The Genom of Homo sapiens.pdf

298 ROGOZIN ET AL.Dm Ag Hs Ce Sc Sp At Pf147156137194184477798735152471971200146307-87933379524471386279224835-Figure 4. Parsimonious evolutionary scenario of introngain/loss for the most likely topology of the eukaryotic phylogenetictree. Intron gains and losses are mapped to each speciesand each internal branch. The designations are as in Fig. 3. Thespecies abbreviations are as in Fig. 1, with the addition ofAnopheles gambiae (Ag) and Plasmodium falciparum (Pf).ancestral introns seem to have survived in plants and animals,but have been lost in yeasts, nematodes, andarthropods.We examined the evolutionary dynamics of introns ingreater detail by using phylogenetic analysis. To this end,intron positions were represented as a data matrix of intronpresence/absence (encoded as 1/0) and the matricesfor all analyzed KOGs were concatenated to produce onealignment, which consisted of 16,577 columns of “ones”and “zeroes” (7,236 columns in conserved portions ofalignments). In this form, the intron presence/absencedata are conducive to evolutionary parsimony analysisand, of the existing parsimony approaches, Dollo parsimonyseems to be most appropriate in this case (Nei andKumar 2000). The Dollo parsimony tree that we reconstructedfrom the matrix of intron presence/absence obviouslydid not mimic the species tree, with humans andArabidopsis forming a strongly supported cluster embeddedwithin the metazoan clade, and another anomalouscluster formed by yeast S. pombe and Plasmodium (notshown). Other phylogenetic approaches, including unweightedmaximum parsimony and several distancemethods, which were applied to the data of Table 2, reproducedthe same, incorrect topology (not shown). Thetopology of these trees supported the notion, already suggestedby the pair-wise comparisons summarized inTable 2, that ancestral introns have been, to a large extent,conserved in plants and vertebrates, but have been extensivelyeliminated in fungi, nematodes, and arthropods.Clustering of Plasmodium with S. pombe, to the exclusionof the other yeast species, S. cerevisiae, seems to bedue to the fact that Plasmodium shared approximately asmany introns with S. pombe as with worm and insects, butthe total number of introns in S. pombe was substantiallyless than in animals (Table 2). Thus, conservaton of intronpositions does not seem to be a good source of informationfor inferring phylogenetic relationships at longevolutionary distances.Having shown that evolution of introns in eukaryoticgenes did not follow the species tree, we applied Dolloparsimony in the opposite direction: Given a species treetopology, construct the most parsimonious scenario forthe evolution of gene structure, i.e., the distribution of introngain and loss events over the tree branches. This approachis completely analogous to the construction of thescenario for gene gain and loss described above.The resulting scenario suggests an intron-rich ancestorfor the crown group, with limited intron loss in the animalancestor, but massive losses in yeasts (particularly, S.cerevisiae), worm, and insects (Fig. 4). The differences inthe relative rates of intron gain and loss in the terminalbranches are remarkable: There is a huge excess of gainsover losses in humans and S. pombe, and equally obviousexcess of losses in insects and S. cerevisiae, whereas C.elegans shows nearly identical numbers of gains andlosses. All introns shared by Plasmodium and any of thecrown-group species (at least 210 as shown by analysis ofcomplete alignments) are assigned to the last common ancestorof alveolates and the crown group, which livedsome 1.5–2.0 billion years ago (Hedges 2002). At present,loss of ancestral introns in Plasmodium cannot bedocumented because Plasmodium is the outgroup to thecrown-group species; neither can losses be assigned tothe internal branch that leads to the ancestor of the crowngroup. Hence, we produced a conservative estimate of thenumber of the most ancient introns in the analyzed geneset, which is likely to be a substantial underestimate,given that Plasmodium is a parasite with a highly degradedgenome and low intron density. Sequencing andanalysis of genomes of other early branching eukaryotesis expected to substantially increase the number of intronsthat have survived since the dawn of eukaryotic evolution.The present analysis pushes the origin of numerousspliceosomal introns back to the stage of eukaryotic evolution,1.5–2.0 billion years ago, which precedes the originof multicellularity, perhaps to the very point of originof eukaryotes. Furthermore, as many as 25–30% of the intronsin vertebrates and plants are apparently inheritedfrom the common ancestor of the crown group. Othercrown-group lineages experienced massive loss of introns.Intron elimination had been extensive already inthe common ancestor of yeasts but was particularly catastrophicin S. cerevisiae, which apparently lost all ancestralintrons. Those few introns that this organism doeshave are not shared with other species and probably havebeen regained secondarily (Table 2 and Fig. 4). More unexpectedly,insects and, particularly, the nematode havealso lost the great majority of the ancestral introns, withthe latter having secondarily acquired numerous new introns(Fig. 4).Why have so many ancestral introns survived nearly 2billion years of evolution? One explanation would involvefunctional roles of introns in specific positions, butthere seems to be no evidence in support of such a possibility.The second and more likely explanation is thatelimination of introns, particularly in highly conservedgenes, which are often essential (such as those included inthe present study), would often lead to gene inactivationand lethality. We found that ancient introns tend to be locatedin more highly conserved portions of genes than recentlyinserted ones (data not shown), which is compati-143