Stable sorghum grain quality QTL were identified using SC35 × RTx430 mapping population

Abstract

Understanding the genetic control and inheritance of grain quality traits is instrumental in facilitating end-use quality improvement. This study was conducted to identify and map quantitative trait loci (QTL) controlling protein, starch, and amylose content in grain sorghum [Sorghum bicolor (L.) Moench] grown under variable environmental conditions. A recombinant inbred line (RIL) population derived from a cross between RTx430 and SC35 was evaluated in six environments across Hays and Manhattan, KS. Significant variation was observed in genotype, environment, and genotype × environment interaction for all three quality traits. Unlike the RILs, the two parental lines did not show significant differences for these traits. However, significant transgressive segregation was observed for all traits resulting in phenotypic performance extending beyond the two parents. A total of seven protein, 10 starch, and 10 amylose content QTL were identified. Chromosomal regions and phenotypic variation (PVE) of QTL were variable across growing conditions. Quantitative trait loci hotspots for all three traits were detected on chromosomes 1 (115.2–119.2 cM) and 2 (118.2–127.4 cM). Candidate gene analysis indicated that these QTL hotspots were conditioned by several transcription factors, such as Cytochrome P450 and basic helix–loop–helix DNA binding protein, which regulate starch and protein accumulation in the grain. The identified genomic regions and underlying candidate genes provide a starting point for further validation and marker-assisted gene pyramiding to improve sorghum grain quality.

Abbreviations
• GWAS
• genome-wide association study
• LOD
• logarithm of the odds
• NIR
• near-infrared reflectance
• PVE
• phenotypic variation
• QTL
• quantitative trait loci
• RIL
• recombinant inbred line
• SNP
• single-nucleotide polymorphism

INTRODUCTION

Sorghum [Sorghum bicolor (L.) Moench] provides nutrition to more than half a billion people in the semi-arid tropics of Africa and Asia (Lindsay, 2010; National Research Council, 1996). It is highly adapted to heat and drought-prone environments (Adotey et al., 2021). More than 80% of global grain sorghum production comes from dryland environments cultivated by subsistence farmers (Blum, 2004; Smith & Frederiksen, 2000). Apart from being a subsistence crop in Asia and Africa, sorghum is largely used as a livestock feed and as raw material for biofuel production in the United States (Murray et al., 2008; Wang et al., 2008). The end-use quality of sorghum is largely influenced by physical and biochemical composition of the grain (Streeter et al., 1990).

Starch and protein are the two main biochemical components that account for ∼70 and 12% of the grain by dry weight, respectively (Bean et al., 2019; Sang et al., 2008). Starch is composed of amylose and amylopectin polysaccharides whose relative abundance and structure determine the nutritional properties and starch-based products of the grain (Bean et al., 2019). Amylose is a linear starch fraction that typically ranges from 0 to ∼30% of starch in grains depending on the genetic background of cultivars (Waniska & Rooney, 2000). Protein content in sorghum is dominated by the storage proteins (kafirins), which form a tight matrix with starch granules thereby reducing both protein and starch digestibility (Duodu et al., 2003). Mutations on the major gene that regulates kafirin level have resulted in lines with high level of lysine when compared with wild types, which avails the opportunity to develop cultivars with high nutritional value to fight malnutrition (Singh & Axtell, 1973).

Grain quality varies among diverse germplasm, and it is highly influenced by the environment and crop management practices (Liu et al., 2013). This provides the opportunity to study and manipulate the genetic and biochemical composition of grain quality traits and design efficient breeding programs for grain quality improvement (Boyles et al., 2018; Smith & Frederiksen, 2000).

Understanding the relative importance of gene, environment, crop management, and their interactions will help select superior genotypes for a desired end use. Employing quantitative genetics and molecular marker technologies will help dissect the genetic control of these complex traits to eventually apply marker-assisted breeding in grain quality improvement. Several previous studies have identified and mapped quantitative trait loci (QTL) for various sorghum grain-quality-related traits. Boyles et al. (2016) reported 10 single-nucleotide polymorphisms (SNPs) significantly associated with 1,000-kernel weight distributed on chromosomes 1–4, and 9, while Tao et al. (2020) reported 10 SNPs for the same trait distributed in all chromosomes except 5 and 9. Similarly, three starch-content QTL were identified on chromosomes 1 and 8 using a diversity panel of 389 individuals tested across 3 yr in one location (Sapkota et al., 2020), while Kimani et al. (2020) reported 14 significant SNPs for starch content on all chromosomes except chromosomes 1 and 6 by testing 196 genotypes in a single year and location. Among the limited QTL mapping reports on protein content, Murray et al. (2008) identified five QTL using a biparental population tested for 2 yr in one location, while Boyles et al. (2017) reported seven QTL on chromosomes 1, 2, 6, 7, and 9 by evaluating 390 diverse genotypes for 2 yr in two locations. Using a large diversity panel of 634 genotypes, Chen et al. (2019) identified 10 significant SNPs for starch and 51 SNPs for amylose content.

Despite the knowledge generated toward grain quality improvement in sorghum, most of the previous studies were carried out in single locations with better control on growth conditions, which makes the wider application of these results less effective. Additionally, while climatic changes can be captured through interannual testing, varying soil types and moisture levels impact phenotypic and genetic outcomes, hence the need for testing across a wider range of environments. The present study evaluated a biparental mapping population across six environments with variable water availability and management practices to identify QTL that are inter- and intra-environment specific. This study further analyzed the candidate genes housed in QTL hotspots to identify the molecular functions driving protein and starch accumulation in the grain.

Core Ideas

• Stable QTL controlling protein, starch, and amylose content were identified through multiple-environment testing.
• Genomic regions conditioning multiple grain quality traits were identified on chromosomes 1 and 2.
• Grain quality QTL were conditioned by transcription factors that regulate starch and protein accumulation.
• NIR spectroscopy enables accurate prediction of grain quality traits in grain sorghum.

MATERIALS AND METHODS Plant material and experimental design

A total of 210 recombinant inbred lines (RILs) derived from a cross between SC35 (postflowering drought tolerant) and RTx430 (postflowering drought susceptible) along with the two parental lines were tested in two locations (Hays and Manhattan, KS) for two growing seasons (2016 and 2017) under six variable water conditions (Chiluwal et al., 2021). Year, location, and moisture condition combinations were taken as separate environments (Table 1). Two water conditions were used in 2016 at Hays: fully irrigated and irrigated until heading, which were designated as ‘Hys2016_FI’ and ‘Hys2016_IH’, respectively. Similarly, in 2017, three water conditions were tested: fully irrigated, irrigated until heading, and rainfed conditions representing three separate environments (Table 1). The rainfed treatment was repeated at Manhattan, KS, in 2017 (Mn2017_RF). Randomized complete block design with two replications was uniformly used across all experiments and environments. Seeds were planted in single rows of 3.6 m long with interrow spacing of 0.75 m in 2016, while two rows with similar interrow spacing was followed in 2017. Seeds were sown on 2 June 2016 and 25 May 2017 at Hays and on 15 June 2017 at Manhattan. Air temperature and precipitation data were recorded using WatchDog data loggers (1000 Series Micro Station, Spectrum Technologies) in each experimental location (for more details see Chiluwal et al., 2021).

TABLE 1 Average temperature and precipitation variations during growing seasons in each of the test environments used to evaluate 210 recombinant inbred lines of sorghum in Kansas

Location	Year	Water condition	Environmenta	Temperature	Precipitation
				°C	mm
Hays, KS	2016	Fully irrigated	Hys2016_FI	31.7–17.2	310.12
Hays, KS	2016	Irrigated until heading	Hys2016_IH	31.7–17.2	310.12
Hays, KS	2017	Fully irrigated	Hys2017_FI	31.3–15.7	229.84
Hays, KS	2017	Irrigated until heading	Hys2017_IH	31.3–15.7	229.84
Hays, KS	2017	Rainfed	Hys2017_RF	31.3–15.7	229.84
Manhattan, KS	2017	Rainfed	Mn2017_RF	30.9–17.8	290.29

a

Hys2016_FI, Hays 2016 fully irrigated; Hys2016_IH, Hays 2016 irrigated until heading; Hys2017_FI, Hays 2017 fully irrigated; Hys2017_IH, Hays 2017 irrigated until heading; Hys2017_RF, Hays 2017 rainfed; Mn2017_RF, Manhattan 2017 rainfed.

Phenotypic data and observations

Protein, amylose, and starch contents were evaluated using near-infrared reflectance (NIR) spectroscopy using Perten DA7250 (Perten Instruments) spectrometer that records NIR absorbance data from 950 to 1,650 nm in 5-nm intervals. Grain samples were taken from three uniform panicles from each replication per trial. Two independent readings or scans were taken within each panicle sample to estimate starch, amylose, and crude protein content using standardized prediction models (Peiris et al., 2019, 2021). The combined average value from three panicles formed a replication. Approximately 20 g of intact grain was sampled and scanned using a Teflon sample cup (60 mm diam. and 10 mm deep). Grains were filled in the Teflon cup and the top was leveled by removing excess grain so that the surface of grains were uniform for optical reading by the instrument. Grain moisture content ranged between 7 and 12%. The final calibrations, using models developed to account for varying levels of grain moisture, had high R² values—0.84 and 0.87 for amylose and starch contents, respectively (Peiris et al., 2021), while that of protein content was >0.83 (Peiris et al., 2019) indicating very high accuracy of predictions.

Phenotypic data analysis

All statistical analyses were conducted using R software (R Core Team, 2021). Correlations between crude protein and starch content were calculated for each environment separately. Analysis of variance was conducted using linear mixed model in lmerTest package (Kunzetsova et al., 2017) and best linear unbiased predictors were generated to run QTL mapping using the following model: Y_ijk = μ + G_i + E_j + GE_ij + e_ijk, where Y_ijk is observed mean, μ is the grand mean, G_i is the effect of the ith genotype, E_j is the effect of the jth environment, GE_ij is the effect of the ith genotype in the jth environment, and e_ijk is the random error. Blocks were nested within environment while genotype (G), environment (E), and G × E were considered as random effects. Broad-sense heritability was calculated from estimated variance components of genotype, genotype × environment interaction, and residual error as follows: H² = $$\delta _{\rm{G}}^2$$/($$\delta _{\rm{G}}^2 + \delta _{{\rm{G}} \times {\rm{E}}/n}^2 + \delta _{{\rm{e}}/nr}^2$$), where $$\delta _{\rm{G}}^2$$, $$\delta _{{\rm{G}} \times {\rm{E}}}^2$$, and $$\delta _{\rm{e}}^2$$ are the estimated genotypic, genotype × environment interaction, and error variances, respectively.

QTL analysis

Marker data and linkage map for this population has previously been published (Bouchet et al., 2017). A total of 5,673 nonmissing SNPs with an average marker density of approximately four markers per 1 cM distance were selected for this study. This same map was used to conduct composite interval mapping based on extended Haley–Knott regression using R/qtl package (Broman et al., 2003) for each trait and environment separately. The logarithm of the odds (LOD) threshold values were determined based on 1,000 permutations. Linkage group and QTL position information from composite interval mapping analysis were used to complete multiple QTL mapping model to estimate percentage of phenotypic variation (PVE) explained by each QTL, and average additive effects using makeqtl() and fitqtl() functions in R/qtl (Arends et al., 2014). Estimated confidence interval (95%) of QTL positions were calculated using lodint() function in R/qtl descending 1.5 LOD value from the QTL peak. Sorghum QTL Atlas database was used to study colocalization of previously identified QTL in sorghum (Mace et al., 2019). Genome regions harboring QTL for the three traits from different environments were considered as QTL hotspots. The QTL names consisted of a combination of acronyms starting with ‘q’ for QTL followed by the first three letters from each trait, chromosome number, and an index number for each QTL (e.g., qAMY1.2 denotes the second amylose content QTL on chromosome 1). The QTL with significant overlap and those that were consistently detected across environments were given identical indices.

Candidate gene identification and pathway analysis

Major QTL regions within 100 kb upstream and downstream of QTL peaks were further explored to identity candidate genes related to the variations. Flanking genomic intervals of the QTL hotspots were used to search candidate genes in the sorghum genome sequence (v2.1) (Goodstein et al., 2012) using bedtools (Quinlan & Hall, 2010). Identified sorghum gene IDs collocating with these QTL hotspots were used to search gene ontology (GO) terms and metabolic pathways were mapped using the AgriGo web-based analysis toolkit (Du et al., 2010; Tian et al., 2017). Candidate genes annotation was based on v2.1 of the sorghum genome. Genes having high similarities (p <.05) with at least three mapping entry (identified candidate genes) were retained in the map and false discovery rate was controlled using Hochberg's approach (Hochberg, 1988).

RESULTS Phenotypic trait variation

Data were checked for normality before ANOVA was conducted (Supplemental Figure S1). Analysis showed significant genotype, environment, and genotype × environment interaction effects for all three traits and means were generated for each genotype and trait. Best linear unbiased predictor means for protein content ranged from 9.7 to 14.9% while that of starch were 65.5–72.3%. Amylose content ranged from 16.5 to 34.0% after adjusting for block and environmental variations (Table 2). The two parental lines did not show any significant difference for the three traits in all six environments (Figure 1). Broad-sense heritability was 83, 72, and 75% for protein, starch, and amylose content, respectively, indicating high proportion of genetic variation among the RILs (Table 2). The RILs showed significant heterosis, which was not surprising given the reproductive nature of sorghum. Starch and protein content showed similar negative correlation across all environments irrespective of water conditions (Figure 2). Similarly, amylose content was negatively correlated with protein content while correlation between starch and amylose content was positive across all environments.

TABLE 2 Descriptive statistics, estimated variance components, and broad-sense heritability of three sorghum grain quality traits tested across six diverse environments in Kansas

Variable	Protein content	Starch content	Amylose content
	%
Minimum	9.72	65.47	16.53
Maximum	14.87	72.32	33.99
Mean	12.66	69.79	26.87
Genotypic variance ($$\delta _{\rm{G}}^2$$)	0.75	1.32	7.84
Genotype × environment variance ($$\delta _{{\rm{G}} \times {\rm{E}}}^2$$)	0.39	1.31	6.97
Error variance ($$\delta _{\rm{e}}^2$$)	0.99	3.39	16.87
Broad-sense heritability (H2)	83	72	75

FIGURE 1. Phenotypic distribution of (a) starch, (b) protein, and (c) amylose content among parental lines and recombinant inbred lines (RILs) tested across six environments: Hys2016_FI, Hays 2016 fully irrigated; Hys2016_IH, Hays 2016 irrigated until heading; Hys2017_FI, Hays 2017 fully irrigated; Hys2017_IH, Hays 2017 irrigated until heading; Hys2017_RF, Hays 2017 rainfed; and Mn2017_RF, Manhattan 2017 rainfed

FIGURE 2. Scatter plot indicating nature of correlation between (a) starch and protein content, (b) starch and amylose content, and (c) protein and amylose content within each environment. Hys2016_FI, Hays 2016 fully irrigated; Hys2016_IH, Hays 2016 irrigated until heading; Hys2017_FI, Hays 2017 fully irrigated; Hys2017_IH, Hays 2017 irrigated until heading; Hys2017_RF, Hays 2017 rainfed; Mn2017_RF, Manhattan 2017 rainfed

QTL mapping Protein content

Composite interval mapping was conducted for each environment separately. Similar analyses were conducted on combined data across all environments. A total of seven protein content QTL (LOD > 3) were detected on chromosomes 1–3, 7, 8, and 10 from all six environments (Table 3). Two major QTL, one on chromosome 1 (115.2–119.2 cM) and another on chromosome 2 (120.3–123.2 cM) were consistently detected in four or five of the six environments and the combined data (Figure 3a,b). Another major QTL on chromosomes 3 (0.7–6.8 cM) was detected at two fully irrigated environments and one rainfed condition (Table 3). Similarly, a QTL on chromosome 7 (103.5–107.3 cM) was consistently detected in irrigated environments and combined multiple environments. A specifically expressed QTL under irrigated-until-heading conditions was detected on chromosome 10 (26.8–50.5 cM) at Hays in 2016 and 2017. With the exception of QTL on chromosome 7, the other four major QTL on chromosomes 1, 2, 3, and 10 gained positive alleles from SC35.

TABLE 3 Chromosomal positions, additive effect, and percentage phenotypic expression of starch, protein, and amylose content quantitative trait loci (QTL) identified across six environments and combined multiple-environment data

QTLa	Environmentb	Chromosome	Marker interval	Position	LOD	Additive effectc	R2
				cM
qPRO1.1	Hys2016_IH	1	S1_15050456–S1_15568205	51.8	4.69	−0.03	1.4
	Hys2017_RF	1	S1_840108–S1_1288629	2.6	6.05	−0.27	2.7
qPRO1.2	Combined	1	S1_56532011–S1_56913169	116.6	6.07	0.56	11.12
	Hys2016_FI	1	S1_56532011–S1_56913169	116.6	6.07	0.55	6.52
	Hys2017_FI	1	S1_56700421–S1_57171649	117.3	6.17	0.86	10.6
	Hys2017_IH	1	S1_1921036–S1_56913169	116.6	4.25	0.52	7.31
qPRO2.1	Combined	2	S2_70733049–S2_70844793	121.7	6.62	0.58	11.71
	Hys2016_FI	2	S2_70733049–S2_70844793	121.7	6.63	0.77	13.57
	Hys2017_FI	2	S2_70571891–S2_71087260	121.7	7.91	0.76	9.83
	Hys2017_IH	2	S2_70571891–S2_71217032	121.7	4.71	0.58	6.41
	Mn2017_RF	2	S2_70571891–S2_71217032	122.3	5.2	0.53	12.66
qPRO3.1	Hys2016_FI	3	S3_899405–S3_1215268	2.7	6.84	0.66	10.49
	Hys2017_FI	3	S3_5100685–S3_5788947	24.2	5.69	−0.15	0.95
	Mn2017_RF	3	S3_1178864–S3_1605688	5.3	3.72	0.41	11.21
qPRO7.1	Combined	7	S7_61576301–S7_62224577	105.5	3.94	−0.36	4.56
	Hys2016_FI	7	S7_61576301–S7_62224577	105.5	3.94	−0.28	3.18
	Hys2017_IH	7	S7_61576301–S7_62224577	104.2	3.55	−0.63	8.45
qPRO8.1	Mn2017_RF	8	S8_5113316–S8_5486677	36.9	7.75	0.47	9.97
qPRO10.1	Hys2016_IH	10	S10_7810139–S10_8599084	40.4	5.92	0.08	0.9
	Hys2017_IH	10	S10_5108607–S10_11083984	49.3	3.49	0.4	2.09
qSTR1.1	Combined	1	S1_62389502–S1_63332584	134.8	7.46	0.28	2.7
qSTR2.1	Hys2016_IH	2	S2_764573–S2_2803343	11	3.2	0.28	2.11
qSTR2.2	Mn2017_RF	2	S2_629250–S2_73065407	125.8	3.08	−0.74	5.34
qSTR3.1	Hys2016_FI	3	S3_4509083–S3_5198092	22	4.55	0.44	6.8
qSTR3.2	Hys2017_IH	3	S3_45640535–S3_52002882	72.2	3.58	0.2	5.1
qSTR5.1	Hys2017_RF	5	S5_4506725–S5_6928694	30.4	3.16	0.31	6.3
qSTR7.1	Hys2017_FI	7	S7_61661923–S7_62224577	105.5	4.17	0.22	5.3
qSTR9.1	Hys2016_FI	9	S9_50222963–S9_50621119	70.5	4.56	0.68	11.03
qSTR10.1	Hys2016_FI	10	S10_7810139–S10_8599084	40.5	3.71	0.72	8.8
	Mn2017_RF	10	S10_4793080–S10_5651346	27	4.02	0.64	3.74
qAMY1.1	Combined	1	S1_62389502–S1_63332584	134.8	9.21	1.53	13.64
	Hys2016_FI	1	S1_62389502–S1_63332584	134.8	9.21	2.13	22.08
	Hys2016_IH	1	S1_12824411–S1_68489248	152.4	5.08	−1	4.5
	Hys2017_FI	1	S1_61831028–S1_63332584	134.8	3.15	1.58	6.09
qAMY2.1	Combined	2	S2_72071413–S2_72806316	125.8	3.56	−1.03	5.67
	Hys2016_FI	2	S2_72071413–S2_72806316	125.8	3.56	−0.17	3.73
	Hys2017_FI	2	S2_70223063–S2_70490233	119.2	4.36	−1.96	7.59
qAMY3.1	Hys2016_IH	3	S3_55465980–S3_56053734	86.9	3.8	0.04	2.5
qAMY3.2	Hys2017_IH	3	S3_9952666–S3_73854306	48.3	3.22	0.6	3.8
qAMY4.1	Hys2017_RF	4	S4_6029059–S4_6233774	30.8	5.2	−1.56	3.8
qAMY4.2	Mn2017_RF	4	S4_47670836–S4_61825776	68.3	3.02	−2.3	6.23
qAMY7.1	Mn2017_RF	7	S7_57751675–S7_57812932	80	3.59	−2.51	6.91
qAMY9.1	Hys2017_FI	9	S9_3102734–S9_3385277	23.5	5.77	−1.43	4.51
qAMY10.1	Mn2017_RF	10	S10_9920177–S10_10591821	48.5	4.01	2.66	9.44
qAMY10.2	Combined	10	S10_56649056–S10_56982210	92.3	3.72	1.02	5.12
	Hys2016_FI	10	S10_56649056–S10_56982210	92.3	3.72	1.14	5.67

a

q, QTL; the following three letters (AMY, PRO, and STR) are abbreviations for traits; QTL that show significant overlap and highly repeatable across environments and traits were indexed using same numbers.

b

Hys2016_FI, Hays 2016 fully irrigated; Hys2016_IH, Hays 2016 irrigated until heading; Hys2017_FI, Hays 2017 fully irrigated; Hys2017_IH, Hays 2017 irrigated until heading; Hys2017_RF, Hays 2017 rainfed; Mn2017_RF, Manhattan 2017 rainfed.

c

The positive alleles came from CS35 (first parent) while the negative alleles were from RTx430 (second parent).

FIGURE 3. Consistently detected amylose, protein, and starch content quantitative trait loci (QTL) and overlapping regions of QTL for the three traits on (a) chromosome 1 and (b) chromosome 2. Hys2016_FI, Hays 2016 fully irrigated; Hys2016_IH, Hays 2016 irrigated until heading; Hys2017_FI, Hays 2017 fully irrigated; Hys2017_IH, Hays 2017 irrigated until heading; Hys2017_RF, Hays 2017 rainfed; Mn2017_RF, Manhattan 2017 rainfed

Starch content

Similar statistical procedures as above were followed to map starch content QTL for the six environments separately and combined across environments. A total of nine starch content QTL with LOD > 3 were identified across six linkage groups explaining 2–11% of phenotypic variation (Table 3). The QTL on chromosome 2 (122–127.4 cM) were specifically detected in rainfed conditions while a QTL on chromosome 9 (11% PVE) was detected under fully irrigated conditions at Hays in 2016. A significant overlap of starch and protein content QTL on chromosome 2 (120.2–127.4 cM) was detected under rainfed conditions in 2017 at the Manhattan location (Figure 4). A cluster of starch content QTL were detected on the long arm of chromosome 1 using best linear unbiased predictors generated using genotype × environment interaction models accounting for environmental variation. This starch content QTL cluster had overlapping regions with protein and amylose content QTL detected in this study.

FIGURE 4. Linkage groups and chromosomal positions of sorghum grain quality quantitative trait loci (QTL) identified in the present study (red, amylose content; green, protein content; and blue, starch content) and previously reported (in black). The intervals between adjacent loci in chromosomes denote the physical distance in mega bases. Previously reported QTL were obtained from https://aussorgm.org.au/sorghum-qtl-atlas/ (Mace et al., 2019)

Amylose content

Among the 10 amylose content QTL identified, two major QTL on chromosomes 1 (134.5–136.1 cM) and 2 (118.2–126.9 cM) were detected under fully irrigated conditions and were confirmed through combined analysis using the multiple-environment data. Two separate QTL on chromosome 3 were detected in environments that were irrigated until heading but not with combined data. Similarly, four separate QTL were detected under rain fed conditions (three QTL at Manhattan 2017 and one at Hays 2017), none of which were overlapping in chromosome position. A region on the long arm of chromosome 2 (118.2–126.9 cM) harbored three major QTL, one for each trait studied (Figure 4).

Candidate genes and gene networks

Emphasis was given to QTL hotspots that were consistently detected across the test environments on specific genome locations (Figure 4). Based on the 200-kb region around the QTL peaks, the two QTL hotspot regions on chromosomes 1 and 2 had a total of 55 and 135 genes, respectively (Supplemental Table 1). Basic helix–loop–helix DNA-binding proteins, histone H2A 12, and histone superfamily proteins were amongst the most prevalent genes in QTL on chromosome 1. The peak of the amylose content QTL on chromosome 1 was closest to UDP-L-rhamnose synthase that is involved in carbohydrate metabolism, nucleotide sugar biosynthesis, and nucleotide rhamnose biosynthesis.

Similarly, QTL regions on chromosome 2 harbored protein kinase (CDK9) superfamily protein, which are involved in primary root development, response to abscisic acid, and osmotic stress tolerance. Genes involved in protein biosynthesis (poly-P/G elongation factor), nucleotide metabolism (adenosine proton symporter ENT3), ribosomal protein (S3Ae), cell cycle regulatory protein (CYCD), and exocytic trafficking (SCAMP) were among the various genes identified in the QTL region of overlap between protein and amylose content on chromosome 2 (Supplemental Table 1). Cytochrome electron carrier (Cytc6a), which is involved in amino–sugar metabolism pathway, was also identified in the QTL hotspot on chromosome 2, justifying the overlap between starch, protein, and amylose content QTL in this hotspot.

Gene ontology analysis showed that protein metabolic (transcription and translation) regulation and oxidative stress response were the main biological processes influenced by genes in these QTL regions (Figure 5a). Similar analysis on cellular functions showed that nucleotide and transmembrane transport activities, peroxide receptors, and heme binding were amongst the main genes that were enriched in the QTL regions identified (Figure 5b).

FIGURE 5. Gene ontology (GO) terms enrichment analysis using AgriGO database. (a) List of enriched GO terms associated with grain quality quantitative trait loci (QTL) including their p value and false discovery rate (FDR). (b) Hierarchical tree graph representing biological processes influenced by the GO terms that are associated with grain quality QTL. The different shades on (b) show level of significance in parenthesis

DISCUSSION

Plant breeding has achieved several breakthroughs in improving yield and disease resistance in grain crops. However, grain quality is one of the persisting challenges that needs to be addressed through the application of modern genetic technologies without compromising grain yield. Compared with other cereals, sorghum is an under-researched crop, where the genetic control of grain quality remains poorly understood (Duressa et al., 2018). In this study, a biparental population was field tested in variable water available environments. Findings indicate that there is a high level of genetic variability in grain quality traits and these traits are governed by several minor- and major-effect QTL, which, in turn, are influenced by the environment.

Genetic hotspots enhancing protein content were identified on chromosomes 1 and 2

Identification of functional genomic regions via QTL mapping enhances our ability to pyramid or clone genes through marker-assisted backcrossing, thus minimizing other confounding genetic and environmental effects. In this study, highly repeatable QTL hotspots were identified, which can be targeted for fine mapping and functional marker development for marker-assisted breeding. Position of protein content QTL identified on chromosome 1 corroborated with previous reports that employed RIL populations (Winn et al., 2009), while QTL positions on chromosome 2 slightly differed from those previously reported (Cuevas et al., 2017; Rhodes et al., 2017), which could be due to differences in mapping strategy (structed vs. unstructured populations) and number of mapping populations employed. Colocalized protein digestibility QTL were reported in the QTL hotspot region on chromosome 1 (Winn et al., 2009), while the QTL hotspot on chromosome 2 was located ∼12.3 Mb upstream of several protein content QTL previously reported using genome-wide association study (GWAS) (Cuevas et al., 2017; Rhodes et al., 2017).

Candidate gene analysis was conducted within 200-kb genomic span centering the QTL peaks to locate and analyze genes in these QTL hotspots. Most of the identified genes were transcription factors related to cell division, cell expansion, and cellular transport (Figure 5a,b), which are related to dry matter (protein, carbohydrates, and lipids) accumulation in cereal grains (Tosi et al., 2009; Watt et al., 2020). Transcription factors such as TaNAC019 in wheat (Triticum aestivum L.) (Gao et al., 2021) and OsNAC20 and OsNAC26 in rice (Wang et al., 2020) were found to coordinately regulate both protein and starch accumulation in the grain. Basic helix–loop–helix transcription factors OsPIL15, and OsPUP7 enhance grain size in rice (Oryza sativa L.) (Ji et al., 2019; Yang et al., 2018). We also found cytochrome electron carrier (Cytc6a) protein that regulates amino–sugar metabolism pathways (Bernroitner et al., 2008), which may explain the overlapping of QTL for the three traits (Mansilla et al., 2018). Targeting transcription factors for future research may be a viable approach to improve starch and protein content in sorghum, which normally are negatively correlated making selection for both traits difficult.

In addition to those two QTL hotspots that were consistently found across multiple environments, some QTL were specifically detected in only a limited number of environments. A QTL on chromosomes 7 (61.6–62.2 Mb) was consistently detected in irrigated environments (Table 3). This region harbors Sobic.007G198600.1 gene, which codes for BTB/POZ-MATH domain family, that are known to have functional roles in stress response, vegetative growth, and reproductive development (Lechner et al., 2011). With the exception of this QTL on chromosome 7, the other major protein content QTL gained positive alleles from SC35 (Table 3). Gene ontology analysis indicated that genes housed in these QTL were mainly involved in five different processes including protein metabolism, redox homeostasis, nucleotide and nucleoside transportation across membrane, and heme and tetrapyrrole binding (Supplemental Table S1). One of the genes in the QTL region on chromosome 2 (Sobic.002G363100) encodes for protein biosynthesis elongation factor, which prevents unwanted ribosome stalling and hence could be a putative target to increase protein content (Neelagandan et al., 2020). In addition, Sobic.002G336600, Sobic.002G336700, Sobic.002G337100, and Sobic.002G336200 genes encoded adenosine proton symporter (ENT3) protein, which is required for nucleoside transport and is expressed under limited nitrogen conditions (Cornelius et al., 2012). Exogenously supplied nucleosides can be transported through these transporters to minimize nitrogen deficiency symptoms such as RNA breakdown, reduced levels of amino acids, chlorophyll breakdown, and anthocyanin accumulation (Cornelius et al., 2012) and therefore play an important role in protein content homeostasis. Cytochrome P450 encoded by gene Sobic.002G336100 is known to increase starch and protein content in seeds of transgenic lines overexpressing cytochrome P450 in Arabidopsis thaliana (L.) Heynh. (Yeon et al., 2021). The same QTL hotspot can be explored to improve grain quality in sorghum.

Genetic control of grain starch and amylose under diverse environments

Chromosomal positions of starch content QTL across environments were less consistent than that of protein and amylose content QTL in this study. Two starch QTL that were detected on the proximal and distal ends of chromosome 2 (2.6 and 7.2 MB) were in repulsion phase gaining positive alleles from parents SC35 and RTx430, respectively (Table 3). There is no starch content QTL reported previously on chromosome 2, while one of the two QTL detected on chromosome 3 was previously reported using RIL populations (Patil et al., 2019) and GWAS (Sukumaran et al., 2012). The starch content QTL on the long arm of chromosome 2 collocated with a protein content QTL reported above, but these starch and protein content QTL were in repulsion phase, with the starch QTL gaining its positive allele from RTx430 and the protein QTL from SC35 (Table 3). These QTL can be suitable targets to improve both traits simultaneously, which is one of the major challenges in quality improvement. Starch content QTL on chromosomes 9 and 10 were previously reported (Chen et al., 2019; Kimani et al., 2020; Patil et al., 2019) except for one QTL on chromosome 10, which was detected under fully irrigated conditions.

Candidate gene analysis in a 200-kb region associated with the amylose content QTL on chromosome 10 identified 20 genes including the waxy (Wx) gene (Sobic.010G226000), which is known for regulating amylose production in cereals (McIntyre et al., 2008). In a similar study, Boyles et al. (2017) reported a significant amylose content QTL on chromosome 10 using RIL and GWAS populations. Among the genes identified in the overlapping regions between protein and amylose content QTL, cell cycle (CYCD) regulatory protein and exocytic trafficking (SCAMP) were the main genes that are ubiquitously expressed in cell recycling (Castle & Castle, 2005). Similarly, QTL on chromosome 9 harbored seven peroxidase genes (Sobic.009G144900, Sobic.009G144600, Sobic.009G145500, Sobic.009G144700, Sobic.009G145600, Sobic.009G144800, and Sobic.009G145700), which are involved in antioxidant (GO:0016209), peroxidase (GO:0004601), and oxidoreductase (GO:0016684) activities (Supplemental Table S1). Maize (Zea mays L.) with high amylose levels in a mutant population was found to have higher antioxidant levels than nonmutant maize (Li et al., 2007).

Breeding for improved grain quality in grain sorghum

Availability of heritable genetic variation is a cornerstone for the success of any breeding program. In this study, the RILs showed high levels of genetic variation in grain quality traits that can be exploited to develop cultivars with desired level of grain quality for various end uses. Similarly, previous studies have also reported high levels of genetic variability for these quality traits in large and genetically diverse sorghum populations (Boyles et al., 2017; Murray et al., 2008; Sukumaran et al., 2012), which provides the opportunity to improve grain quality through selection. Large proportion of the phenotypic variation in this study was contributed by the genotype (additive, dominant, and epistatic genes) as reflected by the high levels of broad-sense heritability, which indicates a potential for grain quality improvement (Table 2). Boyles et al. (2017) found similar level of broad-sense heritability for starch (73%) but lower for protein (65%) and amylose (56%) content, while Murray et al. (2008) reported 80% heritability for protein content in a RIL population. The fact that the parental lines did not significantly differ for all three traits in this study, but showed transgressive segregation, indicated that much of the broad-sense heritability was contributed by dominant and epistatic gene actions, which can be exploited via hybrid breeding (Ayalew et al., 2016; Li et al., 2008). Transgressive segregation for starch and protein content were previously reported in sorghum RIL populations despite the parental lines being statistically similar for these traits (Boyles et al., 2017; Murray et al., 2008). Hybrid breeding will help exploit much of the dominance genetic component, which, otherwise, will be difficult to harness through pure-line breeding using additive gene actions.

CONCLUSION

Findings from this study provide substantial information regarding genetic control of grain quality in sorghum. Both gene and environment or management played significant roles in determining grain quality of sorghum. Much of the genetic component seemed to be controlled by dominant genes, which was reflected in the relatively high broad-sense heritability and heterosis in the RIL population. This genetic variation can be exploited through hybrid breeding harnessing heterosis to improve grain quality improvement through heterosis. The identified QTL hot-spot regions can be further fine mapped to apply marker-assisted selection and gene pyramiding. Future quality improvement research could benefit from studies involving transcription factors, as they were found to simultaneously enhance both starch and protein accumulation in the grain.

ACKNOWLEDGMENTS

We thank the Hays sorghum breeding team for the technical help during experiments. Contribution number 22-186-J is from Kansas Agricultural Experiment Station. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the USDA. The USDA is an equal opportunity provider and employer. This work was supported in part by the USDA Agricultural Research Service.

AUTHOR CONTRIBUTIONS

Habtamu Ayalew: Formal analysis; Investigation; Software; Validation; Visualization; Writing-original draft; Writing-review & editing. Shantha Peiris: Data curation; Methodology; Software; Validation; Writing-review & editing. Anuj Chiluwal: Data curation; Methodology; Writing-review & editing. Ritesh Kumar: Investigation; Visualization; Writing-original draft; Writing-review & editing. Manish Tiwari: Investigation; Methodology; Software; Writing-original draft; Writing-review & editing. Troy Ostmeyer: Data curation; Methodology; Writing-review & editing. Scott Bean: Conceptualization; Methodology; Resources; Software; Supervision; Writing-review & editing. S. V. Krishna Jagadish: Conceptualization; Funding acquisition; Investigation; Methodology; Project administration; Resources; Supervision; Writing-original draft; Writing-review & editing.

CONFLICT OF INTEREST

The authors declare no conflict of interest.