J. Zool., Lond. (1999) 249,49±60 # 1999 The Zoological Society of London Printed in the United Kingdom

A phylogeny of the European lizard genus Algyroides (Reptilia: Lacertidae) based on DNA sequences, with comments on the evolution of the group

D. James Harris, E. Nicholas Arnold* and Richard H. Thomas Department of Zoology, The Natural History Museum, Cromwell Road, London SW7 5BD, U.K. (Accepted 19 November 1998)

Abstract The four species of Algyroides Bibron & Bory, 1833 form part of the relatively plesiomorphic Palaearctic clade of lacertids comprising Lacerta and its allies. An estimate of phylogeny based on DNA sequence from parts of the 12S and 16S rRNA mitochondrial genes con®rms the monophyly of the genus already suggested by several morphological features. The molecular data also indicates that relationships within the clade are: (A. nigropunctatus (A. moreoticus (A. ®tzingeri, A. marchi))); this agrees with an estimate of phylogeny based on morphology that assumes the taxon ancestral to Alygroides was relatively robust in body form, and not strongly adapted to using crevices. Initial morphological evolution within Algyroides appears to involve adaptation to crypsis in woodland habitats. The most plesiomorphic form (A. nigropunctatus) is likely to have originally climbed extensively on tree boles and branches and there may have been two subsequent independent shifts to increased use of litter and vegetation matrices with related anatomical changes (A. moreoticus, A. ®tzingeri), and one to increased use of crevices (A. marchi). Some members of Algyroides are strikingly similar in super®cial morphology to particular species of the equatorial African genus Adolfus. This resemblance results from a combination of many shared primitive features plus a few independently acquired derived ones that are likely to give performance advantage in the relatively similar structural niches that these forms occupy. This study provides evidence that: (1) the use of a combination of molecular and morphological data may sometimes allow the estimation of ancestral anatomical features when these are otherwise unknown; (2) process considerations may permit a choice to be made in cases of character evolution where tree topology means that equally parsimonious alternatives exist; such decisions about character evolution may allow ecological shifts to be similarly assessed; (3) parallel evolution in ecological analogues may involve relatively few characters.

Key words: lizards, Algyroides, DNA sequence, evolution

INTRODUCTION Algyroides is the sister group of Lacerta bedriagae (Harris et al., 1998a). Within Algyroides, morphological Algyroides Bibron & Bory, 1833 is a small clade of four features that might give information about relationships well-differentiated species of lacertid lizards with largely are few with considerable con¯ict in their distributions, disjunct distributions in southern Europe (Fig. 1), that and because of the uncertainties about relationships may be originally associated with woodland habitats within the Lacerta and its allies, it is not possible to (Arnold, 1973, 1987). The genus is part of a large designate the most appropriate outgroups for phyloge- Palaearctic clade of mainly primitive lacertids that netic analysis of the genus with any certainty. includes most present members of the paraphyletic To resolve relationships within Algyroides and con®rm genus Lacerta and the wall lizards Podarcis (Harris, its monophyly, a phylogenetic analysis was conducted Arnold & Thomas, 1998a); it will be referred to here as using DNA sequences derived from sections of two Lacerta and its allies{. Detailed af®nities within this mitochondrial genes. These were then combined with assemblage have not been ®rmly resolved (Arnold, morphological information and employed to assess 1989a; Mayr & Benyr, 1994; Harris, 1997), although changes in anatomy and ecology and the general mtDNA sequence provides restricted evidence that evolution of the genus and to examine parallelism

*All correspondence to: E. N. Arnold, Department of Zoology, The { This assemblage does not include Lacerta jayakari and Lacerta Natural History Museum, Cromwell Road, London SW7 5BD, U.K. cyanura which are most appropriately placed in a separate genus, E-mail: [email protected] Omanosaura Lutz & Mayer 1986 (Haris et al., 1998a). 50 D. J. Harris, E. N. Arnold and R. H. Thomas

C

B

A

C+D D

Fig. 1. Distribution of the species of Algyroides in southern Europe showing their largely allopatric and disjunct ranges. A, A. marchi;B,A. ®tzingeri;C,A. nigropunctatus;D,A. moreoticus (compiled from BoÈhme, 1981). between members of Algyroides and those of the central frequently use crevices as refuges, such as Lacerta oxyce- African genus Adolfus, that show considerable super- phala. Four most parsimonious trees are produced using ®cial resemblance to each other. such an ancestor, and the strict consensus tree derived from these is uninformative. If, on the other hand, alter- native states are entered for these characters, the hypo- EVIDENCE FOR CLADE STATUS OF thetical ancestor resembles a more robustly built Lacerta ALGYROIDES AND INTERRELATIONSHIPS without marked crevice-using specializations, such as WITHIN THE GENUS DERIVED FROM Lacerta laevis. In this case, a single most parsimonious MORPHOLOGY tree is produced (Fig. 3). This latter version of events is more parsimonious overall, given the lack of knowledge The morphological characters mentioned in the text are about detailed relationships within the clade made up of discussed elsewhere (Arnold, 1973, 1989a, b) Lacerta and its allies, as most 0 states are more likely to A number of derived morphological features which, in be primitive in the ancestor of this assemblage. combination, de®ne Algyroides as a clade are likely to have been present in the immediate common ancestor of the genus (Table 1, characters 1±6). Characters varying EVIDENCE FOR RELATIONSHIPS FROM interspeci®cally within Algyroides are also listed in Table ALBUMIN IMMUNOLOGY 1, and their distributions shown in Table 2. To establish a hypothesis of relationships, these data were analysed Two possible phylogenies for the species of Algyroides by maximum parsimony, with exhaustive searches, using have been suggested on the basis of albumin evolution the program PAUP 3.1.1 (Swofford, 1993); the single assessed by immunological means (Mayer & Lutz, 1990). multistate character (number 7) was treated as ordered. One is that the relationships are: A. ®tzingeri (A. marchi Several of the morphological features that vary within (A. nigropunctatus, A. moreoticus)). The alternative is Algyroides also vary in the rest of the clade made up of similar except that A. marchi is excluded and appears to Lacerta and its allies, including characters 8, 9, 10, 11, 20 be most closely related to a putative clade comprising and 25. As the closest outgroups of Algyroides within this Podarcis and Lacerta graeca. The latter arrangement assemblage are not known for certain, this variation would make Algyroides polyphyletic, something that means that the immediate ancestor of Algyroides could con¯icts strongly with the morphological evidence have had a range of different character states. Which ones that the genus is a clade. are assumed affects the supposed pattern of relation- ships produced within Algyroides. At one extreme, the tree can be rooted using a hypothetical ancestor in which EVIDENCE FOR RELATIONSHIPS FROM characters 8, 9, 10 and 11 are scored as 1, and character MITOCHONDRIAL DNA SEQUENCES 20 as 0. In this case the ancestor would resemble some of the delicately built and ¯attened Lacerta species in the Given the problems of choosing appropriate outgroups Palaearctic group that live on rocky surfaces and for Algyroides when assessing the morphological Molecular phylogeny of Algyroides lizards 51

Table 1. Morphological features that de®ne Algyroides and others varying within it. Most of these features are discussed elsewhere (Arnold, 1973, 1989a) and their distribution is shown in Table 2

Features that in combination de®ne Algyroides 1. Small lips on the lobe sulci of the hemipenis (Arnold, 1973) 2. Distinctive dorsal scale micro-ornamentation in which pustullate swellings project among the upturned posterior edges of the strap-shaped cell surfaces that constitute the oberhautchen (Fig. 2). Such pustules occur elsewhere amongst lacertids only in Adolfus africanus 3. Dorsal scales strongly enlarged and markedly keeled 4. 6±8 enlarged dorsal scales in a transverse row between the hind legs 5. Lower number of presacral vertebrae in males than is usual in Lacerta and its allies. The commonest numbers are 25 in A. nigropunctatus, A. moreoticus and A. marchi, and 26 in A. ®tzingeri, overall evidence suggesting the latter ®gure is a reversal 6. Sombre dorsal colouring: usually brownish, often with a darker markings

Features varying within Algyroides 7. Length of adults from snout to vent: > 60 mm (0), < 55 mm (1), < 45 mm (2) 8. Skull: robust (0), lightly built (1) 9. Supraocular osteoderms: intact in adults (0), fenestrated (1) 10. Pterygoid teeth: often present (0), absent (1) 11. Commonest number of presacral vertebrae in females: usually 26 (0), usually 27 (1) 12. On average number of long post-sternal ribs outnumber those in the more posterior series of short ribs by about three in both sexes: no (0), yes (1) 13. Shape of sternal fontanelle: oval (0), sometimes tending to heart-shape (1) 14. Number of postnasal scales: nearly always two superposed scales (0), often a single scale (1) 15. Anterior lateral edge of the parietal scale extending towards edge of the parietal table of skull: no (0), yes (1) 16. Size of lateral scales: distinctly smaller than enlarged dorsals (0), about the same size as the enlarged dorsals (1) 17. Dorsal and lateral scales lanceolate and strongly overlapping posteriorly: no (0), yes (1) 18. Keeling on dorsal scales: relatively weak (0), strong (1) 19. Micro-ornamentation on dorsal scales: raised posterior edges of epidermal cells constituting oberhautchen more or less smooth (0), denticulated (1) 20. Posterior overlap of ventral scales: small (0), more extensive (1) 21. Dark vertebral streak or series of spots: absent (0), sometimes present (1) 22. Tail proportions: relatively slender and evenly tapering (0), robust proximally (1) 23. Obvious sexual dichromatism: no (0), yes (1) 24. Males often with narrow, pale dorsolateral streaks: no (0), yes (1) 25. Hemipenial micro-onamentation: recurved spines (0), crown-shaped tubercles (1)

Table 2 Morphological features varying interspeci®cally within Algyroides, and their distribution in two hypothetical ancestors in the paraphyletic genus Lacerta. Character numbers refer to Table 1

Character numbers

Species 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25

A. nigropunctatus 000 0100000010000001 A. moreoticus 100 0001101111100111 A. ®tzingeri 211 1010001110111000 A. marchi 111 110001000001000± L. oxycephala- like ancestor 0 1 1 1 1 0000000000000? L. laevis- like ancestor 0 0 0 0 0 0000000000000? evidence of relationships, and the con¯icting results MATERIALS AND METHODS FOR DNA ANALYSIS from immunological studies of albumin, additional data are necessary to resolve the relationships of the species Localities and sources of the lizards from which DNA of Algyroides. These data are provided in this paper was extracted are given in the Appendix. Tissue samples by investigation of mitochondrial DNA (mtDNA) consisted of tail tips, stored in 70% ethanol at 4 8C. sequences. Portions of two mitochondrial genes, 12S Voucher specimens are deposited in the collection of the rRNA and 16S rRNA were sequenced for all four Natural History Museum, London. species of Algyroides, for two species of wall lizards, Podarcis muralis and Podarcis taurica, and for the more distantly related Gallotia galloti (Arnold, 1989a; Laboratory procedures Mayr & Benyr, 1994); the later three species provide appropriate outgroups. Total genomic DNA was extracted from 1 or 2 mm3 52 D. J. Harris, E. N. Arnold and R. H. Thomas

Fig. 2. Characteristic micro-ornamentation of dorsal scales in Algyroides, showing raised pustullate swellings projecting among upturned posterior edges of the strap-shaped cells constituting the oberhautchen (6370). The species shown is A. nigropunctatus.

Thermocycling consisted of 30 cycles of 93 8C for 30 s, A. marchi A. fitzingeri 55 8C for 1 min and 72 8C for 1 min, followed by a single 11,16,17,20 cycle at 72 8C for 10 min. PCR products were checked 18R 22 12 by electrophoresis on a 2% agarose gel, and sizes of 15 7(2) A. moreoticus ampli®ed products were estimated using a molecular 21 24 weight marker. Successful PCR bands were cut out and 10 23 puri®ed using a QIAEX II kit (Quiagen). The DNA was 9 19 14 A. nigropunctatus 8 13 then dialysed through a Microcon 30000 MW membrane (Amicon) twice with double-distilled water and the volume made up to 10 ml; 2 ml of this was then 7(1) sequenced on an Applied Biosystems Model 373A DNA Sequencing System, using a PRISM Ready Reaction 6 5 DyeDeoxy Terminator Cycle Sequencing kit. Centrisep 4 3 2 spin columns (Princeton Separations Inc.) were used for 1 excess dye extraction. More detailed procedures are Robustly built given by Harris (1997). Lacerta

Fig. 3. Phylogeny for Algyroides showing principal changes in Alignment morphology. Numbers refer to Table 1. Black rectangles, characters supporting clade status of Algyroides; R, reversal. 12S rRNA sequences were aligned by eye and compared to homologous sequences of Lacerta lepida and Lacerta Characters 11, 16, 17 and 20 found in A. moreoticus and dugesii (Gonzalez et al., 1996) and Meroles. They were A. ®tzingeri are not congruent with the others, and are likely to also aligned against secondary structure models (Neefs have arisen in parallel in these two species (see p. 56). et al., 1990; Hickson et al., 1996). 16S rRNA sequences were similarly aligned using Meroles, other lacertids pieces of tail tissue. The material was ®nely diced (Harris, 1997), various xantusiid lizards (Hedges & and digested with proteinase K (Kocher et al., 1989). Bezy, 1994) and a more limited secondary structure Puri®cation was by phenol/chloroform extractions model (Gutell, 1993). The resulting alignments con- (Sambrook, Fritsch & Maniatis, 1989), followed by tained 315 and 398 sites, respectively. A total of 36 sites centrifugal dialysis through a Centricon 30000 MW that could not be aligned convincingly because they membrane (Amicon). Polymerase chain reaction (PCR) were hypervariable were removed from the analysis. The primers used in both the ampli®cation and the sequen- alignment used is available on request from the cing were 12Sa and 12Sb (Kocher et al., 1989) and 16L corresponding author. Sequences have been deposited and 16H (Hedges, Bezy & Maxson, 1991). These ampli- in Genbank (Accession numbers AF080306, ®ed regions were some 400 bp and 450 bp, respectively. AF019648±019650 and AF111174±111177). Molecular phylogeny of Algyroides lizards 53

0.070 Intraspeci®c variation A. marchi

In all species a single individual was sequenced, as the 0.066 partial gene regions used have very low intraspeci®c 100 variation in lacertids. Where individuals of the same 0.024 species of lacertid have been sequenced independently in A. fitzingeri different laboratories, differences have been minimal 0.018 and largely con®ned to the hypervariable regions 59 removed in this analysis (e.g. L. lepida and G. galloti; Gonzalez et al., 1996; Harris, 1997). 0.055 0.045 A. moreoticus Phylogenetic analysis 93

Phylogenetic analyses were performed using PAUP* 4.0.d52 (Swofford, 1998), employing the 3 main 0.073 methods of inferring phylogenies, namely parsimony, A. nigropunctatus distance methods and maximum likelihood. Branch lengths, transition/transversion ratios (TS/TV) and among-site rate variation have all been shown to in¯u- ence phylogenetic analysis (Hillis, Moritz & Mable, 0.044 1996; Yang, 1996) and were all consequently taken into P. muralis account in the analysis. PAUP* was also used to calculate base composition 0.035 across the combined sequences. If base composition 91 differs substantially between lineages, systematic error 0.049 can be introduced into any analysis that incorrectly P. taurica assume the same equilibrium base frequencies across lineages (Hillis et al., 1996). In fact, base composition in Algyroides is typical of vertebrate mitochondria, and almost identical to that of other lacertids for this region 0.314 (average occurrences in the light strand A 34%, G 20%, G. galloti C 24%, T 22%; Kocher et al., 1989; Harris, 1997). Base Fig. 4. Maximum-likelihood tree (log-likelihood 72243) composition biases within Algyroides and the outgroups based on the combined 12S and 16S partial gene sequences. were tested using the approach proposed by Rzhetsky & The maximum parsimony tree has an identical topology. Nei (1995). Frequencies did not vary signi®cantly Numbers above internal branches are branch lengths for the between species (I = 23.91, P > 0.05). maximum likelihood tree, and re¯ect numbers of nucleotide substitutions per site. Bootstrap (500 replicates) percentages for the maximum parsimony tree are given below the internal RESULTS branches. For details of models used see text. Initially a maximum parsimony analysis was carried out on the combined 12S and 16S data sets. These contained A neighbour joining analysis was also carried out, a total of 677 characters after unalignable ones were using LogDet (Steel, 1994) corrected distances, a excluded, of which 489 were invariant and 91 were method that does not require stationarity of nucleotide uninformative under parsimony criteria. An exhaustive frequencies, the assumption of which can misleadingly search with all characters weighted equally resulted in group sequences of similar nucleotide frequencies one most parsimonious tree (Fig. 4). Support for nodes regardless of true evolutionary relationships (Hasegawa was estimated by bootstrapping (Felsenstein, 1985). & Hashimoto, 1993; Gu & Li, 1996). This analysis gave Bootstrap support is particularly strong for the the same topology as maximum parsimony. monophyly of Algyroides (93%) and for the sister- A maximum likelihood analysis was then performed relationships of A. ®tzingeri and A. marchi (100%). It is on the combined data sets. The optimal TS/TV ratio less robust for the relationship of A. moreoticus to the was estimated using PAUP* 4.0.d52 (Swofford, 1997) to latter two species (59%). Maximum parsimony analyses be 3.10. Among-site rate variation was estimated using a were also carried out on the separate 12S and 16S data discrete approximation to the gamma distribution sets, and bootstrapped 50% majority rule trees produced. (shape parameter 1.132, with four rate categories; Yang, These differed from the combined tree only in that the 1996). Gaps were treated as missing data. Employing tree derived from 12S data alone was not as fully resolved, the HKY85 model (Hasegawa et al., 1985), which thereby justifying the use of an approach in which the two allows for two substitution types and unequal base gene regions were combined (Hillis et al., 1996). frequencies, and estimating the proportion of invariant 54 D. J. Harris, E. N. Arnold and R. H. Thomas

70 alignable positions were used in the analysis, and the Logdet correction applied. The toplogy of the network (Fig. 6) indicates strong support for the relationships 60 shown, which are identical to those produced by the approaches already discussed. 50 To check if there was any signi®cant differences between the estimates of phylogeny derived from the 40 albumin evolution data set and that from the mtDNA data set, the topology of the tree for the DNA 30 Transversions characters was constrained to the topology of the Transitions albumin evolution estimates. The log likelihoods for these hypothesized relationships, using the same model Observed substitutions 20 of evolution as the maximum likelihood estimate, were 72269 (in the estimate derived from albumin evolution 10 in which Algyroides is a clade) and 72276 (in the estimate in which A. marchi is more closely related to 0 Podarcis and Lacerta graeca than to other Algyroides). 0.0 0.1 0.2 0.3 Both of these estimates of phylogeny are signi®cantly LogDet corrected distances less well supported than the unconstrained maximum Fig. 5. Patterns of nucleotide substitutions in combined likelihood tree, according to the likelihood variance test 12S and 16S partial gene sequences. Transitions (*) and of Kishino & Hasegawa (1989) (Table 3), calculated transversions (&) plotted against LogDet corrected sequence using PAUP. divergence for all taxa. The approximately linear relationship in both cases indicates that sites are unlikely to have been saturated by multiple substitutions. PHYLOGENETIC CONCLUSIONS Molecular and morphological data both support the sites (0.607) using maximum likelihood, a single tree of monophyly of Algyroides. The DNA phylogeny corro- log likelihood 72243 was produced (Fig. 4). borates the tree based on morphology in which a Previous studies have shown that relationships may robustly built ancestor without marked crevice speciali- be obscured by saturation of sites by multiple substitu- zations, similar to Lacerta laevis, is assumed. Such an tions, transitions being likely to become saturated assumption does not necessarily mean that Lacerta before transversions (see Hillis et al., 1996). Because of laevis itself is the sister taxon of Algyroides, and DNA this, pairwise proportions of transitions and transver- sequence data provides no evidence for this (Harris sions in the combined sequences were plotted against et al., 1998a). The long branch length of Gallotia on the LogDet corrected sequence divergence calculated using maximum likelihood and splitstree analyses (Figs 4 & 6) PAUP* 4.0.d52 (Swofford, 1996). They show an is consistent with this genus being more distantly related approximately linear relationship (Fig. 5), indicating to Algyroides than Podarcis. This conclusion is also that sites are not likely to have been saturated. supported by evidence from morphological (Arnold, The combined sequences were analysed further using 1989a) and karyological (Olmo, Odierna & Capriglione, the `split-decomposition method' (Bandelt & Dress, 1993) data. 1992a), employing the program Splitstree 1.0 (Huson & In contrast to the similar tree topologies produced Wetzel, 1994). In a typical phylogenetic analysis, data from DNA sequence and morphological data, albumin are forced to ®t a tree topology. The split-decomposition immunology suggests different relationships among the method however allows for con¯icting alternative species of Algyroides, and the possibility of a poly- groupings, exhibiting networks of relationships phyletic origin for the genus (Mayer & Lutz, 1990). As including the more tentative ones that will be overridden two of these three apparently independent sources of on a single tree topology (Bandelt & Dress, 1992b). All phylogenetic evidence corroborate each other closely, it

Table 3. Maximum-likelihood tests (Kishino & Hasegawa, 1989) of alternative tree topologies for Algyroides lizards

Tree topologya Log likelihood D Log likelihood sd P b

DNA sequence (ML) 72243 ± ± ± Albumin evolution (1) 72269 26 9.0 < 0.05 Albumin evolution (2) 72276 33 11.2 < 0.05 a Trees compared were the maximum-likelihood tree based on DNA sequence data (Fig. 3), and alternative hypotheses based on albumin evolution data (1, monophyletic Algyroides; 2, polyphyletic Algyroides). b P is the probability of getting a more extreme t-value under the null hypothesis of no difference between trees. Both these hypotheses are signi®cantly worse than the maximum likelihood tree. Molecular phylogeny of Algyroides lizards 55

A. marchi A. fitzingeri

A. moreoticus

A. nigropunctatus

G. galloti P. muralis

P. taurica

0.105 Fig. 6. Split-decomposition network of the LogDet corrected distances between species of Algyroides, Podarcis and Gallotia. Only alignable positions were included in the analysis. The tree-like topology of the network indicates strong support for these relationships.

Holaspis guentheri Adolfus vauereselli Adolfus africanus result in errors in the estimation of relationships. Unfortunately, with immunological data it is not pos- sible to gauge the degree of statistical support for 18(R) 2 4 particular internal branches. This contrasts with the 2 situation with DNA sequences where the bootstrap 9 technique (Felsenstein, 1985) permits such tests to be made. As already noted, these tests indicate that most 5 8 relationships derived from DNA sequence receive statis- tical support that is often strong, with bootstrap ®gures Adolfus alleni of from 59% to 100% (Fig. 4). The split-tree analysis of sequence data also suggests little homoplasy, with a 77% 22 24 20 21 data ®t and a clear single tree topology, indicating that 16 17 11 12 no other relationship is strongly supported by the DNA 18 data (Fig. 6). 6 Adolfus jacksoni 3 DISTINCTIVE FEATURES OF THE SPECIES OF Gastropholis ALGYROIDES 10 The four species of Algyroides are well de®ned and 15 easily distinguished from each other. The phylogeny 14 supported here allows autapomorphic features to be recognised with some certainty and, with the exception Fig. 7. Phylogeny for the Central African Adolfus and their of A. nigropunctatus, two to ®ve morphological autapo- relatives (from Arnold, 1989b), showing principal changes in morphies are assignable to each species. These are listed morphology that also occur in Algyroides (see Fig. 3). Black below; ®gures in parentheses refer to Table 1. rectangles: characters supporting monophyly of Algyroides. Algyroides moreoticus: sternal fontanelle sometimes Numbers refer to Table 1. weakly heart-shaped (13); often a single postnasal scale (14); micro-ornamentation of dorsal scales with the raised posterior borders of oberhautchen cells denticu- is likely that it is the hypotheses of relationships based lated (19); sexual dichromatism (23); males often with on immunology that are wrong. In the immunological narrow light dorsolateral streaks (24). tree all the internal branches are extremely short, so Algyroides ®tzingeri: very small size (7); more long post- slight variation in the rates of albumin evolution, or sternal ribs (12); mean number of presacral vertebrae in imprecision in the techniques used to assess them, could males particularly high; tail robust (22). 56 D. J. Harris, E. N. Arnold and R. H. Thomas

Algyroides marchi: anterior lateral edge of parietal scale 2. The evolution of the features is concurrent on the extends close to edge of parietal table of skull (15); phylogenies of these taxa with occupation of spatial keeling on dorsal scales relatively weak (18R). niches in which they are likely to confer performance The strong differentiation between species in mor- advantage (see below). There is consequently likely to be phology matches the differentiation found in the DNA strong selection for these features in such environments. sequence tree where external branches are often long 3. There are no residual signs in A. marchi that the relative to internal branches (Fig. 4). lineage leading to it ever possessed the features under consideration.

EVOLUTION WITHIN ALGYROIDES Changes in niche Character evolution (Fig. 3) As noted, the species of Algyroides all seem to have been Combined morphological and molecular considerations originally associated with woodland and woodland-edge suggest that Algyroides is derived from a member of habitats (Arnold, 1987), so this is likely to have been the the Palaearctic assemblage of lacertids comprising ancestral environment of the genus. Although Lacerta and its allies which had features likely to be heliothermic, they tend to occur in relatively cool places primitive for this group, including a robust skull with and often close to shade; Algyroides may also operate at complete suprocular osteoderms and small dorsal rather lower temperatures than syntopic Lacerta and scales. Algyroides nigropunctatus, which constitutes the Podarcis (Arnold, 1973, 1987). If Algyroides is indeed most basal branch of the main lineage of the genus, is derived from a form morphologically similar to Lacerta in many ways the most plesiomorphic form. It is about laevis, its ancestor is likely to have been like this species the same size as most other members of the clade in being partly scansorial on continuous surfaces such consisting of Lacerta and its allies and, although it has as those of rocks and occasionally trees (Zinner, 1967). enlarged keeled scales, these do not overlap very A shift to increased use of tree boles and branches, and extensively posteriorly and are found only on the of fallen timber, would then produce the kind of spatial dorsum and not on the ¯anks, where small scales are niche found in the most basal species, A. nigropunctatus. retained. The three more derived smaller species often occur The remaining three species of Algyroides are close to water or moisture, and a shift to these conditions characterized by relatively small size (7(1)), a derived presumably occurred on the common lineage leading to condition. Algyroides ®tzingeri and A. marchi share them. One of these species, A. marchi, often climbs on several other features that are congruent with the continuous surfaces, like A. nigropunctatus, but makes phylogeny of Algyroides accepted here, including a greater use of narrow crevices than this species and weakly ossi®ed skull (8) with fenestrated supraocular probably other members of the genus too (Arnold, 1987; osteoderms (9) and absence of pterygoid teeth (10); Rubio & Carrascal, 1994). Algyroides moreoticus and these are also the only species to sometimes possess A. ®tzingeri are found more frequently in vegetation a dark vertebral streak or row of spots (21). The matrixes such as brushwood and litter. Because of the phylogeny indicates that weak keeling of the large particular topology of the phylogeny it is not possible to dorsal scales (18R) of A. marchi is secondary and not an decide, just on the basis of the distribution of structural intermediate stage in the development of strong keeling. niche types on this, whether it is more parsimonious to Some derived features are not congruent with the assume the shift to greater use of matrix situations phylogeny, including a possible increase in the number occurred once or twice. However, as already noted, there of presacral vertebrae in females (11), development of are reasons for thinking that some derived morpholo- large scales on the ¯anks (16), dorsal and lateral scales gical features, found in the two Algyroides species that lanceolate and more overlapping (17) and an increase in occur in matrix situations, have developed in parallel. As the overlap of ventral scales (20). These features occur the derived morphological features are likely to confer a in both A. moreoticus and A. ®tzingeri and could result performance advantage in these environments and were from a single origin on the main lineage of the genus, therefore probably promoted by natural selection in followed by reversal in the ancestor of A. marchi. Just as them, a double origin of occupation of the environments parsimoniously, in terms of number of steps on the tree, themselves seems most feasible. Although no detailed they may have developed independently and in parallel comparison of ecology and behaviour has been made, in the two species. The latter alternative is preferred for the marked morphological differences between A. mor- the following reasons: eoticus and A. ®tzingeri suggest that have different 1. The features concerned appear to evolve quite easily. structural niches and life modes while similar are not Within the Lacertidae, characters 16, 17 and 20, and identical. sometimes 11, have appeared together several times, in The marked differences in body size in the species of Psammodromus, Adolfus alleni, Tropidosaura, Ichno- Algyroides suggest they probably eat different sizes of tropis and Ophisops (Arnold, 1989a) and do not appear prey, with a shift to smaller food items eaten by smaller to have reversed in the groups. species that is most marked in A. ®tzingeri. Molecular phylogeny of Algyroides lizards 57

Possible performance advantages of character shift which suggests the possibility that a long-established European lizard fauna comprising members of Lacerta The development in the immediate ancestor of and Algyroides may have been largely displaced by the Algyroides of large keeled dorsal scales with a complex recent expansion of Podarcis (Arnold, 1981). Certainly micro-ornamentation that reduces shine, and sombre the species of rock lizards and of Algyroides tend to dorsal coloration, may have enhanced crypsis in wood- have narrow niches in the presence of species of land and woodland-edge habitats by producing a rough, Podarcis that may be in competition with them. For rather dull surface. Such features have also evolved instance, A. nigropunctatus, which climbs considerably, independently in Adolfus and Takydromus, which both is quite restricted in its habitat on the east Adriatic include species that occur in these environments mainland in areas where it occurs with partly scansorial (Arnold, 1989b). It is uncertain what may have caused Podarcis muralis, being absent for instance from urban the reduction in body size in the three more derived environments. In contrast, it is found in a wide range of members of Algyroides. The morphological parallels situations on Corfu where no climbing Podarcis occurs between A. ®tzingeri and A. moreoticus may be function- (Arnold, 1987). ally connected with a greater occupation of vegetation matrixes such as litter and brushwood. In these habitats, large ¯ank scales and an increased overlap of most body PARALLEL EVOLUTION WITH THE CENTRAL scales provides enhanced mechanical protection for the AFRICAN ADOLFUS CLADE body in spiky situations (Arnold, 1973), and an increase in the number of presacral vertebrae may promote Some tropical African lizards, now placed in the genus ¯exibility, facilitating movement through the often Adolfus, were assigned to Algyroides (Boulenger, 1906, complex twisting interstices of the environments 1920; Loveridge, 1957). These are Adolfus africanus, concerned. Adolfus vauereselli and Adolfus alleni. It was later shown The reduction in the strength of dorsal scale keeling that they differed substantially from Algyroides in A. marchi may be functionally connected with the (Arnold, 1973) and they were consequently placed in the way this species uses narrow crevices more than other Equatorial African clade of the Lacertidae (Arnold, Algyroides species. A smooth ¯exible surface is advanta- 1989a, b). Derived features in which the members of geous to maintain position and facilitate movement in Adolfus differ from Algyroides include: medial loop of crevices (Arnold, 1973) and is provided by the weakly clavicle never interrupted posteriorly, basal autotomic keeled scales. The fact there is not a reversal to the caudal vertebrae always with simple transverse pro- primitive relatively small scales found in most crevice- cesses; hemipenis with a well-developed armature and dwelling lacertids suggests that large scales may be just complexly folded lobes (Arnold, 1986); course of ulnar as effective in this situation or that they are retained as a nerve `varanide' or intermediate between the `lacertide' result of phylogenetic constraint. and `varanide' conditions (Julien & Renous-LeÂcuru, 1972). In addition there are six features that de®ne the Equatorial African clade (see Arnold, 1989a, b for BIOGEOGRAPHY OF ALGYROIDES details). Relationships of the species of Adolfus and their near relatives are shown in Fig. 7. The phylogeny of the extant species of Algyroides The resemblance between Algyroides nigropunctatus suggests their immediate ancestor may have had an and Adolfus africanus is particularly striking. However, eastern distribution and that there was dispersal to the phylogenetic analysis shows that most of this results west. Alternatively there may have been a sequence of from primitive shared characters, although there are a vicariance events that began in the east. few, sometimes very distinctive, features that have been The disjunct distribution of the species of Algyroides, independently derived in the two species. These include: each with small individual ranges, is similar to that of a distinctive micro-ornamentation pattern on the dorsal rock lizards in Europe. As their name suggests, these are body scales (Table 1, character 2); large scales on the mainly saxicolous members of Lacerta that have some- dorsum (3) that are blunt, and strongly keeled (18); few times been placed in their own subgenus, Archaeolacerta enlarged dorsal scales between the hind legs (4); a low Mertens 1921, although there is no evidence that number of presacral vertebrae (5); sombre dorsal colora- they form a clade (Harris et al., 1998a). European tion (6). rock lizards include Lacerta bedriagae, L. monticola, There are also marked resemblances between Adolfus L. bonnali, L. aurelioi, L. aranica, L. horvathi, L. mosor- alleni and Algyroides moreoticus and Algyroides ensis, L. oxycephala and L. graeca. In both the ®tzingeri. Apart from large and strongly keeled dorsal European Rock lizards and in Algyroides, species are scales (3, 18), derived similarities include a probable well de®ned morphologically and often represent long increase in the number of presacral vertebrae, especially external branches in molecular phylogenies, suggesting in females (11), large lateral body scales (16), dorsal and that they have been separated from each other for a lateral body scales lanceolate and imbricate (17) and long time. These features contrast with the situation in ventral scales with marked posterior overlap (20). In the far more widely distibuted wall lizards Podarcis, addition, the presence of narrow light dorsolateral 58 D. J. Harris, E. N. Arnold and R. H. Thomas streaks in at least some individuals (24) is shared with precursor of a clade that is not derivable from morpho- A. moreoticus, while a larger proportion of long logical evidence alone. This is another case where presacral ribs in both sexes (12), the occasional presence combining morphological and molecular approaches to of a dark vertebral streak or series of spots (21) and phylogeny reconstruction provides more insight into the possession of a robust tail (22) are shared with evolution than using either one or the other exclusively. A. ®tzingeri. 2. Where both morphology and molecules indicate a More restricted parallels occur between Holaspis topology in which it is equally parsimonious to assume guentheri, a highly modi®ed member of the Adolfus that some anatomical characters may have either arisen clade (Arnold, 1989b), and Algyroides marchi. Both once and then reversed, or alternatively arisen twice and have ossi®cation of the skull reduced (8), fenestrated not changed, considerations of biological process supraocular osteoderms (9) and ancestral keeling on (Arnold, 1996) may allow a choice to be made between dorsal scales reduced or, in the case of Holaspis, absent these two possiblities. In Algyroides the ease with which (18R). such features seem likely to evolve ± based on the The members of the equatorial clade nearly all existence of parallel cases in other groups, evidence that climb quite extensively and are associated with forest there may be strong selection for them, and the lack of and forest-edge situations, so this is likely to have evidence for previous presence of the features in some been their ancestral life-mode, as it appears to have lineages where they might otherwise be expected ± been in Algyroides. Within the general forest environ- suggests that parallelism has occurred rather than re- ment, Adolfus africanus and Algyroides nigropunctatus versal. Once a case for parallelism in the evolution of both climb on tree boles and branches and fallen derived morphological features has been made, this can timber, although they also spend some time on the be used to assess the number of shifts into the environ- ground (Arnold, 1987, 1989b). Adolfus alleni is excep- ments where these attributes confer performance tional in that the phylogeny indicates it has shifted advantage. from forest situations into montane environments 3. Comparisons of evolution in Algyroides and Adolfus above the tree line, where it lives in and around dense demonstrate that, where ancestral forms are similar, clumps of low vegetation (Arnold, 1989b). It conse- striking parallelism may occur in analogous environ- quently resembles Algyroides moreoticus and ments, although the number of characters involved Algyroides ®tzingeri in using vegetation matrixes and is quite small compared to similarity resulting from litter. Holaspis guentheri and Algyroides marchi also primitive resemblance. have ecological similarities, both climbing extensively and using narrow crevices, especially for refuges (Arnold, 1987, 1989b). Acknowledgements At least some of the derived resemblances occuring in Algyroides and the Adolfus clade may consequently have We are very grateful to S. Peltz and especially H. in den arisen as a result of adaptation to the structurally Bosch for material and we thank J. Mackenzie-Dodds similar habitats they occupy. Possible performance (Natural History Museum, London) for practical advantages of some of these features that could have assistance in the laboratory. David Swofford kindly resulted in their being favoured by natural selection gave us permission to use results derived from the test have already been discussed. version of PAUP* 4.0. D. James Harris was supported Not only do Algyroides and the Adolfus clade show during this work by a PhD studentship from the signi®cant parallels in the structural niches they have Natural History Museum, London. occupied and their morphological evolution, but they are both likely to have been derived from ancestors that were super®cially similar in their gross morphology. As REFERENCES has already been suggested, a robust scansorial ancestor similar to Lacerta laevis is likely for Algyroides, and the Arnold, E. N. (1973). Relationships of the Palaearctic lizards most basal member of the Adolfus clade, A. jacksoni,is assigned to the genera Lacerta, Algyroides and Psammodromus externally very like a robust Lacerta. (Reptilia: Lacertidae). Bull. Br. Mus. (Nat. Hist.) Zool. 29: 289±366. Arnold, E. N. (1981). Competition, evolutionary change and montane distributions. In The evolving biosphere: 217±228. Concluding remarks Forey, P. L. (Ed.). London, Cambridge; British Museum (Natural History) and Cambridge University Press. The use of both morphology and mtDNA sequence to Arnold, E. N. (1986). The hemipenis of lacertid lizards (Reptilia: estimate the phylogeny of Algyroides reveals a number Lacertidae): structure, variation and systematic implications. J. of points of broader interest. Nat. Hist. 20: 1221±1257. 1. The DNA evidence corroborates a particular Arnold, E. N. (1987). Resource partitioning among lacertid lizards in southern Europe. J. Zool. (Lond.) B 1: 739±782. morphological estimate of phylogeny in which an Arnold, E. N. (1989a). Towards a phylogeny and biogeography of ancestor with speci®ed anatomical features is assumed the Lacertidae: relationships within an old-world family of but not known. Molecular evidence thus allows lizards derived from morphology. Bull. Br. Mus. (Nat. Hist.). something to be said about the morphology of the Zool. 55: 209±257. Molecular phylogeny of Algyroides lizards 59

Arnold, E. N. (1989b). Systematics and adaptive radiation of secondary structure for the third domain of animal 12S rRNA. equatorial African lizards assigned to the genera Adolfus, Mol. Biol. Evol. 13: 150±169. Bedriagaia, Gastropholis, Holaspis and Lacerta (Reptilia: Hillis, D. M., Moritz, C. & Mable, B. K. (1996). Molecular Lacertidae). J. Nat. Hist. 23: 525±555. systematics. 2nd edn. Sunderland, MA: Sinauer. Arnold, E. N. (1996). The role of biological process in phyloge- Huson, H. D. & Wetzel, R. (1994). SplitsTree, version 1. netics with examples from the study of lizards. Mem. Soc. Ital. Julien, R. & Renous-Lecuru, S. (1972). Variations du trajet du Sci. Nat. Mus. Civ. Stor. Nat. Milano 27: 9±20. nerf ulnaire (ulnaris) et de l'innervation des muscles dorsaux de Bandelt, H.-J. & Dress, A. W. M. (1992a). Split decomposition: a la jambe chez les lacertiliens (Reptiles: Squamates): valeur new and useful approach to phylogenetic analysis and distance systematique et application phylogenetique. Bulle. Mus. Natl data. Mol. Phylogenet. Evol. 1: 242±252. Hist. Nat. Sect. A Zool. Biol. Ecol. Anim. 29: 207±252. Bandelt, H.-J. & Dress, A. W. M. (1992b). A canonical decom- Kishino, H. & Hasegawa, M. (1989). Evaluation of the maximum postion theory for metrics on a ®nite set. Adv. Math. 92: 47± likelihood estimate of the evolutionary tree topologies from 105. DNA sequence data, and the branching order in Hominoidea. BoÈhme, W. (1981). Handbuch der Reptilien und Amphibien J. Mol. Evol. 29: 170±179. Europas: Weisbaden: Akademische Verlagsgesellschaft. Kocher, T. D., Thomas, W. K., Meyer, A., Edwards, S. V., Boulenger, G. A. (1906). Additions to the herpetology of British PaÈaÈbo, S., Villablanca, F. X. & Wilson, A. C. (1989). Dynamics East Africa. Proc. zool. Soc. Lond. 1906: 570 of mitochondrial DNA evolution in animals: ampli®cation and Boulenger, G. A. (1920). Monograph of the Lacertidae. London: sequencing with conserved primers. Proc. Natl Acad. Sci. 86: Trustees of the British Museum (Natural History). 6196±6200. Felsenstein, J. (1985). Con®dence limits on phylogenies: an Loveridge, A. (1957). Check list of the reptiles and amphibians of approach using the bootstrap. Evolution 39: 783±791. East Africa (Uganda; Kenya; Tanganyika; Zanzibar). Bull. Gonzalez, P., Pinto, F., Nogales, M., Jimenez Asensio, J., Muse. Comp. Zool. Harvard 117: 51±362, xxxvi. Hernandez, M. & Cabrera, V. M. (1996). Phylogenetic relation- Mayer, W. & Benyr, G. (1994). Albumin±Evolution und Phyloge- ships of the Canary islands endemic lizard genus Gallotia nese in der Familie Lacertidae. Ann. Naturhist. Mus. Wien 96: (Sauria: Lacertidae), inferred from mitochondrial DNA 621±648. sequences. Mol. Phylogenet. Evol. 6: 63±71. Mayer, W. & Lutz, D. (1990). Chemosystematische Untersu- Gu, X. & Li, W. H. (1996). Bias-corrected paralinear and LogDet chungen zur Phylogenese der Gattung Algyroides und ihrer distances and tests of molecular clocks and phylogenies under systematischen Position gegenuber der Sammelgattung Lacerta nonstationary nucleotide frequencies. Mol. Biol. Evol. 13: (Reptilia: Sauria: Lacertidae). Zool. Anze. 224: 99±105. 1375±1383. Neefs, J.-M., van de Peers, Y., Hendriks, L. & de Wachter, R. Gutell, R. R. (1993). Collection of small subunit (16S and 16S- (1990). Compilation of small ribosomal subunit RNA struc- like) ribosomal RNA structures. Nucleic Acids Res. 21: 3051± tures. Nucleic Acids Res. 18: 2237±2317. 3054. Olmo, E., Odierna, G. & Capriglione, G. (1993). The karyology Harris, D. J. (1997). Estimating the phylogeny of selected lacertid of Mediterranean lacertid lizards. In Lacertids of the Mediter- lizard groups (Reptilia, Lacertidae). PhD thesis, University of ranean region: 61±84. BoÈhme, W., Perez Mellado, V. & London. Maragou P. (Eds). Athens, Bonn, Alicante: Hellenic Zoolo- Harris, D. J., Arnold, E. N. & Thomas, R. H. (1998a). Relation- gical Society. ships of lacertid lizards (Reptilia: Lacertidae) estimated from Rubio, J. L. & Carrascal, L. M. (1994). Habitat selection and mitochondrial DNA sequences and morphology. Proc. R. Soc. conservation of an endemic Spanish lizard Algyroides marchi Lond. 265: 1939±1948. (Reptilia, Lacertidae). Biol. Conserv. 70: 245±250. Harris, D. J., Arnold, E. N. & Thomas, R. H. (1998b). Rapid Rzhetsky, A. & Nei, M. (1995). Tests of applicability of several speciation, morphological evolution, and adaptation to substitution models for DNA sequence data. Mol. Biol. Evol. extreme environments in South African sand lizards (Meroles) 12: 131±151. as revealed by mitochondrial DNA sequences. Mol. Phylogenet. Sambrook, J., Fritsch, E. F. & Maniatis, T. (1989). Molecular Evol. 10: 37±48. cloning: a laboratory manual. New York: Cold Spring Harbour Hasegawa, M. & Hashimoto, T. (1993). Ribosomal RNA trees Press. misleading? Nature (Lond.) 361: 23. Steel, M. A. (1994). Recovering a tree from the leaf colourations it Hasegawa, M., Kishino, H. & Yano, T. (1985). Dating of the generates under a Markov model. App. Math. 7: 19±24. human±ape splitting by a molecular clock of mitochondrial Swofford, D. L. (1993). PAUP: Phylogenetic Analysis Using DNA. J. Mol. Evol. 21: 160±174. Parismony, Version 3.1. Computer program distributed by the Hedges, S. B. & Bezy, R. L. (1994). Reply: xantusiid lizards and Illinois Natural History Survey, Champaign, Illinois. phylogenetic inference. Mol. Phylogene. Evol. 3: 275±278. Swofford, D. L. (1998). PAUP*. Phylogenetic Analysis Using Hedges, S. B., Bezy, R. L. & Maxson, L. R. (1991). Phylogenetic Parsimony (*and Other Methods), Version 4. Sunderland, relationships and biogeography of xantusiid lizards, inferred Massachusetts: Sinauer Associates. from mitochondrial DNA sequences. Mol. Biol. Evol. 8: 767± Yang, Z. (1996). Phylogenetic analysis using parsimony and 780. likelihood methods. J. Mol. Evol. 42: 294±307. Hickson, R. E. , Simon, C., Cooper, A., Spicer, G. S. , Sullivan, J. Zinner, H. (1967). Herpetological collection trips to the Lebanon & Penny, D. (1996). Conserved sequence motifs, alignment and in 1965 and 1966. Isr. J. Zool. 16: 49±58. 60 D. J. Harris, E. N. Arnold and R. H. Thomas

Appendix. Data for specimens used for DNA extraction

Species Locality Source Dates Algyroides ®tzingeri Sardinia H. in den Bosch Received November 1995 Algyroides marchi Sierra de Cazorla, Spain H. in den Bosch Hatched 1989 Algyroides moreoticus Greece H. in den Bosch Received October 1996 Algyroides nigropunctatus Aghios Stefanos, Corfu E. N. Arnold Collected August 1991 Gallotia galloti Bullolo, Tenerife, Canary Islands H. in den Bosch Collected April 1991 Podarcis muralis Near Cannes, France H. in den Bosch Received November 1995 Podarcis taurica Russian Federation S. Peltz Obtained spring 1994