Intron Size, Abundance, and Distribution within ...

1 Intron Size, abundance , and Distribution within untranslated regions of GenesXin Hong, Douglas G. Scofield, and Michael LynchDepartment of Biology, indiana UniversityMost research concerning the evolution of introns has largely considered introns within coding sequences (CDSs), withoutregard for introns located within untranslated regions (UTRs) of genes. Here, we directly determined Intron size, abun-dance, and Distribution in UTRs of genes using full-length cDNA libraries and complete genome sequences for four spe-cies,Arabidopsis thaliana,Drosophila melanogaster, human, and mouse. Overall Intron occupancy (introns/exon kbp) islower in 5#UTRs than CDSs, but Intron density ( Intron occupancy in regions containing introns) tends to be higher in 5#UTRs than in CDSs. Introns in 5#UTRs are roughly twice as large as introns in CDSs, and there is a sharp drop in intronsize at the 5#UTR-CDS boundary.

2 We propose a mechanistic explanation for the existence of selection for larger intronsize in 5#UTRs, and outline several implications of this hypothesis. We found introns to be randomly distributed within 5#UTRs, so long as a minimum required exon size was assumed. Introns in 3#UTRs were much less abundant than in 5#UTRs. Though this was expected for human and mouse that have Intron -dependent nonsense-mediated decay (NMD)pathways that discourage the presence of introns within the 3#UTR, it was also true forA. thalianaandD. melanogaster,which may lack Intron -dependent NMD. Our findings have several implications for theories of Intron evolution andgenome evolution in since the unexpected discovery of introns (Bergetet al. 1977; Chow et al. 1977; Evans et al. 1977; Goldberget al. 1977), there has been intense debate about their origins,stability, and adaptive significance.

3 Much early attention tothe evolution of introns focused on the timing of their ori-gins. The introns-early or exon theory of genes proposesthat introns are ancient, and that early diversification ofgenes in the progenote, the genome ancestral to all prokar-yotes and eukaryotes, was greatly accelerated by the shuf-fling of exons at Intron -induced boundaries (Blake 1978;Gilbert 1978, 1987). The subsequent loss of introns in pro-karyotes alone then occurred through selection for morestreamlined genes and genomes (Doolittle 1978; Darnelland Doolittle 1986; Senapathy 1986; Roy and Gilbert2005). The introns-late hypothesis maintains that intronsappeared later and at random in early eukaryotic genomes,and that any adaptive role in gene evolution was gained fol-lowing insertion (Orgel and Crick 1980; Cavalier-Smith1985; Palmer and Logsdon 1991; Frugoli et al.)

4 1998). Morerecent theory has emphasized mutational and population-genetic processes that are likely to govern the establishmentand retention of introns (Lynch 2002; Lynch and Richardson2002; Lynch and Kewalramani 2003). A so-called syn-thetic introns-early theory proposes the coexistence of an-cient introns situated to promote exon shuffling, with morerecently gained introns conforming to introns-late expect-ations (Fedorov et al. 2001; de Souza 2003; Fedorova andFedorov 2003).Hypotheses addressing the abundance and locations ofintrons within the protein-coding sequence (CDS) of tran-scripts have figured prominently in early as well as morerecent theories of Intron establishment, maintenance andproliferation. In the introns-late view, initial Intron posi-tions are largely random within transcripts or occur at proto-splice sites that carry short sequences similar toconserved exon sequences flanking introns but that are oth-erwise context free (Cavalier-Smith 1991; Dibb 1991; Choand Doolittle 1997).

5 Intron positions may thus be purelyfortuitous or involve selection for features that influencetranscription or translation, for example, regulation of tran-scription initiation (Fong and Zhou 2001; Le Hir et ), efficiency of mRNA export (Luo and Reed 1999),enhancement of splicing efficiency (Berget 1995; Nissim-Rafinia and Kerem 2002), chromatin assembly (Lauderdaleand Stein 1992; Liu et al. 1995), and recognition of prematuretermination codons via nonsense-mediated decay (NMD)(Kim et al. 2001; Lynch and Kewalramani 2003; Maquat2004a). Under the introns-early view, ancient genes andexons consisted of relatively short polypeptide sequencesof limited secondary or tertiary structural extent. Rearrange-ment of these discrete protein modules at Intron bound-aries are hypothesized to have been the primary means bywhich early proteins acquired diverse structures and functions(Darnell and Doolittle 1986; Gilbert 1987; Gilbert et ).

6 Because introns-early hypotheses predict correlationsbetween gene and protein structure resulting from theserearrangements (de Souza et al. 1996), predictions concern-ing introns found outside the CDS are less 5#and 3# untranslated regions (UTRs) that bracketCDSs are fundamental structural and regulatory regions ofeukaryotic genes (Ptashne and Gann 2001; Larizza et ; Mignone et al. 2002; Wilkie et al. 2003). UTRsare known to contain large numbers of introns (Pesoleet al. 2001), yet Intron abundance and Distribution in UTRshave received little study, and there is a lack of hypothesesspecifically addressing the evolution of introns withinUTRs. Our goal here is to begin to address these gaps inour knowledge. The few summary data available (Pesoleet al. 2001) indicate that 22% 26% of metazoan 5#UTRscarry introns, with lower frequencies in plants (14%) andfungi (5%).

7 The observation that;4 53fewer 3#UTRscarry introns in these taxa is especially curious given that, within taxa, 3#UTRs are generally 2 33longer than 5#UTRs and would thus be expected to form larger targetsfor random Intron insertion. Data from Pesole et al. alsosuggest the possibility of a strong barrier against Intron in 5#UTRs for all taxa, and a similar but lessstringent barrier for introns in 3#UTRs that does not appearto be as consistent among taxa. These patterns stand inKey words: untranslated region, Intron , genome : Biol. (12):2392 2404. 2006 Access publication September 15, 2006 The Author 2006. Published by Oxford University Press on behalf ofthe Society for Molecular Biology and Evolution. All rights permissions, please e-mail: contrast to those for the CDS in most multicellularspecies, the vast majority of which carry multiple introns(Lynch and Conery 2003).

8 Additional differences between UTRs and CDSs mayaffect Intron size, abundance , and Distribution . UTR regionsare under less stringent substitutional constraint (vs. nonsy-nonymous sites) than CDSs, and have a higher indel frequencyand length heterogeneity (Graur and Li 2000; Larizza et ; Shabalina et al. 2004). As a result, introns in UTRsmay experience less stabilizing selection for some traits thanintrons in CDSs, in which case sharp discontinuities in introntraits at CDS-UTR boundaries may be expected. Furthermore,so-called ancient CDS introns that were shared by multipleeukaryotic lineages tended to be found in more conservedregions of the CDS (Rogozin et al. 2003); the more dynamicnature of the UTRs may thus promote Intron loss and resultin lower Intron abundance than in CDSs. An additional con-sideration is that Intron distributions that may promote CDSquality via NMD (Lynch and Kewalramani 2003) may pro-vide no benefit within the 5#UTR, though they may havedirect effects on Intron abundance and Distribution in thedownstream 3#UTR (Nagy and Maquat 1998).

9 With the explosive growth in genome sequencingprojects, a variety of computational methods have beendeveloped to indirectly infer gene structure from genomesequence data, including the detection of Intron exon bound-aries; see Zhang (2002) for a comprehensive review. Thesemethods have reached a high level of performance such thatthey can recognize the large majority of Intron exon bound-aries within the CDS. However, despite significant recentadvances, recognition of Intron exon boundaries withinUTRs, which lack the strong contextual signal provided bya valid open reading frame, is considerably more error prone(Eden and Brunak 2004). The recent availability of large li-braries of full-length cDNA transcripts, when rigorouslyaligned to complete genome sequences for the same species,allows for the direct determination of Intron exon structurewithin UTRs.

10 As a result, we are able to directly examineintron size, abundance , and Distribution in UTRs of thou-sands of transcripts from each of four species,Drosophilamelanogaster,Arabidops is thaliana, human, and and MethodsData SourcesWe obtained publicly available genome and full-length cDNA sequence data forD. melanogaster,A. thali-ana, human, and mouse (table 1). Boundaries betweenUTRs and CDSs in full-length cDNA sequences were de-termined using annotations from GenBank (D. melanogasterandA. thaliana) and the Mammalian Genome Consortium(human and mouse). Intron PositionsIntron positions were determined through the recogni-tion of gaps in alignment of full-length cDNA transcriptswith genomic sequences. In brief, for a single full-lengthcDNA aligned against a contiguous stretch of genomic se-quence, exons were determined as proximal blocks of ho-mologous sequence alignment between full-length cDNAand genomic sequence, whereas introns were determined asgaps between exons consisting solely of genomic first cleaned the full-length cDNA libraries byremoving transcripts with inconsistent annotation and in-complete CDSs.