Triple coding in human SRD5A1 mRNA

Abstract

Background: Nucleotide sequence can be translated in three reading frames from 5' to 3' producing distinct protein products. Many examples of RNA translation in two reading frames (dual coding) have been identified so far.

Results: We report simultaneous translation of mRNA transcripts derived from SRD5A1 locus in all three reading frames that result in the synthesis of long proteins. This occurs due to initiation at three nearby AUG codons occurring in all three-reading frame. Only one of the three proteoforms contains the conserved catalytical domain of SDRD5A1 produced either from the second or the third AUG codon depending on the transcript. Paradoxically, ribosome profiling data and expression reporters indicate that the most efficient translation produces catalytically inactive proteoforms. While phylogenetic analysis suggests that the long triple decoding region is specific to primates, occurrence of nearby AUGs in all three reading frames is ancestral to placental mammals. This suggests that their evolutionary significance belongs to regulation of translation rather than biological role of their products. By analysing multiple publicly available ribosome profiling data and with gene expression assays carried out in different cellular environments, we show that relative expression of these proteoforms is mutually dependent and vary across environments supporting this conjecture. A remarkable feature of triple decoding is its resistance to indel mutations with apparent implications to clinical interpretation of genomic variants.

Conclusion: We argue for the importance of identification, characterisation and annotation of productive RNA translation irrespective of the presumed biological roles of the products of this translation.

Keywords: Overlapping genes; SRD5A1; gene annotation; protein synthesis; ribosome decision graphs; translation control; translation initiation; translon; uORF.

Figure 1. Aggregated Ribo-seq data track provides evidence for translation in three frames.

a Top: open reading frame (ORF) plot with AUG codons in green and stop codons in red. RefSeq and GENCODE SRD5A1 transcript annotations. Coding and non-coding exon regions are indicated with blue panels of different thickness, introns with thin lines. b Ribosome footprint density from aggregated data tracks in RiboCrypt for Ensembl transcript ENST00000274192.7 corresponding to Refseq transcript NM_001047. The line colouring matches the best supported reading frame in the ORF plot below; start codons shown as white bars, stop codons are black. c Frame distribution of Ribo-seq reads mapping to regions of relative overlap between the 3 translon’s ORFs. d Sub-codon footprint density where in-frame peaks for all 3 AUG codons indicate active translation initiation. Density peaks are coloured to match the best supported reading frame below. e Sub-codon coverage plots for two distinct clusters and all samples combined showing differential translation of translons between clusters.

Figure 2. Reporter assay validation of translation initiation at three AUG codons in the SRD5A1 5’ leader.

a An alignment of representative mammalian genomic sequences in the vicinity of the 3 AUGs in SRD5A1 5’ leader showing examples of AUG losses. b Schematic of the reporter where SNAP-tag encoding sequence was fused downstream to the SRD5A1 5’ leader sequence. Binding of SNAP-tag substrate allows visualisation directly in the protein gel. Gel on the right shows all three frames are translated. c Initiation at each AUG codon was monitored in the presence of selected nucleotide mutations, pAUG denotes perfect Kozak context. d eIF1/eIF5 overexpression to assess the stringency effect on start codon selection in SRD5A1 5’ leader in HEK293A and SUSA cell lines; plotted are normalised NanoLuc activities. Top: schematic of the construct where NanoLuc activity reports on each of the frames expression.

Figure 3. Representation of translation at three SRD5A1 RNA transcripts.

aThree RefSeq SRD5A1 transcript isoforms all sharing the first exon but with different AUG codon annotated as a start site. Above is a schematic of the ORF organisation for each coding exon with AUG codons in green and stop codons in red. b RDG representations for each of the three transcripts with putative translons indicated; right: AlphaFold2 structures for each of the predicted proteoforms. Regions of the translons highlighted in purple indicate encoded SRD5A1 enzymatic activity. Rainbow colouring shows N-terminus (blue) to C-terminus (red) orientation.