Molecular Phylogenetics and Evolution Vol. 16, No. 2, August, pp. 199 –211, 2000 doi:10.1006/mpev.2000.0779, available online at http://www.idealibrary.com on
A Phylogeny of the Trimeresurus Group of Pit Vipers: New Evidence from a Mitochondrial Gene Tree Anita Malhotra and Roger S. Thorpe School of Biological Sciences, University of Wales Bangor, Bangor, Gwynedd LL57 2UW, United Kingdom Received June 14, 1999; revised February 9, 2000
The Trimeresurus group is an important radiation of over 40 Asian pit viper species. Once considered congeneric, four genera are generally currently recognized (Trimeresurus sensu stricto, Ovophis, Protobothrops, and Tropidolaemus) but relationships within and between these are still unclear. This study, based on mitochondrial cytochrome b sequences, is the first to include a large number of species (21) and demonstrates that the current taxonomy does not adequately represent either the relationships or the genetic diversity present in the complex. Although many deeper nodes are not strongly supported, the following novel conclusions are all well supported: (1) the paraphyly of Trimeresurus sensu stricto, (2) the presence of several divergent clades within Trimeresurus sensu stricto, (3) the paraphyly of some widespread, medically significant, species, (4) the nonmonophyly of Ovophis, and (5) the monophyly of Protobothrops. Mapping of morphological characters onto the mitochondrial tree further supports the four groups proposed for Trimeresurus sensu stricto. © 2000 Academic Press
INTRODUCTION Pit vipers (Reptilia: Serpentes: Viperidae: Crotalinae) of the Trimeresurus complex are widely distributed across southern Asia and the Indo–Malayan archipelago. The complex represents a major evolutionary radiation, including circa 40 species (McDiarmid et al., 1999; David and Ineich, 1999), which occupy a wide range of habitats and display a range of lifestyles and reproductive modes. New species are also being described on a regular basis. Originally all considered to be congeneric, Trimeresurus sensu lato (s.l.), the species are currently generally arranged in four genera, Trimeresurus sensu stricto (s.s.), Ovophis, Protobothrops, and Tropidolaemus (Tro.). However, there is considerable confusion and disagreement about the content of the four genera and, in particular, the removal of a few species to the latter three genera leaves Trimeresurus s.s. as a large and diverse, possibly paraphyletic, assemblage of over 30 species.
Despite their diversity in ecology and behavior, the species in the group are remarkably conservative in morphology. This marked phenotypic convergence among species is a feature of the crotaline radiation as a whole (Greene, 1992; Ogawa et al., 1995; Kraus et al., 1996). In the Trimeresurus complex, it is particularly well typified by the green pit vipers, or bamboo vipers. All initially considered a single species, Stejneger (1927) and Pope and Pope (1933) described scale and hemipenial characters that could distinguish between them. However, the presence of considerable geographic variation and sexual dimorphism in external color and scalation characters (Malhotra and Thorpe, 1997) contributes to frequent misidentification. All things considered, it is highly likely that a full resolution of their systematics has not yet been achieved and that a number of cryptic species may be undetected. While molecular evidence is not free from the effects of homoplasy, it is likely to play a significant role in resolving the taxonomy of the Trimeresurus complex by providing an independent character system for phylogeny reconstruction. Although a number of molecular studies on crotalines have recently appeared (Kraus et al., 1996; Vidal et al., 1997; Parkinson et al., 1997; Vidal and Lecointre, 1998; Parkinson, 1999), these focus largely on New World species, and few have included more than one species of Trimeresurus. The most comprehensive studies are Kraus et al. (1996) and Parkinson (1999), which included seven and eight species of Trimeresurus s.l., respectively. Both these studies supported the paraphyly of Trimeresurus s.l. (see also Vidal and Lecointre, 1998) and the monophyly of Trimeresurus s.s. (three species) and Protobothrops (two or three species). The monophyly and affinities of Ovophis are still unresolved, as no study to date has included more than one species. All these studies admit that a more comprehensive study of Trimeresurus s.s. is required for a better understanding of the phylogenetic relationships within the Trimeresurus complex, as well as those among all Asian crotaline genera. This study is the first to include a significant number of species in the complex (21), allowing us to rigorously address the monophyly of the genera Trimeresurus s.s,
1055-7903/00 $35.00 Copyright © 2000 by Academic Press All rights of reproduction in any form reserved.
MALHOTRA AND THORPE
Protobothrops, and Ovophis (the genus Tropidolaemus is only represented in this study by one species). MATERIALS AND METHODS Samples and Sequencing An attempt was made to sample as many species as possible. This is a difficult task, not only because of the size of the group, but also because many of the species are secretive in habits and are found in remote areas. For the species in which accurate identification is problematic (i.e., most of the green species), only specimens from known localities, and which could be examined personally to verify the identification, were used. Given the strong possibility of cryptic species being present, we attempted to include multiple samples from across the range of widely distributed species. Comprehensive sampling has been achieved to any great extent only with T. albolabris, but we also have multiple samples for T. popeiorum, T. purpureomaculatus, and T. stejnegeri. Samples from single localities are available for T. cantori, T. erythrurus, T. flavomaculatus, T. gracilis, T. gramineus, T. hageni, T. kanburiensis, T. malabaricus, O. monticola, P. mucrosquamatus, O. okinavensis, T. borneensis, T. trigonocephalus, and Tro. wagleri. In total, 18 nominal species of the Trimeresurus complex, representing all four genera, were sequenced for this study (Table 1). Samples of T. macrops were also available, but failed to amplify with any of the primer pairs used (see below). Samples were in the form of tail tip biopsies preserved in 80% ethanol, liver tissue in 80% ethanol, or 100 –200 l of blood taken from the caudal vein, placed in 1 ml 5% EDTA, and preserved in 2 ml SDS–Tris buffer (100 mM Tris, 3% SDS). Whole genomic DNA was extracted from 0.01– 0.02 g of ethanol-preserved muscle, liver tissue, or 200 –500 l of blood/buffer, using standard protocols (Sambrook et al., 1982). Various fragments of the cytochrome b gene were amplified using a slightly modified version of primer L14841 (5⬘-GCTTCCATCCAACATCTCAGCATGATG-3⬘[Kocher et al., 1989]) and MVZ16 (5⬘GGCAAATAGGAAGTATCATTCTG-3⬘ [Moritz et al., 1992]). Some samples, which did not amplify well using this combination of primers, were amplified with modified versions of the primers Mt-A (5⬘-CTCCCAGCCCCATCCAACATCTCAGCATGATGAAACTTCG-3⬘ [Lenk and Wink, 1997]) and Mt-F (5⬘-AGGGTGGAGTCTTCTGTTTTTGGTTTACAAGACCAATG-3⬘ [Wink, 1995]). Reactions were conducted in 25-l volumes and contained standard reagent concentrations (e.g., Hoelzel and Green, 1994). They were typically subjected to 35 cycles consisting of 30 s denaturation at 94°C, 30 s annealing at 50°C, and 30 s extension at 72°C, followed by a final 5-min extension step at 72°C. A negative (blank) control was always included to exclude the possibility of contamination. Unincorporated nucleotides and primers were
removed using a variety of commercially available kits, e.g., Prep-a-gene (Bio-Rad), Wizard minicolumns (Promega), or QIAquick columns (QIAGen). The doublestranded product was then either manually sequenced using a modification of the Sequenase v2.0 protocol (Perkin–Elmer) or sequenced using dye-labeled terminators (ABI PRISM BigDye Terminator Cycle Sequencing Ready Reaction Kit) and subsequently run on an ABI Prism 377 DNA sequencer. Primer L14841 was primarily used for sequencing. The opposite strand was sequenced using the primers MVZ16 and H15149 (5⬘-CCCTCAGAATGATATTTGTCCTCA-3⬘ [Kocher et al., 1989]), in most cases from several different PCR products. Additional Sequences and Outgroups A further three Protobothrops sequences were obtained from GenBank. These were for the species P. tokarensis (Accession No. AFO38884, from Vidal and Lecointre ), and P. flavoviridis and P. elegans (unpublished, Accession Nos. D31622 and D31623, respectively). Only 548 bp of the P. tokarensis sequence aligned with the cytochrome b section sequenced for this study. The latter two sequences were substantially shorter, being only 286 bp long. The unresolved state of crotaline intergeneric relationships makes the choice of outgroup problematic, and the likely paraphyly of Trimeresurus s.l. necessitates the use of a variety of other crotaline genera to validate this. These included Bothrops moojeni, Bothrops alternatus, Crotalus durissus terrificus (provided by W. Wu¨ster), Crotalus cerastes (GenBank, Accession No. U69733, unpublished), Calloselasma rhodostoma, and Deinagkistrodon acutus (Table 1). Recent studies (Milinkovitch and Lyons-Weiler, 1998) suggest that ingroup topology may be radically altered by use of different outgroups. A number of viperine genera were therefore used as outgroups, represented by Daboia russelii, Macrovipera mauritanica (provided by P. Lenk), and Echis ocellatus, as well as a representative of another family, Naja siamensis (the latter two sequences were provided by W. Wu¨ster). Sequence Analysis The final analysis included 660 bp from 56 terminal taxa (including outgroups). Alignment was trivial, as there were no indels. Sequences were translated into amino acid sequences to check for the unexpected occurrence of stop codons, which might indicate that pseudogenes (Sorenson and Fleischer, 1996; Zhang and Hewitt, 1996) had been amplified. For distance analyses, the substitutional model which best fit the ingroup data was investigated by plotting the observed number of transversions against “time” (crudely measured by genetic distance). This is expected to be linear except at very large evolutionary distances, when sequences may begin to saturate due to multiple substitutions. Conversely, an exponential curve indicates that the
SYSTEMATICS OF PIT VIPERS OF THE Trimeresurus GROUP
TABLE 1 List of Species Included in This Study, Their Origins, and GenBank Accession Nos. Genus
Calloselasma Crotalus Crotalus
rhodostoma cerastes durissus
Deinagkistrodon Echis Macrovipera Ovophis Ovophis Protobothrops Protobothrops Protobothrops Protobothrops Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Trimeresurus Tropidolaemus
acutus ocellatus mauritanica monticola okinavensis elegans flavoviridis mucrosquamatus tokarensis albolabris albolabris albolabris albolabris albolabris albolabris albolabris albolabris albolabris albolabris albolabris albolabris albolabris borneensis cantori erythrurus flavomaculatus gracilis gramineus hageni kanburiensis malabaricus popeiorum popeiorum popeiorum popeiorum purpureomaculatus purpureomaculatus stejnegeri stejnegeri stejnegeri trigonocephalus wagleri
Geographic origin Parana, Brazil Sao Paulo State, Brazil West Malaysia — Sao Paulo State, Brazil Taiwan Garoua, Cameroon Morocco Taiwan Ryu-kyu Islands — — Taiwan — Southeast Thailand Northeast Thailand South Thailand West Thailand North Thailand Hong Kong Nepal West Java East Java Pantar Is., Indonesia Alor Is., Indonesia Flores Is., Indonesia Komodo Is., Indonesia South Thailand Nicobar Islands Rangoon, Myanmar Philippines Taiwan South India South Thailand South Thailand South India West Malaysia South Thailand North Thailand West Thailand West Malaysia Andaman Islands Taiwan Northeast Thailand North Vietnam Sri Lanka West Malaysia
GenBank Accession No.
CAL1 — CROT136
A54 — IB 55601
AF171918 U69733 AF200223
DACUTUS — — TMONT1 B1 — — TMUC1 — TAT1, TAT11 TAT4 TAT5, TAT8 TAT10 T?1, T?2, TAT13 TAHK1 TAN2 TAWJ1, TAWJ3 TAEJ1, TAEJ2 TaPt1 TaAl1 TAF1 TAK1 PUN1 TCANT TERY1 B3 TGRAC TG1 THAG3 TKT3 TMAL2 TPM1 TPS1 TPN1 TPN2 TP1 TPA1 TST4, TST60 TSL18 TPV1 TT1 TWAG1
A223 WW567 — A87 B1 — — A211 — A127, A133 A135 A130, A134 A139 A225, A226, A149 A157 A100 A125, A126 A115, A111 WAM 107876 WAM 107905 A119 A122 A73 A85 A209 B3 A86 A219 A224 A241 A217 A196 A202 A204 A205 A82 A76 A160, A161 A181 ROM 7234 A58 A66
AF171919 AF191579 AJ275714 AF171907 AF171915 D31623 D31623 AF171897 AFO38884 AF171885, AF171893 AF171923, AF171921 AF171920, AF171884 AF171909 AF171886, AF171887, AF171881 AF171882 AF171883 AF171924 AF171912 AF171889 AF171900 AF171916 AF171913 AF171905 AF171911 AF171914 AF171901 AF171888 AF171904 AF171902 AF171906 AF171889 AF171922 AF171896, AF171898 AF171903 AF171890 AF17191
AF171895 AF171894 AF171910, AF171908
Note. Generic names are according to current usage and are listed in alphabetical order. Unless otherwise indicated, the catalogue numbers (CAT) given are for the authors’ personal collection. ROM, Royal Ontario Museum; WAM, Western Australian museum; IB, Instituto Butantan, Sao Paulo, Brazil; WW, Wolfgang Wu¨ster, personal collection.
genetic distance used, which incorporates positions that saturate much more quickly, has not adequately corrected this. For example, a plot of transversions against p-distances is substantially curved upward. Employing a correction such as the Jukes–Cantor (Jukes and Cantor, 1969) distance improves the fit to a
linear regression (as judged by the R 2 values), but it is still obviously nonlinear. This process was repeated with the Kimura two-parameter (K2P) model (Kimura, 1980), the Hasegawa, Kishino, and Yano (HKY) model (Hasegawa et al., 1985), and the General Time-reversible (GTR) model (Lanave et al., 1984; Rodriguez et al.,
MALHOTRA AND THORPE
1990), all with, and without, gamma-distributed rates with various shape parameters. The best substitution model was judged to be the one involving the smallest increase in R 2 between the linear and the quadratic polynomial regressions. This model was then used for subsequent analyses. Some degree of saturation is likely to be present in the sequences. This was investigated by plotting the number of pairwise differences due to transitions (Ts) against the genetic distance produced from the model chosen by the procedure above (Milinkovitch et al., 1996; Krajewski and King, 1996). This was done for transitions at first and second codon positions together and at third positions separately, as the latter are the most likely to show saturation effects. Unless otherwise stated, analyses were performed using the beta test version (b2a) of PAUP* 4.0 (Swofford, 1998). The effect of using different combinations of noncrotaline outgroups was explored using neighbor-joining (NJ) and the model specified by the procedure described above. Once the best combination of outgroups had been selected, parsimony (MP) trees were constructed using a heuristic search. Tree bisection–reconnection (TBR) branch swapping was found to be computationally intractable and instead, subtree pruning–regrafting (SPR) branch swapping was used with 10 random addition replications. Four hundred bootstraps were also performed using the same settings, but with only one random sequence-addition replicate to reduce computational time. This was found to be superior to the alternative in which 1000 bootstraps were performed without swapping (the FASTSTEP option in PAUP* 4.0) in the amount of resolution present in the bootstrap 50% majority-rule consensus tree. Groups supported by less than 50% of the bootstraps were retained if they were compatible with the 50% majority-rule tree. Some form of weighting is often used in parsimony analysis to compensate for homoplasy due to the saturation of rapidly evolving positions. However, this has recently been shown to be ineffective by studies that have tested alternative weighting schemes (Allard and Carpenter, 1996; Philippe et al., 1996; Milinkovitch and Lyons-Weiler, 1998; Vidal and Lecointre, 1998), with much phylogenetic information being discarded in the process. Thus, all the MP analyses presented in this paper are on unweighted data only. A maximum-likelihood (ML) tree was also produced using the evolutionary model and shape parameter of the gamma distribution defined by the procedure described above. Transition:transversion ratios and base frequencies were estimated from the data. Starting trees for heuristic searches were obtained by neighborjoining, with subsequent SPR branch swapping. Only 100 bootstraps were performed because of the computer-intensive nature of this method; in addition, nearest neighbor interchange (NNI) rather than SPR branch
swapping was employed to further reduce the computational time required. All phylogenetic methods incorporate implicit assumptions that, if violated, may lead to incorrect inference. Consequently, these assumptions should be explicitly tested where possible. First, the presence of phylogenetic signal in the data matrix was tested using the g 1 statistic (Hillis and Huelsenbeck, 1992) for the skewness of tree length distributions, estimated from 10,000 random trees. The critical values of g 1 are obtained from the table published in Hillis and Huelsenbeck (1992), and a significant result indicates that the length of the actual tree is significantly shorter than expected from random data (i.e., without any phylogenetic structure). A significant result may, however, also be obtained if the phylogenetic signal is confined to only part of the tree. As cytochrome b is a fast-evolving gene, it is possible that the phylogenetic signal is confined to the tips of the tree, and that at the deeper levels, the signal in this gene has been effectively randomized by saturation. Consequently, this was tested by holding together successive well-supported clades, and recalculating the g 1 statistic. The point at which it ceases to be significant indicates the limit of resolution of the dataset. It is also possible that other sources of structure may exist. For example, two sequences with similar GC content biases will tend to appear to be more closely related in phylogenetic reconstructions. To evaluate this, the GC content of each sequence was calculated and the presence of significant heterogeneity between sequences was tested using a G-test (Sokal and Rohlf, 1981). The possibility of nonneutral evolution of the cytochrome b gene (Ballard and Kreitman, 1995) was tested using McDonald and Kreitman’s (1991) test. This is a goodness-of-fit test that compares ratios of synonymous to replacement substitutions within and between species, with the expectation under neutrality that these should be the same. Eight sequences of T. albolabris were used (from those supported as monophyletic by the phylogenetic analysis), with a variety of outgroups from different levels of the tree. The hypothesis of rate constancy was tested by the likelihood-ratio test (Felsenstein, 1988), which compares the likelihood of a tree using an assumption of clock-like evolution and a tree calculated without this assumption. However, this tests the overall “clockness” of the tree, whereas it would also be of interest to compare rates of substitution in subsets of the data (such as the clades revealed by phylogenetic analyses). Relative rate tests between the various lineages were therefore carried out using the two-cluster test of Takezaki et al. (1993) as implemented in PHYLTEST, which tests the null hypothesis that the average number of substitutions in two lineages, relative to that of an outgroup lineage, will be equal. It is similar to Li and Bosquet’s (1992) test, but allows more than one
SYSTEMATICS OF PIT VIPERS OF THE Trimeresurus GROUP
outgroup sequence to be used. Each of the clades (represented by clusters of monophyletic sequences) was tested against its sister clade, using the clade immediately interior to them as the outgroup (as the power of the test increases when the outgroup is more closely related). Support from Morphology MacClade 3.08a (Maddison and Maddison, 1995) was used to map various morphological characteristics (the type of hemipenis, the condition of the first upper labial and nasal scale, and the number of scale rows at midbody) onto our best-supported phylogeny for Trimeresurus s.s. to explore support for the clades identified in the tree and to resolve any conflicting arrangements. The linear parsimony option was used, with the number of possible reconstructions calculated using the equivocal cycling option, where appropriate. These characters have established taxonomic value (e.g., squamate hemipenes have repeatedly been shown to be valuable taxonomic characters in a number of different groups). Hemipenial morphology. Hemipenes vary considerably throughout the group and provide one of the main characters for distinguishing between some of the green species (Pope and Pope, 1933). The main difference is in the presence or absence of hard spines, but two types of spinose hemipenes (referred to here as Type 1 and Type 2) can also be distinguished on the basis of the arrangement, shape, and number of the spines. A detailed description of these hemipenis types can be found in Pope and Pope (1933). While the ancestor of Trimeresurus s.s. is not well defined, neither Type 2 spinose or smooth hemipenes are found outside this clade, and Type 1 spinose would seem to be the most likely ancestral condition, being widespread in the more basal crotalines, viperines, and nonviperid taxa. The condition of the labial and nasal scale. The first upper labial scale can either be distinctly separated from the adjacent nasal scale by a suture or they may be completely or partially fused (Stejneger, 1927). While the degree of fusion is variable in most species, a species was coded as having the fused state if any degree of fusion is present. The number of scale rows at midbody. This varies between 19 and 29 in the species represented in this study and was coded as a discrete character. The occurrence of green coloration. This was investigated to examine the relationship of the “green pit vipers” to each other. While there is no a priori reason to expect coloration to be a good indicator of evolutionary relationships, as it is likely to be subject to strong selection, many authors use this term as if it implies a close relationship. We distinguished between species
that are never green, those that have the typical green pit viper pattern (a plain green background, usually with a white lateral stripe, which may be edged below with red and sometimes extends onto the head, and a reddish tail), and those that have a green component but with other dominant patterns present (e.g,. purplebrown bands in T. kanburiensis). RESULTS Testing Evolutionary Assumptions DNA extracts amplified using different primer sets always gave identical sequences, as did different PCR products amplified with the same primer sets. No deletions, insertions, or stop codons were found in the data, which suggests that paralogous nuclear insertions have not been amplified. The base frequency distribution (A ⫽ 29.4%, C ⫽ 30.8%, G ⫽ 11.3%, T ⫽ 28.4%) and pattern of substitutions were as expected for coding mitochondrial DNA (mtDNA) sequences. The Ts:tv ratio (estimated by maximum-likelihood from the NJ tree) was 3.72. A conservative test for the significance of the g 1 statistic indicated that there was significant structure in the data (g 1 ⫽ ⫺0.2980). There was no significant difference between the base frequency distribution of the different sequences (G H ⫽ 16.875, P ⬍ 0.99). The McDonald–Kreitman neutrality test failed to reject the null hypothesis of neutral evolution for any comparison. The substitution model which best fit the data (by the procedure described above) was the HKY model with gamma-distributed rates (shape parameter ⫽ 0.30), in which the lines fitted by linear and polynomial regression were exactly coincident. This substitution model was then used for subsequent distance-based analyses. Maximum corrected sequence divergence in the ingroup under this model was 0.4467, between T. gracilis and T. albolabris. However, T. gracilis, T. borneensis, P. mucrosquamatus, and O. okinavensis all have distances over 0.4 with a number of species. When number of transitions are plotted against this distance, they are found to be substantially saturated, but as expected, most of this is due to third position transitions only as transitions at second and third positions lie substantially on the same line as the transversions (Fig. 1). There are few distances between 0.14 and 0.18, and this discontinuity marks the boundary between those distances that are largely unsaturated and those in which saturation is marked. Most of the betweenspecies distances in the dataset are above this boundary, and this indicates the presence of substantial homoplasy at this position, which may bias the parsimony analysis. Testing the influence of the use of different outgroups on NJ trees showed that the tree is sensitive to different combinations, although this variability was
MALHOTRA AND THORPE
FIG. 1. Observed proportion (p-distance) of transversions (squares) plotted against the Hasegawa–Kishino–Yano (HKY) distance with gamma-distributed rates (shape parameter ⫽ 0.3). The equations and R 2 for the linear and quadratic regressions are given. Saturation at third codon positions (diamonds) is indicated by a decline in the rate of accumulation of observed proportion of transitions. Observed proportion of transitional changes at first and second codon positions (not shown) lie on a largely similar line to the transversions (y ⫽ 0.1375⫻, R 2 ⫽ 0.779).
confined to the deeper levels of the tree. Removing one outgroup at a time always changed the relative position of the clades below a certain level of the tree and in some cases also the composition of those clades. However, it seemed clear that the use of both D. russelii and M. mauritanica was redundant, as the results did not change if only one of these was deleted. In addition, using only M. mauritanica and E. ocellatus gave results identical to those using all four outgroups, and so they were retained as the outgroups. Sequence Analysis Among the 54 remaining sequences there were 283 parsimony-informative sites. Unweighted MP produced 16 equally parsimonious trees, which differ in only minor aspects involving closely related (intraspecific) terminal taxa (Fig. 2). A number of species groups have been indicated in Fig. 2 for ease of discussion. However, the bootstrap support for many of the basal nodes defining these species groups is low. The hypothesis that this is due to the randomization of the phylogenetic signal through saturation over greater periods of evolutionary time was tested using the g 1 statistic. This was done by holding together clades which were supported in the bootstrap analysis (see Fig. 2), i.e., the albolabris group, the stejnegeri group, T. popeiroum, the Indian subcontinent species (not including T. borneensis), the Proto-
bothrops species, and a few other highly supported sister taxa relationships between the remaining species, and collapsing the rest of the ingroup branches. Once these relationships are accounted for, the g 1 statistic should be nonsignificant if there is no remaining phylogenetic structure in the rest of the tree, and in this case there would be no justification for attempting further resolution of the tree. However, g 1 is similar to that of the whole dataset at ⫺0.2991 (P ⬍ 0.01), and it would appear that, despite the rapid evolution of cytochrome b, it retains enough signal to justify its use at the intergeneric level. Instead, the low bootstrap support for some interior nodes may be caused by a few “floating” sequences (e.g., T. borneensis, T. kanburiensis, and O. monticola), which are not closely related to any species represented in the tree and therefore are particularly variable in their position. To test this, these taxa were excluded, and the bootstraps repeated to see whether these were responsible for lowering the bootstrap values for nodes which otherwise would be reasonably well supported. Although this does seem to be the case, the increase in support was not large (Fig. 2) and does not increase the overall resolution of the 50% consensus tree. Nevertheless, it is likely that the addition of species, which would reduce the branch lengths lead-
SYSTEMATICS OF PIT VIPERS OF THE Trimeresurus GROUP
FIG. 2. The 50% consensus tree of 16 equally parsimonious trees of length 1782. Full generic names are given for species that do not belong to the Trimeresurus s.l., whereas only the specific name is given for species belonging to Trimeresurus s.s. Finally, species allocated to one of the three other genera contained within Trimeresurus s.l. are indicated by generic abbreviations in parentheses. Filled circles indicate nodes that collapse in the strict consensus tree. The figures below nodes are the bootstrap support values (above 50%) from 400 replications. Figures in parentheses after bootstrap values at three nodes indicate the increase in bootstrap support when some sequences that are particularly variable in position (see text for full list) were eliminated from the analysis. The species groups referred to in the text are indicated on the tree. These are (1) albolabris species group, (2) stejnegeri species group, (3) popeiorum species group, (4) Indian subcontinent species group, (5) Protobothrops/Ovophis species group, and (6) New World species.
ing to these problematic taxa, would help to resolve this problem. The ML (Fig. 3) tree is extremely similar to the MP tree; in fact the likelihood scores of the two trees (⫺ln L ⫽ 8281.13458 and 8282.20308 for the MP and ML
trees, respectively) are not significantly different (P ⫽ 0.9) in a Kishino–Hasegawa test (Kishino and Hasegawa, 1989). Both trees are most consistent above, and including, the Indian subcontinent/T. borneensis clade. This coincides with Trimeresurus s.s. (with the excep-
MALHOTRA AND THORPE
FIG. 3. Maximum-likelihood tree, showing bootstrap support values (above 50%) from 100 bootstrap replications. Species notations are described in the legend to Fig. 2.
tion of T. gracilis; see below). The region of the tree with the best support, as in the MP tree, is the clade containing T. albolabris, T. purpureomaculatus, T. cantori, and T. erythrurus (i.e., the albolabris species group in Fig. 2). Both trees show that the species T. albolabris, as currently defined, is paraphyletic, consisting of at least three clades (an exclusively Indonesian clade with populations from East Java and the Lesser Sunda Islands, a Nepalese clade, and a clade containing all other mainland populations, as well as West Java) which are split by a group of clearly distinct
species (T. erythrurus, T. cantori, and T. purpureomaculatus). However, there are small differences between the topologies produced by the two methods in the relative positions of T. cantori, T. erythrurus and mainland and island populations of T. purpureomaculatus. These form a distinct clade in the ML tree, but are sequential in the MP tree. Also, although the Nepalese T. albolabris falls in exactly the same place in both trees, the branch length is so short in the ML tree that it appears unresolved in the phylogram. The final difference within this subsection of the tree is the po-
SYSTEMATICS OF PIT VIPERS OF THE Trimeresurus GROUP
sition of T. kanburiensis, which falls between the stejnegeri and the albolabris groups in the MP tree, but between the stejnegeri and the popeiorum groups in the ML tree. At deeper levels of the trees, the main groups defined by both methods are still remarkably consistent, and the points of major interest are all drawn from the well-supported aspects of both trees. The species assigned to Protobothrops are supported as a monophyletic group in both trees. However, the Ovophis species represented in this study are not monophyletic; instead, O. okinavensis is grouped together with a species of Trimeresurus s.s., T. gracilis, making both Ovophis and Trimeresurus s.s. paraphyletic. Both trees show that Tropidolaemus is quite distinct from other Trimeresurus s.l., in agreement with all recently published molecular studies (Kraus et al., 1996; Vidal and Lecointre, 1998; Parkinson, 1999). It is worth noting, however, that the monophyly of the New World pit vipers is not strongly supported. It may be that previous studies have not sampled the Asian pit viper radiation sufficiently well to firmly establish their reciprocal monophyly. A likelihood ratio test to test for the action of a molecular clock, which compares a ML tree calculated with and without the assumption of a molecular clock, rejected the null hypothesis of equal rates operating across the tree. However, the relative rate test did not reject the null hypothesis of equal rates between any of the major clades defined above. Most of the rate variation would therefore appear to be within rather than between clades. Support from Morphology
the spines are similar in both T. borneensis and Indian subcontinent species, the sizes of the spines are different. The alternative placement of T. kanburiensis on the MP tree, between the stejnegeri and the albolabris groups, involves the same number of steps; so, this character does not favor one arrangement over the other. Condition of the first upper labial and nasal scale (Fig. 4B). Species having a nasal scale that is fully or partially fused with the first upper labial mostly belong to the albolabris species group, which is a well-supported clade in the mtDNA trees. The exception, in the ML tree, is T. kanburiensis. In this version, two steps are required to explain the distribution of this character (with two possible reconstructions) and it has a RC of 0.4. In fact, the branch length separating the stejnegeri clade and T. kanburiensis in the ML tree is so short that it is effectively a polytomy. The alternative position of this taxon shown in the MP tree (Fig. 2), in which it is basal to the albolabris clade, requires one less step and the RC increases to 1.0. This character, therefore, strongly supports the alternative position for this taxa given in the MP tree.
As the trees are essentially similar for the subsection corresponding to Trimeresurus s.s., only this part of the tree was used for character mapping. The minor disagreements between the two trees were explored by examining each alternative dichotomous resolution rather than by collapsing these into polytomies. We show only the results using the ML tree in Fig. 3, with alternative reconstructions on the MP tree being reported in the text.
The number of scale rows at midbody (Fig. 4C). This character appears to be extremely conservative in the Trimeresurus group, being 21 in a large number of species (which adds to the difficulties of distinguishing between the green pit vipers, as this is character is commonly used in keys). One change to a higher number of scales has occurred in the albolabris species group (in the species T. purpureomaculatus, T. cantori and T. erythrurus). Altogether, five steps are required, and the RC ⫽ 0.6. In the MP tree, in which the species split sequentially from the common ancestor, at least one subsequent reversal would also be required to have occurred, making it the less parsimonious alternative. Evolution of a lower number of scale rows (19) has also occurred in some species, notably in three of four members of the Indian subcontinent/T. borneensis clade.
Hemipenis type (Fig. 4A). There are three types of hemipenes represented within Trimeresurus s.s., which are consistent with the species groups defined by the DNA analysis. Three steps are required with a rescaled consistency index (RC) of 1.0. The albolabris and popeiorum species groups and T. kanburiensis all have smooth hemipenes. The distribution of Type 1 and Type 2 spinose hemipenes also appears very consistent. The former are restricted to members of the stejnegeri group, whereas the latter support the placement of T. borneensis with the Indian subcontinent species group, a clade with low bootstrap support. Note, however, that the latter are not entirely unambiguous; whereas the distribution of
Green coloration (Fig. 4D). In contrast to the previous three characters, the distribution of green coloration shows little consistency between clades. At least nine steps are required, with eight possible reconstructions. The rescaled consistency index is only 0.25. Many species are rather polymorphic; for example, T. flavomaculatus displays all three character states. Again, one less step is required for the T. purpureomaculatus, T. erythrurus, and T. cantori arrangement in the ML tree than in the MP tree. Overall, there are a much larger number of steps (a minimum of nine) required to plot the evolution of this character than required of the other three characters, and the consistency index is correspondingly much lower.
MALHOTRA AND THORPE
FIG. 4. Morphological characters mapped onto the maximum-likelihood tree (with morphologically conservative clades condensed to a single tip) for Trimeresurus sensu stricto only. (A) Hemipenial morphology, (B) fusion of first upper labial scale and nasal scale, (C) number of scale rows at midbody, and (D) green coloration. See text for further explanation.
DISCUSSION Deeper Crotaline Relationships The monophyly of the two species assigned to Ovophis (O. monticola and O. okinavensis) is not supported, and as O. okinavensis is unambiguously more
closely related to T. gracilis, a species currently assigned to Trimeresurus s.s., both genera are clearly paraphyletic as currently defined. In view of the major systematic rearrangement suggested by the mitochondrial gene tree, it is particularly noteworthy that the monophyly of Protobothrops continues to
SYSTEMATICS OF PIT VIPERS OF THE Trimeresurus GROUP
be supported. However, extrapolation beyond the species actually included in this study is unwise. The content of Ovophis and Protobothrops currently owes much to species groups defined in an earlier work (Maslin, 1942), which we have shown elsewhere to be rife with errors. Clearly, more sampling of the spe-
cies involved is required to fully address the status and validity of these genera. In addition, low bootstrap values supporting the critical nodes do not allow firm conclusions to be made about the relationships of these groups at deeper levels. The addition of more sequence data (particularly from slower-
MALHOTRA AND THORPE
evolving mitochondrial genes) may be useful to provide further resolution at this level. Implications for the Taxonomy of Trimeresurus s.s. Trimeresurus s.s. appears to be sufficiently structured to warrant further subdivision into genera or subgenera. However, the unique transmission characteristics of mitochondrial genes compared to nuclear genes imply that it may be unwise to base a major systematic revision on a single mitochondrial gene. Work is in progress on a corroborating nuclear-based phylogeny. However, the congruence between the morphological characters and the mtDNA gene tree not only lends support to the hypothesis of relationships proposed here, but also suggests that we could infer the relationships of other species for which genetic data is not available. For example, T. strigatus, an Indian subcontinent species, has a Type 2 spinose hemipenis in common with the other Indian subcontinent species, and its affinities are almost certainly with this group rather than with Protobothrops, as suggested by Kraus et al. (1996). A combination of two morphological characters, hemipenis type and fusion of labial/nasal scale, defines four major clades within Trimeresurus s.s. It is possible that closer examination will be able to distinguish additional variation in the hemipenes (e.g., there appears to be variation in the presence of soft papillate processes among species with smooth hemipenes). This character may therefore be of crucial importance in resolving the taxonomy of the group, and it is unfortunate that so many new descriptions of species still fail to mention the hemipenis. The number of scale rows at midbody shows more changes (and particularly, more reversals) than the former two characters. However, this character also provides support for clades identified in the analysis where other characters are invariable. Finally, the lack of congruence of coloration with the DNA topology suggest that green coloration is not indicative of close evolutionary relationship but of evolutionary convergence. This convergence has reached a quite remarkable degree of perfection, as indicated by the fact that some species can still be reliably distinguished only by their hemipenial morphology (e.g., T. popeiorum and T. stejnegeri). This coloration is not restricted to Trimeresurus s.s., having apparently been independently derived in the very distinct Tropidolaemus wagleri. All the green pit vipers are predominately arboreal and nocturnal, but other, nongreen, arboreal species exist. Further discussion regarding this convergence must await the resolution of the taxonomy of the bamboo vipers, which is among the most confused in the whole group.
not fully resolved, and there are still cryptic species present. The results show that T. albolabris is clearly paraphyletic, consisting of at least three clades. These correspond to the clades reported in Malhotra and Thorpe (1997), but their paraphyly was not detected in that study, as T. purpureomaculatus was erroneously assumed to be an outgroup. One of these clades, the largest “mainland” clade, appears to contain further substructure (Figs. 2 and 3) which may have systematic implications (currently under investigation). T. purpureomaculatus is also paraphyletic, with mainland populations (T. p. purpureomaculatus) being more closely related to T. erythrurus than to the Andaman island populations (T. p. andersoni). The fact that the only species in which extensive sampling has been carried out across its range shows considerable substructuring and polyphyly raises the possibility that by the time other widespread species have been studied in detail, there may be even more species of Trimeresurus. This, together with the relatively deep divergence detected, even among the very similar green pit vipers, which are commonly not distinguished in the medical and biochemical literature, underlines the fact that arriving at a stable taxonomy for the Trimeresurus group will be of more than esoteric interest. ACKNOWLEDGMENTS We are grateful to the large numbers of people who have assisted us in the field or supplied us with tissue samples for analysis. These include Dr. Jennifer Daltry, and Dr. Wolfgang Wu¨ster, Dr. Nicholas Giannasi, Nicholas Cockayne (University of Wales, Bangor), Professor David Warrell (University of Oxford), Romulus Whitaker, Dr. Indraneil Das, and Gerry Martin (Madras Crocodile Bank), Mr. S. S. Ramachandra Raja (Wildlife Association of Ramnad District, India), Dr. Michihisa Toriba (Japan Snake Institute), Anslem de Silva (Peridinaya University, Sri Lanka), Professor Sangkot Marzuki (Eijkmann Institute, Indonesia), Dr. Aucky Hinting, Rick Hodges, Vincen Khartono, Pak Harwono, and his staff (Surabaya Zoo), Dr. Wen-hao Chou (National Museum of Natural Science, Taiwan), Jean-Jay Mao (Taipei City Zoo, Taiwan), Tanya Chan-ard, Jarujin Nabhitabhata (National Science Museum of Thailand), Dr. Lawan Chanhome (Queen Savoabha Memorial Institute, Thailand), Dr. Kumthorn Thirakhupt and Dr. Peter Paul van Dijk (Chulalongkorn University, Thailand), Dr. Cheelaprabha Rangsiyanon (Chiang Mai University), Merel J. Cox, Jonathan Murray, Galen Valle, Dr. James D. Lazell (the Conservation Agency, USA), Dr. Bob Murphy (Royal Ontario Museum, Canada), Dr. Trinh Xuan Kiem (Cho-Ray Hospital, Vietnam), Michel Guillod, and Vincent Morier (Ophiofalm, Lausanne, Switzerland). We also acknowledge the National Science Council of Thailand for permission to carry out fieldwork in Thailand. We are grateful to Dale Taneyhill for supplying his program for carrying out the G test for heterogeneity. This study was funded by the Leverhulme Trust (F174/I) and the Darwin Initiative (162/6/65), with additional support for fieldwork from the Royal Society, the Percy Sladen Trust, the Bonhote Trust, and the Carnegie Trust.
Paraphyly of T. albolabris and T. purpureomaculatus
This study underlines the fact that, despite decades of study, the taxonomy of the green pit vipers is still
Allard, M. W., and Carpenter, J. M. (1996). On weighting and congruence. Cladistics 12: 183–198.
SYSTEMATICS OF PIT VIPERS OF THE Trimeresurus GROUP Ballard, J. W. O., and Kreitman, M. (1995). Is mitochondrial DNA a strictly neutral marker? Trends Ecol. Evol. 10: 485– 488. David, P., and Ineich, I. (1999). Les serpents venimeux du monde: Systematique et repartition. Dumerilia 3: 3– 499. Felsenstein, J. (1988). Phylogenies from molecular sequences: Inference and reliability. Annu. Rev. Genet. 22: 521–565. Greene, H. W. (1992). The ecological and behavioural context for pitviper evolution. In “Biology of the Pitvipers” (J. A. Campbell and E. D. Brodie, Jr., Eds.), pp 107–117. Selva, Tyler, TX. Hasegawa, M., Kishino, H., and Yano, T. (1985). Dating the human– ape split by a molecular clock of mitochondrial DNA. J. Mol. Evol. 22: 160 –174. Hillis, D. M., and Huelsenbeck, J. P. (1992). Signal, noise, and reliability in molecular phylogenetic analysis. J. Hered. 83: 189 – 195. Hoelzel, A. R., and Green, A. (1994). Analysis of population level variation by sequencing PCR-amplified DNA. In “Molecular Genetic Analysis of Populations: A Practical Approach” (A. R. Hoelzel, Ed.), pp. 159 –188. IRL Press, Oxford. Jukes, T. H., and Cantor, C. R. (1969). Evolution of protein molecules. In “Mammalian Protein Metabolism” (H. N. Munro, Ed.), pp. 21–132. Academic Press, New York. Kimura, M. (1980). Estimation of evolutionary distances between homologous nucleotide sequences. Proc. Natl. Acad. Sci. USA 78: 454 – 458. Kishino, H., and Hasegawa, M. (1989). Evaluation of the maximum likelihood estimate of the evolutionary tree topologies from DNA sequence data, and the branching order in Hominoidea. J. Mol. Evol. 29: 170 –179. Kocher, T. D., Thomas, W. K., Meyer, A., Edwards, S. V., Paabo, S., Villablanca, F. X., and Wilson, A. C. (1989). Dynamics of mitochondrial DNA evolution in animals: Amplification and sequencing with conserved primers. Proc. Natl. Acad. Sci. USA 86: 6196 – 6200. Krajewski, C., and King, D. G. (1996). Molecular divergence and phylogeny—Rates and patterns of cytochrome-b evolution in cranes. Mol. Biol. Evol. 13: 21–30. Kraus, F., Mink, D. G., and Brown, W. M. (1996). Crotaline intergeneric relationships based on mitochondrial DNA sequence data. Copeia 1996: 763–773. Lanave, C., Preparata, G., Saccone, C., and Serio, G. (1984). A new method for calculating evolutionary substitution rates. J. Mol. Evol. 20: 84 –93. Lenk, P., and Wink, M. (1997). A RNA–RNA heteroduplex cleavage analysis to detect rare mutations in populations. Mol. Ecol. 6: 687– 690. Li, P., and Bousqet, J. (1992). Relative rate test for nucleotide substitutions between two lineages. Mol. Biol. Evol. 9: 1185–1189. Maddison, W. P., and Maddison, D. R. (1992). “MacClade, version 3.0”, Sinauer, Sunderland, MA. Malhotra, A., and Thorpe, R.S. (1997). New perspectives on the evolution of South–East Asian pit-vipers (genus Trimeresurus) from molecular studies. In “Venomous Snakes: Ecology, Evolution and Snakebite” (R. S. Thorpe, W. Wu¨ster, and A. Malhotra, Eds.). Symp. Zool. Soc. Lond. 70: 115–118. Maslin, P. (1942). Evidence for the separation of the crotalid genera Trimeresurus and Bothrops, with a key to the genus Trimeresurus. Copeia 1942: 18 –24. McDiarmid, R. W., Campbell, J. A., and Toure, T. A. (1999). “Snake Species of the World: A Taxonomic and Geographic Reference,” Vol. 1. The Herpetologist’s League, Washington DC. McDonald, J. H., and Kreitman, M. (1991). Adaptive protein evolution at the Adh locus in Drosophila. Nature 351: 652– 654.
Moritz, C., Schneider, C. J., and Wake, D. B. (1992). Evolutionary relationships within the Ensatina eschscholtzii complex confirms the ring species interpretation. Syst. Biol. 41: 273–291. Milinkovitch, M. C., and Lyons-Weiler, J. (1998). Finding optimal ingroup topologies and convexities when the choice of outgroups is not obvious. Mol. Phylogenet. Evol. 9: 348 –357. Milinkovitch, M. C., Leduc, R. G., Adachi, J., Farnir, F., Georges, M., and Hasegawa, M. (1996). Effects of character weighting and species sampling on phylogeny reconstruction: A case study based on DNA sequence data in cetaceans. Genetics 144: 1817–1833. Ogawa, T., Kitajima, M., Nakashima, K., Sakaki, Y., and Ohno, M. (1995). Molecular evolution of group II phospholipases A 2. J. Mol. Evol. 41: 867– 877. Parkinson, C. L. (1999). Molecular systematics and biogeographical history of pitvipers as determined by mitochondrial ribosomal DNA sequences. Copeia 1999: 576 –586. Parkinson, C. L., Moody, S. M., and Ahlquist, J. E. (1997). Phylogenetic relationships of the “Agkistrodon complex” based on mitochondrial DNA sequence data. In “Venomous Snakes: Ecology, Evolution and Snakebite” (R. S. Thorpe, W. Wu¨ster, and A. Malhotra, Eds.). Symp. Zool. Soc. Lond. 70: 63–78. Philippe, H., Lecointre, G., Van Le, H. L., and Le Guyader, H. (1996). A critical study of homplasy in molecular data with the use of a morphologically based cladogram, and its consequences for character weighting. Mol. Biol. 13: 1174 –1186. Pope, C. H., and Pope, S. H. (1933). A study of the green pit-vipers of South– eastern Asia and Malaysia, commonly identified as Trimeresurus gramineus (Shaw), with a description of a new species from Peninsular India. Am. Mus. Novit. 620: 1–12. Rodriguez, R., Oliver, J. L., Marin, A., and Medina, J. R. (1980). The general stochastic model of nucleotide substitution. J. Theor. Biol. 142: 485–501. Sambrook, J., Frisch, E. F., and Maniatis, T. E. (1989). “Molecular Cloning: A Laboratory Manual,” 2nd ed., Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. Sokal, R. R., and Rohlf, F. J. (1981). “Biometry,” 2nd ed., Freeman, San Francisco. Sorenson, M. D., and Fleischer, R. C. (1996). Multiple independent transpositions of mitochondrial control region sequences to the nucleus. Proc. Natl. Acad. Sci. USA 93: 15239 –15243. Stejneger, L. (1927). The Green Pit Viper, Trimeresurus gramineus, in China. Proc. United States Natl. Mus. 72: 1–10. Takezaki, N., Rzhetsky, A., and Nei. M. (1995). Phylogenetic test of the molecular clock and linearized trees. Mol. Biol. Evol. 12: 823– 833. Swofford, D. L. (1998). “PAUP*: Phylogenetic Analysis using Parsimony (and Other Methods), version 4.0,” Sinauer, Sunderland, MA. Vidal, N., Lecointre, G., Vie, J. C., and Gasc, J. P. (1997). Molecular systematics of pitvipers: Paraphyly of the Bothrops complex. C. R. Acad. Sci. Paris Life Sci. 320: 95–101. Vidal, N., and Lecointre, G. (1998). Weighting and congruence: A case study based on three mitochondrial genes in pit-vipers. Mol. Phylogenet. Evol. 9: 366 –374. Wink, M. (1995). Phylogeny of Old and New World vultures (Aves: Accipitridae and Cathartidae) inferred from nucleotide sequences of the mitochondrial cytochrome b gene. Z. Naturforsch. 50c: 868 – 882. Zhang, D. X., and Hewitt, G. M. (1996). Nuclear integrations— Challenges for mitochondrial-DNA markers. Trends Ecol. Evol. 11: 247–251.