Disparate Rates of Molecular Evolution in Cospeciating Hosts
Total Page:16
File Type:pdf, Size:1020Kb
Disparate Rates of Molecular Evolution in quence data. In cases where different methods yielded different results, we re Cospeciating Hosts and Parasites tained all host and parasite trees for to pological comparison in order to deter Mark S. Hafner, Philip D. Sudman, Francis X. Villablanca, mine whether the inference of cospecia Theresa A Spradling, James W. Demastes, Steven A Nadler tion is warranted and, if so, whether the inference is sensitive to the method of DNA sequences for the gene encoding mitochondrial cytochrome oxidase I in a group of analysis. rodents (pocket gophers) and their ectoparasites (chewing ~ice) provide evidence for All analyses of the pocket gopher se cospeciation and reveal different rates of molecular evolution in the hosts and their quence data (using different models of parasites. The overall rate of nucleotide substitution (both silent and replacement chang DNA sequence evolution) yielded trees es) is approximately three times higher in lice, and the rate of synonymous substitution that were very similar in overall branch (based on analysis of fourfold degenerate sites) is approximately an order of magnitude ing structure. For example, phylogenetic greater in lice. The difference in synonymous substitution rate between lice and gophers analysis (I 1) of the COl sequence data for correlates with a difference of similar magnitude in generation times. pocket gophers yielded two most-parsimo nious trees of equal length (1423 steps). One of these trees (Fig. ZA) was topolog ically identical to the tree generated by a Chewing lice of the genera Geom:ydoecus in this host-parasite assemblage. We se maximum-likelihood analysis of the same and Thomom:ydoecus are obligate ectopara quenced and compared homologous regions data ( 12). The other most-parsimonious sites of pocket gophers (Fig. 1 ). Because the of the gene encoding the mitochondrial tree showed only minor differences ( 13) entire life cycle of these lice occurs exclu cytochrome c oxidase subunit I (COl) in from the tree shown in Fig. 2A. The sively in the fur of the host, and because both groups (8). Of the 379 nucleotides general structure of the gopher parsimony different host species rarely interact, each sequenced for each taxon, 134 positions tree (Fig. ZA) also was supported by species of louse is normally restricted to a were variable in pocket gophers and 178 Fitch-Margoliash (Fig. ZB) and neighbor single host species (1). As a result, there is positions were variable in chewing lice (Ta joining (14) analyses of genetic distances close correspondence between gopher tax ble 1). ( 12). Differences among the trees gener onomic boundaries and louse taxonomic The cospeciation hypothesis predicts ated by the parsimony, maximum-likeli boundaries (2). When viewed over large that the branching structure of the host hood, Fitch-Margoliash, and neighbor geographic and temporal scales, this re and parasite phylogenies will be similar to joining analyses of the pocket gopher data stricted distributional pattern of chewing a degree beyond that expected to occur by were judged nonsignificant by a likeli Lice on pocket gophers has resulted in phy chance. This prediction can be evaluated hood ratio test (15). Accordingly, all four logenetic histories of lice and gophers that statistically (3, 4). For any particular trees were retained for topological com are remarkably similar (3-5). host-parasite assemblage, confidence in parison with the parasite trees. The basic AJthough well-documented cases of the test of cospeciation can be no stronger structure of these trees and, in particular, host-parasite cospeciation are rare (3 , 6), than confidence in the phylogenies under relations within the genera Onhogeom:ys they are of interest because they permit comparison. Thus, it is essential that the and Geom:ys, also are supported by inde- comparative study of organisms with a long host and parasite phylogenies accurately history of parallel evolution. The temporal estimate the evolutionary history of each Table 1. Obselved percent of difference (mean component of parallel phylogenesis (in group. There are many methods for esti :t1 SD) in various elements oftheCOl nucleic acid sequence from pocket gophers and their acto which lineages of hosts and their parasites mating phylogenies from sequence data parasitic chewing lice. speciate repeatedly at approximately the (9), each of which uses a different model same time) permits examination of relative of nucleotide evolution and potentially Percent of sequence rates of evolution in the two groups by yields a phylogenetic hypothesis (a tree) difference (:t 1 SO)' comparison of the amount of change each that differs from that estimated with other has undergone during their parallel histo methods ( 10). To consider the effects of Gophers Lice ries. Because the life histories of hosts and different evolutionary models on our esti their parasites are often profoundly differ mates of phylogeny, we applied multiple Rrstposition 1.43(0.49) 2.09(0.62) transitions ent, studies of molecular evolution in host methods of analysis (II, 12) to our se F1rst position 0.00(0.25) 0.46(0.34) parasite assemblages can help answer a transversions broad spectrum of questions relating to the Seoond position 0 .24 (0.18) 0.38(0.37} possible effects of generation time, metabol transitions ic rate, and other life history parameters on / -~ Seoond position 0 .00(0.00) 0.16 (0.17} rates of mutation and evolutionary change. transversions ' Third position 8.74 (1.65) 9.59 (1 .51) We examined DNA sequence variation transitions in 14 species of pocket gophers and their Third position 5.01 (1 .67} 7.68 (1.99) chewing lice (7) to test for cospeciation and transversions to investigate rates of molecular evolution Total difference 15.64 (3.20) 20.75 (2.31) Silent nucleotide 14.75 (3.09) 17.66 (2.32) differences Replacement 0.89 (0.71) 2.67 (1.05) nucleotide Fig. 1. Chewrlg lice (Geomydoecus texanus, In differences set) are wingless insects that are obligate ecto· Amino acid 2.40 (1.83) 6.85(2.62) parasites of pocket gophers (Thomomys bottae is differences shown here). The entire life cycle of the chewing ·Meanand standard delliation based on alpaiiWise com louse occurs in the fur of these fossorial rodents. parisons. pendent phylogenetic studies based on and parasites (3), corroborates indepen maximum-likelihood distance matrices for morphology, allozymes, comparative im dent evidence for cospeciarion in several cospedating gophers and lice using Man munology, karyology, and nucleotide se genera of pocket gophers (Orchogeomys, tel's test (25), which showed a high ly sig quence data (3, 5, 16-18). Geomys, and Thomomys) and their lice nificant (P < 0.01) association between All analyses of the chewing Louse se (3-5). genetic distances in corresponding hosts quence data likewise yielded trees with Given evidence for cospecianon, it is and parasites. This test demonstrates that similar branching structures. Phyloge possible to test the null hypothesis that evolutionary rates in gophers and lice are netic analysis of the louse COl data yield pocket gophers and chewing lice have un significantly correlated, regardless of cree ed three most-parsimonious trees of equal dergone equivalent amounts of genetic dif structure. To tesr for equality of rates be length (4208 steps). One of these trees ferentiation during their parallel histories. tween gophers and lice, we compared max (Fig. 2A) was topologically identical to 1c is appropriate that th is test be restricted imum-likelihood branch lengths for all pos the tree generated by a maximum-likeli to hosts and parasites that have cospeciated, sible combinations of cospeciating raxa (24, hood analysis of the same data. The two because time since divergence can be as 26). In all cases, Wilcoxon sign-rank tests remaining parsimony trees showed only sumed to be equal only for host-parasite showed thal louse branches were signifi minor differences (involving one spec1es pairs that show cospeciarion (9 host-para cantly longer than gopher branches (P < in each case) from the tree in Fig. 2A sire pairs in Fig. 2A and 10 host-parasite 0.003 in each case). Given this significant (19). The Fitch-Margoliash analysis (Fig. pairs in Fig. 2B) (24). We first compared difference, we used Model II regression 28) and the neighbor-joining analysis (ZO) yielded louse trees very similar to those generated by the parsimony analy A sis. Differences among the trees generated o. ~ ......... a.~ 0. chetrlli 0 0 0 0 0 0 00 •• 00 0 0 0 0 0 0 G. chetrlli by the parsimony, maximum-likelihood, 0 . IJIId«wocd 0 0 0 0 0 0 0 0 0 0 0 0 0 0 G. AfUtl Fitch-Margoliash, and neighbor-joining L-...;- O. c::e~ ····· ·· ·····- G. ~ analyses of the louse data were judged '--.......,\~- 0. hllplcM ....... ··-··-····· G. dvpW ---.,;::--' nonsignificant by a likelihood ratio tesr 1----- Z lrt;hoput ....•........... G. trlt:ht:f1l (15). Accordingly, all five louse trees '-------- P. bcilltf· ·················-· · G. n.llerl were retained for topological comparison c. ~ ··············· G.....,_ -----, c.mem.trW ·. 0 •• G. «:ff.Jool with the host trees. The basic structure of ...--.:'!:,...-- G. penotWIIII :\···········'· G.,_.. the louse trees and, in particular, relations '------~ G.br~Mc~p~J ...:·., ....,.::". .... G. ewing/ among lice hosted by species of Orchogeo G. bclrNtiu. (a) .:·.,..-: G. ~ mys and Geomys , also are supported by I G.~(b} •/::•._. ...• a.~ independent phylogenetic studies of alloz ..........\.~(;_= ymes (3, 5, 21). K T. bo111N ./..,.,·::·....••....• T. m/rto( • The COMPONENT program (22) de '-- - ----- -!'.';;- T. lll~ ..- ·..............T. b8ttWnN 1-----------' termined if the fit between observed par asite and host trees was significantly bet ter than the fit between the parasite tree ~. 8 and trees drawn at random from the set of 0. /NJttlffdll oo .... oo. G. C05tllli:>IUsll all possible host trees (4). For each of 20 0 . chetrlli 0 0 0 0 0 0 0 0 0 00 0 0 0 --.- G.