104 



W. WUSTER ETAL. 



monophyly of the B. asper-atrox complex, i.e., that the Antillean 

 populations originate from within the B. asper-atrox complex; (iii) 

 non-monophyly of the South American B. atrox complex, i.e., that 

 the Antillean species originate from within the cis-Andean B. atrox 

 complex, paralleling the phylogeography of Corallus (Henderson & 

 Hedges, 1995); and (iv) monophyly of B. caribbaeus, B. lanceolatus, 

 B. asper and northern Venezuelan populations of the B. asper-atrox 

 complex to the exclusion of the cis-Andean B. atrox complex, as 

 implied by the classification of Sandner Montilla (1990). We used 

 Wilcoxon signed-ranks (WSR) tests (Templeton tests - Templeton, 

 1983) to compare the optimal MP tree and MP trees depicting 

 alternative hypotheses, and Shimodaira-Hasegawa (SH) tests 

 (Shimodaira & Hasegawa, 1999) to compare the corresponding ML 

 trees. 



RESULTS 



We aligned a total of 1401 b.p. of mtDNA sequence information 

 (ND4: 693 b.p.; cytfr: 708 b.p.). The sequences included no indels or 

 stop or other nonsense codons, and contained the usual bias towards 

 transitions and substitutions concentrated into third codon positions 

 typical of mitochondrial DNA. We conclude that our sequences 

 represent mtDNA rather than nuclear insertions. Samples are listed 

 in Appendix 1. The 100,000 random trees generated a skewness 

 statistic of gl=-0.599403, rejecting the null hypothesis that the data 

 contain no significant phylogenetic signal (P < 0.01; Hillis and 

 Huelsenbeck, 1992). 



Levels of sequence divergence among the taxa included ranged 

 from 0.3% to 13.65% (unadjusted p-distance). Bothrops caribbaeus 

 and B. lanceolatus differ from each other by 4.3%, and from the B. 

 asper-atrox group by an average of 5.77% and 6. 15% respectively, 

 with an average divergence of 5.9% when the Antillean haplotypes 

 are treated as a single clade. Levels of sequence divergence within 

 the B. asper-atrox clade range from 0.3% to 5.5% 



The MP analysis resulted in a single most parsimonious tree of 

 1030 steps (CI 0.5398; HI 0.4602; RI 0.6465). In this tree, the two 

 Antillean taxa formed a clade, which in turn forms the sister clade of 

 all samples of the Bothrops asper-atrox complex (Fig. 1 ). 



The MODELTEST software identified the GTR+I+G model, a 

 submodel of the general time-reversible model (Yang etal., 1994) as 

 optimal for the data at hand. A ML tree was constructed using the 

 parameters calculated by MODELTEST, and the parameters were 

 recalculated from the resulting tree, and used in a further ML search, 

 which resulted in a tree with the likelihood score -ln(L)= 6652.69 1 22. 



Further estimates of sequence evolution parameters did not result 

 in any change of parameter values or tree likelihood score (Fig. 1). 

 The MP and ML trees differ only in branching order within the cis- 

 Andean B. atrox complex, and in the relative position of the B. 

 jararacussu-brazili clade and B. punctatus. 



The results of our tests of alternative tree topologies are shown in 



Table 1. Neither the WSR nor the SH test significantly reject the 

 possibility that the two Antillean species may be the result of 

 separate colonizations of the Lesser Antilles, although the result of 

 the SH test was almost significant. They do, however, significantly 

 reject the hypothesis that the Antillean species originate from within 

 the cis-Andean radiation of the B. asper-atrox complex, and also 

 reject Sandner Montilla's suggestion of conspecificity between B. 

 lanceolatus, B. caribbaeus, B. asper and northern Venezuelan 

 Bothrops, to the exclusion of other South American populations of 

 the B. atrox group. 



DISCUSSION 



Our results confirm the position of the Antillean species of Bothrops 

 as the sister clade of the Bothrops asper-atrox complex, as suggested 

 by Salomao ef a/. (1997, 1999) and Wustere/a/. (1997, 1999). The 

 monophyly of the Antillean taxa is supported by high bootstrap and 

 Bremer support values, although a tree supporting this arrangement 

 is not significantly longer than the optimal tree constrained not to 

 include this clade. 



The high level of sequence divergence between the Antillean 

 Bothrops and their mainland relatives (5.9%) is consistent with a 

 lineage split dating back to the Miocene or earliest Pliocene. Wiister 

 et al. (in press) suggested a rate of sequence evolution for cytfr and 

 ND4 of between 0.66 and 1.4% My -1 in Neotropical pitvipers. This 

 would date the timing of the split between the Antillean Bothrops 

 clade and the B. asper-atrox clade at 4.2-8.9 Mya, i.e., the late 

 Miocene or earliest Pliocene. Similarly, the split between B. 

 caribbaeus and B. lanceolatus (sequence divergence: 4.3%) can be 

 dated to 3.1-6.5 Mya. Hedges (1996) estimated the divergence of 

 the B. asper-atrox complex to have taken place within the last 4 My, 

 and assumed dispersal to the Antilles to have taken place during that 

 timeframe, whereas our data suggest a slightly earlier lineage split. 

 In any case, it can be concluded that the two Antillean Bothrops 

 species represent two relatively old. independent lineages. Obviously, 

 in view of the errors inherent in any attempt at molecular clock 

 usage, these estimates should be treated as approximations rather 

 than exact timings. 



The notion that these populations are the result of a recent 

 dispersal event from within South America, as is the case in West 

 Indian Corallus (Henderson & Hedges, 1995), is refuted by both 

 tree topology and statistical tree comparison tests. Equally, the 

 notion that the presence of these snakes in the Lesser Antilles is the 

 result of a primitive form of biological warfare among aboriginal 

 people (Dowling, 1965) will have to be abandoned, despite its 

 romantic appeal. 



The colonisation sequence of the two species can be resolved 

 from morphological data, particularly scalation. In terms of dorsal 

 and ventral scale counts, B. caribbaeus is indistinguishable from 

 many populations of the B. asper-atrox complex. On the other hand. 



Table 1 Differences in tree length or likelihood, statistics, and their significance, between the most parsimonious or the most likely trees, and trees 

 constrained to be compatible with alternative phylogenetic or biogeographical hypotheses. 



d(steps) 



Wilcoxon signed-ranks 

 - z 



Shimodaira-Hasegawa 

 d(lnL) P 



Non-monophyly of B. caribbaeus and B. lanceolatus 

 Non-monophyly of B. asper-atrox complex 

 Non-monophyly of cis-Andean B. atrox complex 

 Monophyly of B. caribbaeus, lanceolatus, asper 

 and northern Venezuelan populations 



7 



1.4000 



0.1615 



15.05075 



0.054 



5 



1.1471 - 1.5076 



0.1317-0.2513 



3.15866 



0.181 



5 



2.4019 



0.0163* 



20.06423 



0.018* 



8 



2.9200-3.0870 



0.002-0.0035* 



24.34213 



0.005* 





