Morphometric Analysis of Chameleon Fossil Fragments from the Early Pliocene of South Africa: a New Piece of the Chamaeleonid History

Home , Bradypodion, Chamaeleoninae, Nadzikambia, Palleon

Sci Nat (2015) 102:2 DOI 10.1007/s00114-014-1254-3

ORIGINAL PAPER

Morphometric analysis of chameleon fossil fragments from the Early Pliocene of South Africa: a new piece of the chamaeleonid history

Alexis Y. Dollion & Raphaël Cornette & Krystal A. Tolley & Renaud Boistel & Adelaïde Euriat & Elodie Boller & Vincent Fernandez & Deano Stynder & Anthony Herrel

Received: 25 September 2014 /Revised: 1 December 2014 /Accepted: 3 December 2014 # Springer-Verlag Berlin Heidelberg 2015

Abstract The evolutionary history of chameleons has been pre- fragments are phenotypically dissimilar from the South African dominantly studied through phylogenetic approaches as the fossil endemic genus Bradypodion and are more similar to other cha- register of chameleons is limited and fragmented. The poor state meleon genera such as Trioceros or Kinyongia.However,close of preservation of these fossils has moreover led to the origin of phenetic similarities between some of the fragments and the Sey- numerous nomen dubia, and the identification of many chame- chelles endemic Archaius or the Madagascan genus Furcifer leon fossils remains uncertain. We here examine chameleon fos- suggest that some of these fragments may not contain enough sil fragments from the Early Pliocene Varswater formation, ex- genus-specific information to allow correct identification. Other posed at the locality of Langebaanweg “E” Quarry along the fragments such as the parietal fragments appear to contain more southwestern coast of South Africa. Our aim was to explore genus-specific information, however. Although our data suggest whether these fossil fragments could be assigned to extant gen- that the fossil diversity of chameleons in South Africa was po- era. To do so, we used geometric morphometric approaches tentially greater than it is today, this remains to be verified based based on microtomographic imaging of extant chameleons as on other and more complete fragments. well as the fossil fragments themselves. Our study suggests that the fossils from this deposit most likely represent at least two Keywords Chamaeleonidae . Early Pliocene . Geometric different forms that may belong to different genera. Most morphometrics . Varswater formation . Langebaanweg

Communicated by: Sven Thatje Electronic supplementary material The online version of this article (doi:10.1007/s00114-014-1254-3) contains supplementary material, which is available to authorized users. A. Y. Dollion : A. Herrel (*) R. Boistel : A. Euriat UMR 7179 CNRS/MNHN, Département d’Ecologie et Gestion de la IPHEP, Université de Poitiers UMR CNRS 7262, 6 rue Michel Biodiversité, Muséum National d’Histoire Naturelle (MNHN), 57 brunet, 86073 Poitiers, France rue Cuvier, Case postale 55,75231, Paris Cedex 5, France e-mail: [email protected] E. Boller : V. Fernandez ESRF (European Synchrotron Radiation Facility), 71 avenue des R. Cornette Martyrs, 38000 Grenoble, France UMR 7205 CNRS/MNHN/UPMC/EPHE, Institut de Systématique, Évolution, Biodiversité (ISYEB), 45 rue Buffon, 75005 Paris, France D. Stynder K. A. Tolley Department of Archaeology, Faculty of Science, University South African National Biodiversity Institute, Claremont, of Cape Town, Rondebosch, Cape Town 7701, South Africa Cape Town 7735, South Africa

K. A. Tolley A. Herrel Department of Botany and Zoology, Stellenbosch University, Evolutionary Morphology of Vertebrates, Ghent University, K.L. Matieland 7602, South Africa Ledeganckstraat 35, 9000 Ghent, Belgium 2 Page 2 of 14 Sci Nat (2015) 102:2

Introduction The fossils from “E” Quarry are primarily derived from the Varswater geological formation. Four members are recog- Chameleons are a relatively recent family (Chamaeleonidae) nized in the most recent lithostratigraphic review of this for- of lizards that diverged from their sister group, the mation: the early Late Miocene Langeenheid Sandy Clay Agamidae, in the Late Cretaceous (ca. 90 My; Townsend Member (LSCM) and Konings Vlei Gravel Member et al. 2011; Bolet and Evans 2013; Tolley et al. 2013), (KVGM), and the Late Miocene/Early Pliocene Langeberg whereas the divergence between most other lizard families Quartzose Sand Member (LQSM) and Muishond Fontein Pel- occurred in the Jurassic to Middle Cretaceous (Vidal and letal Phosphate Member (MPPM) (Roberts 2006). The LQSM Hedges 2009). Chameleons are thought to have originated and the MPPM are extremely rich and have thus far produced in Africa and then dispersed to Madagascar, the Seychelles, the bulk of the “E” Quarry fossils, including the chameleon the Comoros Islands, South India, Oman, Yemen, and the specimens studied here. The LQSM is construed as primarily Mediterranean Coast (Tolley et al. 2013) where they are a floodplain and salt marsh deposit, while the MPPM is currently found. Despite the fact that Chamaeleonidae have interpreted as primarily a river channel deposit (Hendey existed for about 90 My, the fossil register of chameleons 1982;Roberts2006; Roberts et al. 2011). Two river channels, only starts in the Lower Miocene (Townsend et al. 2011; beds 3aS and 3aN, have been identified. While the temporal Bolet and Evans 2013). Moreover, the number of fossils relationship between these two beds remains unclear, bed 3aS that can be assigned to genera (or species) within this fam- appears to be the earlier of the two (Hendey 1981, 1984). In ily is rather limited. There are only seven described fossil any event, both river channels are thought to reflect a progres- species and two subfossil species from the Lower Miocene sive northward shift of the lower course of the proto-Berg to the Upper Pleistocene: Chamaeleo bavaricus (Schleich River. 1983;Böhme2003, 2010), Chamaeleo caroliquarti (Moody The LBW chameleon fossils studied here largely consist of and Roček 1980; Čerňanský 2010), Chamaeleo pfeili fragmented cranial bones. Although the taxonomic identifica- (Schleich 1984; Čerňanský 2011), Chamaeleo simplex tion of such fossils is difficult, these fossils were previously (Schleich 1994), Chamaeleo andrusovi (Čerňanský 2010), assumed to belong to the genus Bradypodion, the most wide- Chamaeleo sulcodentatus (Schleich 1994), Trioceros spread extant chameleon genus in South Africa. However, by jacksonii (Leakey 1965), and Chamaeleo chamaeleon the Early Oligocene, all chameleon genera had diverged from (Talavera and Sanchíz 1983). Furthermore, many of these one another (Tolley et al. 2013). Because all extant genera fossils are fragmented and correct identification has proven were present in the Early Pliocene, the fragments could poten- to be difficult, leading to the presence of some nomina tially belong to any African genus. They cannot, however, dubia (Čerňanský 2010, 2011). The only exception to those belong to any Malagasy genus, as the divergence between fragmented fossils is an intact skull from the Early Miocene African and Malagasy genera substantially predates the Early of Kenya dated at about 18 My (Hillenius 1978; Rieppel Pliocene (Tolley et al. 2013). Yet, the likelihood that these et al. 1992) and called Chamaeleo intermedius. fossils effectively belong to Bradypodion is high given that The evolutionary history of chameleons as documented by they are the dominant chameleon fauna across southern Africa the fossil record is fragmentary and scattered, and some pieces and probably have been in place since the mid-Miocene appear to be lacking as illustrated by the time gap of approx- (Tolley et al. 2008). During that period, an episodic global imately 67 My between the estimated origin of the temperature increase (mid-Miocene optimum) (Zachos et al. Chamaeleonidae and the first fossils recorded. Langebaanweg 2001;Böhme2003) probably facilitated the expansion of cha- “E” Quarry (LBW) (32° 58′ S, 18° 7′ E), located on the west meleons across southern Africa. Potentially, the ancestral for- coast of South Africa, is one of the few sites in the world to est habitat of chameleons dominated the landscape at this have produced chameleon fossils. This important fossil local- point in time (Tolley et al. 2008, 2013). Thereafter, declining ity was discovered accidentally during phosphate mining op- temperatures and a decrease in CO2 concentration favored the erations in the early nineteenth century. Although the LBW expansion of C4 vegetation (Rossouw et al. 2009), allowing fossils elicited interest from the scientific community from the for open habitats such as grassland and fynbos (typical shrub- time that they were discovered, their great antiquity was only land or heathland vegetation occurring in the Western Cape realized once systematic analyses began to be carried out on region) to expand, while forests contracted (see Tolley et al. them in the 1960s. Today, LBW is internationally recognized 2013). Bradypodion are known to occur in both forest and as one of the most faunally diverse Mio-Pliocene fossil local- open habitats, and given that the west coast habitat at the start ities in the world. Beyond the fossils of chameleons and other of the Pliocene would have been a mosaic of forest and more squamates, notable fossil representatives of mammalian or- open environments, it may thus have been suitable for ders such as Proboscidea, Carnivora, Artiodactyla, and Bradypodion and potentially other chameleons. Perissodactyla are also present (Hendey 1970a, b; Roberts Until now, the poor state of preservation and the degree of et al. 2011). fragmentation of the LBW chameleon fossils have prevented Sci Nat (2015) 102:2 Page 3 of 14 2 their precise identification to the genus level. Even so, it has Bradypodion or Chamaeleo, as opposed to other genera, be- been previously proposed that fossil fragments may retain cause of the current geographic distribution of these genera in adequate amounts of specific information allowing their cor- southern Africa. rect identification when using geometric morphometric approaches (Bastir et al. 2014; Cornette et al. 2015). To date, no morphometric analysis has been attempted on the LBW Materials and methods chameleon fragments, but this approach may prove to be in- sightful. In fact, geometric morphometric approaches have Fossil and biological material demonstrated their usefulness in many fields of biology (Zelditch et al. 2012) including the taxonomy of fossil taxa The fossil chameleon material studied is part of the microfos- (e.g., Warheit 1992). However, in the majority of the cases sil collections at the Iziko South African Museum, where geometric morphometric approaches were used, fossils Cape Town, South Africa. It is unclear whether they were were complete anatomical pieces such as teeth or mandibles derived from either the LQSM or MPPM as most of the (e.g., Matthews and Stynder 2011). Recent advances in geo- LBW fossils were originally collected from mine dumps. In metric morphometric methods have, nevertheless, demon- any event, all fossils are likely contemporaneous as the LQSM strated the feasibility of using fragmented material for the and MPPM are considered to have been deposited at more or identification and assignment of fossils for both small mam- less the same time. Fossil fragments were identified as belong- mals and amphibians (Bastir et al. 2014; Cornette et al. 2015). ing to Chamaeleonidae based on criteria such as the presence Here, we investigate which extant genus, if any, the chame- of acrodont teeth, parietal bone ornamentation as observed in leon fragments from the fossil site of the Langebaanweg fossil extant Bradypodion, and the shape of the parietal, prefrontal, deposit most closely resemble. To do so, we use 3D geometric and maxilla (Estes et al. 1988). Fossils from this deposit have morphometric methods to investigate phenetic similarity be- been dated to the Early Pliocene (ca. 5.15±0.1 My) based on tween these fossil fragments and the extant genera of chame- paleomagnetic data, global sea level reconstructions, and leons. We hypothesize that these fragments are more similar to biochronology (Roberts et al. 2011). In total, 23 fossil

Fig. 1 Nadzikambia mlanjensis and Bradypodion setaroi skulls illustrating the position of the landmarks used for the study. Top, left lateral view; middle, dorsal view; bottom, left mesial view. Circles represent landmarks placed on the outside of the skull, white circles filled with black represent landmarks placed inside the skull. Landmarks 34–64 and 74–75 areplacedontheright side and follow the same order as those placed on the left side 2 Page 4 of 14 Sci Nat (2015) 102:2

Table 1 Definitions of the landmarks placed on the intact skulls (see Fig. 1) and the fossil fragments (see Figs. 2 and 3). Note that not all the landmarks taken on the right side of the cranium are illustrated

Intact skulls Landmark no. Description Side 1 Snout tip Median 2 Point at the base of the teeth, perpendicular to the tangent passing through landmark 28 Left 3 Landmark at the base of the teeth resulting from the projection of a virtual landmark midway on the line passing through landmarks 26 and 18 4 Posterior ventral tip of the upper jaw 5 Maximum curvature at the contact between the upper temporal arch and the postorbitofrontal 6 Maximum curvature at the contact between the upper temporal arch and the supratemporal 7 Anterior tip of the supratemporal 8 Posterior tip of the supratemporal 9 Posterior extremity of the upper temporal arch along its midline 10 Extremity of the ascending process of the squamosal 11 External upper posterior margin of the temporal fossa 12 Posterior tip of the parietal Median 13 Contact between the postorbitofrontal and the parietal Left 14 Dorsal maximum curvature of the orbit 15 Dorsal surface of the posterolateral edge of the prefrontal 16 Dorsal part of the prefrontal where the prefrontal meets the posterior projection of the maxilla 17 Narrowest part of the anterior projection of the maxilla, external side 18 Narrowest part of the suborbital process of the maxilla, orbital side 19 Maximum of suborbital curvature opposite of landmark 4 20 Anterior extremity of the upper temporal arch along its midline 21 Anteroventral maximum of curvature of the temporal fossa 22 Anterodorsal maximum of curvature of the temporal fossa, and posterior contact between the parietal and the postorbitofrontal 23 Posterodorsal maximum curvature of the temporal fossa 24 Posteroventral maximum curvature of the temporal fossa 25 Anterior tip of the prefrontal forming the dorsal margin of the external naris 26 Narrowest part of the suborbital process of the maxilla, naris side 27 Posteroventral maximum curvature of the naris 28 Center of the naris, basal part 29 Anteroventral maximum curvature of the naris 30 Narrowest part of the anterior projections of the maxilla, naris side 31 Ventral surface of the posterolateral edge of the prefrontal 32 Internal basal edge of prefrontal projection which meets the palatinum 33 External basal edge of prefrontal projection which meets the palatinum 34 Point at the base of the teeth, perpendicular to the tangent passing through landmark 59 Right 35 Landmark at the base of the teeth resulting from the projection of a virtual landmark midway on the line passing through landmarks 57 and 49 36 Posterior ventral tip of the upper jaw 37 Maximum curvature at the contact between the upper temporal arch and the postorbitofrontal 38 Maximum curvature at the contact between the upper temporal arch and the supratemporal 39 Anterior tip of the supratemporal 40 Posterior tip of the supratemporal 41 Posterior extremity of the upper temporal arch along its midline 42 Extremity of the ascending process of the squamosal 43 External upper posterior margin of the temporal fossa 44 Contact between the postorbitofrontal and the parietal 45 Dorsal maximum curvature of the orbit 46 Dorsal surface of the posterolateral edge of the prefrontal Sci Nat (2015) 102:2 Page 5 of 14 2

Table 1 (continued) 47 Dorsal part of the prefrontal where the prefrontal meets the posterior projection of the maxilla 48 Narrowest part of the anterior projection of the maxilla, external side 49 Narrowest part of the suborbital process of the maxilla, orbital side 50 Maximum of curvature opposite of landmark 36 51 Anterior extremity of the upper temporal arch along its midline 52 Anteroventral maximum of curvature of the temporal fossa 53 Anterodorsal maximum curvature of the temporal fossa, and posterior contact between the parietal and the postorbitofrontal 54 Posterodorsal maximum curvature of the temporal fossa 55 Posteroventral maximum curvature of the temporal fossa 56 Anterior tip of the prefrontal forming the dorsal margin of the external naris 57 Narrowest part of the suborbital process of the maxilla, naris side 58 Posteroventral maximum curvature of the naris 59 Center of the naris, basal part 60 Anteroventral maximum curvature of the naris 61 Narrowest part of the anterior projections of the maxilla, naris side 62 Ventral surface of the posterolateral edge of the prefrontal 63 Internal basal edge of prefrontal projection which meets the palatinum 64 External basal edge of prefrontal projection which meets the palatinum 65 Landmark on the external side of the parietal across from landmark 70 (70=internal Median vs. 65=external) 66 Center between landmarks 65 and 67 67 Landmark on the external side of the parietal across from landmark 71 (71=internally vs. 67=externally) 68 Middle between landmarks 14 and 45 69 Contact between the nasals on the midline where the nasals touch the maxilla 70 Posterior extremity of the internal descending process of the parietal Inside/median 71 Anterior extremity of the internal descending process of the parietal 72 Landmark at the center of the dorsal surface of the posterior nasal process of the Inside/left maxilla where it meets the palatinum 73 Landmark at the center of the ventral surface of the posterior nasal process of the maxilla where it meets the palatinum 74 Landmark on the dorsal surface of the internal maxillary process where it meets the palatinum, Inside/right center of the posterior nasal process of the maxilla 75 Landmark on the ventral surface of the internal maxilla process where it meets the palatinum, center of the posterior nasal process of the maxilla Fossil fragments Fragment Landmark no. Description Side Maxilla 45628 1 Posteroventral maximum curvature of the naris Right 2 Center of the naris, basal part 3 Point at the base of the teeth, perpendicular to the tangent passing through landmark 2 4 Landmark at the base of the teeth at the level of landmark 5 but on the external side 5 Landmark on the dorsal surface of the internal maxillary process where it meets the palatinum, center of the posterior nasal process of the maxilla 6 Landmark on the ventral surface of the internal maxilla process where it meets the palatinum, center of the posterior nasal process of the maxilla Maxilla 31549-1 1 Narrowest part of the suborbital process of the maxilla, orbital side Right 2 Narrowest part of the suborbital process of the maxilla, naris side 3 Posteroventral maximum curvature of the naris 4 Center of the naris, basal part 5 Anteroventral maximum curvature of the naris 6 Narrowest part of the anterior projections of the maxilla, naris side 2 Page 6 of 14 Sci Nat (2015) 102:2

Table 1 (continued) 7 Narrowest part of the anterior projection of the maxilla, external side 8 Point at the base of the teeth, perpendicular to the tangent passing through landmark 4 9 Landmark at the base of the teeth resulting from the projection of a virtual landmark midway on the line passing through landmarks 2 and 1 10 Landmark on the dorsal surface of the internal maxillary process where it meets the palatinum, center of the posterior nasal process of the maxilla 11 Landmark on the ventral surface of the internal maxilla process where it meets the palatinum, center of the posterior nasal process of the maxilla Maxilla 4680-3 1 Narrowest part of the posterior projection of the maxilla, orbital side Left 2 Narrowest part of the posterior projection of the maxilla, naris side 3 Posteroventral maximum curvature of the naris 4 Landmark at the base of the teeth resulting from the projection of a virtual landmark midway on the line passing through landmarks 2 and 1 5 Landmark on the dorsal surface at the center of the posterior nasal process of the maxilla 6 Landmark on the ventral surface at the center of the posterior nasal process of the maxilla Maxilla 46949-1 1 Narrowest part of the posterior projection of the maxilla, orbital side Right 2 Narrowest part of the posterior projection of the maxilla, naris side 3 Posteroventral maximum curvature of the naris 4 Center of the naris, basal part 5 Anteroventral maximum curvature of the naris 6 Point at the base of the teeth, perpendicular to the tangent passing through landmark 4 7 Landmark at the base of the teeth resulting from the projection of a virtual landmark midway on the line passing through landmarks 2 and 1 8 Landmark on the dorsal surface at the center of the posterior nasal process of the maxilla 9 Landmark on the ventral surface at the center of the posterior nasal process of the maxilla Maxilla 70488-1 1 Posteroventral maximum curvature of the naris Right 2 Center of the naris, basal part 3 Anteroventral maximum curvature of the naris 4 Point at the base of the teeth, perpendicular to the tangent passing through landmark 2 5 Landmark at the base of the teeth on the external side at the level of landmarks 6 and 7 6 Landmark on the dorsal surface of the internal maxillary process where it meets the palatinum, center of the posterior nasal process of the maxilla 7 Landmark on the ventral surface of the internal maxilla process where it meets the palatinum, center of the posterior nasal process of the maxilla Prefrontal 46522 1 Internal basal edge of prefrontal projection which meets the palatinum Left 2 External basal edge of prefrontal projection which meets the palatinum 3 Ventral surface of the posterolateral edge of the prefrontal 4 Dorsal surface of the posterolateral edge of the prefrontal Parietal (70488-5 and 70488-6) 1 Landmark on the external side of the cranium on the posterior tip of the parietal Median 2 Anterodorsal maximum curvature of the temporal fossa, and posterior contact Left between the parietal and the postorbitofrontal 3 Landmark on the external side of the cranium opposite of landmark 7 Median 4 Anterodorsal maximum curvature of the temporal fossa, and posterior contact Right between the parietal and the postorbitofrontal 5 Center between landmarks 1 and 3 Median 6 Posterior extremity of the internal descending process of the parietal Median 7 Anterior extremity of the internal descending process of the parietal Median Sci Nat (2015) 102:2 Page 7 of 14 2 fragments were retrieved and identified as cranial fragments. morphology, we here group it with other Brookesia for our The fragments were scanned using micro-computer tomogra- analyses. To test whether geometric morphometric approaches phy (μCT) at the Central Analytical Facility, Stellenbosch are indeed appropriate to distinguish the different chameleon University (Table S1). For this study, we focused on eight genera, we added an Uromastyx (Uromastyx acanthinura)to fragments corresponding to three bones: maxilla (N=5), pari- our initial analysis, because this genus belongs to the etal (N=2), and prefrontal (N=1). These fragments were se- Agamidae, the sister taxon to Chamaeleonidae (Pyron et al. lected based on the fact that they could be unambiguously 2013). Our dataset incorporates both males and females for all assigned to a region on the skull and thus compared to extant species of Bradypodion. For the other species, males or fe- chameleon genera. males were included depending on the availability of speci- To compare our fossils to extant chameleon genera, we mens for scanning (see Supplementary Table 1). obtained μCT scans of 64 preserved chameleons (see Supple- mentary Table 1). Specimens were obtained from the Port Microtomography Elizabeth Museum (Bayworld), the Museum of Comparative Zoology at Harvard, the Muséum National d’Histoire All material was scanned using high-resolution X-ray Naturelle, the Royal Museum for Central Africa, and the per- microtomography setups including a ViscomX8050-16 CT scan- sonal collection of Anthony Herrel. These individuals repre- ner (Viscom, Hannover, Germany), a Metris X-Tek HMX ST sent a sample of all extant genera, including all described 225 CT scanner (Nikon Metrology, Tring, South East England, species of the genus Bradypodion. We include the species UK), and a v|tome|x L 240 CT scanner (GE control and mea- Brookesia nasus which has recently been assigned to the ge- surements, General Electric©, Fairfield, CT, USA). Automatic nus Palleon (Glaw et al. 2013). Given the recent description of and semi-automatic segmentation and surface rendering was per- this genus and its similarity to Brookesia in terms of skull formed using Avizo® 7 (FEI™, France).

Fig. 2 Maxillary fragments used in the present study with landmarks indicated. The number associated with each landmark does not correspond to the landmark numbers of the analysis on the whole skull but refer to the landmarks placed on each fragment. To the right,a Bradypodion pumilum skull is illustrated showing the regions concerned by the fragments 2 Page 8 of 14 Sci Nat (2015) 102:2

Geometric morphometric analyses fragment characterized a slightly different aspect/position of the cranium (Figs. 2 and 3; Table 1). Uromastyx was ex- We used geometric morphometric techniques (Zelditch et al. cluded from the analysis of fossils because the difference in 2012) as they allow the separation of the form of an object shape forced most of the variation onto the first principal into size and shape components (Needham 1950). To do so, component. 3D reconstructions were loaded into the software package Landmark coordinates were extracted and used for subse- “Landmark” (Wiley et al. 2005), and 75 landmarks (Fig. 1; quent analyses. First, a generalized Procrustes analysis Table 1) were placed on each cranium of all individuals (GPA) was performed to remove variation due to size, trans- from extant species to investigate morphological variation. lation, and rotation (Rohlf and Slice 1990). The Procrustes Landmarks were placed to accurately describe cranial shape residuals were then used to perform a principal component using homologous landmarks (contact points between two analysis (PCA) in order to reduce the dimensionality of the bones or maxima and minima of curvature; see Table 1; dataset (Baylac and Friess 2005). We used the first 13 prin- Bookstein 1991). Next, landmarks were put onto the eight cipal components (PCs) as variables because they represent fossil fragments. It was not possible to standardize the num- 90 % of the total shape variability in our dataset. Because ber of landmarks between the fossil fragments because each size is an important component of the form of an organism,

Fig. 3 Parietal and prefrontal fragments used in the present study with landmarks. The number associated with each landmark does not correspond to the landmark numbers of the analysis on the whole skull but refers to the landmarks placed on each fragment. To the right,a Bradypodion pumilum skull is illustrated showing the regions concerned by the fragments Sci Nat (2015) 102:2 Page 9 of 14 2 we tested for allometric effects in the crania of the extant however, potential for overlap in morphological space that is genera using a multivariate analysis of covariance not revealed by this analysis. On the first four PCs, accounting (MANCOVA) on the shape data with genus as factor and for 69.45 % of the variability, Uromastyx is clearly distinct size as a covariate. We next constructed a neighbor-joining from all chameleon genera, but there is a lack of distinction tree based on the distance matrix including both centroid size between some chameleon genera (Fig. 4). Chamaeleo and as well as the first 13 principal components. Shape variation Furcifer are scattered in the morphospace (Fig. 4), potentially was visualized using multivariate regressions along the axes due to the fact that the specimens used in our study represent of the PCA (Monteiro 1999). extremes in terms of morphology for both genera. Indeed, We next averaged PC scores by genus to remove intrageneric Chamaeleo calyptratus and Furcifer oustaleti display well- shape variability and constructed a neighbor-joining tree using developed casques in comparison to the other species from the distance matrix to evaluate phenotypic proximity between the the same genus that were sampled (Chamaeleo dilepis and fossil fragments and the extant genera. We excluded size from the Furcifer campani). When Uromastyx is excluded (Supple- analysis of fossil fragments because the extant species used for mentary Fig. 1), most of the chameleon genera can be distin- comparison included both subadult and adult specimens. More- guished from one another. Interestingly, this shows that over, the size of the individuals that gave rise to the fossil frag- Kinyongia and Nadzikambia are similar to one another, yet ments is unknown. All statistical analyses were performed with distinctive from other chameleons, reflecting the prior taxo- R version 2.15.3 (R Development Core Team 2013) including nomic classification of Nadzikambia within the Bradypodion the libraries MASS (Venables and Ripley 2002), ade4 (Dray et al. based on anatomical features (Tilbury et al. 2006). Moreover, 2007), and Rmorph (Baylac 2013). Archaius appears very different from other chameleons, reflecting its long evolutionary history in isolation on the Sey- chelles (Townsend et al. 2010). The visualizations associated with the shape space based on the multivariate regressions Results suggest that features such as the relative length of the snout, the relative width and height of the cranium, the shape of the Morphometric analyses on extant genera casque, and its relative size allow the discrimination of the different chameleon species and genera (Fig. 5). The PCA performed on the whole cranium of the chameleons The results of the MANCOVA (genus: P<0.001; size: including the outgroup Uromastyx revealed that chameleons P<0.001; genus×size: P>0.001) indicate a strong allometric occupy a different part of morphological space compared to effect for the extant chameleon genera. The neighbor-joining Uromastyx. Moreover, the morphometric analysis of the cra- tree which takes into account size (Supplementary Fig. 2) nium allowed a characterization of most of the genera. Be- broadly reflects early classifications based on morphological cause of the limited sample size for each genus, there is, features such as hemipenis (Klaver and Böhme 1986)orlung

Fig. 4 Principal component analyses performed on the overall associated with each axis is indicated. On the first four principal chameleons including Uromastyx as an outgroup and indicated by a components, Uromastyx always falls outside of the chameleon shape solid black arrow. a Scatterplot of PC1 and PC2 accounting for space. Within chameleons, Archaius appears to be morphologically 49.20 % of the shape variability; b scatterplot of PC3 and PC4 quite distinct (PC3 and PC4) which may be due to the long geographic accounting for 20.25 % of the shape variability; the shape variability isolation of this genus endemic to the Seychelles 2 Page 10 of 14 Sci Nat (2015) 102:2

Fig. 5 Morphological variation associated with the five first principal components of the PCA performed on the extant chamaeleonid genera without the outgroup Uromastyx. Red and blue shapes represent the extremes along the first five axes, with red shapes representing the positive side of the axis and blue shapes the negative side (Monteiro 1999). Shape variations associated to PC1 concern the width and height of the cranium; shape variations associated with PC2 and PC3 mostly depict variation in the shape of the casque and the rostrum; PC4 depicts variation in cranium height and rostrum shape; PC5 depicts variation in cranium width

morphology (Klaver 1981). This tree shows that Chamaeleo is chameleon genera with a wider geographic distribution but most- morphologically similar to Trioceros, Calumma similar to lyfoundinEastAfrica,Trioceros (Fig. 6a)andRhampholeon Archaius, Rieppeleon similar to Rhampholeon,and (Fig. 6b). However, fragment 70488-1 seems to be morpholog- Bradypodion similar to Kinyongia and Nadzikambia. More- ically most similar to the genus Archaius (Table S2), an endemic over, this phenetic tree also reflects to some degree the most genus from the Seychelles Islands. Fragment 31549-1 is morpho- recent molecular phylogeny (Tolley et al. 2013). The size of the logically most similar to the genus Nadzikambia (Table S2), specimens used in our analysis plays clearly a role in structuring while fragment 46949-1 is most similar to the genus Kinyongia the tree. Indeed, specimens from large-bodied chameleons (e.g., (Table S2). Note that these fragments describe the anterior part of Furcifer) are differentiated from specimens of smaller chame- the maxilla and that both present the posterior nasal process of leon genera (e.g., Brookesia; Supplementary Fig. 2). the maxilla (Fig. 2). Thus, both describe the nasal fossa better, compared to the other maxillary fragments. Moreover, fragment Maxillary fragments 31549-1 was the only one presenting an anterior nasal process. The maxillary fragments 45628 and 4680-3 both present Two different forms can be inferred from the analyses on six landmarks but do not describe the same part of the maxilla maxillary fragments. One group of fragments is morphologi- (Fig. 7). They are morphologically most similar to the chame- cally most related to the chameleon genera Kinyongia and leon genera Trioceros and Rhampholeon. Whereas fragment Nadzikambia (Fig. 7), and the other group is more similar to 45628 is placed as having a shape similar to chameleons in the Sci Nat (2015) 102:2 Page 11 of 14 2

Fig. 6 Neighbor-joining tree performed on the shape dataset. The trees place the different maxillary fragments as morphologically related to chameleon genera found in East Africa, a Trioceros (Euclidian distance=0.25) for fragment 45628 and b Rhampholeon (Euclidian distance=0.11) for fragment 4680-3. One fragment, fragment 70488-1 (c), is morphologically most similar to the genus Archaius (Euclidian distance=0.20). Bold lines on the neighbor-joining tree represent the shortest distance existing between the fragment and the extant genus to which the fragment is morphologically most similar

genus Trioceros (Table S2), fragment 4680-3 is most similar analysis suggests that the fragment is morphologically most to Rhampholeon (Table S2). This fragment was described by similar to chameleons of the genus Furcifer, an endemic genus six landmarks mostly describing the posterior nasal process. from Madagascar (Fig. 8; Table S2).

Prefrontal fragment (46522) Parietal fragments (70488-5 and 70488-6)

This fragment of the prefrontal is the fragment described by Two parietal fossil fragments are described by the same seven the fewest number of landmarks (four) in our sample. The landmarks. Interestingly, our analyses reveal that these two landmarks placed on it describe the thickness of this structure fragments are morphologically divergent. Fragment 70488-5 and the contact between the prefrontal and the palatine. Our has a morphology most similar to that observed in chameleons

Fig. 7 Neighbor-joining tree performed on the shape data. The two trees and b Kinyongia (Euclidian distance=0.12) for fragment 46949-1. Bold place the different maxillary fragments as morphologically related to lines on the neighbor-joining tree represent the shortest distance existing chameleons previously classified as Bradypodion (Tilbury et al. 2006) between the fragment and the extant genus to which the fragment is including a Nadzikambia (Euclidian distance=0.13) for fragment 31549 morphologically most similar 2 Page 12 of 14 Sci Nat (2015) 102:2

Fig. 9 Neighbor-joining tree performed on the shape dataset for the parietal fragments. Whereas one of the fragments is morphologically most similar to Kinyongia (Euclidian distance=0.22), the other is more similar to Trioceros (Euclidian distance=0.18). Bold lines on the neighbor-joining tree represent the shortest distance existing between Fig. 8 Neighbor-joining tree performed on the shape dataset for the the fragment and the extant genus to which the fragment is prefrontal. The tree places the different fragments as morphologically morphologically most similar most similar to Furcifer (Euclidian distance=0.19). Bold lines on the neighbor-joining tree represent the shortest distance existing between the fragment and the extant genus to which the fragment is morphologically most similar lower, even small skull fragments such as parts of the maxillary and the parietal are able to discriminate between chame- of the genus Trioceros, and fragment 70488-6 is most similar leon genera even though sampling was not exhaustive to chameleons of the genus Kinyongia based on the geometric (Figs. 6, 7, 8,and9). The parietal seems to be the best candi- morphometric descriptors (Fig. 9;TableS2). date for fragment reattribution because this structure is variable across genera (Fig. 5). Maxilla fragments are also good candidates, although not all of them. In fact, one of the maxilla fragments suggests affinity to an endemic genus from the Sey- chelles, Archaius, in contrast to all other fragments. Yet, as this Discussion fragment does not present any dorsal projection, the information contained by this fragment may be rather limited. In gen- Our study suggests greater potential chameleon diversity dur- eral, our analyses suggest that the information contained in ing the Early Pliocene (±5.15 My) of South Africa with the fossil fragments has its limitations. As such, interpretations presence of two large morphological groups. Firstly, there is a based on these approaches need to be verified based on other group that contains fossil fragments most similar to chame- and ideally more complete fragments. leons of the genera Nadzikambia and Kinyongia, and second- None of the fossils in our study appears to belong to the ly, there is a group of fossil fragments that is Trioceros-like. genus Chamaeleo or Bradypodion, currently present in South However, one fragment is also similar to extant species of the Africa. In contrast, our analyses suggest similarity to forms genus Furcifer from Madagascar. As this fragment presents living mostly in closed habitats, many of which are currently the lowest number of landmarks and thus presumably the least associated with forests. The expansion of the C4 vegetation genus-specific information, this assignment should be (Rossouw et al. 2009) during the Late Miocene which led to interpreted with caution. Consequently, the confidence in the the decrease of forest habitats should have favored chameleon reattribution with this kind of fragment is lower than those genera adapted to both closed and open habitats such as suggesting similarity to Nadzikambia, Kinyongia,or Bradypodion (Tolley et al. 2008). However, none of the frag- Trioceros with more landmarks to describe their shape. This ments was identified as being most similar to this genus. In result also demonstrates the importance of having enough contrast, several were suggested to be morphologically close landmarks to describe the fragments, as the taxonomic infor- to the genera Nadzikambia and Kinyongia. These genera mation contained in a fragment is dependent on the accuracy were until 2006 (Tilbury et al. 2006) classified within the of the description of the structure. Our analyses based on genus Bradypodion because of their similar morphology. extant taxa do tend to show that geometric morphometric Given that Nadzikambia and Bradypodion diverged around techniques are able to describe and identify extant chameleon 43–44 My ago as suggested by molecular analyses, we can genera. Although the information contained in fragments is assume that they should have been morphologically distinct Sci Nat (2015) 102:2 Page 13 of 14 2 from one another 5.15 My ago. Although fragments found as Acknowledgments We would like to thank Graham Avery, Kerwin van being most similar to Nadzikambia or Kinyongia may be Willingh, and Romala Govender from the Iziko South African Museum for logistical assistance with regard to allowing access to and analyses of hard to assign with precision, our data suggest that these their microfossil collection from Langebaanweg. We would also like to may have been present in South Africa in the Early Pliocene thank Andrej Čerňanský and two anonymous reviewers for constructive or perhaps the Late Miocene. and helpful comments on a prior version of our manuscript. Several Although our dataset comprises at least one representative museums provided comparative specimens: Port Elizabeth Museum- Bayworld (W. Conradie), the Muséum National d’Histoire Naturelle of each described chameleon genus (including a specimen of (Ivan Ineich), the Royal Museum for Central Africa (D. Meirte), and the recently described genus Palleon), most genera were on- the Museum of Comparative Zoology at Harvard (J. Rosado). μCT scans ly represented by a few species, and the genus Calumma was of fossils and extant specimens were performed at the ESRF Grenoble, represented by a subadult of a small species. This clearly the IPHE at the Université de Poitiers, the Laboratoire Navier at the Ecole des Ponts-ParisTech (France), and the Central Analytical Facilities from renders our characterization of the genus-level diversity in Stellenbosch University. We thank Antoine Du Plessis, Nicolas Lenoir, morphology less complete and may have affected the assign- Paul Tafforeau, and Peter Cloetens for help with scanning. We also ac- ment of fossils to extant genera. Some of the extant genera knowledge the UMS CNRS/MNHN2700, “Outils et méthodes de la ” show a great diversity in skull form that is clearly not cap- systématique intégrative for access to the morphometrics platform and the CEMIM platform for access to their facilities. This study was sup- tured by the inclusion of only one or two species as is the ported by a European Synchrotron Radiation Facility Project through case in our dataset. Despite the fact that we included all allocation of beam time. μCTscanning at Navier was supported by grants species within the genus Bradypodion as well as C. dilepis, from the Région Ile-de-France. one of the only two Chamaeleo species found in South Af- rica at present, none of the fragments showed a morphology most similar to that of these taxa. The parietal appears to be a References good candidate for fragment reattribution because our morphometric analysis performed on the entire cranium showed Bastir M, Böhme M, Sanchiz B (2014) Middle Miocene remains of Alytes that the parietal is variable within the group as a whole. As (Anura, Alytidae) as an example of the unrecognized value of fossil such, this structure likely contains both generic and species- fragments for evolutionary morphology studies. J Vertebr Paleontol 34:69–79 specific information. The maxilla showed different patterns Baylac M (2013) Rmoprh: a geometric and multivariate morphometrics of reattribution according to the fragmentation pattern. Our library. Available from the author: [email protected] analysis moreover suggests that the nasal process of the max- Baylac M, Friess M (2005) Fourier descriptors, Procrustes superimposi- illa combined with the basal curvature of the nasal fossa tion, and data dimensionality: an example of cranial shape analysis in modern human population. Modern morphometrics in physical contains at least some genus-specific information. This is anthropology. Kluwer Academic Plenum Publishers, New York confirmed by the visualizations associated with PCs of the Böhme M (2003) The Miocene climatic optimum: evidence from ecto- PCA performed on the intact skulls, showing that the ante- thermic vertebrates of Central Europe. Palaeogeogr Palaeoclimatol rior part the maxilla comprising the two nasal processes Palaeoecol 195:389–401 varies greatly between chameleon genera. Böhme M (2010) Ectothermic vertebrates (Actinopterygii, Allocaudata, Urodela, Anura, Crocodylia, Squamata) from the Miocene of In summary, our analyses show that several types of cha- Sandelzhausen (Germany, Bavaria) and their implications for envi- meleons were likely present at the start of the Pliocene on the ronment reconstruction and paleoclimate. Paläontol Z 84:3–41 western margin of South Africa. Whereas chameleons of the Bolet A, Evans SE (2013) Fossil history of chameleons. In: Tolley KA, genus Bradypodion may not have been present at Herrel A (eds) The biology of chameleons. University of California Press, Berkeley, pp 175–192 Langebaanweg, other taxa similar in morphology to current- Bookstein FL (1991) Morphometric tools for landmark data: geometry day Kinyongia, Nadzikambia,orTrioceros may have occurred and biology. Cambridge University Press, Cambridge in the area, or further upriver. The loss of forms similar to Čerňanský A (2010) A revision of chamaeleonids from the lower these taxa may be associated with the continued forest frag- Miocene of the Czech Republic with description of a new species – mentation and desertification of the western part of South of Chamaeleo (Squamata, Chamaeleonidae). Geobios 43:605 613 Čerňanský A (2011) A revision of the Chamaeleo pfeili Schleich Africa, but this remains to be explored further. Although our (Squamata; Chamaeleonidae) with description of a new material of approach using geometric morphometrics to assign fossil ma- chamaeleonids from the Miocene deposits of southern Germany. terial to extant taxa appears promising, further analyses using Bull Geosci 86:275–282 additional and ideally larger fragments are needed to better Cornette R, Herrel A, Stoetzel E, Moulin S, Hutterer R, Denys C, Baylac M (2015) Specific information levels in relation to fragmentation understand the evolutionary history of chameleons in the re- patterns of shrew mandibles: do fragments tell the same story? J gion. Similar approaches including a larger sample of extant Archaeol Sci 53:323–330 species of different families, and incorporating fossil frag- Dray S, Dufour AB, Chessel D (2007) The ade4 package-II: two-table ments from Europe (Moody and Roček 1980; Talavera and and K-table methods. R News 7(2):47–52 Sanchíz 1983; Schleich 1983, 1984, 1994; Čerňanský 2010, Estes R, De Queiroz K, Guathier J (1988) Phylogenetic relationships within Squamata. In: Estes R, Pregill G (eds) Phylogenetic relation- 2011), would be important to better understand the diversity of ships of the lizard families: essays commemorating Charles L. chameleons in the Early Pliocene. Camp. Stanford University Press, Stanford, pp 99–118 2 Page 14 of 14 Sci Nat (2015) 102:2

Glaw F, Hawlitschek O, Ruthensteiner B (2013) A new genus name for Rohlf FJ, Slice DE (1990) Extensions of the procrustes method for the an ancient Malagasy chameleon clade and a PDF-embedded 3D optimal superimposition of landmarks. Syst Zool 39:40–59 model of its skeleton. Salamandra 49:237–238 Rossouw L, Stynder DD, Haarhof P (2009) Evidence for opal phytolith Hendey QB (1970a) A review of the geology and palaeontology of the preservation in the Langebaanweg ‘E’ Quarry Varswater Formation Plio/Pleistocene deposits at Langebaanweg, Cape Province. Ann and its potential for palaeohabitat reconstruction. South Afr J Sci South Afr Mus 56:75–117 105:223–227 Hendey QB (1970b) The age of the fossiliferous deposits at Schleich HH (1983) Die mittlemioän Fossil-Lagerstätte Sandelhausen 13. – Langebaanweg, Cape Province. Ann South Afr Mus 56:119 131 Chamaeleo bavaricus sp. Nov., ein neuer Nachweis aus dem Hendey QB (1981) Geological succession at Langebaanweg, Cape Jungtertiär Süddeutschlands. Mitteilungen der Beyerischen Province, and global events of the late Tertiary. South Afr J Sci Staatssammlung für. Paläontol Hist Geol 23:77–81 – 77:33 38 Schleich HH (1984) Neue Reptilienfunde aus dem Tertiär Deutschlands Hendey QB (1982) Langebaanweg. A record of past life. South African 2. Chamaeleo pfeili sp. Nov., Von der untermiozänen Museum, Cape Town Fossilfundstelle Rauscheröd/Niederbayern (Reptilia, Sauria Hendey QB (1984) Southern African late Tertiary vertebrates. In: Klein Chamaeleonidae). Mitt Bayer Staatssamml Paläontol Hist Geol 24: RG (ed) Southern African prehistory and palaeoenvironments. AA 97–103 Balkema, Rotterdam, pp 81–106 Schleich HH (1994) Neue Reptilienfunde aus dem Tertiär DeutschlandS Hillenius D (1978) Notes on chameleons. IV: A new chameleon from the 15. Neue funde fossiler Chamäleonen aus dem Neogen Miocene of Fort Ternan, Kenya (Chamaeleonidae, Reptilia). Süddeutschlands. Courrier Forschungsinstitut Senckenb 173:175– Beaufortia 28:37–55 195 Klaver CJJ (1981) Lung morphology in the Chamaeleonidae (Sauria) and its bearing upon phylogeny, systematics and zoogeography. Z Zool Talavera R, Sanchíz F (1983) Restos pliocénicos de Camaleón común, – Chamaeleo chamaeleon (L.) de Málaga. Bol Real Soc Esp Hist Nat Syst Evol 19:36 58 – Klaver CJJ, Böhme W (1986) Phylogeny and classification of the (Geolo) 81:81 84 Chamaeleonidae (Sauria) with special reference to hemipenis mor- Tilbury CR, Tolley KA, Branch WR (2006) A review of the systematics phology. Bonn Zool Monogr 22:1–64 of the genus Bradypodion (Sauria: Chamaeleonidae), with the de- Leakey LSB (1965) Olduvai Gorge 1951–1961. 1. A preliminary report scription of two new genera. Zootaxa 1363:23–38 on the geology and fauna. Cambridge University Press, Cambridge Tolley KA, Chase BM, Forest F (2008) Speciation and radiations track Matthews T, Stynder D (2011) An analysis of two Myosorex species climate transitions since the Miocene climatic optimum: a case study (Soricidae) from the Early Pliocene site of Langebaanweg (West of southern African chameleons. J Biogeogr 35:1402–1414 Coast Fossil Park, South Africa) using the geometric morphomet- Tolley KA, Townsend M, Vences M (2013) Large-scale phylogeny of rics, linear measurements and non-metrics characters. Geobios 44: chameleons suggests African origins and Eocene diversification. 87–99 Proc R Soc B 280:20130184 Monteiro LR (1999) Multivariate regression models and geometric mor- Townsend TM, Tolley KA, Glaw F, Böhme W, Vences M (2010) phometrics: the search for causal factors in the analysis of shape. Eastward from Africa: palaeocurrent-mediated chameleon dispersal Syst Biol 48:192–199 to the Seychelles islands. Biol Lett 7:225–228 č Moody S, Ro ek Z (1980) Chamaeleo caroliquarti (Chamaeleonidae, Townsend TM, Mulcahy DG, Noonan BP, Sites JW Jr, Kuczynski CA, Sauria) a new species from the Lower Miocene of central Europe. Wiens JJ, Reeder TW (2011) Phylogeny of iguanian lizards inferred ě ř – V stnik Úst edniho Ustavu Geologického 55:85 92 from 29 nuclear loci, and comparison of concatenated and species- Needham AE (1950) The form-transformation of the abdomen of the tree approaches for an ancient, rapid radiation. Mol Phylogenet Evol female pea-crab, Pinnotheres pisum Leach. Proc R Soc Lond Ser 61:363–380 – B137:115 136 Venables WN, Ripley BD (2002) ‘Modern applied statistics with R’, Pyron RA, Burbink FT, Wiens JJ (2013) A phylogeny and revised clas- fourth edition. Springer ISBN 0-387-95457-0 sification of Squamata, including 4161 species of lizards and snakes. Vidal N, Hedges BS (2009) The molecular evolutionary of lizard snakes, BMCEvolBiol13:93 and amphisbaenians. C R Biol 332:129–139 R Development Core Team (2013) R: a language and environment for Warheit KI (1992) The role of morphometrics and cladistics in the tax- statistical computing. R Foundation for Statistical Computing, – Vienna, Austria. ISBN 3-900051-07-0 onomy of fossils: a paleornithological example. Syst Biol 41:345 Rieppel O, Walker A, Odhiambo I (1992) A preliminary report on fossil 369 chamaeleonine (Reptilia: Chamaeleoninae). J Zool 243:351–380 Wiley DF, Amenta N, Alcantra DA, Ghosh D, Kil YJ, Delson E, Roberts DL (2006) Lithostratigraphy of the Varswater Formation. South Harcourt-Smith W, F. J. Rohlf, St. John K, Hamann B (2005) Afr Comm Stratigr-Lithostratigraphic Ser 9:27–31 Evolutionary morphing. Proceedings of IEEE Visualization 2005. Roberts DL, Matthews T, Herries AIR, Boulter C, Scott L, Dondo C, Zachos J, Pagani M, Sloan L, Thomas E, Billups K (2001) Trends, Mtembi P, Browning C, Smith RMH, Haarhoff P, Batema MD rhythms, and aberrations in global climate 65 Ma to present. (2011) Regional and global context of the Late Cenozoic Science 292:686–693 Langebaanweg (LBW) palaeontological site: West Coast of South Zelditch ML, Swiderski DL, Sheets HD (2012) Geometric morphomet- Africa. Earth Sci Rev 106:191–214 rics for biologists, second edition: a primer. Elsevier, San Diego