Differentially Regulated Orthologs in Sorghum and the Subgenomes of Maize

Abstract

Identifying interspecies changes in gene regulation, one of the two primary sources of phenotypic variation, is challenging on a genome-wide scale. The use of paired time-course data on cold-responsive gene expression in maize (Zea mays) and sorghum (Sorghum bicolor) allowed us to identify differentially regulated orthologs. While the majority of cold-responsive transcriptional regulation of conserved gene pairs is species specific, the initial transcriptional responses to cold appear to be more conserved than later responses. In maize, the promoters of genes with conserved transcriptional responses to cold tend to contain more micrococcal nuclease hypersensitive sites in their promoters, a proxy for open chromatin. Genes with conserved patterns of transcriptional regulation between the two species show lower ratios of nonsynonymous to synonymous substitutions. Genes involved in lipid metabolism, known to be involved in cold acclimation, tended to show consistent regulation in both species. Genes with species-specific cold responses did not cluster in particular pathways nor were they enriched in particular functional categories. We propose that cold-responsive transcriptional regulation in individual species may not be a reliable marker for function, while a core set of genes involved in perceiving and responding to cold stress are subject to functionally constrained cold-responsive regulation across the grass tribe Andropogoneae.

Figure 1.

Gene Level and Expression Level Conservation between Sorghum, Maize1, and Maize2. (A) The overlap between syntenic orthologous gene pairs conserved between maize1/sorghum and maize2/sorghum. (B) Comparison of average control condition expression levels (log₂ transformed FPKM) for either maize1/sorghum or maize2/sorghum gene pairs. (To improve readability, a random sample of 1/3 of all gene pairs is displayed for each category.)

Figure 2.

Effects of Cold Stress on Maize, Sorghum, and Related Species. (A) to (C) Representative seedling phenotypes for maize and sorghum. Control conditions (A), 24 h of stress at 6°C (B), and 14 d at 6°C and 2 d recovery under greenhouse conditions (C). (D) Normalized relative CO₂ assimilation rates for six panicoid grass species with differing degrees of sensitivity or tolerance to cold stress. Individual data points were jittered (adding random noise to data in order to prevent overplotting in statistical graphs) on the x axis to avoid overlap and improve readability.

Figure 3.

Combined DEG Analysis of Maize and Sorghum. (A) An illustration of the DEG-based gene pair classification model and a comparison of expected and observed values for gene pairs classified as differentially expressed in response to cold in zero, one, or both species. Expected distributions were calculated based on a null hypothesis of no correlation in gene regulation between maize and sorghum (see Methods). DE0, gene pairs classified as differentially expressed in response to cold in neither species; DE1, gene pairs classified as differentially expressed in response to cold in one species but not the other; DE2, gene pairs classified as differentially expressed in response to cold in both species. Observed number of gene pairs in maize1/sorghum: DE1 maize = 850, DE2 = 836, DE1 sorghum = 1507, DE0 = 12,038. Observed number of gene pairs in maize2/sorghum: DE1 maize = 508, DE2 = 460, DE1 sorghum = 986, DE0 = 7599. Expected number of gene pairs in maize1/sorghum: DE1 maize = 1427, DE2 = 259, DE1 sorghum = 2084, DE0 = 11,461. Expected number of gene pairs in maize2/sorghum: DE1 maize = 822, DE2 = 146, DE1 sorghum = 1300, and DE0 = 7285. (B) Comparison of fold change in gene expression between the treatment and control groups for pairs of orthologous genes in maize and sorghum. Log₂-transformed treatment/control expression ratios are shown.

Figure 4.

Patterns of Gene Expression across a Cold-Stress Time Series in Maize and Sorghum. (A) Changes in classification of individual gene pairs as DE0, DE1 maize, DE1 sorghum, and DE2 across adjacent time points. (B) The proportion of genes identified as differentially expressed in both species in excess of the number of gene pairs expected in this category in the absence of either conservation of gene regulation or parallel evolution of gene regulation. True discovery proportion is defined as (observed positives − estimated false positives)/observed positives. The expected number false positive DE2 gene pairs was calculated from the proportion of all genes classified as DEGs in maize and sorghum using the null model described in Figure 3A.

Figure 5.

Conceptual Illustration of the Differentially Regulated Ortholog Model. (A) Illustration of the different classification outcomes that can be produced for a given gene pair using both a DEG-based analysis (testing whether the expression pattern of each gene changes significantly between conditions) and a DRO-based analysis (testing whether the pattern across the two conditions is significantly different between copies of the same gene in both species). (B) Two models, additive and multiplicative, for predicting what a conserved pattern of gene regulation should look like when the underlying level of expression changes. (C) Relationship between prediction error (log₁₀ transformed) for expression under cold stress using a multiplicative model to predict expression between maize1/maize2 gene pairs or an additive model to predict expression between maize1/maize2 gene pairs. Maize1: Predictions for the expression pattern of maize2 genes using data from their maize1 homoeologs. Maize2: Predictions for the expression pattern of maize1 genes using data from their maize2 homoeologs. Blue dots mark cases where the additive model was the better predictor; red dots mark cases where the multiplicative model was the better predictor.

Figure 6.

Characteristics of Genes in Different DEG Groups at Different Time Points. (A) The proportion of gene pairs classified as DROs between maize and sorghum in different DEG groups at each of the six time points examined. (B) and (C) Median ratios of nonsynonymous substitutions to synonymous substations in coding sequences for maize and sorghum for gene pairs classified as DE0, DE1, or DE2 at each of six time points. Time points where there is a statistically significant difference in Ka/Ks ratio between DE2 and any of the other three categories are marked with either + (if P < 0.05) or ++ (if P < 0.01). Color of the + indicates the category to which DE2 is being compared. Time points where there is a statistically significant difference in Ka/Ks ratio between DE0 and either DE1 maize or DE1 sorghum categories are marked with either * (if P < 0.05) or ** (if P < 0.01). Color of the asterisk indicates the category to which DE0 is being compared. Enrichment of genes annotated as transcription factor genes among DE2 gene pairs relative to all syntenic gene pairs indicated by the black line and the right-hand axis. Double white triangles mark time points where the enrichment is statistically significant (P < 0.01). (D) Frequency of CNS within the promoters of genes classified as DE0, DE1 maize, DE1 sorghum, DE2, DRO, or CRO at each of the six time points. Black lines within the box plot mark the average number of CNS per gene for each category.

Figure 7.

Chromatin Patterns Associated with Different Groups of Genes in Maize and Sorghum. Patterns of MNase HS regions around the transcriptional start sites of genes classified based on their pattern of gene regulation in the 24-h stress time point. Maize1 sorghum gene pairs and maize2 sorghum gene pairs were aggregated to increase statistical power. The lighter band around the DE2 line indicates a 2 sd confidence interval. Black bars at the bottom of the graph indicate individual base pair positions where the amount of open chromatin associated with DE2 genes is significantly different from that of each of the other four categories displayed with a P value < 0.01 for each comparison. Pairwise comparisons were performed using Fisher’s exact test.