Supporting Material Response to: Alternative splicing at NAGNAG acceptors: Simply noise or noise and more?

In their letter Hiller et al. mention the conservation test that was performed in [1] and the more recent results of Akerman and Mandel-Gutfreund [2]. To explain our interpretation of these results we first briefly describe our rough hypothesis for the origin of the bulk of NAGNAG splice variations. We believe that in general the splicing machinery splices invariantly at the first AG that follows the polypyrimidine tract. This is supported by the fact that the large majority of invariant NAGNAGs splice indeed at the first NAG. Variant NAGNAGs can occur when, due to a combination of selection on the protein level and neutral evolution, a second NAG is introduced immediately downstream of the splice site. Especially when the major splice site undergoes some mutations that weaken its affinity for the spliceosome this may lead to a situation in which the spliceosome will sometimes ’slip’ beyond the first NAG and splice at the second (exonic) NAG. In cases where this variation is not strongly deleterious, this ’slippery’ splice site may persist for a significant amount of evolutionary time. In short, we hypothesize that most variant NAGNAGs have their major splice site at the first NAG. Under this hypothesis these NAGNAGs should show roughly the same conservation as invariant splice sites that splice at the first NAG. In contrast, invariant NAGNAGs that always splice at the second NAG would be expected to have a lower amount of selection on the first (always intronic) NAG. In [1] Hiller et al. took variant NAGNAGs, i.e. those that show splice variation, and classified them as ’intronic’ (extra NAG in the intron) or exonic (extra NAG in the exon) based on the RefSeq annotation of the transcript. For intronic NAGNAGs they observed a somewhat higher level of conservation of the A and G nucleotides at positions -5 and -4 than the A and G nucleotides at -5 and -4 from transcripts that do not have an AG dinucleotide at those positions. Such enhanced conservation was not observed for the class of exonic NAGNAGs. We believe that these observations are precisely consistent with our hypothesis above: for many of these ‘intronic’ variant NAGNAGs the major splice site is really at the first NAG, but the RefSeq sequence happens to contain the minor form which splices at the second NAG. For these intronic NAGNAGs there will be a strong selection on maintaining the major splice site, i.e. the A and G at positions -5/-4. In contrast, the ‘exonic’ NAGNAGs have their splice site (almost always correctly) assigned to the first NAG, and these will not show more conservation than invariant NAGNAGs that splice at the first NAG. We can in fact validate this explanation using our splice site selection model [3]. Our model predicts that we can distinguish NAGNAGs in which the first NAG is the major splice site, from those in which the second NAG is the major splice site, by the likelihoods of the sequences around the two putative acceptors for the acceptor site weigh matrix (ASWM). If the first NAG has much higher likelihood than the second we predict it to be the major splice site. Conversely, if the second NAG has much higher likelihood than the first we predict the second NAG to be the major splice site. We collected all NAGNAG acceptors from our data [3] and calculated, for each, the difference in log-