INTRODUCTION

The Cordylidae, girdled lizards, form a monophyletic group composed of more than 50 species arranged in four nominal genera (Loveridge, 1944; Branch, 1998): Platysaurus, a group of 15 species most notable for their extremely flattened shape; Pseudocordylus, a group of seven species most notable for their apparent intermediacy in flatness between Platysaurus and deeper-bodied Cordylus; Cordylus, a group of 31 species of diverse morphology from extremely spiny to relatively smooth scaled; and finally, Chamaesaura, a group of three serpentiform species with reduced limbs. The Cordylidae are restricted to the southern subcontinent of Africa, except for a few species that extend into central and East Africa.

Knowledge of cordylid phylogeny is in its infancy, even though its near outgroups and its monophyly have been known for a considerable length of time (McDowell and Bogert, 1954; Estes et al., 1988). Most of the systematic work on cordylid lizards has been focused primarily on species limits and geographic variation (e.g., FitzSimons, 1943; Loveridge, 1944), although recent work has also attempted to place species within a phylogenetic context (e.g., Mouton, 1990; Mouton and Van Wyk, 1989, 1994, 1995; Harvey and Gutberlet, 1995). The most comprehensive attempt at phylogeny reconstruction within the Cordylidae is that of Lang (1991), and although that paper provides considerable information on morphological variation within the Cordylidae and its first taxonomic outgroup, Gerrhosauridae, it only provides a beginning to our understanding of phylogenetics within the group. Genera of unsubstantiated monophyly were used as terminal taxa in Lang's revision, even though evidence to the contrary existed then in the form of considerable interspecific variation in many characters showing coherent distributions across generic boundaries. Unfortunately, a data matrix for species was not provided even though the apomorphy and change lists provided in appendices suggest that such a matrix exists. Many of the characters discussed by Lang provide tantalizing evidence for the paraphyly of such nominal genera as Cordylus and Pseudocordylus, even though his evidence, reified through the assumption of generic monophyly, necessarily appears to support monophyly of those taxa.

Our initial intentions in this project were (1) to use the morphological data provided by Lang (1991) and Harvey and Gutberlet (1995) as external criteria to evaluate differences in phylogenetic results due to sensitivity to alignment parameters (i.e., to search for topologies that minimize incongruence among data sets due to alignment parameters—sensu Wheeler, 1995) and (2) to perform a character congruence study to enhance our understanding of cordylid relationships. With respect to Lang's (1991) study, this has proven impossible for reasons of incompatible taxon sampling as well as our inability to extract species-level information from that revision with confidence. Coupled with our doubt regarding the monophyly of at least two of the four genera (i.e., Pseudocordylus and Cordylus) assumed to be monophyletic by Lang, these difficulties have proven insurmountable. With respect to the scale surface ultrastructure study of Harvey and Gutberlet (1995) there were very few characters addressed (and, to foreshadow our results, only three characters in conflict with the molecular evidence). The third author of this paper, P. Mouton, has for some time been engaged in a survey of the taxonomy, morphology, and anatomy of the group (e.g., Mouton, 1985, Mouton et al., 1992; Brody et al., 1993; Mouton and Van Wyk, 1994, 1997). A more inclusive publication of the morphological data will be provided by Mouton at a later date. Therefore, the purpose of this contribution is not to report on the morphology of the cordylids, but to report on an analysis of DNA sequences with comments on what these data suggest with respect to reported morphological diversification within the Cordylidae. We will mention the morphological evidence of Lang (1991) and Harvey and Gutberlet (1995), but only within a narrative context.

METHODS

RESULTS AND DISCUSSION

Topologies for gap, transversion, and transition costs of 1:1:1 and 1:2:1 are provided in figures 1 and 2, respectively. Length for the two trees was, respectively, 1501 and 2257, with consistency and retention indices of, respectively, 0.22 and 0.80, and 0.25 and 0.80. Bremer (= decay) indices (Bremer, 1994) are provided with the relevant stems on figures 1 and 2. A strict consensus of these two topologies is provided in figure 3. The genus-terminal tree of cordylids provided by Lang (1991, his fig. 17) and a summary of our results are provided in figure 4 for comparison.

Cost parameters showed some effect with a difference of 2× in the weighting of transversions over transitions and gaps. Only one stem with a Bremer index of 10 was affected by this change (affecting the relative position of Cordylus coeruleopunctatus), and monophyly of the putative sister-taxon, Gerrhosauridae, is not supported by the molecular data. This is a surprising result and may have been caused by our limited sampling. However, the Bremer indices are high enough that reevaluation of the previously published evidence for gerrhosaurid monophyly would be prudent.

Monophyly of the Cordylidae sensu Lang (1991), always noncontroversial, is affirmed. Lang (1991) summarized a number of morphological synapomorphies for the group and it must be one of the most highly corroborated taxa in lizard systematics.

Only one species of Chamaesaura is included in the analysis, but the extraordinary morphology of this taxon strongly suggests monophyly of the three included species. The suggestion of the molecular data—that Chamaesaura rests within nominal Cordylus—is a novel hypothesis and at variance with the previous suggestion by Lang (1991) that Chamaesaura is the sister taxon of other cordylids. Beyond a large number of generic apomorphies, Lang (1991) suggested five characters for which Chamaesaura was plesiomorphic with respect to the remaining cordylids, but he also mentioned several other features of Chamaesaura that were found in some Cordylus (e.g., his S1-position of nasal scales, S3-condition of postnasal scale). Further, Chamaesaura is so highly apomorphic, correlated with its serpentiform lifestyle, that our molecular data suggest that careful evaluation of the morphological data is required. Note, however, that the molecular data placing Chamaesaura within “Cordylus” are corroborated by at least one anatomical-ecophysiological characteristic. Mouton and Van Wyk (1997) noted that outgroup comparison would judge ovoviviparity within Chamaesaura, Cordylus, and Pseudocordylus as a synapomorphy and that oviparity in Platysaurus (the only ovoviparous cordylid) would be counted as a plesiomorphy, a hypothesis congruent with our molecular results.

Although we have only two representatives of Platysaurus (P. monotropis and P. intermedius rhodesianus) we have no reason to doubt its monophyly. In Lang's (1991) cladogram, Platysaurus is diagnosed by seven apomorphies. The molecular data place Platysaurus as the sister-taxon of the remaining cordylids, including Chamaesaura. This, like the placement of Chamaesaura within Cordylus based on molecular evidence, is surprising inasmuch as what appears to be a morphocline—from the smooth-scaled, flattened crevice-dwellers (Platysaurus) through the somewhat less flattened Pseudocordylus to variably spiny Cordylus—had initially appeared to us as a phylogenetic transition (with the Chamaesaura morphotype independently derived from the Cordylus morphotype). The molecular data, like the ecophysiological data of Van Wyk (1989) and Mouton and Van Wyk (1997), suggest that this is not the case, although this placement within Cordylus resolves the problem of reversal from ovoviviparity (Chamaesaura, Cordylus, and Pseudocordylus) to oviparity (Platysaurus and cordylid outgroups) required by Lang's placement.

No evidence of monophyly of Pseudocordylus is provided by the molecular data. Two clades of Pseudocordylus are recognized by our analysis, the first composed of Ps. microlepidotus and Ps. melanotus, and the second composed of Ps. capensis and Ps. nebulosus. Regardless of their similarity, there is no molecular support for a special relationship of Pseudocordylus with Platysaurus and considerable evidence against it. Lang (1991) supported the monophyly of Pseudocordylus with three characters: (a) nasal scales in contact (his S1.1; even though he suggested that Platysaurus, the putative sister taxon, was highly variable in this characteristic), (b) anterior parietal scales in contact (his S11.1; although he suggested that Platysaurus, the putative sister taxon in his analysis, showed contact in most species and variability in Cordylus was high), and (c) lingual exposure of the angular (his C24.0, although he noted on p. 185 that it is variable among unspecified species of Pseudocordylus). In summary, we view the evidence for monophyly of Pseudocordylus to be equivocal and consider Pseudocordylus a candidate for being polyphyletic. Further, at least in this study, evidence tying either group of Pseudocordylus to Platysaurus appears to be lacking.

No evidence of monophyly of Cordylus is suggested by the molecular evidence. Indeed, there is very strong evidence of paraphyly with respect to Pseudocordylus and Chamaesaura. That Cordylus is paraphyletic with respect to Pseudocordylus is not surprising. Lang (1991) suggested five synapomorphies of Cordylus: S21.1 (keeled scales on bottom of foot; not scorable in Chamaesaura and variable in Platysaurus and Pseudocordylus), S24.1 (keeled subdigital lamella; not scorable in Chamaesaura and variable in Pseudocordylus), C28.0 (posteriorly tapered retroarticular process [a character influenced by degree of skull flattening]), P1.2 (intercentra fused to posterior vertebrae [sutured in Chamaesaura]; and M3.0 (osteoderm distribution, osteoderms apparently always reduced in flattened crevice dwellers). Nevertheless, even though Cordylus was assumed to be monophyletic on the basis of these characters (and for purposes of his intergeneric analysis), Lang (1991) noted at least nine features that are variable among species of Cordylus and that extend to other cordylid taxa.

Differences due to transversion weighting are evident, particularly in the placement of the Cordylus vittifer group, C. tropidosternum, C. warreni, C. coeruleopunctatus, and the relative placement of Chamaesaura within “Cordylus” clade.

Clades that remain the same among our analyses within nominal Cordylus are: (a) a clade of Cordylus oelofseni, C. cordylus, C. niger, C. macropholis, and C. minor; (b) a Cordylus vittifer, C. peersi, C. lawrenci clade; (c) a clade composed of C. jordani and C. polyzonus. These associations are hardly surprising. Mouton (1990) first noted the association of C. oelofseni, C. cordylus, and C. niger. Loveridge (1944) considered these three species to be parts of a single larger species, C. cordylus, but he had also included a number of other populations as well, including C. lawrenci, judged to be distant from this group on the basis of our data. The close association of C. jordani and C. polyzonus has also been long recognized (Loveridge, 1944). Nevertheless, the scale ultrastructure data presented by Harvey and Gutberlet (1995), and discussed below, suggest that continued careful evaluation of relationships is warranted.

The scale surface ultrastructure data of cordylids were cast by Harvey and Gutberlet (1995) into 10 characters of transformation and mapped upon the cladogram of Lang (1991), although Harvey and Gutberlet discussed some additional refinement within Cordylus. When the accuracy of Lang's (1991) generic arrangement is not assumed, very little discordance with Harvey and Gutberlet's data and our results is discovered and what is found has no impact on our hypothesis of Cordylus paraphyly with respect to Pseudocordylus and Chamaesaura. Indeed, when their data are analyzed independently they show considerable concordance with our molecular results (Stems A and B, fig. 5; strict consensus of a sample of 16383 trees; length = 11, CI = 0.9091; without uninformative characters 0.8750). What discordance exists is solely in regard to the relative placement of various species of Cordylus (fig. 5, stems C and D). The presence on dorsal scales of the macrohoneycomb oberhautchen (character #1 of Harvey and Gutberlet), thought to be lost in Platysaurus when mapped on Lang's arrangement is plesiomorphically absent in our cladogram in Platysaurus and its presence is a synapomorphy of the remaining cordylids. (In other words, this character corresponds in placement with the transition from oviparity to ovoviviparity.) Where their evidence does show discordance with the molecular evidence is in their character #6 (flaplike free margins associated with the cell-ridge system; fig. 5, Stem C), and their character #7 (ridges in the center of oberhautchen cells in C. polyzonus, C. niger, and C. minor; fig. 5, stem D). Although discordance between ultrastructural character evidence and the molecular evidence within Cordylus bears further scrutiny, in this case the molecular evidence is rather strong, suggesting the existence of more homoplasy in scale surface ultrastructure than previously suspected.

Clearly a study of character congruence of both morphology and molecules of species (not of higher taxa as was done by Lang, 1991) would provide a stronger basis on which to optimize transversion, transition, and gap cost parameters and would further elucidate the relative relationships within cordylids. Nevertheless, on the basis of the molecular evidence alone, it is evident that “Cordylus” is paraphyletic with respect to Chamaesaura and Pseudocordylus, that Pseudocordylus is dubiously monophyletic, and that the oviparous Platysaurus is the sister taxon of the other, ovoviviparous, cordylids. On this basis we regard Pseudocordylus and Chamaesaura to be junior synonyms of Cordylus. This affects only 10 species names and formulates a conservative taxonomy consistent with the state of knowledge of cordylid relationships.