RNA-Seq Unveils Cross-Talk among Stress Response Mechanisms during Leaf Color Transformation in ALS Resistant Sorghums

Abstract

Acetolactate synthase (ALS) inhibitor herbicides are among widely marketed herbicide chemistries that act both against grass and broad-leaved weeds. Sorghum (Sorghum bicolor (L.) Moench) variants carrying resistance to ALS inhibitor herbicides were developed as a post-emergence weed control solution in sorghum. However, some ALS-resistant lines exhibit noticeable interveinal chlorosis at seedling stage, leading to reduced vigor. Although the plants eventually recover at an advanced growth stage, this may be a source of concern for growers and can undermine adoption of the technology. This study was initiated to identify mechanisms related to the manifestation of this phenotype. Two ALS-resistant genotypes, one displaying a yellow phenotype and the other a normal green phenotype, were cultivated, and tissue samples were collected at four time intervals, with the final sampling occurring after the genotypes had fully re-greened. RNA was extracted from the tissue samples and subjected to RNA-Seq analysis. Differential gene expression analysis was carried out using DESeq2, and a selected set of genes were confirmed via qRT-PCR. Gene Ontology enrichment and SorghumCyc pathway analysis uncovered notable regulatory changes in genes associated with chloroplasts, plant defense responses, and hormonal networks in the yellow genotypes. The pattern of gene expression strongly mimicked responses under abiotic stresses. In addition, the findings offer new insights into the potential for sorghum genotypes resistant to environmental stresses to also exhibit tolerance to a range of additional stresses.

1. Introduction

The United States accounts for about 9% of the world’s sorghum (Sorghum bicolor (L.) Moench) acreage and contributes 25% to the global grain sorghum production with over half of this being produced in Kansas [1]. However, the lack of viable post-emergence grass weed control has seriously undermined sorghum production in the U.S. Apart from causing considerable economic loss, the lack of effective weed control tools has forced farmers to switch to less resilient crops such as corn. Sorghum has not benefitted from the glyphosate resistance technology due to the concern of the resistance gene escaping in to the wild sorghum population [2]. Thus, sorghum acreage in the U.S. has steadily declined over the past few decades with much of the lost acreage being picked up by corn. According to local farmers in Kansas, the major driver for the switch was a better weed control option that corn offers. This agrees with a nationwide survey conducted in 2007 where the lack of effective post-emergence weed control was ranked as the major constraint to sorghum production. The discovery of the acetolactate synthase (ALS) resistance brought some optimism for addressing this important need of the grower community by providing a viable post-emergence weed control option [3].

ALS herbicides are among widely marketed herbicide chemistries that act on both grass and broad-leaved weeds. In 2007, a source of ALS resistance carrying a double mutation in the ALS gene that comprises amino acid substitutions from Val-560-Ile and Trp-574-Leu was identified in a wild sorghum (shattercane) population in Kansas. Of these mutations, only Trp-574 is a conserved residue; thus strongly associated with resistance to ALS inhibitor herbicides [4,5]. The effort to incorporate this gene into adapted backgrounds via repeated back crosses has been successful over the past few years. Hence, a large number of sorghum germplasm and parent lines with strong resistance to all classes of ALS inhibitor herbicides have been developed and released by the Kansas State University sorghum breeding program. Sorghum carrying the resistant gene can tolerate herbicide concentrations that are 6 to 10× the normal use rate [4]. However, many ALS-resistant lines tend to show variable degrees of leaf yellowing and reduced seedling vigor at early stages of growth, and this phenotype seems to be not due to herbicide injury given that it occurs before herbicide spray [6]. Even though the plants eventually re-green and effectively grow out of these symptoms after a few weeks of emergence, such a yellow phenotype with low seedling vigor may be disturbing to growers and undermine the adoption of the technology. The expression of such a phenotype seems to be background dependent since some backgrounds tend to be more affected (suffer from extreme yellowing and stunting), while others are only moderately affected or, not affected at all. Leaf chlorophyll scoring of several families derived from different backgrounds confirmed this variability [6]. Thus, it appeared that the variable phenotype emerged as a result of interaction between the mutant ALS gene and other genes in the backgrounds.

Conventional phenotyping approaches based on field evaluations carried out over the years have not been able to provide substantial evidence to conclude on plausible causes for this variable phenotype. Hence, a new approach was sought to allow the detection of genes, associated metabolic pathways, and plant mechanisms contributing to this altered phenotype. The availability of the whole genome sequence of sorghum and the advent of molecular techniques such as RNA sequencing (RNA-Seq) have facilitated the development of better tools to address complex questions similar to this one. RNA-Seq is a recently developed robust deep-sequencing technology that utilizes the capabilities of next-generation sequencing to allow deep sampling of transcriptomes. Herein, our effort was to adapt RNA-Seq technology in combination with the sorghum genome sequence [7] and Gene Ontology (GO) analysis followed by metabolic pathway analysis using the SorghumCyc database to characterize the sorghum transcriptome to elicit gene expression differences in seedlings of ALS-resistant sorghum sister lines (near-isogenic lines) differing in seedling phenotype (yellow and green/normal). This investigation was aimed at identifying genetic factors that trigger the expression of the interveinal chlorosis in the affected genotypes. This study is expected to provide clues on genetic mechanisms resulting in chlorotic leaves and facilitate breeding of new ALS resistant varieties/hybrids that suffer less from this phenomenon.

2. Materials and Methods

2.1. Genetic Materials

A pair of ALS inhibitor herbicide-resistant sister lines derived from a lineage “Brhan x (Macia//Macia/Tw)—3” that represents the two classes of seedling phenotypes (green and yellow) were selected for this study. The sister lines were pooled from an F6 family of a cross that involved tropically adapted varieties “Brhan” and “Macia”, and the ALS-resistant gene donor “Tailwind” (Tw). The two sister lines exhibit similar adult plant phenotype but maintain clear differences in seedling phenotypes with one of the lines (PR12/13-764-2) strictly expressing yellow leaf color and the green (normal) sister line (PR12/13-763-2) consistently remaining green. The use of such genetically related genotypes was expected to eliminate the background noise and reduce complexities during expression analysis. The seeds were pre-treated with Maxim, Apron XL, Concep III were grown at the Kansas State University Agronomy Research Farm at Ashland Bottoms near Manhattan, Kansas. The field was preferred to greenhouse cultivation as it offered better expression of the yellowing trait and good contrast between the yellow and green genotypes. Three plants from each sister line were randomly tagged for data and sample collection.

2.2. Experimental Design, Tissue Collection, and RNA Extraction

The RNA-Seq experimental design was set up to compare gene expression from three biological replicates of the yellow and green genotypic backgrounds. Samples were collected at four growth stages, namely: S0, the stage prior to ALS herbicide application (i.e., 14 days after planting), and S1 through S3, corresponding to the three stages after herbicide application (i.e., 21, 28, and 36 days after planting, respectively) (Figure 1). Since the variation in seedling vigor and leaf chlorosis continued until four weeks after emergence, sampling during this time was suggested to capture most of the genes that may be differentially expressed between green and yellow sister lines. At 15 days after planting, the plots were sprayed with ALS inhibitor herbicide Accent^® (Dupont Pioneer, Johnston, IA, USA) at the rate of 1.5 oz ai/acre. Accent contains nicosulfuron, which is a sulfonylurea that works mainly by contact action where susceptible plants die within two weeks after application. Approximately 100 mg of leaf tissue samples were collected from each tagged plant. The tissues were immediately frozen in liquid nitrogen to prevent mRNA degradation and stored under −80 °C. Total RNA extraction of the collected tissue samples was performed using the RNeasy Plant Mini isolation Kit (Qiagen Inc., Valencia, CA, USA), and the extracted total RNA was treated with Amplification Grade DNase I (Invitrogen Corporation, Carlsbad, CA, USA) before further analysis. The RNA samples were diluted with RNase free water to obtain samples with the required concentration (100-200ng/ul). The RNA integrity and quantity were checked using the Agilent 2100 Bioanalyzer (Agilent Technologies Genomics, Palo Alto, CA, USA). Additionally, several other ALS-resistant genotypes showing variability for leaf color were grown alongside the two sister lines used for RNA-Seq, and the leaf chlorophyll contents were measured using a SPAD 502 chlorophyll meter (Spectrum Technologies, Aurora, IL, USA) at the S0 and S3 sampling stages to monitor the phenotypic changes as additional evidence (Supplementary Table S1).

2.3. cDNA Library Construction and Sequencing

cDNA libraries were constructed using the Illumina TruSeq RNA sample preparation kit according to the manufacturer’s protocol (Illumina Inc., San Diego, CA, USA). RNA from each genotype was subjected to two rounds of enrichment for poly-A mRNAs using “oligodT” attached magnetic beads. The purified mRNA was fragmented and converted to single-stranded cDNA according to the manufacturer’s protocol (Illumina Inc., San Diego, CA, USA). cDNA samples from each genotype were separately barcoded with adapter indexes and pooled. Sequencing was performed on a HiSeq 2000 platform (Illumina Inc., San Diego, CA, USA) using 100 bp single-end sequencing runs at the Genome Sequencing Facility of Kansas University Medical Center.

2.4. Differential Gene Expression Analysis and Gene Clustering

Single-end sequencing reads obtained from HiSeq 2000 runs were subjected to adapter trimming and quality filtering with “Cutadapt”, which is a stand-alone adapter trimmer [8]. The Sorghum bicolor reference genome (Sbicolor_v1.49) (https://phytozome-next.jgi.doe.gov/: accessed on 26 June 2024) was used to perform read alignment using the Genomic Short-read Nucleotide Alignment Program (GSNAP) [9]. Read counting per gene in each sample was conducted using an in-house script. Differential gene expression among yellow and normal genotypes was analyzed using “DESeq2”, which employs a method based on the negative binomial distribution, with the variance and mean linked by local regression. A q-value [10] was determined for each gene to account for multiple tests. To control the false discovery rate (FDR) at 5%, the differentially expressed genes were required to have q-values smaller than 0.05. Additionally, we only included genes shown at least two fold-change in the list of significantly differentially expressed genes. The RPKM value per gene in each sample represents read counts per kilobase of the transcribed region per million reads [11]. The analysis was further extended to test the null hypothesis that there was no interaction between genotypes and sampling stages in order to identify the pattern of changes in differential gene expression among the sampling stages between two genotypes using DESeq2. The 5% FDR was used as the threshold to obtain a set of genes with significant interaction between genotypes and sampling stages. The Log₂ expression ratios between green and yellow genotypes of this set of genes were used as the input for cluster analysis with the R package “mclust” [12] using the model “VVV”.

2.5. qRT-PCR Validation of RNA-Seq Results

RNA samples from the sister lines used in RNA-Seq were used for qRT-PCR in order to validate the RNA-Seq results. Hence, a selected set of seven sorghum genes were subjected to qRT-PCR. The selected genes were basically involved in mechanisms such as chlorophyll metabolism and defense response mechanisms. The sorghum actin (Sb Actin) gene (accession No. SVSOAC1), which showed constant expression in seedlings [13,14], was chosen as a reference gene for the normalization of gene expression using primers Sb_Ac_F (5′-ACA TTG CCC TGG ACT ACG AC-3′) and Sb_Ac_R (5′-TGA TGA CCT GTC CAT CAG GA-3′). From the DNase1-treated total RNA samples used for RNA-Seq, 5 ug was diluted to a final concentration of 100 ng/uL for use in qPCR. Gene-specific primers were designed for each of the selected genes to span a selected exon while avoiding the splice junctions using the Beacon designer software (PREMIER Biosoft, Palo Alto, CA, USA). cDNA synthesis was performed using the iScript cDNA Synthesis Kit followed by qPCR reactions that were set up on cDNA samples using an iQ SYBR Green SuperMix Kit according to the manufacturer’s protocols (Bio-Rad Laboratories, Hercules, CA, USA). For each yellow and green comparison, expression measurements were performed using three biological and two technical replications. The CFX96 Touch Real-Time PCR Detection System (Bio-Rad Laboratories, Hercules, CA, USA) was used for the qPCR analysis. The results in the form of fluorescence curves and cycle threshold (Ct) values were analyzed using CFX Manager Software Version 3.1 (Bio-Rad Laboratories, Hercules, CA, USA). Sample cycle threshold (Ct) values were determined from plots of relative fluorescence units (RFU) vs. PCR cycle numbers during exponential amplification. The relative expression for each gene was normalized against a sorghum actin gene so that sample measurement comparisons were possible. Relative expression ratios were calculated on the basis of the 2^(−ΔΔCt) method that utilizes group means for the target gene versus the reference gene, and group ratio results were tested for significance using t-tests.

2.6. GO Enrichment

Gene Ontology (GO) enrichment analysis was performed to identify over-represented GO terms in our differentially expressed gene lists using an R software package, goseq. GO functional annotations for sorghum gene products were downloaded from Agrigo (http://bioinfo.cau.edu.cn/agriGO/: accessed on 2 January 2015). GO categories were considered significantly enriched based on the P-value cutoff of 0.05. Based on the results for the GO analysis, the genes related to significant GO terms were extracted, and the expression pattern of related genes in terms of Log₂ fold change at each stage was visualized using a heatmap generated via the R package Heatplus.

2.7. SorghumCyc Pathway Analysis Followed by Visualization via Mapman

Metabolic pathway enrichment using SorghumCyc genome database was performed for each differentially expressed gene list using the Z-score method suggested by [15] in order to derive functional annotations to infer the metabolic pathways of sorghum [16]. Pathways were considered significantly enriched if the following criteria were met: Z-score ≥ 2 and the expected number of genes for a family > 1. Mapman has the capability of lightening the redundancy that occurs in other commonly used ontologies. Hence, Mapman was used to collect and classify the calculated fold change values into a set of hierarchical functional categories called “bins”, which then were organized and displayed according to our choice. Herein, using Mapman alone was not preferred, as it may not provide a holistic view of the significant pathways in order of significance as would a Z-core method. However, Mapman provides a better graphical output of the expression under a certain cell component/pathway of choice. Thus, a combination of two methods was used for visualization of the results.

3. Results and Discussion

3.1. Mapping of Transcriptome to the Sorghum Genome

The leaf chlorophyl content between the yellow and normal genotypes monitored using SPAD 502 was significantly different between the sister lines during first three stages of sampling, while no difference was observed at the last sampling stage (Supplementary Table S2 and Figure 2). Apart from our previous published reports and reports by Kansas state research and extension [6,17], there was no literature support on color transformation in sorghum that is caused by a genetic disorder. However, there is plenty of documentation on the yellowing of sorghum due to other factors such as herbicide injury, pesticide injury, fungal attacks, pest attacks, and nutrient deficiencies [3,18,19,20,21,22]. On the other hand, there are few reports on scenarios similar to those in our study, but on other crops such as Nicotiana [23], cucumber [24], and citrus [25]. However, none of these reports relate to genotypes conferring herbicide resistance.

In the present study, the expression of 27,608 unique gene models was resolved out of 34,496 sorghum gene models reported earlier [7]. The number of gene models found in our study was much higher as compared with a previous microarray analysis of Sorghum bicolor [26], which reported the activity of 12,982 unique genes. Overall, approximately 512.6 million reads were generated across all three biological replicates for yellow and green genotypes. Of those, 497.2 million (~95–98%) passed quality filtering standards, and out of those, 461.8 million (~88–91%) confidently mapped to the sorghum genome. At the same time, a quite high percentage of uniquely mapped reads were observed as compared with other earlier studies that resulted in 67.1% [15,27]. The read mapping summary for all the yellow and normal samples used in the study can be found in

Table 1. The read mapping summary for yellow and normal samples at each stage of sampling.

Sample	Trimmed Reads	Mapped Reads	Mapped %	Confident Mapped Reads	Confident Mapped %
G-S0-rep1	17,100,000	16,717,481	97.8	15,489,374	90.6
G-S0-rep2	11,386,966	10,314,346	90.6	9,478,054	83.2
G-S0-rep3	8,138,406	7,919,657	97.3	7,302,557	89.7
G-S1-rep1	12,544,645	11,972,499	95.4	11,094,164	88.4
G-S1-rep2	11,304,313	11,009,385	97.4	10,188,836	90.1
G-S1-rep3	12,367,994	12,098,903	97.8	11,261,965	91.1
G-S2-rep1	12,715,191	12,151,177	95.6	11,236,344	88.4
G-S2-rep2	11,345,117	10,851,960	95.7	10,078,756	88.8
G-S2-rep3	15,031,984	14,694,783	97.8	13,570,020	90.3
G-S3-rep1	12,185,823	11,838,370	97.1	10,975,077	90.1
G-S3-rep2	10,549,889	10,096,828	95.7	9,352,468	88.6
G-S3-rep3	10,053,833	9,774,332	97.2	9,048,876	90
Y-S0-rep1	12,932,513	12,585,942	97.3	11,656,006	90.1
Y-S0-rep2	14,913,289	14,551,466	97.6	13,512,155	90.6
Y-S0-rep3	14,578,395	14,244,619	97.7	13,222,902	90.7
Y-S1-rep1	8,850,648	8,598,795	97.2	7,995,057	90.3
Y-S1-rep2	11,699,313	11,428,736	97.7	10,566,808	90.3
Y-S1-rep3	11,271,245	10,928,415	97	10,133,689	89.9
Y-S2-rep1	14,450,551	14,051,613	97.2	13,064,471	90.4
Y-S2-rep2	13,050,379	12,460,776	95.5	11,593,536	88.8
Y-S2-rep3	11,872,101	11,542,378	97.2	10,755,236	90.6
Y-S3-rep1	11,241,136	10,989,004	97.8	10,182,955	90.6
Y-S3-rep2	11,541,148	11,235,620	97.4	10,434,927	90.4
Y-S3-rep3	12,012,777	11,625,966	96.8	10,759,674	89.6

. Quantile normalization for each run showed an improvement in the quality of the mapped reads, and Spearman correlations observed between biological replicates were quite high and comparable to previous RNA-Seq work [15,27,28], which attested high reproducibility of the results generated. For an example, the correlations for two of the samples from stage S0 are shown in Figure 3A,B. Additionally, P-value histograms for the normalized read counts for each comparison showed acceptable read count distributions. For an example, Figure 3C shows the P-value distribution for the comparison at the S0 stage.

3.2. Determination of Differentially Expressed Genes and Validation via qRT-PCR

We performed comparisons between the yellow and normal genotypes separately for each growth stage from S0 through S3. The sampling was conducted during the period when the highest phenotypic changes occurred in the plants; thus, gradual re-greening of the yellow color plants was markedly visible throughout the sampling stages. The differential expression analyses resulted in 7510, 6787, 5709, and 3575 differentially expressed genes from the comparisons of yellow vs. green genotypes, at S0, S1, S2, and S3, respectively. In total, GO term enrichment based on the lists of differentially expressed genes identified for four sampling stages resulted in 136–180 GO categories at 0.05 significance level. The significant number of up- and down-regulated GO categories in each stage primarily accounted for the phenotypic changes and defense response mechanisms involved in the recovery process. Thus, essentially, the enriched GO categories included chloroplast-related genes and structural components within the chloroplast, the response to biotic and abiotic stresses, sugar metabolism, and numerous pathways related to toxic catabolite detoxification (Supplementary file S1). Pathway enrichment using SorghumCyc annotations revealed unique regulation of metabolic pathways ranging from 31 to 62 based on the Z-score analysis, which helped to further refine the results obtained via the GO term analysis. This facilitated us to filter out only the pathways that exhibited high-confidence differentially expressed genes (Supplementary file S2). Pathways with significant regulatory activity basically accounted for defense response, phenotypic changes in leaf tissues, hormonal networks, and other processes that resulted due to the internal stress created due to the significant loss of chlorophylls. For an example, betanidin degradation, glutathione (GSH) mediated detoxification, gamma glutamyl cycle, nicotine degradation, chlorophyll a degradation, phospholipases, etc., were common in almost all the sampling stages. However, few pathways that had a considerable number of genes grouped under both the up and down categories limited our capacity to make meaningful conclusions. The selected set of seven genes tested using qRT-PCR agreed with the RNA-Seq data with for the expression level for a few genes that could be attributed to the small sample size of the selected set of genes for qRT-PCR (Figure 4 and Supplementary Table S3).

3.3. Transcript Analysis in Response to Herbicide Effect

Gene expression levels between the yellow and green genotypes were compared at S0, in order to monitor any expression level changes that occurred prior to the herbicide application and from S1 through S3 to monitor the expression after herbicide application. GO categories that showed enrichment for up- and down-regulated genes at the stages before (S0) and immediately after herbicide application (S1) showed a large overlap and thus further confirmed our early observation that leaf yellowing occurred due to a reason other than the herbicide effect (Figure 5). Thus, at both stages, GO categories that showed enrichment for up- and down-regulated genes commonly comprised the chloroplast, the thylakoid membrane, the thylakoid lumen, electron carrier activity, response to high light intensity, the sucrose biosynthetic process, and several terms related to toxin catabolic processes, etc. (Supplementary file S1). In total, the activity of 308 genes was resolved as in relation to chloroplast, out of which 274 genes occupied in chlorophyll metabolism showed differential expression in at least one of the sampling stages (Figure 5C). However, the variability in the expression profiles of chloroplast-related genes reduced from S0 to S3 stage (Figure 2A,B). Figure 5A,B compare gene bins that were up- and down-regulated between S0 and S3 during the initial step of photosynthesis (photosynthesis light reaction) showing a marked difference in gene expression levels between the two stages. However, a considerable number of GO terms linked to the chloroplast, which were both up- and down-regulated at S0 and S3, implied possible complexities in the regulation of chloroplast genes and proteins targeting the chloroplast. However, GO term analysis in combination with pathway analysis helped to identify pathways with higher regulatory activity; thus, it was evident that a considerable number of genes linked to the chlorophyll degradation pathway were significantly up-regulated, while both up- and down-regulated genes were observed in the chlorophyll biosynthesis pathway. Apart from GO terms related to the chloroplast, each stage comprised a large number of biotic and abiotic stress-related GO terms such as stress response hormonal networks, oxidative stress due to reactive oxygen species (ROS), and toxin catabolic processes involving glutathione S-transferase (GST) activity (Figure 6). Pathway analysis further supported this through significant activity in the gamma glutamyl cycle (which involves GSTs), at all sampling stages (Supplementary file S1). Although GSTs have been shown to catalyze the conjugation of GSH with potentially dangerous xenobiotics such as herbicides [29], the increased regulation of gamma glutamyl cycle even before the herbicide application suggested possible abiotic stress created due to altered gene regulation within the yellow genotype. Furthermore, the evidence on important roles of GSH in the defense response in plants such as protection against oxidative stress from free radicals owing to its redox-active thiol group [30,31] and regeneration of antioxidants [32] and as a regulator of gene expression [33,34] further supported our observations of possible stresses due to altered gene regulation within the yellow genotypes, which was also reflected through increased activity in other stress response pathways.

3.4. Activated Hormonal Networks in Yellow Plants

As revealed through GO enrichment and pathway analysis, apart from the up-regulated activity of detoxification processes, there was a lot of variability in transcript abundance in hormonal networks related to defense responses (Figure 7). Among those were abscisic acid (ABA), cytokinins (CKs), brassinosteroids (BRs), jasmonate (JA), and ethylene (ET), which showed significant Z-scores under pathway analysis (Supplementary File S2) with possible implications that any kind of mechanism that causes altered regulation in chloroplast-related genes should have triggered defense response mechanisms within the plant causing a “domino effect”. A rapid increase in endogenous ABA levels causing activation of specific signaling pathways leading to modifications in gene expression levels has been observed in response to abiotic stresses [35,36,37], and up to 10% of protein-encoding genes are transcriptionally regulated by ABA [38]. Also, altered endogenous CK levels have been evident during various kinds of stress in plants [39]. The significant number of down-regulated genes in the CK biosynthesis pathway observed in this study suggest that the hormone was involved in stress responses created due to the loss of chlorophylls during early sampling stages (Supplementary file S2). In agreement with our observation, there are several previous studies that have shown increased abiotic stress resistance at low CK levels and mutants lacking functional CK receptors expressing higher resistance to abiotic stresses [40,41]. Furthermore, there is evidence that CK levels in plants are usually co-regulated with the salicylic acid SA levels under biotic stress conditions such as pathogen attacks [42,43,44]. Generally, either deficiency or very high levels of SA are attached to increased plant susceptibility to abiotic stress [45]. Thus, non-differentially expressed SA levels detected in our study further confirmed its role in response to abiotic stress tolerance. On the other hand, SA is required for inducing stress resistance proteins, such as antioxidants and heat shock protein (HSP); thus, SA-deficient plants cannot create an effective abiotic stress defense system [46,47]. Thus, the up-regulatory activity of heat shock proteins in our study samples further confirmed the presence of SA activity, which is commonly observed in plants subjected to oxidative stress conditions (Figure 7). At the same time, nucleotide binding receptor—leucine rich repeat (NBS-LRR) genes that have been identified to have a well-established role in disease resistance reaction were also overexpressed with potential evidence on cross-talk in certain defense response mechanisms under both biotic and abiotic stress.

Brassinosteroids, on the other hand, are known to induce plant tolerance to a variety of biotic and abiotic stresses [48], and the prospective role of BRs to increase plant tolerance to abiotic stress has been reviewed recently [49]. A unique activity in brassinosteroid biosynthesis in most of the sampling stages was apparent in our study, attesting its role in resistance to early season stress created due to the excessive loss of chlorophylls. In fact, BRs have been described as a booster of net photosynthetic rate in a number of plant species, and BR activity is generally more pronounced in plants under various abiotic stressors such as drought, high or low temperature, salinity, or heavy metals [50].

3.5. Defense Responses Closely Mimic Gene Regulation under Drought and Salinity Stresses

Plant responses to different types of stresses have been shown to be associated with the generation of ROS. Thus, ROS may function as a common signal in pathways of plant stress responses [48]. Enriched GO terms such as singlet oxygen, response to oxidative stress, response to hydrogen peroxide, and significant up-regulated activity of antioxidant compounds such as betanidin, which carries high antioxidant activity [51], observed in our study suggest activated plant defense responses due to the early chlorophyll breakdown. Moreover, there is evidence on elevated hydrogen peroxide levels resulting in BR-induced stress tolerance [48]. Chloroplasts detoxify the elevated hydrogen peroxide generated due to the stress reactions via ascorbate peroxidases [52]; thus, significant regulation of pathways related to ascorbate shows possible photo-oxidative damage to the leaf tissues due to the loss of chlorophylls, as was also evident through the resulted significant GO terms related to the chloroplast and thylakoids. Increased activity in the degradation of betanidin, which is a strong antioxidant, and Cytochromoe 450, which has an oxidative, peroxidative, and reductive role in the degradation of endogenous xenobiotics, further supported this interpretation. On the other hand, protein kinases [53,54] and glycein betain [55,56], which are well known to play an important role in acquired tolerance to drought and salinity stress, showed significant up-regulatory activity during all sampling stages.

Apart from the discussed defense response pathways, ROS-driven up-regulation of breakdown of membrane phospholipids was observed in the yellow genotype due to altered regulation of phospholipid biosynthesis, phospholipases, and triacylglycerol degradation. At the same time, increased degradation of starch reserves and sucrose biosynthesis, which has shown strong relationships with plant stress response under drought [57], were also evident. Overall, observations on alerted gene regulation in the yellow genotype suggested cross-talk among pathways related to biotic and abiotic stress responses basically mimicking drought and salinity stresses.

3.6. Clustering Pattern of Differentially Expressed Genes

Out of a total of 27,608 gene models resolved in our study, 5321 were resolved with significant interactions between the genotype and the sampling stages. Clustering performed based on the Log₂ expression ratio between the green and yellow genotypes identified 11 major genes clusters (groups) (Figure 8). The resulted clustering patterns varied from stable expression ratios to irregular patterns. However, among all of the clusters, our major consideration was drawn toward gene clusters that showed alterations in expression ratios from increased to decreased and vice versa along the four sampling stages. Thus, clusters 9, 10, and 11, which contained 327, 46, and 28 genes, that comprised stable, increasing, and decreasing gene expression ratios, were excluded from further analysis. The rest of the clusters were divided into two major groups based on their patterns. Thus, clusters 1, 5, and 7, which showed more or less similar expression ratio pattern were considered as cluster set1, while clusters 2, 3, 4, 6, and 8, which showed more or less the opposite pattern of variability to the first set of clusters, were considered as the second set of clusters. The variability seen in the genes captured in clusters 1, 5, and 7 reflected up-regulated activity during early sampling stages that was down-regulated at the late sampling stages in yellow genotypes. Common functions of the genes captured in clusters 1, 5, and 7 amalgamated with significant GO terms and pathways in previous analysis revealed that, they were related to photosystem I and II reaction centers; chlorophyll-binding proteins; chloroplast precursors; signal transduction involving calmodulin; hormones such as auxin, cytokinin, and ethylene; oxidative stress response genes involving glutamate cycle genes; heat shock proteins; cytochrome P450; oxidoreductases; and specifically chlorophyll catabolic genes such as pheophorbide a oxygenase and drought induced proteins (Supplementary file S3). In contrast, clusters 2, 3, 4, 6, and 8, which showed increased initial expression ratios that decreased at later stages, generally involved expansins that are primarily involved in channeling water, which are also known to express in leaves under drought stress [58]; anthocyanins that reduce levels of photo-oxidative stress on the degenerating chloroplasts with improving rates of nitrogen recovery from senescing leaves [59]; aquaporins that show water stress-dependent expression with individual gene regulation profiles [60]; and UDP-glucosyltransferases that help in sucrose synthesis. A few other genes that were commonly found in both sets of clusters included heat shock proteins, cytochrome P450, oxidoreductases, and peroxidase precursors. The occurrence of considerable overlap between genes with similar functions in two groups of clusters could be explained in two ways. First, this kind of observation may be possible when certain pathways show both up- and down-regulated genes at all or some sampling stages. Second, genes in a particular pathway may comprise variability in expression profiles; thus, some genes could be switched on when others are switched off to assist in maintaining a defense response or resistance at all stages until the plant is fully recovered from the stress condition. However, a large number of genes clustered could not be classified under a specific group due to their intermediary involvement in protein synthesis, transcription, cellular transport, signal transduction, and other cellular processes that are not highlighted here due to their complex behavior tied to several pathways. As well, a considerable number of the clustered genes being un-annotated or annotated only as “expressed proteins” limited their usefulness in drawing further conclusions.

4. Conclusions

The whole process of developing a cultivar and taking it to the farmer is as important as earning farmers’ satisfaction and cultivar stabilization. Our attempt in this study to understand the genetic basis of the unusual but self-corrective yellow phenotype in some ALS-resistant cultivars showed that survival mechanisms under the stress created due to decreased chlorophyll contents in leaf tissues seemed to have primarily accompanied by the defense response mechanisms and hormonal networks that mimicked salinity and drought stress response reactions. At the same time, the significant number of DE genes that largely consisted of putative uncharacterized and un-annotated genes limited their usefulness during interpretations with emphasizing the need for further refinement and confirmation of sorghum genome annotations. The findings provide useful insight on possible cross-talk among biotic and abiotic defense response mechanisms in sorghum. Thus, the outcome of this study opened up a new window for future research by putting forward the idea that inherent tolerance to abiotic stresses that could be embedded in a certain germplasm or parental line would alternatively provide plant tolerance to multiple stressors regardless of the source condition that initiated the stress.