Genetic mapping of sorghum resistance to an Illinois isolate of Colletotrichum sublineola

Abstract

Anthracnose leaf blight (ALB) is an economically important disease of sorghum [Sorghum bicolor (L.) Moench] caused by the fungal pathogen Colletotrichum sublineola Henn. ex Sacc. & Trotter. Although qualitative and quantitative resistance have been identified for ALB, the usefulness of resistance loci differs depending on the pathogen pathotype. Identifying resistance effective against unique pathogen pathotypes is critical to managing ALB, as the disease is managed primarily through the deployment of host resistance. We isolated C. sublineola from ALB‐infected leaves collected in Illinois and found that the strain was a novel pathotype, as it produced a unique combination of virulence against a set of differential lines. Using this isolate, we inoculated 579 temperate‐adapted sorghum conversion lines in 2019 and 2020. We then conducted a genome‐wide association study (GWAS) and a metabolic pathway analysis using the Pathway Associated Study Tool (PAST). We identified 47 significant markers distributed across all chromosomes except chromosome 8. We identified 32 candidate genes based on physical proximity with significant markers, some of which have a known role in host defense. We identified 47 pathways associated with ALB resistance, indicating a role for secondary metabolism in defense to ALB. Our results are important to improve the understanding of the genetic basis of ALB resistance in sorghum and highlight the importance of developing durable resistance to ALB.

Core Ideas

• We isolated and characterized a novel pathotype of C. sublineola from Illinois.

• We identified 47 significant markers for anthracnose resistance using GWAS mapping.

• There were significant loci distributed across all chromosomes except chromosome 8.

• We identified 32 candidate genes based on physical proximity with significant GWAS markers.

• We identified 47 metabolic pathways associated with resistance.

Abbreviations

ALB

anthracnose leaf blight

AUDPC

area under the disease progress curve

BLUP

best linear unbiased predictor

FarmCPU

fixed and random model circulating probability unification

FDR

false discovery rate

GWAS

genome‐wide association study

MLM

mixed linear model

PAST

Pathway Associated Study Tool

QTL

quantitative trait loci

SCP

sorghum conversion panel

SNP

single‐nucleotide polymorphism

1. INTRODUCTION

Sorghum [Sorghum bicolor (L.) Moench] is one of the top five cereal crops produced in the world; however, production is limited by biotic stresses. Anthracnose, caused by Colletotrichum sublineola Henn. ex Sacc. & Trotter, is one of the most economically damaging diseases of sorghum worldwide, especially in humid and warm areas (Chala et al., 2010; Frederiksen & Odvody, 2000). Grain yield losses resulting from anthracnose can be over 50% for susceptible lines (Cota et al., 2017; Harris et al., 1964). Anthracnose management in sorghum is based almost exclusively on host resistance (Stutts & Vermerris, 2020; Tesso et al., 2012). The pathogen population is heterogeneous, and multiple pathotypes of C. sublineola have been reported based on their differential ability to cause disease on a common set of sorghum lines (Ali & Warren, 1987; Cardwell et al., 1989; Costa et al., 2015; Moore et al., 2008; Prom et al., 2012; Xavier et al., 2018). To improve sorghum for anthracnose resistance, it is important to dissect the genetic architecture of resistance using multiple pathotypes.

Anthracnose has three stages: leaf blight, stalk rot, and panicle and grain rot (Tesso et al., 2012). The leaf blight stage, also known as anthracnose leaf blight (ALB), is the most common form and most destructive phase of the disease, as it decreases seed weights and is responsible for most of the grain yield losses from anthracnose (Ali, 1987). The symptoms of ALB include chlorotic flecks on leaves, necrotic spots on leaf lamina, leaf sheath lesions, and midrib infection. Foliar symptoms often vary because of differential host reactions, the physiological condition of the host, and specific isolates (Ferreira & Warren, 1982).

The anthracnose resistance response of sorghum lines is variable depending on the C. sublineola population (Ali & Warren, 1987). The pathogen population origin or genotype does not always indicate the pathotype (Prom et al., 2012). Different resistance loci have been reported for different pathotypes of C. sublineola even when evaluating the same sorghum population (Cruet‐Burgos et al., 2020). In order to find durable resistance to C. sublineola, sorghum germplasm must be evaluated using multiple pathotypes.

Both qualitative and quantitative resistance have been reported for ALB, and there are some indications of mechanisms that confer resistance (Abreha et al., 2021; Stutts & Vermerris, 2020). Quantitative trait loci (QTL) have been reported on all 10 chromosomes, indicating that resistance is highly polygenic (Abreha et al., 2021; Stutts & Vermerris, 2020). In addition to the mapped QTL, major‐effect genes have also been reported. ANTHRACNOSE RESISTANCE GENE1 (ARG1) is a nucleotide‐binding leucine‐rich repeat receptor gene effective against anthracnose (Lee et al., 2022). A major dominant resistance gene, Cg1, was identified in SC748 and mapped to the distal end of chromosome 5 (Mehta et al., 2005; Perumal et al., 2009). Furthermore, 3′‐deoxyanthocyanidins, phytoalexins produced by sorghum, are involved in anthracnose resistance (Ibraheem et al., 2010; Lo et al., 1999; Nicholson et al., 1987).

Genome‐wide association mapping studies (GWAS) are popular to identify genotype–phenotype associations for complex traits in humans, plants, and animals. Various models, including the mixed linear model (MLM) (Yu et al., 2006), multilocus mixed model (Segura et al., 2012), fixed and random model circulating probability unification (FarmCPU) (Liu et al., 2016), and K_chr model (Chen & Lipka, 2016; Rincent et al., 2014), have been developed to identify variants associated with phenotypes. Different approaches are needed for different traits because of underlying biological complexities. Several challenges are associated with GWAS including multiple testing correction for the large number of markers tested, low effect sizes of some causal variants, allelic heterogeneity, and population structure (Tibbs Cortes et al., 2021). False negatives, as well as false positives, might mislead the interpretation of the biological meaning of the results. Thus, it is recommended to examine the dataset using different methods.

Core Ideas

• We isolated and characterized a novel pathotype of C. sublineola from Illinois.

• We identified 47 significant markers for anthracnose resistance using GWAS mapping.

• There were significant loci distributed across all chromosomes except chromosome 8.

• We identified 32 candidate genes based on physical proximity with significant GWAS markers.

• We identified 47 metabolic pathways associated with resistance.

Pathway analysis can increase the power to capture true associations between variants and the trait of interest. Pathway analysis uses sets of genes grouped according to their shared function in a specific pathway as opposed to testing each variant individually (Wang et al., 2007). More recently, this approach has been applied to association mapping datasets to search for signal from metabolic pathways as opposed to signal from individual variants (Thrash et al., 2020). In maize (Zea mays L.), this approach has been used to identify pathways associated with biotic stress resistance (Tang et al., 2015; Warburton et al., 2018) as well as oil biosynthesis (Li et al., 2019). Pathway analysis combined with association mapping provides better insights into the biological significance of the results vs. GWAS alone.

The sorghum conversion program sought to convert diverse sorghum landraces to early, combine‐height lines adapted to temperate areas including the United States (Stephens et al., 1967). The sorghum conversion lines were developed by introgressing genetic loci for dwarfism and photoperiod insensitivity from a dwarf temperate elite donor, BTx406, into hundreds of diverse exotic sorghum landraces (Stephens et al., 1967). The sorghum conversion lines are closely related to their exotic progenitor lines except for plant height and photoperiod‐sensitivity loci that were introgressed from the elite donor (Thurber et al., 2013). The sorghum conversion panel (SCP) is composed of sorghum conversion lines and consists of lines belonging to five morphologically and genetically defined races or subpopulations of sorghum (Thurber et al., 2013). The allelic richness but consistent flowering times and plant heights of the sorghum conversion lines means that the SCP serves as a useful resource for mapping disease resistance (Zhang et al., 2020). A subset of the SCP has been included in mapping studies for ALB resistance using isolates from Puerto Rico and Texas (Cuevas et al., 2018; Prom et al., 2019). There are, however, no studies on anthracnose resistance in response to isolates from Illinois or pathogen variability in Illinois.

Our overall goal was to map loci for anthracnose resistance in sorghum using a pathogen strain collected in Illinois. Thus, we evaluated SC lines in response to an Illinois isolate of C. sublineola with a novel pathotype to identify genetic loci conferring ALB resistance. Our objectives were to (a) characterize the Illinois isolate of C. sublineola, (b) identify genetic loci and candidate genes associated with ALB resistance in the sorghum conversion panel in response to an Illinois isolate, and (c) identify metabolic pathways involved with ALB resistance by conducting a pathway analysis. This work contributes to the understanding of ALB resistance in sorghum.

2. MATERIALS AND METHODS

2.1. Plant materials and experimental design

We evaluated the SCP (Thurber et al., 2013) for resistance to ALB in inoculated field trials in 2019 and 2020. A total of 579 and 552 sorghum lines were planted in the 2019 and 2020 field seasons, respectively, in Urbana, IL. The lines were planted in two replications per year with an incomplete block design. There were 25 and 24 blocks for each replication in 2019 and 2020, respectively. Each block was augmented with a susceptible (BTx623) and a resistant (SC748) check line. Field experiments were designed using the package ‘agricolae’ (de Mendiburu & de Mendiburu, 2019) in R (R Core Team, 2021). The seeds were treated with mefenoxam (Apron; Syngenta) and fluxofenim (Concep; Syngenta) before planting. Each line was planted as a single‐row plot and standard agronomic practices for central Illinois were used. The plots were machine planted with 46 seeds per plot in 3.65‐m‐long rows with 0.91‐m alleys. The 2019 field experiment was planted on 4 June, while the 2020 field experiment was planted on 2 July because of flooding in early June.

2.2. Characterization of the Illinois Colletotrichum sublineola isolate

We isolated a single spore of C. sublineola (19CS002.1) from diseased sorghum leaves collected in 2016 in Urbana, IL. The morphology and sequence of the internal transcribed spacer region confirmed that the isolate was C. sublineola. We tested the virulence of the isolate using eight sorghum differential lines including SC326‐6, SC414‐12E, BTX378, TAM428, TX2536, SC328C, QL3, and BTX398 (Cardwell et al., 1989; Moore et al., 2008) in the Plant Care Facility, University of Illinois at Urbana‐Champaign, IL. Lines were arranged in a completely randomized design with three replications. The seed for the differential lines was obtained from Dr. William Rooney at Texas A & M University. Seeds were planted in a seedling tray with soilless media composed of coarse‐grade peat moss, perlite, dolomitic and calcitic limestone, and a nonionic wetting agent (BM6, Berger). Seedlings were transplanted after 11 d into one‐gallon pots filled with general‐purpose potting mix. A 12‐h light duration with 26 vs. 21 °C, day vs. night temperature was maintained in the greenhouse.

The differential lines were inoculated with the 19CS002.1 isolate 30 d after planting using two methods including (a) spraying a spore suspension (1 × 10⁶ spores ml⁻¹) until runoff and (b) inoculating 0.5 ml of spore suspension (1 × 10⁶ spores ml⁻¹) in the whorl. We modified the protocol described by Prom et al. (2009) for inoculum preparation. Briefly, a spore suspension was prepared by culturing the isolate on potato dextrose agar media with a 12‐ vs.12‐h light vs. dark cycle for 12–14 d, flooding the potato dextrose agar plate with 15–20 ml of distilled water, dislodging the spores with a glass rod, and filtering through sterile double‐layered cheesecloth. Plants were misted for 24 h after inoculation.

Disease symptoms were rated 2 wk after inoculation and subsequently in weekly intervals for 3 wk. For the plants that were spray inoculated, a 1‐to‐5 scale was used for ratings (Prom et al., 2012). To clearly demonstrate whether each line used for the pathotyping had a susceptible or resistant reaction and report the results in a table format, the rating scale of 1‐to‐5 was converted to resistant and susceptible by considering ratings of 1 and 2 as resistant and ratings of 3–5 as susceptible. A binary scale (susceptible or resistant) depending on the presence or absence of necrosis with acervuli was used for whorl inoculated plants. The avirulence or virulence of the isolate 19CS002.1 was determined based on the resistance and susceptibility of the differentials to the isolate. The spray inoculations were repeated two times to confirm the result.

2.3. Inoculum preparation and field inoculation

The single‐spore Illinois isolate (19CS002.1) was used to prepare solid inoculum to inoculate lines in the field. The solid sorghum substrate was prepared by soaking sorghum grains in deionized water in plastic pans for 24 h, autoclaving 1.5 kg of the soaked sorghum grains in mushroom growing bags (0.2 Micron Filter, 3 Mil Polypropylene, Large Size “8 by 5 by 20 inches,” Root Mushroom Farm, USA) at 121 °C twice each for 60 min. A spore suspension was prepared in the same way as for the greenhouse trial with a final concentration of 1 × 10⁶ spores ml⁻¹. The autoclaved mushroom growing bags were inoculated with 45 ml of the spore suspension. Each inoculated bag was incubated at room temperature on shelves with fluorescent lighting with a 12‐h light v. dark cycle for 10 d. Bags were shuffled every other day to promote uniform colonization. The solid inoculum was stored in a cold room at 4 °C until required for field inoculation. Sorghum plants in the field were inoculated by placing 2 g of the solid inoculum in the whorl ∼30 d after planting or when the plants were at the V5–V7 growth stage.

2.4. Phenotyping and data analysis

The field plots were visually rated for ALB based on disease incidence and severity (diseased leaf area) 3 wk after inoculation on a biweekly basis for a total of three ratings. Plots were rated using the 0‐to‐9 scale reported by Burrell et al. (2015) with minor modifications as indicated in Supplemental Table S1. The area under the disease progress curve (AUDPC) values were calculated using the disease rating data and the ‘audpc’ function with default settings in the package ‘agricolae’ (de Mendiburu & de Mendiburu, 2019) in R (v4.0.5) (R Core Team, 2021). The following MLM was fit for the dataset for each year separately using the lme4 package (Bates et al., 2015) in R (v4.0.5) (R Core Team, 2021):

$$Y_{ijk}=\mathrm{\mu }+L_i+R_j+{B_k}_{(j)}+{\mathrm{\epsilon }}_{ijk}$$

Where 𝑌_ijk is AUDPC for the 𝑖th line of 𝑗th replication and 𝑘th block; μ is the grand population mean; L_i is the random effect of the 𝑖th line, R_j is the random effect of the 𝑗th replication; B_{k ( j )} is the random effect of the 𝑘th block nested in jth replication, and ε _ijk is the random error term.

The combined 2019 and 2020 data was also fitted and included the same factors as the individual years but with the inclusion of a random line × year interaction term. Best linear unbiased predictors (BLUPs) were predicted for each year individually and for both years combined (Supplemental File S1).

2.5. Genome‐wide association study

The genotypic dataset for the SCP was obtained from Zhang et al. (2020). Single‐nucleotide polymorphisms (SNPs) were filtered for a minor allele frequency of 0.05 and heterozygosity of 10% using TASSEL (Bradbury et al., 2007). The resulting 312,222 SNPs were included in further analysis. Three different methods were used for genome‐wide association mapping: (a) MLM (Yu et al., 2006), (b) FarmCPU (Liu et al., 2016), and (c) the K_chr method (Chen & Lipka, 2016; Rincent et al., 2014). All models were run using Genome Association and Prediction Integrated Tool (GAPIT) (Lipka et al., 2012) in R (v4.0.5) (R Core Team, 2021). Based on the scree plot, no principal components were included in the analysis. The kinship matrices for each method were calculated by GAPIT. For the K_chr method, GAPIT calculated kinship individually for each chromosome to increase the statistical power to detect any significant associations (Chen & Lipka, 2016; Rincent et al., 2014). The Benjamini and Hochberg (1995) procedure was used to obtain an adjusted P value that controlled the genome‐wide false discovery rate (FDR) at 10%. We used this threshold so that more associations that might be potentially interesting would be included. The number of false positives is higher with this threshold, but likely there are some true positives. Linkage disequilibrium between pairs of significant SNPs on the same chromosome were estimated in TASSEL v5.0 using squared allele‐frequency correlations (r ²) (Bradbury et al., 2007). To conduct logistic regression, we converted the quantitative data to binary data. Lines with a BLUP value <0 were considered resistant and >0 were susceptible. We then conducted logistic regression using EMMAX with kinship included (Kang et al., 2010).

We identified candidate genes based on the physical proximity of significant associations to genomic features in the reference sorghum genome v3.1.1 available at the JGI Phytozome 12.5.1 website (McCormick et al., 2018). Genes that encompassed the significant SNP, or the closest genes to the significant SNP, were regarded as candidate genes.

2.6. Pathways associated with the ALB resistance

We investigated metabolic pathways associated with ALB resistance using the package Pathway Associated Study Tool (PAST) (Thrash et al., 2020) in R (R Core Team, 2021). Phenotypic and genotypic data used for the analysis were the same as that used for GWAS. We performed association mapping using the BLUPs from each year with a MLM in TASSEL v5.0 (Bradbury et al., 2007) and generated the association and effect files used in the PAST analysis. The association analysis was conducted in TASSEL because PAST is designed to accept TASSEL output files. Prior to conducting the association analysis, all heterozygous markers were removed, as the PAST package does not include loci with heterozygous markers. No principal components were included in the MLM. Linkage disequilibrium analysis was conducted in TASSEL v5.0 (Bradbury et al., 2007) to find the correlation between each pair of SNPs on the chromosome and associate the SNP with genes. The annotated gene file in GFF format was obtained from JGI Phytozome 12.5.1 website (McCormick et al., 2018). The metabolic pathway file for sorghum was downloaded from the sorghum functional genomics database (Tian et al., 2016). A Type I error rate of α = .05 was used to declare statistical significance at a pathway‐wise level.

3. RESULTS

3.1. Characterization of an Illinois isolate of Colletotrichum sublineola

We tested the virulence of the Illinois isolate 19CS002.1 on eight common sorghum differentials to characterize its pathotype. Previously, these eight sorghum differentials were used to identify eight pathotypes from 12 isolates collected from Texas, Georgia, and Puerto Rico (Cardwell et al., 1989) and 13 pathotypes from 87 isolates collected from Arkansas (Moore et al., 2008). The isolate 19CS002.1 showed different virulence patterns on these eight differentials than the previously reported pathotypes, indicating that the Illinois isolate was a novel pathotype. The isolate 19CS002.1 was virulent on all differentials except SC413‐12E and R.TAM428 (Table 1).

TABLE 1.

Virulence pattern of an Illinois isolate (19CS002.1) on eight sorghum host differentials

Experiment	Inoculation method	SC326‐6	SC414‐12E	B.Tx378	R.TAM428	R.Tx2536	SC328	QL3 (India)	B.Tx398
I	Spray	S	R	S	R	S	S	S	S
II	Spray	S	R	S	R	S	S	S	S
III	Whorl	S	R	S	R	S	S	S	S

Note. S, susceptible; R, resistant.

3.2. Germplasm characterization

We observed a wide range of phenotypic variation in the SCP (Figure 1) with rating values ranging from 1 to 9. The check lines BTx623 and SC748 performed as expected with higher disease severity in BTx623 in both years (Figure 2). This indicates that there was sufficient disease pressure to conduct genetic mapping. The AUDPC distribution in 2019 was bimodal, and the AUDPC distribution in 2020 was right skewed. There was less disease in 2020 than in 2019 (Figure 2). This likely is due to the late planting and decreased precipitation in 2020 vs. 2019. Rainfall is known to have a strong influence on the incidence and severity of anthracnose (Erpelding & Prom, 2004; Prom et al., 2015). The 2019 field season had more rain than the 2020 field season, which likely influenced the amount of disease present as well. Furthermore, the 2020 experiment was planted ∼1 mo later than in 2019 and, thus, temperatures were warmer when the plants were inoculated. The data were not normal even with a box‐cox transformation, therefore untransformed values were used for the analysis.

FIGURE 1.

Sorghum plants showing symptoms when infected by Colletotrichum sublineola. (a) Necrosis on leaf lamina and veins. (b) Necrotic symptoms in most of the leaves of a plant. (c) Necrosis on leaf lamina, vein, and midrib. (d) Necrosis on all the plants of a plot. The pictures of sorghum conversion panel lines with symptoms were taken in the field 3 wk after inoculation

FIGURE 2.

Distribution of the area under disease progress curve (AUDPC) and best linear unbiased predictors (BLUPs) data for each year. (a) Distribution of AUDPC in 2019. (b) Distribution of BLUPs in 2019. (c) Distribution of AUDPC in 2020. (d) Distribution of BLUPs in 2020. (e) Distribution of AUDPC in combined dataset. (f) Distribution of BLUPs in combined dataset

Line, which was highly significant (<.001) for both individual years and the combined dataset, explained most of the phenotypic variation. Blocks were significant (<.05) for the individual year and combined datasets (Table 2). Replications were significant for the year 2020 dataset and the combined dataset (Table 2). On average, kafir was the most susceptible race and guinea the most resistant (Supplemental Figure S3).

TABLE 2.

Significance (P values) of factors in a linear model for resistance to Colletotrichum sublineola in the sorghum conversion panel

Source of variation	2019	2020	2019 and 2020 combined
Lines	<.001	<.001	<.001
Replication	.06	<.001	<.001
Replication/block	.001	.02	<.001
Lines × year	–	–	<.001
Year	–	–	<.001

There was a significant genotype × environment interaction for ALB resistance, which likely is due to differing environmental conditions and planting dates between the 2 yr. Because the 2019 and 2020 environments were different, we conducted a GWAS for the individual years as well as the combined dataset. We identified environment‐specific associations as well as stable associations. Significant genotype × environment interactions and environment‐specific QTL for ALB resistance in sorghum have been reported previously in a biparental mapping population (BTx623 × SC414‐12E) (Patil et al., 2017). Together, this suggests that ALB resistance is dependent not only on the C. sublineola pathotype but also on the environment.

3.3. Detection of significant loci associated with the ALB resistance and implicated candidate genes

We conducted GWAS using three different models: MLM, K‐chr, and FarmCPU. Log quantile–quantile plots indicated that population structure and relatedness were well accounted for in all the models (Supplemental Figure S1). We detected significant associations (FDR adjusted p values < .1) with the K_chr and FarmCPU models on chromosomes 1, 2, 3, 4, 5, 6, 7, 9, and 10 (Figures 3 and 4; Table 3). There were 14 significant associations using the combined dataset, of which, only one was genic. Of particular interest was a significant SNP on chromosome 2 that was near a gene (Sobic.002G205300) encoding a shikimate O‐hydroxycinnamoyltransferase. This transferase is involved in the phenylpropanoid biosynthesis pathway. In the individual year analyses a total of 12 (five genic) and 21 (10 genic) significant SNPs were detected in 2019 and 2020, respectively, for ALB resistance (Table 3). Nine significant SNPs in high linkage disequilibrium were detected in a 6.4‐kb region ranging from 44,378,042 to 44,384,499 bp on chromosome 6 with the K_chr model and the 2020 data (Figure 5). Among them, six SNPs were in the genic regions of Sobic.006G077900 and three SNP were within 1 kb of Sobic.006G077900 (Table 3). The SNP Chr06_44382216 on chromosome six was consistently detected with multiple models in 2020. Sobic.006G077900 is annotated as a cysteine synthase.

FIGURE 3.

Genome‐wide association study results of resistance to anthracnose leaf blight using different models on the sorghum conversion panel in 2019 and 2020. (a) Manhattan plot for 2019 data with mixed linear model (MLM) (b) Manhattan plot for 2020 data with MLM. (c) Manhattan plot for 2019 data with K_chr model. (d) Manhattan plot for 2020 data with K_chr model. (e) Manhattan plot for 2019 data with fixed and random model circulating probability unification (FarmCPU) model. (f) Manhattan plot for 2020 data with FarmCPU model

FIGURE 4.

Genome‐wide association study results for resistance to anthracnose leaf blight using different models on the sorghum conversion panel in the combined dataset. (a) Manhattan plot for combined data with mixed linear model (MLM). (b) Manhattan plot for combined data with fixed and random model circulating probability unification (FarmCPU) model. (c) Manhattan plot for combined data with K_chr model

TABLE 3.

Significant single‐nucleotide polymorphisms (SNPs) (false‐discovery‐rate [FDR]‐adjusted P‐values < .1) associated with the anthracnose leaf blight in sorghum conversion panel

SNP	Method	Year	FDR a	Location	Closest gene b	Gene description
Chr01_2524716	FarmCPU	2019	0.00025	Genic	Sobic.001G033300	Similar to expansin‐A7 precursor
Chr01_14456242	FarmCPU	2019	0.00079	Intergenic	Sobic.001G172400	Similar to cytochrome P450 85A1
Chr01_19587975	FarmCPU	2020	9.33 × 10−5	Intergenic	Sobic.001G212451	Weakly similar to cotton fiber expressed protein‐like protein
Chr01_74280158	FarmCPU	Combined	1.4 × 10−6	Intergenic	Sobic.001G469966	Na
Chr01_75664616	FarmCPU	2020	2.58 × 10−5	Intergenic	Sobic.001G485901	Na
Chr01_78260987	FarmCPU	2019	0.04	Intergenic	Sobic.001G515800	Similar to No apical meristem protein, expressed
Chr01_78260987	FarmCPU	Combined	6.86 × 10−8	Intergenic	Sobic.001G515800	Similar to No apical meristem protein, expressed
Chr01_78641366	FarmCPU	2019	0.022	Genic	Sobic.001G521100	Similar to DNA polymerase delta small subunit
Chr02_3403011	FarmCPU	2019	0.0009	Genic	Sobic.002G036150	Tify 5a related
Chr02_59661346	FarmCPU	Combined	1.28 × 10−6	Intergenic	Sobic.002G205300	Shikimate O‐hydroxycinnamoyltransferase
Chr02_63548861	FarmCPU	2020	9.33 × 10−5	Genic	Sobic.002G247500	Homolog of Ethylene‐insensitive protein 3
Chr02_63553895	FarmCPU	Combined	5.75 × 10−9	Intergenic	Sobic.002G247500	Homolog of Ethylene‐insensitive protein 3
Chr03_7878769	FarmCPU	Combined	1.06 × 10−6	Intergenic	Sobic.003G090400	Ubiquitin‐protein ligase SIN‐like
Chr03_65910297	FarmCPU	2019	0.04	Genic	Sobic.003G335800	Na
Chr03_70351964	FarmCPU	Combined	5 × 10−10	Intergenic	Sobic.003G392550	Na
Chr03_70351964	Kchr	Combined	2.46 × 10−6	Intergenic	Sobic.003G392550	Na
Chr03_70365478	Kchr	2019	0.072	Intergenic	Sobic.003G392550	Na
Chr03_70365478	Kchr	Combined	2.29 × 10−6	Intergenic	Sobic.003G392550	Na
Chr03_70366405	Kchr	Combined	8.73 × 10−6	Intergenic	Sobic.003G392550	Na
Chr03_70419242	Kchr	Combined	3.23 × 10−6	Genic	Sobic.003G393100	Structural maintenance of chromosomes protein
Chr03_70455886	FarmCPU	2019	6.57 × 10−10	Intergenic	Sobic.003G393400	Homolog of Exocyst complex protein EXO70
Chr04_9955735	FarmCPU	Combined	4.88 × 10−7	Intergenic	Sobic.004G105500	Predicted protein
Chr04_12099460	FarmCPU	2020	0.0068	Intergenic	Sobic.004G116050	Na
Chr05_1116575	FarmCPU	Combined	1.38 × 10−6	Intergenic	Sobic.005G012100	Zing finger
Chr05_1721684	FarmCPU	2020	0.00078	Intergenic	Sobic.005G019000	Domain of unknown function
Chr05_3541269	FarmCPU	2020	0.0043	Genic	Sobic.005G038700	Similar to BZIP transcription factor
Chr05_69044814	FarmCPU	2019	0.04	Intergenic	Sobic.005G063200	Na
Chr05_70180152	FarmCPU	2020	0.06	Intergenic	Sobic.005G216000	Na
Chr06_44378042	Kchr	2020	0.04	Intergenic	Sobic.006G077900	Cysteine synthase
Chr06_44378875	Kchr	2020	0.08	Genic	Sobic.006G077900	Cysteine synthase
Chr06_44380096	Kchr	2020	0.08	Genic	Sobic.006G077900	Cysteine synthase
Chr06_44380463	Kchr	2020	0.08	Genic	Sobic.006G077900	Cysteine synthase
Chr06_44381679	Kchr	2020	0.03	Genic	Sobic.006G077900	Cysteine synthase
Chr06_44381902	Kchr	2020	0.04	Genic	Sobic.006G077900	Cysteine synthase
Chr06_44382216	FarmCPU	2020	2.50 × 10−5	Genic	Sobic.006G077900	Cysteine synthase
Chr06_44382216	Kchr	2020	0.01	Genic	Sobic.006G077900	Cysteine synthase
Chr06_44384384	Kchr	2020	0.08	Intergenic	Sobic.006G077900	Cysteine synthase
Chr06_44384499	Kchr	2020	0.05	Intergenic	Sobic.006G077900	Cysteine synthase
Chr07_2120687	FarmCPU	2019	0.04	Intergenic	Sobic.007G023600	Similar to AAA‐type ATPase family protein
Chr07_62300664	FarmCPU	2019	0.01	Intergenic	Sobic.007G059600	Similar to Benzothiadiazole‐induced somatic embryogenesis receptor kinase 1
Chr09_3234591	FarmCPU	Combined	4.41 × 10−7	Intergenic	Sobic.009G035300	Putative uncharacterized protein
Chr09_6001550	FarmCPU	2020	0.025	Intergenic	Sobic.009G057900	Galactose oxidase/Kelch repeat type 1
Chr10_2458553	FarmCPU	2020	0.001	Intergenic	Sobic.010G030100	Weakly similar to Homeodomain‐leucine zipper transcription factor tahdzipii‐1
Chr10_2761316	FarmCPU	2020	0.062	Intergenic	Sobic.010G034000	Carboxypeptidase
Chr10_54116425	FarmCPU	2019	0.014	Genic	Sobic.010G198000	Tryptophan synthetase
Chr10_55609499	FarmCPU	2020	6.60 × 10−5	Genic	Sobic.010G213100	RNA‐dependent RNA polymerase
Chr10_5786947	FarmCPU	Combined	5.67 × 10−7	Intergenic	Sobic.010G071700	Similar to EREBP‐like protein

Note. FarmCPU, fixed and random model circulating probability unification.

^a False discovery rate adjusted p values.

^b Candidate genes based on the physical proximity of significant associations to genomic features on the reference sorghum genome, v3.1.1.

FIGURE 5.

Plot of association results from the K_Chr analysis of the 2020 data and linkage disequilibrium estimates (r² ) in the 44‐Mb region of chromosome 6. The −log₁₀ of the false discovery rate adjusted P values (left side of y axis; bars) based on the 2020 phenotypic data and the r² values (right side of y axis; triangles) are plotted against physical position for a 44‐Mb region on chromosome 6. Nine significant markers (P < .1) on chromosome 6 were detected using the K_chr method and the neighboring three markers in both directions are included

In total, we identified 32 candidate genes based on the proximity of significant SNPs. Most of the associations with FDR‐adjusted P values between .05 and .10 involve genes that are also implicated by associations that have an FDR‐adjusted P value <.05. A total of 10 genes had significant SNPs within the gene (Table 3). Several of these candidate genes are involved in plant defense.

In addition to using the GWAS using the quantitative data, we converted the data to a binary trait and ran a logistic regression model using EMMAX. Nothing was significant with an FDR or Bonferroni correction (data not shown).

3.4. Identification of pathways associated with the ALB resistance

We investigated the metabolic pathways associated with ALB resistance in sorghum using the GWAS results and the R package PAST (Thrash et al., 2020). We identified 10 pathways significant at α < .01 (Table 4) and a total of 47 pathways significant at α < .05 (Supplemental File S2). Of particular interest is the aromatic polyketides biosynthesis pathway, which is involved in secondary metabolite production (Flores‐Sanchez & Verpoorte, 2009). The 1,4‐dihydroxy‐2‐naphthoate biosynthesis pathway was significant, which is a branch point metabolite leading to the biosynthesis of phylloquinone and pigments. The citrulline‐nitric oxide cycle was significant. Nitric oxide is a gaseous reactive oxygen species and is involved in biotic stress responses (Domingos et al., 2015). Several of the significant pathways are involved in photosynthesis, including the plastoquinol‐9 biosynthesis pathway and the phylloquinol superpathway. We did not find overlap between the candidate genes implicated from GWAS and the genes involved in the significant metabolic pathways (Tables 3 and 4). This is because there is cumulative signal across the pathways and thus, individual genes do not meet threshold for detection but the pathway as a whole does. In summary, several of these pathways are of interest and have data to support a role in biotic stress response. Genes in these pathways can be considered candidate genes (Warburton et al., 2018).

TABLE 4.

Significant metabolic pathways identified for sorghum anthracnose leaf blight resistance using the 2019, 2020, and combined datasets. Only pathways significant at p <.01 are included in this table

Pathway ID	Pathway name	p value	Dataset
PWY‐6316	Aromatic polyketides biosynthesis	.00707812	Combined
PWY‐5337	Stachyose biosynthesis	.00772482	Combined
PYRUVDEHY	Pyruvate decarboxylation	.00216618	2019
PWY‐5271	Phaseic acid biosynthesis	.00549644	2019
PWY‐4983	Citrulline‐nitric oxide cycle	.00713992	2019
PWY‐6352	3‐phosphoinositide biosynthesis	.00882664	2019
PWY‐5863	Superpathway of phylloquinol	.0042388	2020
PWY‐5791	1,4‐dihydroxy‐2‐naphthoate biosynthesis	.00511809	2020
PWY‐1581	Plastoquinol‐9 biosynthesis	.00637594	2020
VALDEG‐PW	Valine degradation	.00845254	2020

4. DISCUSSION

Colletotrichum sublineola is a diverse pathogen with numerous reported pathotypes, and thus, identifying and deploying durable resistance is critical. We characterized the Illinois C. sublineola isolate 19CS002.1 and evaluated the SCP in response to this isolate. We identified loci associated with ALB resistance on all chromosomes except chromosome 8. Based on the physical proximity of the genes to the significant SNPs, 32 candidate genes were implicated for ALB resistance. Further work is needed to confirm these associations before they can be applied to breeding programs to improve for ALB resistance. A total of 47 metabolic pathways were significantly associated with resistance to C. sublineola in sorghum. The combined GWAS and PAST analysis show that there is likely a role for secondary metabolites in resistance to anthracnose in sorghum. These pathways provide insight into components of resistance that should be targets of further investigation. Together, our study provides a better understanding of the genetic basis of ALB resistance in sorghum and informs further investigation into the candidate genes and pathways.

There were 47 SNPs significantly associated with ALB resistance. Previous studies have identified anthracnose resistance on chromosome 5 including a major gene that is effective against different pathotypes of C. sublineola (Burrell et al., 2015; Cuevas et al., 2018, 2014; Patil et al., 2017). The significant SNP detected on chromosome 5 (Chr05_69Mb and Chr05_70Mb) and on chromosome 10 (Chr10_54Mb and Chr10_55Mb) in this study were within 5 Mb of that reported by Cuevas et al. (2018) and Prom et al. (2019), respectively. Since both studies evaluated the sorghum association panel, which includes a subset of the SCP, it is highly likely that the locus on chromosome 5 identified in our study is the same as that previously reported (Cuevas et al., 2018; Prom et al., 2019). Similar to Cuevas et al. (2018) guinea was on average the most resistant subpopulation, indicating that lines in this subpopulation are useful for anthracnose breeding, and kafir was the most susceptible. For the evaluation of anthracnose resistance Cuevas et al. (2018) used isolates from Puerto Rico and Prom et al. (2019) used isolates from Texas. This suggests that the anthracnose resistance identified on chromosomes 5 and 10 is effective for multiple pathotypes and diverse environmental conditions.

Several candidate genes of interest were identified. The rice (Oryza sativa L.) ortholog of the candidate gene Sobic.005G063200, a candidate based on the physical proximity of the significant SNP on chromosome 5, is annotated as a Bowman–Birk type bran trypsin inhibitor (BBI) precursor. The BBI genes from wheat (Triticum aestivum L.) and rice are involved in resistance to plant pathogenic fungi (Chilosi et al., 2000; Qu et al., 2003). The Arabidopsis thaliana (L.) Heynh. ortholog of Sobic.010G198000, a candidate gene on chromosome 10, is annotated as a tryptophan synthase. The tryptophan pathway is involved in defense responses in rice (Ishihara et al., 2008) and increased resistance to Verticillium dahliae in cotton (Gossypium spp.) via the salicylic acid pathway (Miao et al., 2019). Furthermore, sorghum lines with enhanced levels of tryptophan were more resistant to C. sublineola (Tugizimana et al., 2019). This suggests that the genes Sobic.005G063200 and Sobic.010G198000 are strong candidate genes for further investigation.

We identified a significant SNP on chromosome 2 (Chr02_63548861) and three SNPs on chromosome 3 (Chr03_65910297, Chr03_70365478, and Chr03_70455886) that were located within 5 Mb of that reported by previous studies in response to Texas isolates (Ahn et al., 2019; Prom et al., 2019). The genes nearest to the significant SNP with highest probability of association on chromosomes 2 and 3 were Sobic.002G247500 (homolog of Arabidposis ETHYLENE‐INSENSITIVE PROTEIN 3) and Sobic.003G393400 (homolog of Arabidoposis exocyst complex protein), respectively. The orthologs have been implicated in defense against bacterial and fungal pathogens in Arabidopsis (Chen et al., 2009; Pečenková et al., 2011; Stegmann et al., 2012). This suggests that these genes might also have a role in ALB resistance in sorghum and might be effective against multiple pathotypes. These are also important regions for breeding for durable resistance, as they have been identified in multiple studies using diverse C. sublineola isolates.

There were a cluster of nine significant SNPs in high linkage disequilibrium on chromosome 6. A gene Sobic.006G077900, annotated as the ortholog of a rice cysteine synthase, encompassed several significant SNPs in the genic region and three SNPs adjacent to the gene were significant. Cysteine synthase is an enzyme required to synthesize cysteine, which is important for sulfate reduction in plant metabolism (Wirtz & Hell, 2006). An antifungal protein rich in cysteine was found to increase host defense in radish against Alternaria longipes (Terras et al., 1995).

Other sorghum candidate genes identified in this study and evidently involved in host defense include Sobic.005G038700 (encoding a bZIP transcription factor) and Sobic.007G023600 (encoding an AAA‐type ATPase family protein). A pepper bZIP transcription factor gene CabZIP2 is induced upon infection with the bacterial pathogen Xanthomonas campestris and overexpression in Arabidopsis enhanced resistance to Pseudomonas syringae (Lim et al., 2015). In rice, LESION MIMIC RESMBLING encodes an AAA‐type ATPase and, aside from its role in lesion mimicry, the gene is also involved in resistance to rice blast caused by the fungus Magnaporthe oryzae and bacterial blight caused by Xanthomonas oryzae pv. oryzae (Fekih et al., 2015). Since several of the candidate genes detected in this study have been implicated in defense against pathogens in other systems, they might also be associated with ALB resistance in sorghum.

There is no single best GWAS method (Tibbs Cortes et al., 2021), and thus we performed GWAS using three different models including a genome scan of individual SNPs and multilocus methods. We obtained different significant associations with different models, with some overlap on chromosome 6 from the FarmCPU and K_chr models (Table 3). However, we did not have any significant associations using the MLM. This might be because the unified MLM model tends to overcorrect for false positives and includes all SNPs to calculate the kinship matrix, so the significant SNPs in high linkage disequilibrium regions would not be detected (Kaler et al., 2020; Rincent et al., 2014). Meanwhile, the K_chr model excludes the SNPs present in the chromosome being tested while calculating the kinship matrix (Chen & Lipka, 2016; Rincent et al., 2014). The FarmCPU model uses the associated markers as covariates in the fixed‐effect model and those markers are separately optimized in a random effect model (Liu et al., 2016). Our results suggest that it is useful to use multiple models while conducting GWAS to minimize missing any significant associations with the trait of interest.

A metabolic pathway analysis was conducted to understand the molecular mechanisms underlying ALB resistance and identify functional pathways responsible for resistance (Thrash et al., 2020). The pathway analysis implicated basic metabolic pathways including photosynthesis and also those that produce the precursors to compounds with known roles in defense. While there was no overlap in the genes implicated by the GWAS and those implicated by the PAST analysis, some of the same functions were implicated such as flavonoids and other secondary metabolites. The hypothesis that these compounds lead to a more resistant plant should be tested and might be useful in a breeding context. More investigation is needed to fully understand how these pathways might play a role in resistance to anthracnose in sorghum and to offer new opportunities for engineered resistance strategies.

The phenylpropanoid pathway synthesizes an array of aromatic compounds with key roles in cell wall biosynthesis, defense against herbivory and pathogenesis, and stress response (Boerjan et al., 2003; Dixon & Paiva, 1995). The metabolites of the phenylpropanoid and flavonoid pathways are phenolic compounds with an important role in plant–fungal interactions and plant defense (Treutter, 2006; Weisshaar & Jenkins, 1998). In sorghum, genes involved in flavonoid and phenylpropanoid biosynthetic pathways were shown to be differentially regulated in response to C. sublineola vs. mock inoculated plants (Wang et al., 2019). Furthermore, metabolomic analysis in response to C. sublineola was used to identify phenylpropanoid and flavonoid pathways as the central hub of the network consisting of metabolites from several interconnected metabolic pathways (Tugizimana et al., 2019). Phenolic compounds were shown to be produced faster and in greater amounts in disease‐resistant sorghum cultivars (Lo et al., 1999). The aromatic polyketides biosynthesis pathway found to be significant in the PAST analysis includes enzymes such as chalcone synthase, benzalactone synthase, and stilbene synthase, which are involved in the synthesis of flavonoids and other defense‐related compounds (Borejsza‐Wysocki & Hrazdina, 1996). The association of the shikimate O‐hydroxycinnamoyltransferase with sorghum anthracnose resistance in this study indicates that there is functional genetic variation in these pathways, including the phenylpropanoid pathway, that have the potential to improve disease resistance. Conferring quantitative resistance through biochemical compounds might lead to a more durable resistance response (Cowger & Brown, 2019; Yadav et al., 2020).

5. CONCLUSIONS

To conclude, we identified 47 significant markers, which implicate 32 candidate genes, as well as 47 metabolic pathways associated with ALB resistance in the SCP. The significant loci on chromosomes 2, 3, 5, and 10 overlapped with previously reported QTL identified in response to different pathotypes. Therefore, these loci may be important to achieve durable ALB resistance and would be useful loci to incorporate into breeding programs. The genes Sobic.005G063200, Sobic.002G247500, Sobic.005G038700, Sobic.007G023600, and Sobic.006G077900, selected based on their physical proximity to significant SNPs, have been implicated in host defense in several crops including sorghum, thus making them important candidates for ALB resistance and, potentially, for use in marker‐assisted selection. We also identified metabolic pathways that might have an important role in ALB resistance in sorghum. The genes in these pathways can be considered candidate genes and provide direction for further research into biotic stress resistance in sorghum. Overall, our findings improve our understanding of host resistance to C. sublineola and are useful for improving ALB management in sorghum.

AUTHOR CONTRIBUTIONS

Ashmita Khanal: Formal analysis; Investigation; Methodology; Software; Visualization; Writing – original draft; Writing – review & editing. Pragya Adhikari: Investigation; Writing – review & editing. Christopher Kaiser: Investigation; Methodology; Writing – review & editing. Alexander E. Lipka: Formal analysis; Software; Writing – review & editing. Tiffany M. Jamann: Conceptualization; Funding acquisition; Methodology; Project administration; Resources; Supervision; Writing – review & editing. Santiago X. Mideros: Conceptualization; Funding acquisition; Methodology; Project administration; Resources; Supervision; Writing – original draft; Writing – review & editing.

CONFLICT OF INTEREST

The authors declare no conflicts of interest.

Supporting information

Supplemental Figure S1. QQ plots of the GWAS results using three models for each year. A) QQ plot for GWAS with MLM for 2019 B) QQ plot for GWAS with MLM for 2020 C) QQplot for GWAS with K_chr for 2019 D) QQ plot for GWAS with K_chr for 2020 E) QQ plot for GWAS with FarmCPU for 2019 F) QQ plot for GWAS with FarmCPU for 2020

Supplemental Figure S2. QQ plots of the GWAS results using three models for the combined dataset.

Supplemental Figure S3. BLUP distribution by subpopulation.

Supplemental Table S1. The rating scale used for anthracnose leaf blight evaluations (modified from Burrell et al., 2015).

Supplemental File S1. BLUP values for the sorghum conversion panel. Tab names indicate dataset.

Supplemental File S2. PAST analysis pathways significant at α < 0.05.

Khanal, A. , Adhikari, P. , Kaiser, C. , Lipka, A. E. , Jamann, T. M. , & Mideros, S. X. (2022). Genetic mapping of sorghum resistance to an Illinois isolate of Colletotrichum sublineola. The Plant Genome, 15, e20243. 10.1002/tpg2.20243