Maize pan-transcriptome provides novel insights into genome complexity and quantitative trait variation

Abstract

Gene expression variation largely contributes to phenotypic diversity and constructing pan-transcriptome is considered necessary for species with complex genomes. However, the regulation mechanisms and functional consequences of pan-transcriptome is unexplored systematically. By analyzing RNA-seq data from 368 maize diverse inbred lines, we identified almost one-third nuclear genes under expression presence and absence variation, which tend to play regulatory roles and are likely regulated by distant eQTLs. The ePAV was directly used as "genotype" to perform GWAS for 15 agronomic phenotypes and 526 metabolic traits to efficiently explore the associations between transcriptomic and phenomic variations. Through a modified assembly strategy, 2,355 high-confidence novel sequences with total 1.9 Mb lengths were found absent within reference genome. Ten randomly selected novel sequences were fully validated with genomic PCR, including another two NBS_LRR candidates potentially affect flavonoids and disease-resistance. A simulation analysis suggested that the pan-transcriptome of the maize whole kernel is approaching a maximum value of 63,000 genes, and through developing two test-cross populations and surveying several most important yield traits, the dispensable genes were shown to contribute to heterosis. Novel perspectives and resources to discover maize quantitative trait variations were provided to better understand the kernel regulation networks and to enhance maize breeding.

Figure 1. ePAV candidates played key roles in distant-regulation.

(a) The ratio of local- (green) and distant- (orange) eQTLs among ePAV, non-ePAV and ePAV + non-ePAV together, expressed as percentages. (b) The effects of local eQTL were larger than those of distant eQTL both for ePAV (P = 7.05E-22; student test) and non-ePAV genes (P = 1.92E-135). The eQTL effects for ePAV genes were greater than those for non-ePAV genes in both local (P = 1.34E-18) and distant (P = 7.18E-56) types. (c) Top 10 GO enrichment terms in biological processes of ePAV (red) and non-ePAV (blue) are displayed. The left y-axis represents the percentage of genes belonging to each GO term. The colored circles and right y-axis represent the significance level (FDR). Red, blue and black colors means ePAV, non-ePAV and reference levels, respectively. The corresponding GO term description for each GO number could be available in Supplementary Table S1 online. (d) ePAV candidates as distant-eQTL affecting expression of Non-ePAV genes. “Diff” means the eQTL is located on a different chromosome with its regulating gene and “Same” represents both are located on the same chromosome (expressed as %). And even for the “Same” cases, the eQTLs tend to be located far away with their regulated targets (the colored rectangles represent the different distance windows, and the width represents the corresponding ratio.

Figure 2. ePAV candidates contributed to both maize cob color and various kinds of flavonoids.

(a) Manhattan plot of the association of three ePAV candidates, maize cob color and several flavonoids. Different shapes represent different traits, and points with different color represent different kinds of ePAV candidates: Blue: pericarp color1 (p1, GRMZM2G084799); Red: p2, another copy of R2R3 Myb-like transcription factor (GRMZM2G057027); Green: anthocyanidin 3-O-glucosyltransferase (GRMZM2G162755); Grey: other ePAVs. Black dashed horizontal line was the cut-off (P = 7.47E-5) of significant level. (b) The three ePAV candidates were also significantly associated with expression of related genes within maize flavonoid pathway. Nodes in red are the three ePAV candidates above, green nodes represent several identified genes located in the maize flavonoid pathway, purple nodes are other genes encode enzymes, light blue were other genes encoding non-enzyme proteins (such as transporters), and grey nodes had no annotation. The blue arrow edges link the ePAV candidates and its associated targets and the a3g links to itself meaning self-regulation in expression level, while thicker lines represented more significant associations.

Figure 3. Two novel NBS-LRR genes showed significant association with flavonoid metabolites and with expressed genes involved in flavonoid pathway.

Read distribution and predicted conserved domains of novel reference gene Unigene_678 (a) and Unigene_705 (b) and sequence alignments for all presence genotypes. (c) Q-Q plot of association mapping for different flavonoids. (d) The two novel NBS-LRR genes were also significantly associated with other genes with expression presence-absence variation. (e) Validation of the PAV of the two novel genes. Green represents consistency between experiment and prediction. Yellow means the gene was absent in our prediction but exists in the genome.

Figure 4. PAV status of novel genes and ePAV both correlated with heterosis of most yield-related traits.

(a) Correlation between mid-parent heterosis and the number of complementary novel genes exhibiting PAV between the parents of the F1 population. Six panels represent different yield-related traits. KWd: Kernel Width; KT: Kernel Thick; RPE: Rows Per Ear; EL: Ear length; CW: Cob Weight; KPE Kernels Per each row of Ear. (b) Boxplot in different colors represents different traits ordered by mid-parent heterosis (the left y axis). The points in red and black represent Pearson’s r² of correlation between mid-parent heterosis and the number of complementary novel genes and ePAVs showing presence-absence variation between parents of the F₁ population.