Molecular Phylogenetics and Evolution

Transcription

1 Moleular Phylogenetis and Evolution 58 (2011) Contents lists available at SieneDiret Moleular Phylogenetis and Evolution journal homepage: Between a rok and a hard polytomy: Rapid radiation in the rupiolous girdled lizards (Squamata: Cordylidae) Edward L. Stanley a,b,, Aaron M. Bauer a, Todd R. Jakman a, William R. Branh,d, P. Le Fras N. Mouton e a Department of Biology, Villanova University, 800 Lanaster Ave., Villanova, PA 19085, USA b Rihard Gilder Graduate Shool, Amerian Museum of Natural History, NY 24, USA Department of Herpetology, Bayworld (Formerly Port Elizabeth Museum), P.O. Box 13147, Humewood 6013, South Afria d Researh Assoiate, Department of Zoology, P.O. Box 77000, Nelson Mandela Metropolitan University, Port Elizabeth 6031, South Afria e Department of Botany and Zoology, University of Stellenbosh, Matieland, South Afria artile info abstrat Artile history: Reeived 27 April 2010 Revised 21 August 2010 Aepted 24 August 2010 Available online 8 September 2010 Keywords: Taxonomi revision Rapid radiation Likelihood branh support Phyas New genera Southern Afria Girdled lizards (Cordylidae) are sub-saharan Afria s only endemi squamate family and ontain 80 nominal taxa, traditionally divided into four genera: Cordylus, Pseudoordylus, Chamaesaura and Platysaurus. Previous phylogeneti analysis revealed Chamaesaura and Pseudoordylus to be nested within Cordylus, and the former genera were sunk into the later. This taxonomi revision has reeived limited support due to the study s poor taxon sampling, weakly supported results and possible temporary nomenlatural instability. Our study analyzes three nulear and three mitohondrial genes from 111 speimens, representing 51 ingroup taxa. Parsimony, likelihood and Bayesian analyses of onatenated and partitioned datasets onsistently reovered a omb-like tree with 10, well-supported, monophyleti lineages. Our taxonomi reassessment divides the family into 10 genera, orresponding to these well-supported lineages. Short internodes and low support between the non-platysaur lineages are onsistent with a rapid radiation event at the base of the viviparous ordylids. Ó 2010 Elsevier In. All rights reserved. 1. Introdution The girdled lizards (Cordylidae) are a family of distintively armored siniform lizards endemi to Sub-Saharan Afria. There are 80 named taxa within the group, most of whih are highly adapted for a rok-dwelling lifestyle, though the family enompasses a wide variety of morphologies, life histories and behaviors. The Cordylidae has traditionally been divided into four nominal genera (Loveridge, 1944; Lang, 1991): Platysaurus are highly flattened, lightly armored rok speialists ourring hiefly in southeast Afria, their extreme bauplan allowing them to use the narrowest of raks as retreat sites; Chamaesaura are serpentine lizards with redued limbs and greatly elongated tails that oupy grasslands of southern and eastern Afria; Pseudoordylus are rag-dwelling, moderately armored lizards that our in the Cape Fold and Drakensberg mountain ranges of South Afria, Lesotho and Swaziland; and Cordylus are a morphologially and eologially diverse group of heavily armored lizards that range from South Afria to Angola and Ethiopia. The family has a turbulent Corresponding author. Address: Rihard Gilder Graduate Shool, Amerian Museum of Natural History, Central Park 79th Street, New York, NY 24, USA. address: estanley@amnh.org (E.L. Stanley). taxonomi history, and evolutionary relationships between the members of the Cordylidae remain poorly understood, despite being the subjet of several modern taxonomi treatments (e.g. Lang, 1991; Frost et al., 2001). Gray (1845) plaed all fully-limbed ordylids then known in the family Zonuridae, reognizing the genera Cordylus Gronovius 1763, Zonurus Merrem 1820, Pseudoordylus Smith 1838, Hemiordylus Smith 1838, and Platysaurus Smith 1844, along with what are now gerrhosaurids, laertids of the genus Takydromus, and several anguids. Boulenger s (1884) reassessment of Zonuridae retained only Zonurus (inorporating Cordylus, Pseudoordylus and Hemiordylus) and Platysaurus, adding Chamaesaura (Fitzinger, 1843). The genus Pseudoordylus was resurreted one year later (Boulenger, 1885). This taxonomy remained relatively stable until Stejneger (1936) demonstrated that Cordylus Laurenti 1768 was a senior synonym of Zonurus Merrem Subsequent taxonomi works have onsistently employed Cordylidae as the familial name of the group (Mertens, 1937; FitzSimons, 1943; Loveridge, 1944) although gerrhosaurs have been inluded in the same family by some authors (MDowell and Bogert, 1954; Romer, 1956; Townsend et al., 2004). Lang (1991) presented the first phylogeneti analysis of the Cordyliformes (Gerrhosauridae + Cordylidae) using morphologial data. The resulting phylogeny reovered the serpentiform Cham /$ - see front matter Ó 2010 Elsevier In. All rights reserved. doi: /j.ympev

2 54 E.L. Stanley et al. / Moleular Phylogenetis and Evolution 58 (2011) aesaura as the earliest diverging taxon, with Cordylus sister to a lade omprising Platysaurus and Pseudoordylus. This arrangement was onsistent with previous works (Boulenger, 1884; Loveridge, 1944), whih postulated that extensive armor is plesiomorphi in ordylids and that there was a redution in osteoderms and spinose sales from Cordylus, through Pseudoordylus, to Platysaurus (Fig. 1a). However, this topology nests the oviparous Platysaurus deep within the otherwise viviparous members of the family, a phenomenon otherwise extremely rare among squamates (Lynh and Wagner, 2009). Furthermore, despite employing broad taxon sampling and using a large number of haraters, Lang (1991) did not attempt to resolve speies level relationships and assumed the monophyly of all of the ordylid genera, although he aknowledged that this assumption might prove to be inorret. Frost et al. (2001) analyzed sequene data from two mitohondrial genes, 12S and 16S, for 22 speies in the first moleular phylogeny of ordylids. Their maximum parsimony analysis reovered a very different topology to that of Lang (1991). Platysaurus was shown to be the earliest diverging ordylid lade, whih obviated the need to invoke a reversal in reprodutive strategy, and was onsistent with impliations from ordylid life histories (Mouton and van Wyk, 1997). Additionally, Frost et al. (2001) identified Cordylus and Pseudoordylus as paraphyleti and polyphyleti, respetively, orroborating preliminary results from other studies (Herselman et al., 1992a; Mouton and van Wyk, 1997). Frost et al. s (2001) analysis retrieved a step-like phylogeny, with Chamaesaura and two separate lineages of Pseudoordylus nested among the 15 speies of Cordylus sampled (Fig. 1b). Rather than implement a major taxonomi revision on the basis of an inompletely sampled phylogeny with limited resolution, the authors proposed that Pseudoordylus and Chamaesaura be synonymized with Cordylus. Some authors have adopted this arrangement (du Toit et al., 2002; MConnahie and Whiting, 2003; Cooper, 2005), but the traditional, four-genus taxonomy remains widely employed (Moon, 2001; Curtin et al., 2005; Costandius and Mouton, 2006; Menegon et al., 2006; Alexander and Marais, 2007; Eifler et al., 2007). Moreover, adoption of Frost et al. s (2001) proposal would result in a number of nomenlatural onflits with the neessity for replaement names e.g. Pseudoordylus nebulosus Mouton and Van Wyk (1995) would have beome a junior homonym of Cordylus nebulosus A. Smith 1838 (=Cordylus ataphratus Boie 1828). Some subsequent works have implemented a ombination of the old and new taxonomies (Broadley, 2006; Vitt and Caldwell, 2008), maintaining Pseudoordylus in the synonymy of Cordylus but treating Chamaesaura as a valid genus. These arrangements have no phylogeneti basis but rather reflet the strong preferene of many workers to reflet the morphologial and eologial distintiveness of these attenuate grass-swimmers. (a) Gerrhosauridae Chamaesaura Cordylus Pseudoordylus Platysaurus (b) Gerrhosaurus typius Platysaurus rhodesianus Platysaurus monotropis Cordylus ataphratus Cordylus jordani Cordylus polyzonus Chamaesaura anguina Cordylus warreni Cordylus giganteus Cordylus oeruleopuntatus Pseudoordylus apensis Pseudoordylus nebulosus Cordylus lawreni Cordylus vittifer Cordylus peersi Pseudoordylus melanotus Pseudoordylus mirolepidotus Pseudoordylus namaquensis Cordylus tropidosternum Cordylus maropholis Cordylus minor Cordylus niger Cordylus oelofseni Cordylus ordylus Fig. 1. Phylogeneti relationships of the Cordylidae as proposed by (a) Lang (1991) and (b) Frost et al. (2001).

3 E.L. Stanley et al. / Moleular Phylogenetis and Evolution 58 (2011) Although Frost et al. s (2001) taxon sampling was limited for Chamaesaura (1 of 6 taxa) and Platysaurus (2 of 26 taxa), the monophyly of these two genera has never been alled into question, and in the ase of Platysaurus it has subsequently been orroborated (Sott et al., 2004). However, the partial taxon sampling within Cordylus (15 of 38 taxa) and Pseudoordylus (3 of 10 taxa) limited the study s ability to aurately represent the speies relationships of the family as a whole, and may have ompounded weaknesses in the analytial methods (Heath et al., 2008). Furthermore, the genes utilized in this study may not, by themselves, be apable of resolving relationships at all depths of the tree, as mitohondrial DNA tends to evolve rapidly and beomes saturated at deeper nodes (Overton and Rhoads, 2004). Four primary fators affet the ability of a moleular phylogeneti analysis to aurately estimate true historial relationships among speies: (1) taxon sampling; (2) seletion of geneti markers; (3) amount of sequene data analyzed, and (4) hoie of analytial methods (Swofford et al., 1996). Our study aims to improve on previous analyses in all four of these areas to reover a robust and well-resolved phylogeny of the family Cordylidae. If a strong phylogeneti signal an be reovered by amending these ompounding fators, a new lassifiation may be proposed that is onsistent with traditional groupings based on morphology and life history while still refleting generi monophyly. 2. Materials and methods 2.1. Taxon sampling Sequene data were obtained from 111 speimens, representing 51 ordylid taxa from all four genera. Taxon sampling was partiularly dense in Cordylus and Pseudoordylus (Table 1), as previous studies have reovered these groups to be non-monophyleti. Wherever possible, speimens from separate loalities were inluded for eah taxon. Inreased sampling was employed for widely distributed speies (Cordylus polyzonus and C. ordylus), or speies with geographially disrete, well separated populations (C. oelofseni and C. vittifer) so as to inlude representatives from the speies entire geographi range (Table 2). Following previous phylogeneti assessments (MDowell and Bogert, 1954; Romer, 1956; Lang, 1991; Townsend et al., 2004) three speies of gerrhosaurid, Gerrhosaurus validus, G. nigrolineatus and Cordylosaurus subtessellatus, were used as outgroups for the study Gene sampling We analyzed sequene data from three nulear and three mitohondrial genes. The mitohondrial genes 12S, 16S and ND2 were utilized in previous studies of ordylids (Frost et al., 2001; Odierna et al., 2002; Daniels et al., 2004) and were inluded in this study to failitate omparisons with these prior analyses. Three rapidly evolving nulear genes were seleted: Prolatin reeptor gene PRLR (Townsend et al., 2008), Myosin Heavy hain 2 MYH2 (Whiting et al., 2006) and Kinesin Family Member 24 KIF24 (Portik et al., Table 1 The total number of Cordylidae taxa, the number of taxa inluded by Frost et al. (2001) and the number of taxa present in this study. Genus Total named taxa Taxa in Frost et al. (2001) Cordylus Pseudoordylus Chamaesaura Platysaurus Taxa in this study in press). All three genes are protein oding and MYH2 also ontains a rapidly evolving, non-oding intron Moleular data Total genomi DNA was isolated from the liver or skeletal musle of speimens preserved in 95% ethanol using the Qiagen DNeasy tissue kit (Valenia, CA, USA). DNA from fresh tissues of members of the Cordylus warreni omplex and several other key speies was isolated at the Leslie Hill Laboratories, South Afrian National Biodiversity Instituite (SANBI), Kirstenbosh, South Afria, using a the protools of Tolley et al. (2004). Target genes were amplified using double-stranded Polymerase Chain Reation (PCR). 2.5 ll of the extrated genomi DNA was ombined with 2.5 ll forward primer (8.p.p.m), 2.5 ll reverse primer (8.p.p.m), 2.5 ll dinuleotide pairs, 2.5 ll 5 buffer, 2.5 ll MgCl 10 buffer, 0.18 ll Taq polymerase and 8.92 ll H 2 0. PCR yling was exeuted on an Eppendorf Masteryler gradient thermoyler and eah primer-set was initially amplified under the following onditions: initial denaturation for 2 min at 95 C followed by 95 C for 35 s, annealing at 50 C for 35 s, and extension at 72 C for 95 s (Greenbaum et al., 2007). Produts were visualized with 1.5% agarose gel eletrophoresis. If neessary, annealing temperatures were modified aordingly (Table 3). Target produts were treated with AMPure magneti bead solution (Agenourt Biosiene, Beverly, MA, USA) to remove byproduts of the PCR proess. The leaned PCR produt was then prepared for sequening with the DYEnami ET Dye Terminator Kit (GE Healthare, Pisataway, NJ, USA). Sequening reations were purified with CleanSeq magneti bead solution (Agenourt Biosiene, Beverly, MA, USA) and analyzed with an ABI 3700 automated sequener. All genes were sequened from both 3 0 and 5 0 ends separately and internal primers were used for genes over 800 bp long. The omplimentary and ontiguous sequenes were aligned using the program Genious (Drummond et al., 2008). Ambiguous or onfliting bases were oded as heterozygotes. Multiple sequene alignment was performed with Musle (Edgar, 2004) and visualized with Malade (Maddison and Maddison, 2000) to onfirm the amino aid reading frame and to hek for stop odons Phylogeneti analysis We employed a pluralisti approah for phylogeneti analysis, performing separate analyses with the three most ommonly employed optimality riteria, Maximum Parsimony (MP), Maximum Likelihood (ML) and Bayesian Inferene (BI). Separate analyses were performed on eah individual gene as well as onatenated sets of mitohondrial genes (ND2, 16S and 12S) nulear genes (PRLR, MYH2 and KIF24), and all genes. The maximum Parsimony analyses were run using PAUP* (Swofford, 2002) under the following onditions: 0 random addition repliates, tree bisetion-reonnetion branh swapping, zero-length branhes ollapsed to yield polytomies, and gaps treated as missing data. Nodal support for this analysis was estimated using non-parametri bootstrapping (0 iterations) and Goodman-Bremer support (Bremer, 1994). ModelTest (Posada and Crandall, 1998) was run to identify the most likely model of evolution for eah of the individual genes. Both ML and BI analyses were partitioned by odon and gene, eah partition employing the appropriate model estimated by ModelTest. The Maximum Likelihood analyses were performed using RAx- ML HPC (Stamatakis et al., 2008) on the CIPRES server. The analysis was performed using a GTR gamma model with gaps

4 56 E.L. Stanley et al. / Moleular Phylogenetis and Evolution 58 (2011) Table 2 Vouher numbers, loalities and Genbank aession numbers for samples. ( = Republi of South Afria). Taxon Vouher Loality Coordinates 16s 12s ND2 PRLR KIF24 MYH2 Chamaesaura aenea 1 QP0041 Free State, S, E Chamaesaura aenea 2 QP0043 KwaZulu Natal, S, E Chamaesaura aenea 3 QP0042 Freestate, S, E Chamaesaura anguina 1 PEM195 Eastern Cape, S, E Chamaesaura anguina 2 SU2 Western Cape, S, E Chamaesaura anguina 3 SU1 Western Cape, S, E Chamaesaura a. tenuior PEMR Arusha, Kenya S, E Cordylus aridus 1 PEMR16376 Western Cape, S, E Cordylus aridus 2 PEMR16371 Western Cape, S, E Cordylus beraduii WRB0037 Mtera, Tanzania S, E Cordylus breyeri MBUR00320 Limpopo, S, E Cordylus ampbelli 1 MCZ27028 Namibia S, E Cordylus ampbelli 2 MCZ27028 Namibia S, E Cordylus ataphratus 1 SU1 Western Cape, S, E Cordylus ataphratus 2 MBUR01792 Northern Cape, S, E Cordylus oeruleopuntatus KTH329 Western Cape, S, E Cordylus oeruleopuntatus QP0046 Western Cape, S, E Cordylus oeruleopuntatus SU Western Cape, S, E Cordylus oeruleopuntatus QP0044 Western Cape, S, E Cordylus ordylus 1 AMB8168 Eastern Cape, S, E Cordylus ordylus 2 PEMR9714 Eastern Cape, S, E Cordylus ordylus 3 PEMR16382 Western Cape, S, E Cordylus ordylus 4 PEMR17466 Eastern Cape, S, E Cordylus ordylus 5 PEMR17464 Eastern Cape, S, E Cordylus ordylus 6 PEMR17467 Eastern Cape, S, E Cordylus ordylus 7 PEMR13511 Western Cape, S, E Cordylus ordylus 8 AMB8865 Western Cape, S, E Cordylus giganteus 1 MJC 5403 Free State, S, E Cordylus giganteus 2 MJC 6638 Free State, S, E Cordylus giganteus 3 MJC 6640 Free State, S, E Cordylus giganteus 4 MJC 6642 Free State, S, E Cordylus imkeae 1 MBUR01795 Northern Cape, S, E Cordylus imkeae 2 MBUR01796 Northern Cape, S, E Cordylus jonesi 1 AMB8396 Limpopo, S, E Cordylus jonesi 2 AMB8310 Limpopo, S, E Cordylus jordani 1 AMB5876 Namibia S, E Cordylus jordani 2 MCZ27023 Namibia S, E HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ HQ167421

5 E.L. Stanley et al. / Moleular Phylogenetis and Evolution 58 (2011) Table 2 (ontinued) Taxon Vouher Loality Coordinates 16s 12s ND2 PRLR KIF24 MYH2 Cordylus lawreni PEM285 Northern Cape, S, HQ HQ HQ HQ HQ E Cordylus mahadoi 1 KTH09059 Humpata, S, HQ HQ HQ HQ HQ HQ Angola 59 E Cordylus mahadoi 2 KTH09080 Humpata, S, HQ HQ HQ HQ HQ HQ Angola 17 E Cordylus maropholis 1 AMB8874 Western Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus maropholis 2 AMB8873 Western Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus mlahlani 1 AMB8855 Western Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus mlahlani 2 SU1 Western Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus meulae 1 PEMR16203 Meula, S, HQ HQ HQ HQ HQ Mozambique 21 E Cordylus meulae 2 PEMR16202 Meula, S, HQ HQ HQ HQ HQ Mozambique 21 E Cordylus meulae 3 PEMR16165 Meula, S, HQ HQ167 - HQ HQ HQ Mozambique 19 E Cordylus meulae 4 PEMR16164 Meula, S, HQ HQ HQ HQ HQ Mozambique 19 E Cordylus minor SU Northern Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus mossambius PEMR5227 Mozambique S, HQ HQ HQ HQ HQ HQ E Cordylus namaquensis 1 AMB6848 Namibia S, HQ HQ HQ HQ HQ E Cordylus namaquensis 2 AMB6849 Namibia S, HQ HQ HQ HQ HQ E Cordylus niger 1 AMB8875 Western Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus niger 2 SU1 Western Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus oelofseni 1 SU1 Western Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus oelofseni 2 SU2 Western Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus oelofseni 3 AMB8851 Western Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus oelofseni 4 AMB8860 Western Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus oelofseni 5 AMB8862 Western Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus polyzonus 1 A38345 Namibia S, HQ HQ HQ HQ HQ HQ E Cordylus polyzonus 2 JM1117 Northern Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus polyzonus 3 PEMR17462 Eastern Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus polyzonus 4 SU1 Western Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus peersi MB20710 Northern Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus pustulatus visser Namibia S, HQ HQ HQ HQ HQ HQ E Cordylus regius AMB6171 Eastern S, HQ HQ HQ HQ HQ HQ Zimbabwe 16 E Cordylus rhodesianus 1 ELSPET4 aptive Unknown HQ HQ HQ HQ HQ HQ Cordylus rhodesianus 2 ELSPET5 aptive Unknown HQ HQ HQ HQ HQ HQ Cordylus tasmani 1 PEMR17394 Eastern Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus tasmani 2 PEMR15012 Eastern Cape, S, HQ HQ HQ HQ HQ HQ E Cordylus tropidosternum 1 WRB0038 Tanzania Unknown HQ HQ HQ HQ HQ Cordylus tropidosternum 2 WRB0042 Tanzania Unknown HQ HQ HQ HQ HQ Cordylus ukingensis WRB0039 Uzungwe Mts, S, HQ HQ HQ HQ HQ Kenya 43 E Cordylus vandami 1 AMB8292 Mpumalanga, S, HQ HQ HQ HQ HQ HQ E Cordylus vandami 2 AMB8195 Limpopo, S, HQ HQ HQ HQ HQ HQ E Cordylus vandami 3 AMB8193 Limpopo, S, HQ HQ HQ HQ HQ HQ E Cordylus vittifer 1 AMB6073 Mpumalanga, S, E HQ HQ HQ HQ HQ HQ (ontinued on next page)

6 58 E.L. Stanley et al. / Moleular Phylogenetis and Evolution 58 (2011) Table 2 (ontinued) Taxon Vouher Loality Coordinates 16s 12s ND2 PRLR KIF24 MYH2 Cordylus vittifer 2 AMB8274 Limpopo, S, HQ HQ HQ HQ HQ HQ E Cordylus vittifer 3 AMB8603 Swaziland S, HQ HQ HQ HQ HQ HQ E Cordylus warreni warreni ELS012 Mpumalanga, S, HQ HQ HQ HQ HQ HQ E Cordylus w. barbertonensis RCBS2133 Mololotja, S, HQ HQ HQ HQ HQ HQ Swaziland 59 E Cordylus w. depressus 1 MFB141 Limpopo, S, HQ HQ HQ HQ HQ HQ E Cordylus w. depressus 2 MCZF38871 Limpopo, S, HQ HQ HQ HQ HQ HQ E Gerrhosaurus validus AMB 6090 Limpopo, S, HQ HQ HQ HQ HQ E Gerrhosaurus nigrolineatus AMB 8339 Limpopo, S, HQ HQ HQ HQ HQ E Cordylosaurus AMB 4649 Northern Cape, S, HQ HQ HQ HQ HQ subtessellatus 1 23 E Cordylosaurus AMB 6928 Namibia S, HQ HQ HQ HQ HQ subtessellatus 2 07 E Platysaurus broadleyi AMB 4944 Northern Cape, S, HQ HQ HQ HQ HQ HQ E Platysaurus apensis AMB 4524 Northern Cape, S, HQ HQ HQ HQ HQ HQ E Platysaurus minor AMB8411 Limpopo, S, HQ HQ HQ HQ HQ HQ E Platysaurus mithelli MEEP XS173 Malawi S, HQ HQ HQ HQ HQ HQ E Platysaurus i. intermedius NMBR Limpopo, S, HQ HQ HQ HQ HQ HQ E Platysaurus i. nigresens PEMR Mashatu Unknown HQ HQ HQ HQ HQ HQ sample Botswana Pseudoordylus apensis 1 AMB8859 Western Cape, S, HQ HQ HQ HQ HQ HQ E Pseudoordylus apensis 2 AMB8857 Western Cape, S, HQ HQ HQ HQ HQ HQ E Pseudoordylus apensis 3 PEMR16378 Western Cape, S, HQ HQ HQ HQ HQ HQ E Pseudoordylus langi 1 NMBR8556 KwaZulu Natal, S, HQ HQ HQ HQ HQ HQ E Pseudoordylus langi 2 NMBR8555 KwaZulu Natal, S, HQ HQ HQ HQ HQ HQ E Pseudoordylus melanotus AMB8210 Limpopo, S, HQ HQ HQ HQ HQ HQ E Pseudoordylus SU4 Western Cape, S, HQ HQ HQ HQ HQ HQ mirolepidotus 1 00 E Pseudoordylus SU3 Western Cape, S, HQ HQ HQ HQ HQ HQ mirolepidotus 2 00 E Pseudoordylus SU1 Western Cape, S, HQ HQ HQ HQ HQ HQ nebulosus 1 54 E Pseudoordylus SU2 Western Cape, S, HQ HQ HQ HQ HQ HQ nebulosus 2 00 E Pseudoordylus sp. PEMR2701 Eastern Cape, S, HQ HQ HQ HQ HQ HQ Transkei 47 E Pseudoordylus spinosus 1 NMBR8572 KwaZulu Natal, S, HQ HQ HQ HQ HQ HQ E Pseudoordylus spinosus 2 NMB R8572 KwaZulu Natal, S, HQ HQ HQ HQ HQ HQ E Pseudoordylus spinosus 3 NMB R8570 KwaZulu Natal, S, HQ HQ HQ HQ HQ HQ E Pseudoordylus subviridis 1 NMBR8558 KwaZulu Natal, S, HQ HQ HQ HQ HQ HQ E Pseudoordylus subviridis 2 NMBR8561 KwaZulu Natal, S, HQ HQ HQ HQ HQ HQ E Pseudoordylus MFB Limpopo, Unknown HQ HQ HQ HQ HQ HQ transvaalensis 1 Pseudoordylus transvaalensis 2 NMBR8548 Limpopo, S, E HQ HQ HQ HQ HQ HQ treated as missing data, 25 distint rate ategories, and run for 0 rapid bootstrap iterations. Likelihood deay index values were also alulated for eah node of the tree, in addition to bootstrap support. This was ahieved by the following method. A maximum likelihood tree was reated from the partitioned dataset using GARLI (Zwikl, 2006) under a GTR+gamma model, with the default geneti algorithm settings. This tree was used to reate a ommand file from

7 E.L. Stanley et al. / Moleular Phylogenetis and Evolution 58 (2011) Table 3 Primer information for the genes utilized in this study. The PCR olumn denotes the number of repeated yles/annealing temp ( C) used in the PCR. Primer Gene Referene Sequene PCR 16Sa 16S Simon et al. (1994) 5 0 CGCCTGTTTATCAAAAACAT /52 16Sb 16S Simon et al. (1994) 5 0 CCGGTCTGAACTCAGATCACGT /52 12sf700 12S This study 5 0 AAACTGGGATTAGATACCCCACTAT /52 12sr600 12S This study 5 0 GAGGGTGACGGCGGTGTGT /52 L4437 ND2 Maey et al. (1997) 5 0 AAGCTTTCGGGCCCATACC /52 H5540 ND2 Maey et al. (1997) 5 0 TTTAGGGCTTTGAAGGC /52 R102 ND2 This study 5 0 CAGCCTAGGTGGGCGATTG 3 0 / PRLRf1 PRLR Townsend et al. (2008) 5 0 GACARYGARGACCAGCAACTRATGCC /54 PRLRr1 PRLR Townsend et al. (2008) 5 0 GACYTTGTGRACTTCYACRTAATCCAT /54 Kif24f Kif24 Portik et al. (2010) 5 0 WGGCTGCTGRAAYTGCTGGTG /50 Kif24r Kif24 Portik et al. (2010) 5 0 SAAACGTRTCTCCMAAACGCATCC /50 MYH2f MYH2 This study 5 0 GAACACCAGCCTCATCAACC /52 MYH2r MYH2 This study 5 0 TGGTGTCCTGCTCCTTCTTC /52 the Partition Branh Support program, TreeRot v.3 (Sorenson and Franzosa, 2007). The output ommand file ontains a series of onstraint trees that represent every node on the original ML tree and eah of the onstraint trees was onverted to serve as a negative onstraint file. Eah onstraint file was inorporated into a GARLI analyses and run on the original dataset under the original onditions. The Likelihood deay index is the differene between the log likelihood sore of eah onstrained tree and the log likelihood sore of the unonstrained tree for eah node. The Bayesian analyses were onduted using MrBayes 3.1 (Ronquist and Huelsenbek, 2003; Huelsenbek and Ronquist, 2001) with default priors. Two runs were performed for 10,000,000 generations and the Markov hains were sampled every 0 generations. If adequate onvergene had not ourred after 5 million generations, additional generations were run until the average standard deviation of split frequenies was less than The log likelihood was plotted against generation to identify the onvergene point and the burn-in disarded. Topologial onvergene was tested for using the program Are We There Yet (Nylander et al., 2008). Nodes that returned lade posteriors above 0.95 were onsidered signifiantly supported. To ompare the similarity of phylogeneti signal between the different genes, partition homogeneity tests (with repliations) were onduted aross the entire data set and between all individual partitions (Baker and DeSalle, 1997). The topologies of individual gene trees were ompared to hek for any obvious onflits and a Partitioned Branh Support analysis was arried out using TreeRot V.3 (Sorenson and Franzosa, 2007). Finally, the dataset was analyzed using Phyas, version (Lewis et al., 2009), whih allows unresolved tree topologies to be sampled during the ourse of a phylogeneti analysis in addition to fully-resolved tree topologies. A GTR+I+C model was used on the onatenated dataset, as Phyas does not urrently allow partitioning. The prior on the gamma shape parameter was set as an exponential distribution with a mean of 0.5. Polytomies were allowed, and an exponential distribution with mean 1.0 (e) was set as the polytomy prior, as suggested by Lewis et al. (2005). Two MCMC hains were run for 500,000 yles eah, with trees sampled every 10 yles (one yle is equivalent to over generations in MrBayes). The first 5000 trees were disarded as burn-in and the remaining trees summarized with TreeAnnotator V ( beast.bio.ed.a.uk). 3. Results 3.1. Gene suess All six genes were suessfully reovered with PCR amplifiation. However, for several speies ND2 onsistently failed to amplify (Chamaesaura anguina tenuior, Cordylus lawreni, C. namaquensis, C. beraduii, C, tropidosternum C. ukingensis and C. meulae), so separate analyses were run on the onatenated dataset with these taxa exluded. These runs reovered the same topology as the full dataset, with omparable support for all lades. All the sequened gerrhosaurid speies ontained a large insertion in MYH2 making it impossible to align with the ingroup sequenes. Platysaurus represented the outgroup for remaining ordylids in analyses of MYH2. The onatenated dataset totaled 4503 base pairs, ontaining 532 autapomorphi haraters and 1975 parsimony-informative haraters with a total of 2983 unique patterns in the data matrix (Table 4) Phylogeneti relationships of the Cordylidae Dense taxon sampling and analysis of multiple nulear and mitohondrial geneti markers reovers a phylogeny that ontains 10 well-resolved lineages (lades A J in Figs. 2 and 3). The parsimony analysis reovered 32 trees with a length of 11,348. The best ML tree had a log likelihood sore of The same 10 monophyleti ordylid lineages were also onsistently reovered from MP, ML, and BI analyses of the individual genes. Table 4 Gene lengths in base pairs, number of informative sites, perentage of informative sites, perentage of nodes >0.95 posterior probability from Bayesian analysis, and appropriate model of evolution from ModelTest (Posada and Crandall, 1998). Gene Length (bp) Informative sites % Informative % Sig. nodes Model used 16S GTR+I+C 12S GTR+I+C ND GTR+I+C PRLR TVM+C Kif K81uf+C MYH GTR+I+C mtdna GTR+I+C ndna K81uf+C All GTR+I+C

8 60 E.L. Stanley et al. / Moleular Phylogenetis and Evolution 58 (2011) Fig. 2a. Maximum likelihood phylogram of the Cordylidae, based on a onatenated dataset of six genes. Likelihood bootstrap support shown above branhes, Posterior probabilities below branhes and olored irles on eah node represent the likelihood deay index values. Well-supported lineages, A I, shown with photos of representative speies: A = Platysaurus intermedius, B=Cordylus giganteus,c.cordylus oeruleopuntatus, D=Chamaesaura anguina, E=Pseudoordylus m. melanotus, F=Cordylus ataphratus, G = Cordylus jordani, H= Cordylus namaquensis and I = Pseudoordylus apensis.

9 E.L. Stanley et al. / Moleular Phylogenetis and Evolution 58 (2011) Clades A I, figure 2a Log likelihood differene (LLD) values = 30 = = = 5 10 = < Cordylus vittifer Cordylus vittifer Cordylus vittifer 3 Cordylus jonesi Cordylus jonesi Cordylus mahadoi Cordylus mahadoi Cordylus rhodesianus Cordylus rhodesianus 2 Cordylus beraduii 92 Cordylus ukingensis Cordylus tropidosternum Cordylus tropidosternum Cordylus meulae Cordylus meulae 2 Cordylus meulae 3 Cordylus meulae 4 Cordylus maropholis Cordylus maropholis 2 64 Cordylus mlahlani Cordylus mlahlani 2 79 Cordylus imkeae Cordylus imkeae Cordylus minor 1.0 Cordylus aridus Cordylus aridus 2 Cordylus niger Cordylus niger 2 Cordylus oelofseni Cordylus oelofseni Cordylus oelofseni Cordylus oelofseni Cordylus oelofseni Cordylus ordylus 3 Cordylus ordylus 4 Cordylus ordylus Cordylus ordylus 1 Cordylus ordylus 5 Cordylus ordylus 6 Cordylus ordylus 7 Cordylus tasmani 1 Cordylus ordylus 8 Cordylus tasmani 2 J Fig. 2b. Maximum likelihood phylogram of the Cordylidae, based on a onatenated dataset of six genes. Likelihood bootstrap support shown above branhes, Posterior probabilities below branhes and olored irles on eah node represent the likelihood deay index values. Well-supported lades shown with photos of representative speies: Cordylus tropidosternum, Cordylus maropholis and Cordylus ordylus. Parametri bootstrapping, log likelihood differene, Goodman- Bremer support and posterior probabilities of the nodes within these lineages and at the base of the tree are generally high, while support for the nodes between the lineages is onsistently lower. Analysis of subsets of the genes returned the same pattern of support as the onatenated dataset and, although some variation is seen in gene topologies, the majority of onflit ours around the weakly supported nodes at the base of the non-platysaurordylids. The six Platysaurus speies are onsistently reovered as a monophyleti group (group A) that is sister to all other ordylids. The northern speies P. mithelli is reovered at the base of this lade with good support. A lade omprising the southwestern speies, P. broadleyi and P. apensis, is sister to the remaining three speies. Platysaurus intermedius is shown to be paraphyleti, with P. i. nigresens sister to a lade of P. minor and P. i. intermedius. The nine remaining major ordylid lineages are separated by short, often poorly supported internodes in all analyses. The lineage that ontains Cordylus giganteus and members of the C. warreni omplex is sister to all other non-platysaurs. The remaining ordylids fall into two sublades, the first ontaining the speies of robust Pseudoordylus, Chamaesaura and Cordylus oeruleopuntatus, the seond ontaining the two speies of graile Pseudoordylus and the remaining Cordylus speies. Strong support is found for the monophyly of Cordylus giganteus and the members of the Cordylus warreni lade (group B). Four major lineages are reovered within the Cordylus warreni omplex: (1) the Highveld and bushveld speies Cordylus vandami and C. breyeri; (2) the Soutpansberg girdled lizard C. warreni depressus; (3) the hiefly lowveld forms C. warreni warreni and C. w. barbertonensis and (4) two speies from Zimbabwe and Mozambique, C. mossambius and C. regius. The three Chamaesaura taxa form a monophyleti group (D), with Chamaesaura anguina anguina sister to Chamaesaura a. tenuior. Analysis of the nulear genes MYH2 and Kif24 and the onatenated dataset reover Chamaesaura with Cordylus oeruleopuntatus, while analysis of the onatenated mtdna genes plaes C. oeruleopuntatus sister to group J. The large-bodied Pseudoordylus form a well-supported monophyleti group (E). Strong support for the basal position of Pseudoordylus langi is reovered for all analyses. The remaining robust

10 62 E.L. Stanley et al. / Moleular Phylogenetis and Evolution 58 (2011) Goodman-Bremer support = > 21 = = 4 10 = 2 3 = Cordylosaurus subtessellatus 1 Cordylosaurus subtessellatus 2 Gerrhosaurus validus Gerrhosaurus nigrolineatus Platysaurus mithelli Platysaurus apensis Platysaurus broadleyi Platysaurus i. nigresens Platysaurus i. intermedius Platysaurus minor Cordylus giganteus 4 Cordylus giganteus 1 Cordylus giganteus 2 Cordylus giganteus 3 Cordylus w. depressus 1 Cordylus w. depressus 2 Cordylus w. barbertonensis Cordylus w. warreni B Cordylus regius Cordylus mossambius Cordylus breyeri Cordylus vandami 3 Cordylus vandami 1 Cordylus vandami 2 Cordylus lawreni Cordylus peersi Cordylus namaquensis 1 Cordylus namaquensis 2 H Cordylus pustulatus Cordylus ampbelli 1 Cordylus ampbelli 2 Chamaesaura aenea 2 Chamaesaura aenea 1 Chamaesaura aenea 3 Chamaesaura a. tenuior D Chamaesaura anguina 1 Chamaesaura anguina 2 Chamaesaura anguina 3 Pseudoordylus langi 1 Pseudoordylus langi 2 Pseudoordylus sp. Transkei Pseudoordylus mirolepidotus 1 Pseudoordylus mirolepidotus 2 Pseudoordylus melanotus Pseudoordylus transvaalensis 1 Pseudoordylus transvaalensis 2 Pseudoordylus m. subviridis 2 Pseudoordylus m. subviridis 1 Pseudoordylus spinosus 1 Pseudoordylus spinosus 2 Pseudoordylus spinosus 3 Pseudoordylus nebulosus 1 Pseudoordylus nebulosus 2 Pseudoordylus apensis 1 Pseudoordylus apensis 3 Pseudoordylus apensis 2 Cordylus ataphratus 1 Cordylus ataphratus 2 Cordylus jordani 1 Cordylus jordani 2 Cordylus polyzonus 1 Cordylus polyzonus 3 Cordylus polyzonus 2 Cordylus polyzonus 4 Cordylus oeruleopuntatus 1 Cordylus oeruleopuntatus 2 Cordylus oeruleopuntatus 3 Cordylus oeruleopuntatus 4 Cordylus mahadoi 1 Cordylus mahadoi 2 Cordylus jonesi 1 Cordylus jonesi 2 Cordylus rhodesianus 1 Cordylus rhodesianus 2 Cordylus vittifer 1 Cordylus vittifer 2 Cordylus vittifer 3 Cordylus beraduii Cordylus ukingensis Cordylus tropidosternum 1 Cordylus tropidosternum 2 Cordylus meulae 1 Cordylus meulae 2 Cordylus meulae 3 Cordylus meulae 4 Cordylus maropholis 1 Cordylus maropholis 2 Cordylus mlahlani 1 Cordylus mlahlani 2 Cordylus imkeae 1 Cordylus imkeae 2 Cordylus minor Cordylus aridus 1 Cordylus aridus 2 Cordylus niger 1 Cordylus niger 2 Cordylus oelofseni 1 Cordylus oelofseni 2 Cordylus oelofseni 3 Cordylus oelofseni 4 Cordylus oelofseni 5 Cordylus ordylus 7 Cordylus ordylus 1 Cordylus ordylus 3 Cordylus ordylus 4 Cordylus ordylus 2 Cordylus ordylus 6 Cordylus ordylus 5 Cordylus tasmani 1 Cordylus ordylus 8 Cordylus tasmani 2 G A F J I C E Fig. 3. Maximum Parsimony ladogram of the Cordylidae based on a onatenated dataset of six genes (50% majority rule from 32 trees). Non-parametri bootstrap support shown above branhes and shaded irles on eah node represent the Goodman-Bremmer support values. Pseudoordylus are divided into two lineages: P. mirolepidotus and a putative speies from the Transkei (Branh, 1998) represent a southern lade and the speies of the Drakensberg and eastern esarpment form a seond lineage. Within the later group, P. melanotus subviridis is polyphyleti; one individual lusters with P. spinosus, with the remaining speimen sister to this lade. Pseudoordylus transvaalensis and P. melanotus form a sublade that is sister to the P. spinosus-subviridis group. Cordylus polyzonus and C. jordani are unambiguously reovered as sister taxa (group G). Relationships within C. polyzonus are well resolved, with the Namibian sample (C. polyzonus 1, Figs. 2a and 3) sister to the South Afrian forms. Analysis of the fully onatenated dataset reovers Cordylus ataphratus (group F) as sister to the polyzonus lade. The support for this pairing in the onatenated tree results entirely from the strong signal of the mtdna. No evidene of this relationship is seen in the individual or the onatenated analyses of the nulear genes. Separate analyses of the genes Kif24 and MYH reover Cordylus ataphratus as sister to all nonplatysaurs exlusive of C. giganteus and the members of the C. warreni omplex.

11 E.L. Stanley et al. / Moleular Phylogenetis and Evolution 58 (2011) A monophyleti group of ordylids from Namaqualand and entral Namibia is returned by all analyses (group H). This group ontains two well-supported, deeply divergent lineages: a Namibian lade, omprising Cordylus pustulatus, C. ampbelli and C. namaquensis and a Northern Cape lade, of C. peersi and C. lawreni. There is high support for a lade ontaining the two reiproally monophyleti speies of graile Pseudoordylus (group I). A deep divergene is observed between P. apensis from the Northern Langeberg and those from the western esarpment. The final lineage (group J) is split into two geographially disjunt lades. One ontains the small Cordylus that our mainly from Mpumalanga, South Afria northwards. Within this, the southernmost speies, C. vittifer, is reovered as the basal member and is represented in our sample by three well-separated lineages. Cordylus ukingensis is sister to C. tropidosternum and C. meulae, with Cordylus beraduii reovered outside this lade. Cordylus jonesi is sister to C. mahadoi plus C. rhodesianus. The seond lade ontains the small Cordylus speies found in the Cape Fold Mountains and throughout the Western Cape. This group ontains a well-supported sublade of dwarf taxa, C. imkeae, C. aridus and C. minor, with C. mlahlani sister to this group. Cordylus ordylus is sister to three distint lineages of C. oelofseni, and these together form a monophyleti group with Cordylus niger. Cordylus tasmani is nested within C. ordylus. Likelihood and Bayesian analysis of the onatenated dataset reover Cordylus maropholis at the base of the lade of dwarf Cordylus, though maximum parsimony analysis plaes C. maropholis at the base of the entire southwestern lineage, and the Phyas analysis reovers a polytomy between the dwarf group, the C. ordylus group and C. maropholis Gene onflits Partition Homogeneity Tests reveal signifiant heterogeneity between all genes exept ND2 and Kif24. Partitioned Branh Support analysis reovers gene onflit at several nodes but in the majority of ases the onfliting Goodman-Bremer support values were low. The only instane of onflit between mtdna and ndna with high Goodman-Bremer support was the pairing of Cordylus polyzonus with Cordylus ataphratus, reovered with strong support by mtdna but not ndna. This relationship had low support from parametri bootstraps and deay indies in the likelihood and parsimony analyses of the onatenated dataset. When a polytomy prior was added in the Phyas analysis there was a signifiant redution in the posterior probabilities of the nodes at the base of the non-platysaur ordylids, with two nodes ollapsing to polytomies. however, differ signifiantly between the phylogenies, although neither analysis resulted in strong support for these nodes. The onsistently weak support and short internodes at the base of the non-platysaur lineages for all analyses is onsistent with a rapid radiation event. In this senario, all nine lineages (B J) diverged over a short period of time. The members of eah lineage onform tightly to a morphologial and eologial phenotype, and the large amount of variation seen within the family ours between these lades, not within them. There are some notable exeptions, e.g. the obligate terriolous lizards Cordylus maropholis and C. ukingensis have onvergent morphologies, but were found at different plaes within the generally rupiolous lade of small, typial Cordylus Taxonomi revision of the Cordylidae If the rapid radiation onstitutes a hard polytomy, several lassifiation options are available. The lassifiation proposed by Frost et al. (2001) is still fully ompatible with our revised phylogeny. Their two-genus lassifiation aptures the lear evolutionary distintion between the oviparous Platysaurus and viviparous Cordylus lineages. However, the onservative nature of this arrangement has reeived limited support, primarily beause it plaes the serpentine Chamaesaura in the synonymy of Cordylus and does not reflet the morphologial and eologial variation seen within the Cordylidae. Given the strong support for eah of the ten lineages, we adopt an alternative lassifiation that reognizes less inlusive, but morphologially and eologially distintive groups as genera. Consequently, we ontinue to reognize the existing genera, Cordylus Laurenti 1768, Platysaurus Smith 1844, Pseudoordylus, Smith 1838, Chamaesaura, Shneider 1801, and Hemiordylus Smith 1838, and propose five new genera to aommodate the well-supported lades within this family. In the generi Content (below), an asterisk indiates taxa unsampled in this study; however, based on morphologial synapomorphies we onsider generi alloation unambiguous. Cordylidae Mertens, Type Genus: Cordylus Laurenti Content: Platysaurinae subfam. nov and Cordylinae Mertens, Definition: Short, distally divided tongue overed in long papillae. Large square parietal plates present. Cranial osteoderms invariably present. Body sales large and in regular transverse rows or granular. Large retangular ventral sales. Spiny or strongly keeled audal sales arranged in whorls. Body often depressed. Femoral pores present. 4. Disussion The ten ordylid lades reovered by this study orrespond well to geographi distributions and morphology. There is great heterogeneity aross the Cordylidae, but the phenotype of eah sublineage is relatively onservative. For this reason, many of the relationships reovered by this phylogeny have been previously reognized on the basis of morphology. However, novel and unpredited relationships were also revealed and permit a reassessment of generi lustering within the family. Despite limited sampling, the study by Frost et al. (2001) inluded representative taxa from eah of the 10 lineages reovered by our analysis. Some groupings remain onsistent between both treatments, inluding the basal position of Platysaurus, the lose relationship of Cordylus ataphratus to C. polyzonus + C. jordani (mtdna support only) and the reovery of two separate lineages of Pseudoordylus. The relationships among the major lineages, Platysaurinae subfam. nov. Type Genus: Platysaurus Smith Content: Platysaurus Smith Definition: As for sole onstituent genus (see below). Clade A: Platysaurus Smith Type speies: Platysaurus apensis Smith 1844 by monotypy. Content: P. apensis Smith 1844, P. guttatus* Smith 1849, P. torquatus* Peters 1879, P. intermedius intermedius Matshie 1891, P. i. wilhelmi* Hewitt 1909, P. i. rhodesianus* FitzSimons 1941, P. i. natalensis* FitzSimons 1948, P. i. nyasae* Loveridge 1953, P. i. subniger* Broadley 1962, P. i. parvus* Broadley 1976, P. intermedius nigresens Broadley 1981, P. i. inopinus* Jaobsen 1994, P. minor FitzSimons 1930, P. orientalis orientalis* FitzSimons 1941, P.o. fitzsimonsi* Loveridge 1944, P. mithelli Loveridge 1953, P. pungweensis pungweensis* Broadley 1959, P. p. blakei* Broadley 1964, P. impera-