Abstract
Alternative splicing (AS) is a key modulator of development in many eukaryotic organisms. In plants, alternative splice forms of non-coding RNAs (ncRNAs) are known to modulate flowering time in Arabidopsis and fertility in rice. Here we demonstrate that alternative splicing of coding and long non-coding RNAs occurs during rice seed development by comparing AS in immature seeds vs. embryo and endosperm of mature seeds. Based on computational predictions of AS events determined from a Bayesian analysis of junction counts of RNA-seq datasets, differential splicing of protein-coding, and non-coding RNAs was determined. In contrast to roots, leaves, flowers, buds, and reproductive meristems, developing seeds had 5.8-57 times more predicted AS. Primers designed to span introns and exons were used to detect AS events predicted by rMATs in cDNA derived from early (milk) seed, embryo, and endosperm. Comparing milk seed vs. mature embryo and endosperm, AS of MORC7 (a gene implicated in epigenetic gene silencing), was markedly different. Long non-coding RNAs (lncRNAs) also underwent AS during the transition from milk seed to mature embryo and endosperm, with a complex gene structure, and were more extensively processed than predicted by current genome annotation. Exon retention of lncRNAs was enhanced in embryos. Searching all 5,515 lncRNAs in the NCBI genome annotation uncovered gene families based on highly conserved regions shared by groups of 3-35 lncRNAs. The homologies to other lncRNAs, as well as homologies to coding sequences, and the genomic context of lncRNAs provide inroads for functional analysis of multi-exonic lncRNAs that can be extensively processed during seed development.