Heritable temporal gene expression patterns correlate with metabolomic seed content in developing hexaploid oat seed

Abstract

Oat ranks sixth in world cereal production and has a higher content of health-promoting compounds compared with other cereals. However, there is neither a robust oat reference genome nor transcriptome. Using deeply sequenced full-length mRNA libraries of oat cultivar Ogle-C, a de novo high-quality and comprehensive oat seed transcriptome was assembled. With this reference transcriptome and QuantSeq 3' mRNA sequencing, gene expression was quantified during seed development from 22 diverse lines across six time points. Transcript expression showed higher correlations between adjacent time points. Based on differentially expressed genes, we identified 22 major temporal co-expression (TCoE) patterns of gene expression and revealed enriched gene ontology biological processes. Within each TCoE set, highly correlated transcripts, putatively commonly affected by genetic background, were clustered and termed genetic co-expression (GCoE) sets. Seventeen of the 22 TCoE sets had GCoE sets with median heritabilities higher than 0.50, and these heritability estimates were much higher than that estimated from permutation analysis, with no divergence observed in cluster sizes between permutation and non-permutation analyses. Linear regression between 634 metabolites from mature seeds and the PC1 score of each of the GCoE sets showed significantly lower p-values than permutation analysis. Temporal expression patterns of oat avenanthramides and lipid biosynthetic genes were concordant with previous studies of avenanthramide biosynthetic enzyme activity and lipid accumulation. This study expands our understanding of physiological processes that occur during oat seed maturation and provides plant breeders the means to change oat seed composition through targeted manipulation of key pathways.

Keywords: heritability; oat; temporal gene expression; transcriptome assembly.

Figure 1

Results of aligning the assembled oat seed transcriptome against reference proteomes of oat relatives, scaffolds of the hexaploid oat genome v1.0 and the UniRef100.

Figure 2

PCA plot of 326 samples with more than 0.5 million mapped reads based on the 500 transcripts with highest variance.

Figure 3

Pairwise correlation of transcript expression between time points (a) and numbers of differentially expressed transcripts between adjacent time points (b). In (b), total stacked bar height indicates the number of transcripts whose transcription level changed significantly over the time interval. Coloured stack components indicate transcripts with significant change in an interval in common with a previous interval. For example, in red, 4805 transcripts significantly changed expression over both the 13–18 DAA and 8–13 DAA intervals.

Figure 4

Top 20 expression patterns plus two main expression patterns of 13DAA (three‐steps‐up‐at‐13DAA, Top‐31) and 18DAA (two‐steps‐up‐at‐18DAA, Top‐24). Expression pattern plots are named, and ranking numbers and total number of transcripts of each expression patterns are given at the bottom. Median gene expression profiles of individual transcripts across the 22 oat lines were depicted in grey lines, and average expression profiles for each pattern are depicted in blue (if up‐regulated) or red (if down‐regulated).

Figure 5

Box plots of heritabilities estimated from GCoE sets of the 22 TCoE sets (red) against box plots of heritabilities estimated from permuted TCoE sets (blue). The TCoE sets were ordered as in Figure 4.

Figure 6

Transcript expression patterns of oat avenanthramides (a) and fatty acids (b) biosynthetic genes based on individual oat lines. Transcript expression values of individual samples were depicted in coloured dots, and a LOWESS (Locally Weighted Scatterplot Smoothing) curve through all expression values of each genotype was drawn in coloured smooth lines. Different colours represent different oat lines. Oat lines with too many missing values were excluded.