S 12.89 3.85 69.13 1.32 BGEM-0058-N 11.98 4.25 69.36 1.34 BGEM-0059-S 13.19 4.19 68.14 1.34 BGEM-0060-S 12.67 4.23 68.81 1.34 BGEM-0061-N 12.15 3.74 69.96 1.32 BGEM-0063-N 12.26 4.16 69.28 1.33 BGEM-0064-N 12.95 3.99 68.81 1.33 BGEM-0065-N 12.66 3.95 69.07 1.33 BGEM-0066-N 12.49 3.87 69.44 1.32 BGEM-0067-S 13.36 4.22 68.09 1.34 BGEM-0068-S 13.44 4.46 67.83 1.34 BGEM-0069-S 13.22 4.2 68.39 1.34 BGEM-0070-S 12.92 4.51 68.12 1.33 BGEM-0071-S 12.87 4.06 68.74 1.34 BGEM-0072-S 12.75 4.02 68.91 1.33 BGEM-0073-S 12.55 4.12 69.03 1.33 BGEM-0074-S 12.94 4.63 67.9 1.34 BGEM-0075-S 14.14 4.36 67.23 1.36 BGEM-0076-S 11.01 4.06 70.69 1.32 BGEM-0077-S 13.1 4.09 68.56 1.32 BGEM-0078-S 11.82 4.1 69.64 1.32 BGEM-0079-S 12.52 4.28 68.75 1.31 BGEM-0080-S 12.78 3.93 69.04 1.32 BGEM-0081-S 13.52 4.37 67.9 1.33 BGEM-0082-S 14.11 4.35 67.37 1.36 BGEM-0084-S 11.57 4.18 69.8 1.33 BGEM-0085-N 10.93 3.89 68.92 1.27 BGEM-0086-N 11.62 4.19 69.75 1.32 BGEM-0087-N 11.94 4.24 69.38 1.33 BGEM-0088-N 13.27 4.43 67.96 1.35 BGEM-0089-N 12.71 4.63 68.34 1.33 BGEM-0090-N 12.97 4.23 68.49 1.34 BGEM-0091-N 11.48 3.9 70.31 1.34 BGEM-0092-S 12.05 4.08 69.5 1.32 BGEM-0093-S 11.45 3.98 70.1 1.31 BGEM-0094-S 12.44 4.06 69.1 1.35 BGEM-0095-S 13.63 4.45 67.75 1.36 BGEM-0096-N 11.06 3.83 70.83 1.32 BGEM-0097-S 12.44 3.84 69.54 1.34 BGEM-0098-S 11.86 3.98 69.98 1.33 BGEM-0099-S 12.32 4.33 68.97 1.34 BGEM-0100-S 12.62 3.96 69.23 1.33 BGEM-0101-S 11.17 3.88 70.32 1.3 BGEM-0102-N 11.59 3.79 70.4 1.33 BGEM-0105-N 10.77 4.15 69.65 1.29 BGEM-0106-N 10.47 4.1 70.35 1.29 BGEM-0107-N 12.06 3.92 69.73 1.32 BGEM-0108-S 13.81 3.98 67.83 1.3 BGEM-0109-N 12.35 3.85 68.49 1.32 BGEM-0110-N 12.4 3.61 69.69 1.32 BGEM-0111-S 15.09 4.6 65.97 1.28 BGEM-0112-S 13.95 4.58 67.11 1.28 BGEM-0113-S 14.35 4.6 66.66 1.29 BGEM-0114-S 15.26 4.52 65.94 1.29 BGEM-0115-S 15.1 4.62 65.95 1.29 BGEM-0116-S 14.77 4.64 66.34 1.29 BGEM-0117-S 14.11 4.66 66.76 1.28 BGEM-0118-S 14.14 4.67 66.83 1.28 BGEM-0119-S 12.54 3.99 69.03 1.26 BGEM-0120-N 13.16 4.07 68.45 1.31 BGEM-0121-N 12.87 4.06 68.7 1.33 BGEM-0122-N 10.91 3.81 70.99 1.33 BGEM-0124-N 11.49 4.03 70.11 1.32 BGEM-0125-N 12.37 4.42 68.87 1.34 BGEM-0126-N 11.6 3.98 70.02 1.33 BGEM-0127-N 12.9 3.98 68.94 1.32 BGEM-0128-N 12.96 4.05 68.67 1.33 BGEM-0129-N 12 4.18 69.4 1.35 BGEM-0130-N 12.28 4.21 69.08 1.33 BGEM-0131-N 12.46 4.16 69.05 1.32 BGEM-0132-N 14.11 4.27 67.33 1.34 BGEM-0133-S 14.23 4.03 67.65 1.35 BGEM-0134-S 13.95 4.61 67.07 1.35 BGEM-0135-S 12.83 4.31 68.29 1.32 BGEM-0136-S 13.89 4.79 66.95 1.35 BGEM-0137-S 12.49 4.17 69.01 1.33 BGEM-0138-S 14.73 4.47 66.65 1.36 BGEM-0139-N 12.57 4.04 68.93 1.33 BGEM-0140-N 10.34 3.99 71.23 1.31 BGEM-0141-N 12.82 3.95 69.14 1.34 BGEM-0143-N 11.74 4.3 69.67 1.35 BGEM-0144-N 12.72 4.23 68.76 1.33 BGEM-0145-N 12.65 3.8 69.33 1.33 BGEM-0146-N 12.7 4.06 68.95 1.33 BGEM-0147-S 12.61 4.1 68.87 1.32 BGEM-0148-S 13.85 3.92 68.12 1.34 BGEM-0149-S 12.25 3.95 69.51 1.33 BGEM-0150-S 13.41 4.12 68.16 1.34 BGEM-0151-N 12.59 3.93 69.22 1.32 BGEM-0152-N 11.76 3.79 70.17 1.33 BGEM-0153-S 13.77 4.12 67.9 1.34 BGEM-0154-S 13.43 4.1 68.27 1.34 BGEM-0155-S 12.86 4.12 68.71 1.31 BGEM-0157-N 12.41 3.81 69.58 1.36 BGEM-0158-S 13.15 4.14 68.57 1.34 BGEM-0159-S 12.11 4.37 69.18 1.33 BGEM-0160-S 14.13 4.56 66.95 1.32 BGEM-0161-S 13.22 4.36 68.02 1.32 BGEM-0162-S 12.14 4.08 69.27 1.33 BGEM-0163-S 11.84 4.35 69.05 1.28 BGEM-0164-S 13.21 4.34 68.05 1.33 BGEM-0165-S 13.65 4.19 68.09 1.34 BGEM-0166-S 12.05 4.15 69.18 1.26 BGEM-0167-S 13 4.11 68.7 1.35 BGEM-0168-S 12.5 4.23 68.74 1.32 BGEM-0169-S 12.96 4.32 68.33 1.34 BGEM-0170-S 13.22 4.35 68.17 1.34 BGEM-0171-S 12.05 4.61 69.05 1.34 BGEM-0172-S 13.38 4.25 68.15 1.34 BGEM-0173-S 11.45 3.94 70.21 1.32 BGEM-0174-N 11.88 3.96 69.67 1.31 BGEM-0175-S 11.68 3.91 70.12 1.33 BGEM-0176-S 13.13 4.21 68.46 1.33 BGEM-0177-S 14.23 4.74 66.62 1.36 BGEM-0181-N 11.52 4.08 70.27 1.33 BGEM-0182-N 12.03 4.36 69.19 1.33 BGEM-0183-N 11.62 3.8 70.41 1.33 BGEM-0184-N 12.47 4.25 69.01 1.36 BGEM-0185-N 12.21 4.06 69.4 1.33 BGEM-0186-S 13.04 4.23 68.41 1.33 BGEM-0187-S 12.89 4.23 68.39 1.34 BGEM-0188-S 12.55 4.16 68.36 1.31 BGEM-0189-S 10.97 4.29 70.35 1.31 BGEM-0190-N 11.36 4.03 70.22 1.32 BGEM-0191-N 10.04 4.12 71.45 1.34 BGEM-0192-N 11.26 4.15 70.06 1.34 BGEM-0193-N 10.91 3.99 70.57 1.33 BGEM-0194-N 12.25 3.87 69.6 1.34 BGEM-0196-N 12.13 4.06 69.21 1.32 BGEM-0198-S 13.9 4.32 67.43 1.36 BGEM-0199-S 12.65 4.21 68.61 1.32 BGEM-0200-S 12.31 4.12 69.27 1.34 BGEM-0201-N 13.2 4.03 68.52 1.36 BGEM-0202-N 11.61 4.29 69.67 1.35 BGEM-0203-N 13.17 4.08 68.38 1.35 BGEM-0204-N 12.59 4.03 69.08 1.35 BGEM-0205-N 12.22 3.78 69.63 1.35 BGEM-0206-N 11.51 4.08 70.08 1.33 BGEM-0207-N 12.98 4.36 68.4 1.35 BGEM-0208-N 13.14 4.35 68.25 1.34 BGEM-0209-N 11.18 4 70.41 1.32 BGEM-0210-N 13.49 4.06 68.13 1.37 BGEM-0211-N 11.61 4.12 70 1.33 BGEM-0213-S 12.18 4.37 68.74 1.34 BGEM-0214-S 14.04 4.13 67.78 1.35 BGEM-0215-N 12.95 4.29 68.37 1.32 BGEM-0216-N 11.3 4.26 70.09 1.34 BGEM-0217-S 11.81 4.04 69.81 1.33 BGEM-0218-S 13.04 4.87 67.32 1.31 BGEM-0219-S 11.11 4.34 69.8 1.3 BGEM-0220-S 13.5 4.21 67.99 1.32 BGEM-0221-S 12.3 4.21 69.14 1.32 BGEM-0222-S 12.7 4.31 68.57 1.33 BGEM-0223-N 13.27 3.81 68.71 1.32 BGEM-0224-N 12.71 4.13 68.8 1.33 BGEM-0225-N 12.51 3.79 69.4 1.3 BGEM-0226-S 12.53 4.46 68.61 1.33 BGEM-0227-N 13.43 3.88 68.64 1.34 BGEM-0228-N 12.02 4.11 69.68 1.36 BGEM-0229-N 12.68 4.12 68.99 1.34 BGEM-0230-N 11.75 4.18 69.82 1.35 BGEM-0231-N 12.35 4.28 69.03 1.33 BGEM-0232-N 13.43 4.19 68.16 1.35 BGEM-0233-S 12.66 4.2 68.66 1.33 BGEM-0234-N 12.45 3.88 69.57 1.34 BGEM-0235-N 12.81 4.06 68.83 1.34 BGEM-0236-S 13.34 4.55 67.77 1.33 BGEM-0237-N 12.31 3.69 69.85 1.33 BGEM-0238-N 12.58 3.77 69.43 1.32 BGEM-0239-N 11.93 3.74 70.12 1.33 BGEM-0240-N 12.78 3.89 69.24 1.33 BGEM-0241-N 11.6 3.9 70.13 1.31 BGEM-0242-N 11.83 3.91 69.94 1.32 BGEM-0243-S 12.41 4.13 69.27 1.34 BGEM-0244-S 12.47 4.13 69.23 1.33 BGEM-0245-S 11.58 4.43 69.44 1.33 BGEM-0246-N 11.02 3.96 70.58 1.31 BGEM-0247-N 12.65 4 69.09 1.35 BGEM-0248-N 12.77 4.47 68.48 1.33 BGEM-0249-N 12.92 4.02 68.86 1.34 BGEM-0250-S 12.24 4.2 69.24 1.32 BGEM-0251-S 12.22 4.3 69.05 1.33 BGEM-0252-S 11.94 4.15 69.57 1.34 Supplemental Table S4. List of significant markers found through association mapping using two stability measures and the mean.

Method Trait SNP Chr Position Effect p -value S1_264529186 1 264549186 0.26 1.04E-06 Protein S10_34900374 10 34900374 -0.36 9.93E-07 S1_89240615 1 89240615 -0.06 1.80E-06 S1_259175617 1 259175617 -0.07 1.77E-07 S1_287722726 1 287722726 0.07 2.98E-09 Oil S2_41022257 2 41022257 0.06 1.62E-06 S5_134191041 5 134191041 0.08 2.52E-09 S6_163668041 6 163668041 -0.07 1.20E-07 Intercept S9_135891509 9 135891509 -0.06 2.28E-08 S6_109072099 6 109072099 0.24 2.80E-06 Starch S6_164789563 6 164789563 -0.24 1.87E-06 S6_160623346 6 160623346 -0.01 1.51E-08 S7_171931880 7 171931880 -0.01 3.11E-08 Desnity S8_160916175 8 160916175 0.01 8.64E-08 S10_4911495 10 4911495 0.01 5.65E-07 S10_28866488 10 28866488 0.01 2.51E-07 S2_18205663 2 18205663 -0.09 6.95E-07 S3_189984261 3 189984261 -0.1 2.58E-06 S4_199801614 4 199801614 0.13 1.00E-07 Protein S4_239209342 4 239209342 0.15 3.39E-08 S8_88275245 8 88275245 -0.18 7.07E-07 S8_150046222 8 150046222 -0.17 5.63E-09 S10_138426620 10 138426620 0.12 1.95E-06 S1_45941725 1 45941725 -0.21 1.96E-06 S3_209553302 3 209553302 -0.25 4.14E-07 Slope S4_228746780 4 228746780 0.26 1.27E-06 Oil S5_814310 5 814310 -0.25 1.29E-06 S5_205228239 5 205228239 0.23 2.09E-06 S6_128483729 6 128483729 -0.27 2.13E-07 S2_8502029 2 8502029 -0.33 6.67E-08 S3_166390887 3 166390887 -0.2 8.02E-07 Starch S8_159888337 8 159888337 0.32 7.36E-07 S10_145465426 10 145465426 0.3 5.41E-07 S3_219472117 3 219472117 -0.11 1.96E-06 Density S10_120647899 10 120647899 0.13 2.59E-06 Protein S4_2435702 4 2435702 0.05 1.64E-06 Delta S7_164995787 7 164995787 0.07 5.76E-09 Starch S8_134783872 8 134783872 -0.07 1.94E-07 Supplemental Table S5. List of candidate genes within 0.5 Mb of significant SNPs identified through GWAS and found to be in agreeance with previously identified QTL.

Trait JRA term Chr Pos SNP Gene Name Distance GRMZM5G812580 229218 GRMZM2G141834 204380 GRMZM5G820904 75130 GRMZM2G164084 64616 GRMZM2G033619 18058 GRMZM2G033787 14163 GRMZM2G036505 8913 GRMZM2G036640 4740 Intercept 1 264529186 S1_264529186 GRMZM2G037152 324 GRMZM2G037164 3700 GRMZM2G031022 21165 GRMZM2G073668 62804 GRMZM5G883222 102489 GRMZM2G002499 189565 GRMZM5G810275 209782 GRMZM2G002687 214408 GRMZM2G002699 224532 GRMZM2G019783 207501 GRMZM2G019863 203426 GRMZM2G019876 200872 GRMZM2G020500 130053 GRMZM2G175504 73412 GRMZM2G174807 70321 GRMZM2G047347 1305 2 18205663 S2_18205663 GRMZM2G349344 - GRMZM2G388585 35931 GRMZM2G087678 41827 GRMZM2G163110 102349 GRMZM2G462986 111228 GRMZM2G036996 144898 GRMZM2G172795 211707 Protein GRMZM2G173647 206463 GRMZM2G173654 203880 GRMZM2G702078 172688 GRMZM2G173658 168069 GRMZM2G022934 153859 GRMZM2G321839 131219 3 189984261 S3_189984261 GRMZM2G116700 81847 GRMZM2G116670 45504 AC190560.2_FG002 10099 Slope GRMZM2G116632 942 GRMZM2G105224 78690 GRMZM2G061515 101114 GRMZM2G056983 179540 AC187910.5_FG002 248330 GRMZM2G134717 238409 GRMZM2G060516 154713 AC196465.3_FG002 157442 GRMZM2G060541 152766 GRMZM2G359237 144308 AC196465.3_FG006 136701 GRMZM2G014346 125680 GRMZM2G014066 116914 8 150046222 S8_150046222 GRMZM2G160990 61783 GRMZM2G115766 44272 GRMZM2G115828 31892 GRMZM2G179146 - GRMZM2G163159 82047 GRMZM2G163157 98674 GRMZM2G163141 104463 GRMZM2G068462 202845 GRMZM2G369485 215017 GRMZM2G068586 231896 GRMZM2G170851 249630 GRMZM2G073429 243310 GRMZM2G073508 235524 GRMZM2G073511 229736 GRMZM2G046430 205995 GRMZM2G070322 138151 GRMZM2G070381 125707 GRMZM2G052471 87892 GRMZM2G074356 17801 GRMZM2G074404 7809 1 287722726 S1_287722726 GRMZM2G132971 3424 GRMZM2G133023 289 GRMZM2G448927 25929 GRMZM2G448925 35549 Intercept GRMZM2G371271 72041 GRMZM2G068936 80047 GRMZM2G068952 95515 GRMZM2G068989 101735 GRMZM2G376757 159245 GRMZM2G078200 165909 GRMZM2G110279 134379 GRMZM2G168968 60046 GRMZM2G029001 13889 GRMZM2G029356 2770 2 41022257 S2_41022257 GRMZM2G325131 1486 GRMZM2G030426 7592 GRMZM2G094532 202492 GRMZM2G394321 222085 GRMZM2G375762 193884 GRMZM2G076049 184196 GRMZM2G375782 158890 GRMZM5G885274 143447 GRMZM2G090419 79785 GRMZM2G090505 76025 Oil GRMZM2G090609 71582 GRMZM5G898898 47665 GRMZM5G811908 45912 GRMZM2G385999 307 GRMZM2G090981 1976 GRMZM5G872068 8780 5 205228239 S5_205228239 GRMZM2G181522 47108 GRMZM2G181519 52863 GRMZM5G832989 74785 GRMZM2G109787 92178 GRMZM2G481730 100684 GRMZM2G181507 104637 GRMZM2G181505 110197 GRMZM2G357923 134851 Slope GRMZM2G059496 140576 GRMZM2G059526 145473 GRMZM2G303509 158298 AC203430.3_FG007 160631 AC195458.3_FG001 247678 GRMZM2G044442 174538 GRMZM2G044358 166881 GRMZM2G341155 160562 GRMZM2G458776 61849 GRMZM2G352618 27829 GRMZM2G323888 17642 GRMZM2G019180 15721 GRMZM2G165865 12446 6 128483729 S6_128483729 GRMZM2G008252 - AC207798.2_FG003 65 GRMZM2G002473 37093 GRMZM2G027723 76888 GRMZM2G028110 97099 GRMZM2G137849 154600 GRMZM2G137820 179274 GRMZM2G438378 238318 GRMZM2G013016 237568 GRMZM2G419328 187995 GRMZM2G045433 144189 GRMZM5G863784 136981 AC196009.2_FG008 70904 GRMZM2G059108 11241 3 166390887 S3_166390887 GRMZM2G359397 6292 GRMZM2G058954 - GRMZM2G115000 40942 GRMZM2G411956 55375 GRMZM2G114873 57845 AC203806.3_FG006 137043 GRMZM2G406951 205782 GRMZM2G353236 238666 GRMZM2G353234 238068 GRMZM2G700742 237365 GRMZM2G051604 233204 GRMZM2G461666 228281 GRMZM2G118515 206830 GRMZM2G423917 187544 GRMZM2G122607 180536 GRMZM2G423959 177787 GRMZM2G395429 98382 Slope GRMZM2G095510 91793 GRMZM2G095538 90548 GRMZM2G095541 86470 GRMZM2G302035 8998 GRMZM2G002135 1924 GRMZM2G002147 - 8 159888337 S8_159888337 GRMZM2G302045 2199 GRMZM2G002304 10961 GRMZM2G002315 18863 GRMZM2G302074 20418 GRMZM2G302080 32439 Starch GRMZM2G460566 62919 GRMZM2G460576 65795 GRMZM2G460581 66883 GRMZM2G579057 67962 GRMZM2G161168 68822 GRMZM2G010202 91289 GRMZM2G010176 100197 GRMZM2G010065 103550 GRMZM2G009919 125381 GRMZM2G461186 177903 GRMZM2G579497 178294 GRMZM2G116557 210305 GRMZM2G351484 222947 GRMZM2G351482 221387 GRMZM2G049176 217391 GRMZM2G049077 208482 GRMZM2G037444 166305 GRMZM2G037411 162613 GRMZM2G009538 42103 GRMZM2G009320 38946 GRMZM2G038338 4537 GRMZM2G038303 - GRMZM2G038066 2796 Delta 7 164995787 S7_164995787 GRMZM2G093902 14940 GRMZM2G393671 18822 GRMZM2G577405 23128 GRMZM2G157598 33972 GRMZM2G121022 101082 GRMZM2G169636 160888 GRMZM2G116426 200167 GRMZM2G552376 204716 GRMZM2G417072 205853 GRMZM2G030731 247764 GRMZM2G030578 249979 Supplemental Table S6. Linkage disequilibrium between candidate genes with known effects on kernel composition and SNP markers identified with association mapping.

Trait Chr Pos SNP Gene Genename Start Stop Distance r 2 1 264529186 S1_264529186 o10 GRMZM2G346263 264953283 264959862 424097 0.47 2 18205663 S2_18205663 czog1 GRMZM2G168474 18460858 18463062 255195 0.63 Protein 3 189984261 S3_189984261 mrpi2 GRMZM2G105224 190062951 190065367 78690 - 8 150046222 S8_150046222 emp16 GRMZM2G060516 149887725 149891509 154713 0.29 1 287722726 S1_287722726 gdh1 GRMZM2G178415 287289676 287296815 425911 0.34 2 41022257 S2_41022257 fl1 GRMZM2G094532 41224749 41226347 202492 0.47 Oil 5 205228239 S5_205228239 gln4 GRMZM5G872068 205237019 205240533 8780 0.98 6 128483729 S6_128483729 ga3ox1 GRMZM2G044358 128315121 128316848 166881 0.72