New Maximum Likelihood Estimators for Eukaryotic Intron Evolution

The evolution of spliceosomal introns remains poorly understood. Although many approaches have been used to infer intron evolution from the patterns of intron position conservation, the results to date have been contradictory. In this paper, we address the problem using a novel maximum likelihood method, which allows estimation of the frequency of intron insertion target sites, together with the rates of intron gain and loss. We analyzed the pattern of 10,044 introns (7,221 intron positions) in the conserved regions of 684 sets of orthologs from seven eukaryotes. We determined that there is an average of one target site per 11.86 base pairs (bp) (95% confidence interval, 9.27 to 14.39 bp). In addition, our results showed that: (i) overall intron gains are ~25% greater than intron losses, although specific patterns vary with time and lineage; (ii) parallel gains account for ~18.5% of shared intron positions; and (iii) reacquisition following loss accounts for ~0.5% of all intron positions. Our results should assist in resolving the long-standing problem of inferring the evolution of spliceosomal introns.Synopsis: When did spliceosomal introns originate, and what is their role? These questions are the central subject of the introns-early versus introns-late debate. Inference of intron evolution from the pattern of intron position conservation is vital for resolving this debate. So far, different methods of two approaches, maximum parsimony (MP) and maximum likelihood (ML), have been developed, but the results are contradictory. The differences between previous ML results are due predominantly to differing assumptions concerning the frequency of target sites for intron insertion. This paper describes a new ML method that treats this frequency as a parameter requiring optimization. Using the pattern of intron position in conserved regions of 684 clusters of gene orthologs from seven eukaryotes, the authors found that, on average, there is one target site per ~12 base pairs. The results of intron evolution inferred using this optimal frequency are more definitive than previous ML results. Since the ML method is preferred to the MP one for large datasets, the current results should be the most reliable ones to date. The results show that during the course of evolution there have been slightly more intron gains than losses, and thus they favor introns-late. These results should shed new light on our understanding of intron evolution.