The code within the code: microRNAs target coding regions

Abstract

The coding sequence of a protein must contain the information required for the canonical amino acid sequence. However, the redundancy of the genetic code creates potential for embedding other types of information within coding regions as well. In a genome-wide computational screen for functional motifs within coding regions based on evolutionary conservation, highly conserved motifs included some expected motifs, some novel motifs and coding region target sites for known microRNAs, which are generally presumed to target 3' untranslated regions (UTRs) (www.SiteSifter.org). We report here an analysis of published proteomics experiments that further support a functional role for coding region microRNA binding sites, though the effects are weaker than for sites in the 3' UTR. We also demonstrate a positional bias with greater conservation for sites at the end of the coding region, and the beginning and end of the 3' UTR. An increased effectiveness of microRNA binding sites at the 3' end of transcripts could reflect proximity to the poly(A) tail or interactions with the 5' terminal 7mGpppN "cap", which is physically adjacent to this region once the message is circularized. The effectiveness of 3' UTR sites could reflect a cooperative role for RNA binding proteins. Finally, increased microRNA conservation near the stop codon suggests to us the possible involvement of proteins that execute nonsense-mediated decay, since this process is activated by tagging of termination codons with factors that induce transcript degradation.

Figure 1. Coding region microRNA target sites are functional, but less effective than 3’ UTR sites

Protein levels were determined for 2634 proteins in HeLa cells that were either mock-treated or treated with a LNA that targets the let-7 microRNA using pulse SILAC labeling followed by mass spectrometry. The cumulative fraction of proteins with a given ratio of protein level in mock-treated versus let-7 LNA-treated cells is plotted for proteins with no target site, a 3’ UTR target site, an evolutionarily conserved 3’ UTR target site, a coding region site or an evolutionarily conserved coding region site. The number of genomes in which a site was conserved was summed across all occurrences of the let-7 target site in each region, and the highest-scoring 25% were designated as "conserved." In order to define error bars for the no site distribution, a subset of 649 proteins without a let-7 binding site was randomly selected 100 times and the 5^th percentile highest and lowest values were plotted as the error. Coding region target sites were clearly functional, as the distribution of effect sizes is statistically significantly different from the distribution for proteins with no recognition site (Wilcoxon p-value < 10⁻⁵). However, the presence of a recognition site within the coding region confers less repression than the presence of a 3’ UTR site on average. For both the coding region and the 3’ UTR, evolutionarily conserved sites tend to confer a stronger repressive effect than all sites.

Figure 2. Target site conservation by location within the coding region or 3’ UTR

The location of microRNA binding sites was determined for all highly conserved microRNA families (as listed in TargetScan version 5.0) using a modified version of SiteSifter (www.sitesifter.org). Target sites were grouped based on their location within the coding region or 3’ UTR such that target sites within the first 10% of the coding region or 3’ UTR were considered together, target sites within the second 10% of the coding region or 3’ UTR were considered together, etc. For target sites within each group, the fraction that are well-conserved (black bars), defined as 14+ genomes for coding region target sites and 13+ for 3'UTR target sites, and the fraction that are poorly-conserved (empty bars), defined as the remainder of sites, are plotted. Conservation was determined based on a 17-genome alignment available from the University of California at Santa Cruz. A statistically significant enrichment for evolutionarily conserved sites is observed at the end of the coding region and at the beginning and end of the 3’ UTR. In the middle of the 3’ UTR, sites are more likely not to be evolutionarily conserved. Statistically significant differences are indicated with asterisks and reflect a p value less than 0.01 using a chi-squared test after correcting for multiple hypothesis testing.