Massively parallel sequencing of the polyadenylated transcriptome of C. elegans

Abstract

Using massively parallel sequencing by synthesis methods, we have surveyed the polyA+ transcripts from four stages of the nematode Caenorhabditis elegans to an unprecedented depth. Using novel statistical approaches, we evaluated the coverage of annotated features of the genome and of candidate processed transcripts, including splice junctions, trans-spliced leader sequences, and polyadenylation tracts. The data provide experimental support for >85% of the annotated protein-coding transcripts in WormBase (WS170) and confirm additional details of processing. For example, the total number of confirmed splice junctions was raised from 70,911 to over 98,000. The data also suggest thousands of modifications to WormBase annotations and identify new spliced junctions and genes not part of any WormBase annotation, including at least 80 putative genes not found in any of three predicted gene sets. The quantitative nature of the data also suggests that mRNA levels may be measured by this approach with unparalleled precision. Although most sequences align with protein-coding genes, a small fraction falls in introns and intergenic regions. One notable region on the X chromosome encodes a noncoding transcript of >10 kb localized to somatic nuclei.

Figure 1.

Relative coverage of WormBase genes by position and strand. (A) Results of analyzing 7571 full-length transcripts, normalized for length and depth of coverage, show almost uniform coverage along the body of the gene, with a pronounced overrepresentation at the 5′ end and decreasing depth of coverage at the 3′ end. (YA) Young adult. (B) Average depth of coverage at the 5′ end by strand, measured in bases from the start of the transcript. This plot confirms the overrepresentation at the 5′ end and demonstrates that it is almost entirely restricted to the coding strand. (C) Average depth of coverage at the 3′ end, by strand, measured from the transcript end. Coverage on both strands falls approaching the 3′ end, with loss of the sense (s) strand coverage followed by decrease of the antisense (as) to less than half the average depth of coverage. Only data for L2 and L3 are shown for clarity.

Figure 2.

Heat map of the 2000 most variably expressed transcripts in the four C. elegans stages. The 15,000 most abundantly expressed transcripts, based on their mean expression, were evaluated for their variation in expression, based on the ratio of the variance to the mean. We arbitrarily selected the 2000 most variable transcripts (Supplemental Table S8), which were then clustered based on their coverage levels as reflected in the dcpm statistic. The columns represent the four stages, L2, L3, L4 and young adult (YA), respectively, and each row represents a different transcript. Transcripts absent in L2 and abundant in L4 and young adult include the major sperm protein (msp-), vitellogenin (vit-), and other transcripts related to the germline.