Introduction

Myliobatiformes (stingrays) is a derived clade of batoid fishes known for having a characteristic serrated caudal spine. They are found worldwide and most species are living near the coast (Nelson 2006). Extant members of Myliobatiformes have been studied extensively and relationships among the derived taxa in the clade are well resolved based on morphology (Compagno 1973, 1977; Heemstra and Smith 1980; Maisey 1984; Nishida 1990; Lovejoy 1996; McEachran et al. 1996; Compagno 1999; Carvalho et al. 2004; González-Isáis and Domínguez 2004). Myliobatidae is a highly nested clade within Myliobatiformes. Its members consist of pelagic and sometimes coastal stingrays that are widely dispersed in modern oceans (e.g., Lovejoy 1996; Nelson 2006). The presence of powerful crushing jaws with several rows of pavement-like teeth is characteristic of these durophagous fishes (Summers 2000; Dean et al. 2005). There are seven extant myliobatid genera (Myliobatis, Aetobatus, Aetomylaeus, Manta, Mobula, Pteromylaeus, and Rhinoptera) containing approximately 37 species (Nelson 2006). Myliobatinae (sensu Nelson 2006) is a subset of taxa within Myliobatidae including Myliobatis, Aetobatus, Aetomylaeus, and Pteromylaeus. Evidence from phylogenetic analyses indicates that Myliobatinae is paraphyletic (Lovejoy 1996; Carvalho et al. 2004).

The myliobatid fossil record is extensive, with taxa known primarily from isolated dentitions. Approximately 150 extinct species have been identified and several have first appearances in the Late Cretaceous (Berg 1940; Cappetta 1987). To investigate further the phylogeny of Myliobatidae and the survivorship of the clade across the Cretaceous—Paleogene (K/T) boundary we expand upon previously published studies of extant Myliobatiformes and Myliobatidae by including fossilized dentitions and new characters derived from dental morphology.

Our survey of museum collections and the literature reveals that thousands of isolated dentitions have been attributed to Myliobatis for over a century (Woodward 1888); however, the monophyly of this expanded concept of Myliobatis has never been tested phylogenetically. To test the monophyly of Myliobatis, it is necessary to integrate taxa known exclusively on the basis of dentitions into a phylogenetic matrix that includes characters from the entire skeletal system (e.g., Carvalho et al. 2004). Such a taxon sample introduces both new data and missing data (empty cells) into the phylogenetic analysis, but the inclusion of data from previously unsampled species is essential for completing a total phylogeny of the clade (Kearney 2002; Wiens 2003, 2006; Kearney and Clark 2003; Shimada 2005). Furthermore, the inclusion of fossils facilitates the study of ghost lineages, which is critical for constructing phylogenetically informed hypotheses of species diversity in any given time horizon (Smith 1994).

Ours is the first examination of a diverse assemblage of extinct species of Myliobatis and their impact on relationships of all myliobatid taxa. We describe new fossil myliobatid dentitions from three Maastrichtian (Late Cretaceous) localities in the Iullemmeden Basin of Mali that were collected by a Centre Nationale de la Recherche Scientifique et Technologique-Stony Brook University expedition in 1999. We test relationships of Myliobatidae by adding these specimens as well as other fragmentary specimens to the published phylogenetic matrix of Carvalho et al. (2004). In contrast to prior investigations, we use species as operational taxonomic units (OTUs) to test the monophyly of myliobatid genera. By investigating the placement of the Malian and other fossil taxa within Myliobatidae we further examine the survivorship of the clade across the K/T boundary. By working at the species level our study can also be readily expanded upon by future systematists (see discussion by Prendini 2001).

Institutional abbreviations.—AMNH, American Museum of Natural History, New York, USA; ANSP, Academy of Natural Sciences, Philadelphia, USA; CNRST-SUNY, Centre Nationale de la Recherche Scientifique et Technologique, Bamako, Republic of Mali-Stony Brook University, Stony Brook, USA; FMNH, Field Museum of Natural History, Chicago, USA; MCZ, Museum of Comparative Zoology, Harvard University, Cambridge, USA; NHM, The Natural History Museum, London, UK; SMF, Senkenberg Museum, Frankfurt, Germany; SMNS Staatliches Museum für Naturkunde, Stuttgart, Germany; TNHC, Texas Natural History Collection, Austin, USA; YPM, Yale Peabody Museum, New Haven, USA; YPM-PU, Yale Peabody Museum, Princeton Collection, New Haven, USA; ZMB, Museum für Naturkunde, Berlin, Germany.

Other abbreviation.—CI, consistency index; HI, homoplasy index; OTU, operational taxonomic unit; RC, rescaled consistency index; RI, retention index; TL, tree length.

Geological setting

The Iullemmeden Basin of northern Mali and Niger (Fig. 1) preserves Cretaceous and Paleogene marine and continental strata deposited in and along the shores of the epeiric Trans-Saharan Seaway. This seaway inundated parts of central West Africa in the Early Cretaceous, and then again from the Late Cretaceous to middle Eocene (Petters 1979; Reyment and Dingle 1987; Tapanila et al. 2004, 2008). Relatively thin sedimentary packages derived from each of these two cycles are found in northern Mali, along the margin of the Adrar des Iforas Mountains (Radier 1959; Bellion et al. 1989; Moody and Sutcliffe 1991; Tapanila et al. 2004, 2008). The strata occur in fining-upward sequences (Fig. 2) associated with transgressive-regressive cycles of the Trans-Saharan Seaway, and they preserve diverse invertebrate and vertebrate faunas that demonstrate the seaway periodically served as a faunal conduit between the Tethys and South Atlantic Ocean (Moody and Sutcliffe 1993).

Several fossiliferous sections (localities Mali-7, -8, and -10) spanning the K/T boundary in the vicinity of Ménaka were measured and placed within a rigorous stratigraphic and sedimentological framework (Tapanila et al. 2004, 2008; Gaffney et al. 2007; Hill et al. 2008). Detailed facies analysis of the Ménaka area and other localities revealed the presence of five repeated facies, including sandstone and siltstone (Facies 1), paper shale (Facies 2), mollusk and echinoderm packstone (Facies 3), mudstone and wackestone (Facies 4), and phosphate conglomerate (Facies 5; Tapanila et al. 2008). Those facies associations are interpreted respectively as tidally-influenced shoreline and deltaic environments (Facies 1); shallow, normal-to-restricted marine lagoons and open platform settings (Facies 2); small patch oyster reefs and storm beds associated with shallow, sublittoral marine settings under normal salinity (Facies 3); low-energy shallow, sublittoral open marine settings with water depths <50 m (Facies 4); and shallow marine-to-brackish water phosphorites associated with periods of amalgamation and concentration by storm activity during periods of marine transgression (Facies 5; Tapanila et al. 2008).

Fig. 1.

Map of Mali indicating three localities discovered in the 1999 CNRST-SUNY expedition. Boundary between the Illummeden and Tauodeni basins in northern Mali is outlined in light gray. Mali-8 marks localities yielding fossils of Myliobatidae. Dark Gray marks exposed basement rocks in the Adrar des Iforas Mountains; white marks Proterozoic structure that connected the two light gray basin periodically during the Cretaceous— Paleogene.

Fig. 2.

Composite stratigraphic sections of localities Mali-7, -8, and -10. Relative stratigraphic positions of index fossils and inferred depositional settings supporting age of Myliobatis wurnoensis (Mali-8). Index fossils from Mali-7, -8, and -10. Lower gray line is the inferred KT boundary in this section and the upper gray line is the inferred position of the Paleocene—Eocene boundary in this section. Abbreviations: CG, conglomerate; LS, limestone; MS, shale; SS, sandstone.

The section at Mali-8 in particular (Figs. 1 and 2) consists of approximately 35 m of interbedded sandstones (Facies 1), shales (Facies 2), marly limestones (Facies 3 and 4), and a thick phosphate conglomerate (Facies 5), which were deposited as part of at least one complete sea level cycle (T5 Cyclothem of Greigert [1966]) in the Trans-Saharan Seaway (see Tapanila et al. [2008] for complete discussion). The Mali-8 section was well-correlated with another section located approximately 15 km away (∼45 m thick), designated as Mali-7 (see also Hill et al. [2008: fig. 2] for detailed stratigraphic correlations). The base of each section is defined by Facies 1, characterized by large, inclined, fine-grained sandstone beds with glauconite and petrified wood of delta front origin. The next 20 m of each section represents deposits of the T5 transgressive systems tract, dominated by thinly laminated shales and marls of Facies 2 and 3, which are overlain by phosphatic limestones and a distinctive phosphate conglomerate (Facies 5) representing the maximum flooding surface (condensed section) of the T5 cyclothem. Above that level are open marine, quiet water mudstones and wackestones (Facies 4) and storm generated molluscan and echinoderm packstones (Facies 3) associated with the overlying T5 highstand systems tract (Tapanila et al. 2008: fig. 3).

Several myliobatid specimens were found in these sections, which also contain a variety of index fossils allowing us to use biostratigraphy to make a biochronological age assessment of the section. The base of the section (Fig. 2) contains latest Cretaceous index fossils, including the sawfish shark, Schizorhiza stromeri (which occurs in Maastrichtian deposits in Africa, Iraq, and North and South America; Cappetta [1987]), and the regular sea urchin, Echinotiara perebaskinei (Smith and Jeffery 2000). Furthermore, specimens of another Maastrichtian index fossil, Cretalamna maroccana, a lamniform shark (Case and Cappetta 1997; Shimada 2007) were also recovered from shale beds stratigraphically above the ray specimens. Approximately 10 meters higher in the section, specimens of the Paleogene echinoderm Oriolampas michelini bracket the approximate K/T contact in northern Mali.

Myliobatid specimens described here were collected at Mali-8 from a distinctive bone, pebble, and coprolite phosphatic conglomerate (Facies 5) at the 26 m level. Immediately above this interval, multiple specimens of the pseudoceratitic ammonite Libycoceras crossensi (Fig. 2) were recovered from Facies 4 mudstones and wackestones at localities Mali-7 and -8 demonstrating a Maastrichtian age for this part of the sequence, and establishing the first Mesozoic record of the morphologically characteristic myliobatid dentition. The ammonitebearing mudstones and wackestones are interpreted as lowenergy marine depositional environments, indicating that these fossils are in situ and have not been reworked from older deposits. Sequence stratigraphic analysis by Tapanila et al. (2008) also indicated that deposits at Mali-7 and -8 sections correspond to the second Libycoceras transgression (T5) documented throughout the Trans-Saharan Seaway. This stratigraphic interpretation is comparable with other sequence stratigraphic investigations by Greigert (1966) in the Iullemmeden Basin to the southeast and by Bellion et al. (1989) to the west in the Taoudeni Basin. A wealth of other faunal and geologic investigations on the Maastrichtian transgressive sequence (T5 Cyclothem) directly below the K/T boundary, also support the stratigraphic interpretations for Mali-7 and-8 (Radier 1959; Moody and Sutcliffe 1993; Dikouma et al. 1994; Colin et al. 1996; Zaborski and Morris 1998).

Systematic paleontology

Fig. 3.

The fossil stingray Myliobatis wurnoensis White, 1934 from Maastrichtian of Mali. A, B. Partial upper dental plates. A. CNRST-SUNY-5 in posterior (A₁), occlusal (A2), and basal (A₃) views. B. CNRST-SUNY-37 in posterior (B₁), occlusal (B₂), and basal (B₃) views. C. Partial lower dental plate, CNRSTSUNY-3 in occlusal (C1) and basal (C₂) views. Anterior is to top of page for all images except A₁ and B₁, which are in posterior view. Scale bars 10 mm.

Phylogenetic analysis

Taxonomic sample.—Roughly 90 fossil dentitions, some fragmentary, representing extinct species were examined and compared with dentitions of 25 specimens of extant myliobatids (Appendix 1). The data matrix expands on that presented by Carvalho et al. (2004), which included 23 taxa that were scored at the genus-level. Our final combined matrix includes 40 terminal taxa. Thirty-eight are myliobatiform ingroup taxa and two, Raja and Rhinobatos, are outgroup taxa (Appendix 2). We retained the genus level OTUs of the Carvalho et al. (2004) analysis with the exception of species in Myliobatidae. We decompose Myliobatis into eight species-level OTUs (four extant and four extinct), and Aetobatus and Rhinoptera into three species-level OTUs each, so that we might assess the monophyly of Myliobatis, Aetobatus, Rhinoptera, and Myliobatidae. For the 13 taxa represented by dentition only, initial identifications made for this study were based on criteria detailed by Cappetta (1987). Most taxa were examined from original material; however, original material could not be accessed for all characters. In those cases information was scored from the literature.

Character sample.—We compiled a total of 65 morphological characters (Appendices 2 and 3) using Mesquite 2.0 (Maddison and Maddison 2006) and MorphoBank (O'Leary and Kaufman 2007). The entire data matrix is retrievable with supporting images at MorphoBank.org. The data matrix includes 43 unmodified characters from Carvalho et al. (2004). An additional 22 new characters (numbers 44–65) are exclusively dental and are described in this text. Characters and states for the entire matrix are listed in Appendix 3.

Phylogenetic methods.—We treated all characters as unordered and equally weighted. Multistate characters were treated as uncertain. The character matrix was analyzed using PAUP* 4.0b10 (Swofford 2002) using the maximum parsimony optimality criterion. We employed heuristic searches with 1000 replicates of random stepwise addition (branch swapping: tree-bisection-reconnection) holding one tree at each step. Branches were collapsed to create soft polytomies if the minimum branch length was equal to zero (amb- option); afterwards, we explored agreement subtrees (Cole and Hariharan 1996). We calculated Bremer support (Bremer 1994) for nodes retained in the strict consensus tree. This was done manually in PAUP* using constraint trees generated in MacClade 4.08 for OS X (Maddison and Maddison 2005) from the Decay Index PAUP* File command. We report unambiguous optimizations for particular nodes of interest retained in the strict consensus (optimizations were performed on individual most parsimonious trees). We calculated ghost lineages (Norell 1992; Cavin and Forey 2007) by mapping part of our tree onto the stratigraphic record, using First Appearance Data as described in Cappetta (1987) and the new Cretaceous record of M. wurnoensis as described in our study.

Results

All characters were parsimony informative and the matrix had 27.6% missing data. The heuristic search resulted in eight most parsimonious trees. The strict consensus tree is depicted in Fig. 4 with Bremer support values given for all nodes. Unambiguous character changes are mapped for Myliobatidae in Fig. 5. Ghost lineages are drawn in Fig. 6. The strict consensus tree depicts a paraphyletic Myliobatinae (Fig. 4). Myliobatis is also paraphyletic. A revision of the taxonomy of the species of Myliobatis is outside the scope of this paper.

The strict consensus tree topology (Fig. 4) is congruent with the consensus tree recovered by Carvalho et al. (2004) and the non-Myliobatidae portion of our consensus tree is identical to that portion in theirs (Fig. 4A). The optimal agreement subtree removed only five (Plesiobatis, “Himantura”, Dasyatis, Pteroplatytrygon, and the extinct taxon Asterotrygon) of 40 taxa, all of which were outside the ingroup, Myliobatidae. From here on we describe in detail only the hypothesized relationships within Myliobatidae (Figs. 4B, 6). These differ from the relationships hypothesized by Carvalho et al. (2004).

Hypolophites is the immediate sister taxon to a monophyletic Myliobatidae (Figs. 4B and 5: node-A); these clades share broad, six sided, pavement-like teeth. The majority of extinct taxa we studied are interspersed among extant myliobatids. Apocopodon is the sister taxon to all remaining taxa within Myliobatidae (Figs. 4B and 5: node-B). Myliobatidae is distinguished from outgroup taxa by their expanded median teeth with a polyaulacorhizous root morphology that interlock by a tongue and groove mechanism. There is a polytomy at node-C (Figs. 4B and 5) consisting of Myliobatis freminvillii, Myliobatis goodei, and the unnamed clade including node-D. No unambiguous character changes were mapped to the polytomy at node-C. No unambiguous character changes could be mapped to the polytomy at node-D, comprising the extinct taxon Myliobatis striatus, Myliobatis aguila, Myliobatis californica, and the unnamed clade including node-E (Figs. 4B, 5). These two polytomies were present in all eight most parsimonious trees as an effect of the amb- option during tree searches.

There were no unambiguous character changes mapped for the clade including node-E. This clade, including node-E, consists of the extinct taxon Myliobatis toliapicus, which is sister to the clade including node-F (Figs. 4B, 5). The clade including node-F is distinguished from M. toliapicus by possessing upper teeth that are curved. At node-F, the extinct taxon, Weissobatus micklichi is sister taxon to the clade including node-G. Node-G is distinguished from W. micklichi by non-dental morphology, a pelvic girdle that is arched.

At node-G, an Aetobatus-clade splits from the remainder of Myliobatidae (Fig. 4B). Node-H includes five extinct taxa and leads towards Rhinoptera, Mobula, and Manta. That clade is distinguished from Aetobatus by possessing domed tooth crowns, as opposed to deep crowns. At node-H, the extinct taxon Myliobatis dixoni is sister taxon to the clade including node-I. The clade including node-I possesses wide, blocky, and irregularly spaced root laminae. The extinct taxon Myliobatis wurnoensis possesses a single autapomorphy, pinched lateral margins of the median teeth, distinguishing it from the clade including node-J. Node-J is further distinguished from Myliobatis wurnoensis by possession of teeth that are differentially expanded, loosely interlocking, and connected by a bulbous tongue and groove joint. At node-J, the extinct taxon Igdabatis is the sister taxon to Rhinoptera + (Brachyrhizodus + [Mobula + Manta]) (Figs. 4B and 5: node-K).

Fig. 4.

Strict consensus of eight most parsimonious trees (MPT). A. Tree from full analysis with Myliobatidae condensed as single terminal taxon in gray box labeled “B”. B. Expanded Myliobatidae portion of tree, which is identical on all eight MPTs. TL = 141, CI = 0.6312, HI = 0.3688, RI = 0.8844, RC = 0.5583. Bold face in B denotes extinct taxa.

Rhinoptera is monophyletic (Figs. 4B and 5: node-L, Rhinopterinae sensu Nelson 2006). Rhinoptera davisei is more closely related to Rhinoptera quadriloba than either taxon is to Rhinoptera bonasus (Fig. 5: node-M). The relative position of these species of Rhinoptera in the consensus tree and the morphological variation reported for each species (Appendix 3) is consistent with R. quadriloba being a valid species.

Brachyrhizodus is the sister taxon to Mobula + Manta (Fig. 5: node-L), contrary to the hypothesis proposed by Cappetta (1987) that Brachyrhizodus is closely related to Rhinoptera. We recover Mobulinae (Fig. 4B; node-O, Mobulinae sensu Nelson 2006): Rhinoptera as the sister taxon of Brachyrhizodus + (Mobula + Manta). Rhinoptera is distinguished from Mobulinae by possessing regularly spaced, fine edged root laminae, which are narrower than the grooves dividing them, while Mobulinae lack any curvature to their teeth. Brachyrhizodus is a stem-mobuline and Mobula + Manta represent crown-Mobulinae (Fig. 4B: node-O). Rhinoptera + Mobulinae share a straight tooth crown. Mobula + Manta reversed their tooth morphology to possess minute teeth with a low crown. At node-P, Aetobatus is monophyletic. Aetobatus irregularis + Aetobatus narinari are more closely related to one another than either is to Aetobatus arcuatus. Aetobatus is distinguished from the clade including node-H by eight unambiguous character changes (Fig. 5).

Fig. 5.

Summary of unambiguous character transformations across Myliobatidae (node-B) that were optimized on all most parsimonious trees. Black boxes have a CI = 1.0 and white boxes have a lower CI value. Bold face denotes extinct taxa.

Considering both tree topology and the first appearances of a clade and its sister taxon, we can make inferences about ghost lineages (Norell 1992). New fossils of Myliobatis wurnoensis demonstrate for the first time that this taxon occurred on both sides of the K/T boundary (Fig. 6). The new M. wurnoensis material recovered from lagoonal/shallow subtidal deposits of the Cretaceous of Iullemmeden Basin represents a temporal extension for the species of several million years. At least two other genera within Myliobatidae have a fossil record prior to the K/T boundary, Igdabatis and Brachyrhizodus (Romer 1942; Cappetta 1972; 1987; Cappetta and Case 1975; Prasad and Cappetta 1993). These taxa are more highly nested than all species of Myliobatis. Thus, the relative positions of all Cretaceous taxa within Myliobatidae unambiguously extend the record of the clade across a major extinction boundary. The present phylogeny shows that all major lineages of Myliobatidae examined existed in the Mesozoic, making the Mesozoic diversity count much greater than a direct tally of stratigraphic occurrences alone.

Fig. 6.

Phylogenetic relationships and stratigraphic distribution of Myliobatidae. Epochs are not drawn to scale.

Dental character analysis and polarity

The characters that are new or modified for this study are described here with reference to figures illustrating these states (Figs. 7, 8). A complete character list is available in Appendix 3 and at Morphobank.org. Characters 44–46 are modified from character 19 by Carvalho et al. (2004: 84), which described “Arrangement of teeth in both upper and lower jaws” as a combination of tooth type, shape, and relative position. We split this character because tooth type, shape, and position do not vary together.

(44) Tooth type in both upper and lower jaws: (0) minute; (1) broad; (modified from character 19, Carvalho et al. [2004]). In outgroup taxa and most non-myliobatid stingrays, teeth are usually minute (state 0; Fig. 7A, B). The derived mobuline taxa Mobula and Manta, have secondarily minute cusps (Fig. 7E–G). The alternative tooth type is a broad, flattened one (state 1; Figs. 7C, D, H, 8B–F). The extinct Hypolophites, which has enormous individual teeth, demonstrates the derived state (Fig. 8A).

(45) Arrangement of teeth in both upper and lower jaws: (0) arranged in separate diagonal rows or ribbons; (1) horizontal conveyor or pavement-like arrangement; (modified from character 19, Carvalho et al. [2004]). The many individual teeth in non-myliobatid taxa are lined up in separate, diagonal rows or criss-crossing ribbons (state 0; Figs. 7A, B, 8A). In all myliobatid taxa, however, a more horizontal conveyor or pavement-like arrangement is present (state 1; Figs. 7C–H, 8C–F). The horizontal alignment of the teeth is retained in Mobula and Manta, despite their relatively minute size (Fig. 7F₂, G₂).

(46) Tooth shape: (0) square to rounded; (1) hexagonal, six distinct sides; (2) rectangular with posteriorly deflected lateral margins; (modified from character 19, Carvalho et al. [2004]). The minute teeth of non-myliobatid taxa are square to round and they may slightly overlap one another (state 0; Fig. 7A, B). Teeth with six recognizable sides are present in the extinct, non-myliobatid taxon, Hypolophites, as well as all myliobatids except for Aetobatus (state 1; Figs. 7C–G, 8A). Aetobatus, which almost always consists only of a single row of tooth plates, has teeth that are more rectangular and not six-sided (state 2; Fig. 7H). The appearance of six-sided teeth co-occurs with the appearance of broad and flattened teeth. These states are retained in the secondarily minute-toothed taxa, Mobula and Manta, and lost in the greatly expanded-toothed taxon, Aetobatus.

(47) Lateral teeth: (0) present; (1) absent. Lateral teeth are present in most taxa examined (state 0; Figs. 7A–G, 8A–E). Lateral teeth were observed to be absent from almost all specimens of Aetobatus (state 1; Figs. 7H, 8F). In a single specimen of Aetobatus narinari (TMM-M-7010) five lateral tooth plates were present among all 19 intact median tooth plates. We chose to score Aetobatus with the derived state, because this observation was isolated to a single specimen.

(48) Differentiation of median teeth from lateral teeth: (0) median and lateral teeth are similar; (1) median teeth relatively expanded. Median teeth can be similar to laterally adjacent teeth as they are in all non-myliobatid taxa and in Manta (state 0; Figs. 7A, B, G, 8A). Median teeth can otherwise be expanded in the lateral direction, i.e., they are broader than they are anteroposteriorly long (state 1; Figs. 7C, D, F, 8B–E). Mobula is the only taxon examined to possess expanded median teeth despite having small teeth. Aetobatus is scored as not applicable (-) because it possesses no lateral teeth.

(49) Differentiation among lateral teeth: (0) lateral teeth unexpanded; (1) some lateral teeth expanded. Lateral teeth are primitively unexpanded relative to median teeth and other lateral teeth (state 0; Figs. 7A, B, 8A). In Rhinoptera, Mobula, Brachyrhizodus, and Igdabatis, some lateral teeth are expanded, and resemble the longer teeth of the median tooth row (state 1 ; Fig. 7C, F). Lateral teeth in Manta are considered secondarily unexpanded (Fig. 7G) Aetobatus is scored as not applicable (-) because it possesses no lateral teeth.

(50) Relative amount of curvature in expanded lower teeth: (0) straight and uncurved; (1) moderately curved; (2) strongly curved into distinct chevron. Individual teeth are often straight, or uncurved in the extant myliobatid taxa (Fig. 7B–G) Rhinoptera, Myliobatis freminvillii, Myliobatis goodei, Myliobatis aguila, Myliobatis californica, Mobula, and Manta and the extinct taxon Myliobatis striatus (state 0; Fig. 8E). Median and sometimes lateral teeth (when expanded) may be slightly curved (state 1). This state is observed in the majority of extinct taxa examined, namely, Myliobatis toliapicus, Myliobatis dixoni, Myliobatis wurnoensis, Weissobatis, Brachyrhizodus, and Igdabatis (Figs. 3C, 8B–D). A third condition is a strongly curved chevron-shaped tooth plate, which is present only in Aetobatus (state 2; Figs. 7H, 8F). Non-myliobatid taxa were scored as not applicable (-) because they possessed no expanded teeth.

(51) Upper tooth curvature: (0) uncurved; (1) curved. The amount of curvature in upper median teeth is always less than in the lower median teeth, even in Aetobatus. We observed only two states for upper median teeth: uncurved (state 0) or curved (state 1). Curved upper teeth were present in the extinct taxa Myliobatis dixoni, Myliobatis wurnoensis, Weissobatis, Igdabatis, and Aetobatus arcuatus (Figs. 3A, B, 8D). All other myliobatid taxa had no distinct curvature to their expanded teeth. Non-myliobatid taxa were scored as not applicable (-) because they possessed no expanded teeth. Brachyrhizodus was scored as missing (?) because we consider the teeth we sampled to be lower dentition.

(52) Direction of tooth curvature: (0) concave; (1) horizontal; (2) convex. This character is observable on upper and lower dental plates and is distinct from characters 50 and 51. A concave curvature occurs in the taxa that have median teeth with lateral margins that are directed anteriorly or out of the mouth when articulated with the jaw, i.e., extinct taxa Brachyrhizodus, Myliobatis wurnoensis (state 2; Figs. 3, 8B). In taxa without expanded tooth curvature, the condition is scored as horizontal (state 1). The third condition, a convex curvature, occurs when the lateral margins of median teeth are directed posteriorly, or into the mouth (state 2 as it is in Aetobatus (Figs. 7H, 8F).

Fig. 7.

Comparative extant taxa of Myliobatidae. A, B, D. Articulated jaws and tooth rows. C. Disarticulated jaws and articulated tooth rows. E–G. Articulated tooth rows. A. Raja sp., AMNH 92321b, in labial view. B. Dasyatis sp., FMNH 15625, in labial view. C. Rhinoptera quadriloba (LeSueur, 1817), FMNH 82986, in occlusal view. D. Myliobatis californica Gill, 1865, MCZ 424, in lingual view. E. Mobula hypostoma (Bancroft, 1831), AMNH 44124, in occlusal view, photograph (E1), line drawing (E₂); F. Mobula rochebruni (Vaillant, 1879), FMNH 38450, in occlusal view, photograph (F1), line drawing (F₂). G. Manta hamiltoni (Walbaum, 1792), FMNH 41385, in occlusal view, photograph (G1), line drawing (G₂). H. Aetobatus narinari (Euphrasen, 1790), FMNH 10985, in labial view.

(53) Tooth association: (0) loosely interlocking; (1) sometimes loosely interlocking or tightly interlocking; (2) tightly interlocking. Fossilized myliobatid dentitions are found in isolation or as whole dental plates. Isolated teeth imply a loosely interlocking association among articulated teeth (state 0). Multiple teeth found in articulation that must be forcibly disarticulated in order to observe them in isolation are considered derived (state 2). We observed a third state in one taxon, Aetobatus irregularis, which we tentatively consider an intermediate state (state 1). Occasionally dentitions were present as loosely interlocking isolated teeth or as tightly interlocking teeth as indicated by two or more plates associated with each other. It was usually the case that upper teeth of A. irregularis were found disarticulated while lower teeth were in articulation.

(54) Tooth interlocking mechanism: (0) overlapping; (1) tongue and groove; (2) no direct contact. The way that teeth interlock anteroposteriorly is not a direct indication of how tightly those teeth interlock and therefore it is scored separately. Teeth may overlap as is the case for non-myliobatid stingrays (state 0). In most myliobatid stingrays, they may fit together by means of a tongue and groove (state 1). Teeth in Mobula and Manta, which are secondarily minute, as well as the extinct taxon, Brachyrhizodus, have no direct contact (state 2).

(55) Shape of interlocking mechanism: (0) bulbous; (1) short shelf; (2) long shelf. The tongue of the tongue-and-groove interlocking mechanism may exist as a bulbous ridge, which is the condition in Rhinoptera and the extinct taxon Igdabatis (state 0; Fig. 8D₂, D₄). In Myliobatis and Weissobatis, there is a short posteriorly directed shelf (state 1; Fig. 8C₂, E₄). In Aetobatus, there is a long, posteriorly directed shelf created by the roots that begins slightly anterior to the posterior margin of the tooth crown and extends far posteriorly, past the posterior margin of the tooth crown (state 2; Fig. 7F₄). Taxa without a tongue and groove articulation were scored as not applicable (-) in the data matrix.

(56) Crown height: (0) crown height exceeds root depth on unworn teeth; (1) crown height does not exceed root depth on unworn teeth. Crown height was scored using median teeth for Myliobatidae. Although crown height is variable within many batoid species seasonally, the relative height of the crown compared to the depth of the underlying roots is consistent in specimens observed. In most taxa, the crown height is high, exceeding the root depth in unworn teeth (state 0; Figs. 3A₁, B1, 8E₃). Aetobatus, Mobula, and Manta, all have a relatively low crown, where the height does not exceed root depth on unworn teeth (state 1; Fig. 7F₃).

(57) Occlusal surface: (0) cusped; (1) smooth; (2) depressed. The occlusal surface of teeth, prior to wear, is cusped with one or more peaks ancestrally (state 0; Fig 7A). The surface is otherwise smooth, with no cusps or depressions in most taxa (state 1: Figs. 3B₂, C₂, 7B–H, 8A₁, B₁, C₁, D₁, F₁). In the extinct taxon, Rhinoptera davisei, and Manta, the occlusal surface is depressed in the center (state 2). Dasyatis is polymorphic for smooth and depressed occlusal surfaces. Mobula is polymorphic for smooth and cusped occlusal surfaces (Fig. 7E, F).

(58) Crown shape in anterior or posterior view: (0) straight; (1) domed; (2) deep. Crown shape was scored using median teeth for Myliobatidae. The occlusal surface of a tooth is not a good indication of the overall morphology of the tooth crown. When viewed anteriorly or posteriorly, the crown may be straight or uniformly thick (state 0), as it is in the extinct taxon, Brachyrhizodus, as well as Myliobatis freminvillii, Myliobatis goodei, Aetobatus narinari, Aetobatus irregularis, Rhinoptera, Mobula, and Manta. The crown may also be domed with a bulbous, outward or lingually curving surface (state 1), as is the condition in Myliobatis aguila, Myliobatis dixoni, Myliobatis wurnoensis, and Igdabatis (Figs. 3A₁, 8D₂). The third observed condition is a deep crown, where the surface is not greatly domed, but the contact with the roots bulges towards the jaw cartilages (state 2), as it does in Myliobatis californica, Myliobatis striatus, Myliobatis toliapicus, and Aetobatus arcuatus (Fig. 8E₃, F₃).

(59) Lateral margins: (0) not pinched; (1) pinched. When viewed anteriorly or posteriorly, the lateral margins of the median teeth are often the same height as middle of the tooth or gradually-sloped so that the distal ends of the teeth form a peak, much lower than the middle of the tooth (state 0). In the extinct taxon, Myliobatis wurnoensis, lateral margins were pinched relative to the center of the tooth (state 1; Fig. 3A₁, B1). This character state is autapomorphic for M. wurnoensis.

(60) Root type: (0) holaulacorhizous; (1) polyaulacorhizous. Roots in the stingrays are either holaulacorhizous (state 0; Fig. 8A₃) or polyaulacorhizous (state 1; Figs. 3A₃, B₃, C₁, 8B₂, C₃, D₃, E₂, F₂); see Cappetta (1987) for review.

(61) Number of roots: (0) 2 roots; (1) 3 to 4 roots; (2) 5 roots or greater. Number of root laminae in non-myliobatid stingrays is 2 (state 0; Fig. 8A₃). In Brachyrhizodus there are 3 or 4 roots (state 1; Fig. 8B₂). All other taxa examined from Myliobatidae have 5 or more roots (state 2; Figs. 3A₃, B₃, C₁, 8C₃, D₃, E₂, F₂).

(62) Roots in basal view: (0) triangles; (1) wide blocks; (2) narrow blocks; (3) fine edges. The shape of the root laminae in basal view is triangular in non-myliobatid stingrays (state 0; Fig. 8A₃). The extinct taxa, Apocopodon, Brachyrhizodus, Igdabatis, and Myliobatis wurnoensis possess wide blocks (state 1; Figs. 3A₃, B₃, C₁, 8B₂, C₃, D₃). Myliobatis striatus, Myliobatis freminvillii, Myliobatis aguilla, Myliobatis toliapicus, Myliobatis goodei, and Weissobatis have narrow blocks (state 2; Fig. 8E₂). Rhinoptera and Aetobatus have extremely thin, fine, comb-like edges to their laminae (state 3; Fig. 8F₂). The condition of this character is unknown in Manta and Mobula and scored as “?” in the data matrix.

(63) Distance between root laminae: (0) narrower than root laminae; (1) broad, groove wider than root laminae. Ancestrally, the grooves dividing each lamina are narrower than the root itself (state 0). In Rhinoptera, Myliobatis californicus, and Dasyatis, however, the groove between each lamina is wider than the lamina itself (state 1).

Fig. 8.

Comparative extinct taxa of Myliobatiformes; known ages mapped onto Fig. 6. A. Hypolophites myliobatoides Stromer, 1910, NHM P18781; A₁, occlusal view, anterior to top; A₂, lateral view, anterior to left; A₃, root view, anterior to top. B. Brachyrhizodus wichitaensis Romer, 1942, NHM P89095; B₁, occlusal view; anterior undetermined; B₂, root view; anterior undetermined. C. Apocopodon sericius, NHM P24670, C₁, occlusal view, anterior to top; C₂, lateral view, anterior to left; C₃, root view, anterior to top. D. Igdabatis sigmodon, TMM 45892-1; D1, occlusal view, anterior to top; D₂, posterior view; D₃, root view, anterior to bottom; D₄, lateral view, anterior to left. E. Myliobatis striatus, NHM P.66859; E₁, occlusal view, anterior to top; E₂, root view, anterior to top; E₃, posterior view; E₄, lateral view, anterior to left. F. Aetobatus arcuatus, SMNH 12656-3; F₁, occlusal view, anterior to top; F₂, root view, anterior to top; F₃, anterior view; F₄, lateral view, anterior to left. Scale bars 10 mm.

(64) Inclination of roots: (0) no inclination; (1) offset and step-like; (2) long and strongly inclined. When viewed from the side, the laminae of the root in the majority of taxa have a vertical slope, with no posterior offset or inclination (state 0; Fig. 8C₂, D₄). In some taxa, the roots are slightly offset and step like as in Myliobatis striatus, Myliobatis toliapicus, Rhinoptera quadriloba, and R. davisei (state 1; Fig. 8E₄). Autapomorphic to Aetobatus is the presence of long and strongly posteriorly inclined roots (state 2; Fig. 8F₄).

(65) Root groove position: (0) regularly spaced between laminae; (1) irregularly spaced between laminae. The spacing of grooves between laminae is often at regular intervals (state 0). However, in Brachyrhizodus, Igdabatis, and Myliobatis wurnoensis those grooves can be irregularly spaced (state 1; Figs. 3A₂, B₃, C₁, 8B₂, D₃).

Discussion

The comprehensive phylogenetic study of Myliobatiformes by Carvalho et al. (2004) was based on several extant taxa and a number of relatively complete and exquisitely preserved fossils from the late early Eocene Green River Formation of Wyoming. Those authors found that the Eocene fossils were relatively basal among Myliobatiformes and that including those fossils in their analysis helped to resolve relationships among all Myliobatiformes. Not studied, however, were the less complete fossil specimens of taxa within Myliobatidae that include a plethora of fragmentary, although characteristic, pavement-like dentitions.

Here we test for the first time the relationships of several extinct taxa within Myliobatidae. We integrate taxa known from fossilized dental plates, including specimens recovered from the Cretaceous of Mali, into a phylogenetic analysis that combines data from prior studies, including non-dental character systems scored for extinct and extant taxa. We redescribe through the character analysis dental morphology for Myliobatiformes and offer new characters and states for distinguishing and relating species in Myliobatidae. The results of the phylogenetic analyses here corroborate the topologies of several genus-level morphological analyses that were conducted for the Myliobatidae (e.g., Nishida 1990; Lovejoy 1996; McEachran et al. 1996; Shirai 1996; Carvalho et al. 2004). We also find that synapomorphies for major nodes within Myliobatidae are often dental features, which has not been previously observed (Fig. 5).

Extinct taxa are widely distributed across the tree presented here (Fig. 4B). Hypolophites is not most closely related to other members of “Dasyatidae” (such as Dasyatis) as was previously hypothesized (Cappetta 1987), but is instead the immediate sister taxon to a monophyletic Myliobatidae (Fig. 4). Our results, however, are consistent with those from previous studies (Carvalho et al. 2004) that demonstrate that “Dasyatidae” is not monophyletic. Apocopodon is sister to all remaining taxa within Myliobatidae (Fig. 5: node-B) and shares the pavement-like dentition characteristic of the clade. Igdabatis also shares the elongate median teeth characteristic of Myliobatidae, and is the sister taxon to Mobulinae + Rhinoptera. The relative position of the extinct species in Rhinoptera is interesting because R. davisei is more closely related to R. quadriloba than either is to R. bonasus (Fig. 5: node-P). Previously R. quadriloba was considered a junior synonym of R. bonasus (Eschmeyer 1998), therefore, we would have predicted that these species would be sister taxa. The variation recognized in the specimens examined for this study (Appendix 3), however, provides evidence that R. quadriloba is a valid species. Brachyrhizodus is the sister taxon to Mobula + Manta (Fig. 5: node-L), contrary to the hypothesis proposed by Cappetta (1987) that Brachyrhizodus is closely related to Rhinoptera.

The new Malian Myliobatis wurnoensis material represents the first well-preserved upper and lower dentitions of this species. In addition, the new fossils demonstrate for the first time that Myliobatis wurnoensis occurred on both sides of the K/T boundary. Previously, the earliest documented occurrence of the species was the Paleocene—Landenian (de Geyter et al. 2006; early Eocene, sensu White 1934). Several myliobatid taxa have Mesozoic ranges, including Igdabatis and Brachyrhizodus, and collectively, our strict consensus shows that there are a number of ghost myliobatid lineages that cross the K/T boundary. Thus a number of myliobatid taxa are inferred to have survived this mass extinction event and species counts of the Late Mesozoic myliobatids are much higher than a simple count of fossils recovered alone. Aetobatus and Rhinoptera are each monophyletic. Aetobatus is the sister taxon to a clade that includes Rhinoptera + Mobulinae. Our expanded character and taxon sample, however, does not support a monophyletic Myliobatis. The relative position of Aetobatus and Myliobatis on the strict consensus tree also indicates that “Myliobatinae” (sensu Nelson 2006) is paraphyletic (Fig. 4B). “Myliobatinae” was erected based on conventional phenetic similarities, and we have corroborated Nelson's (2006) prediction that the subfamily would be found to be paraphyletic if more species were included in a cladistic analysis. Based on our results, “Myliobatinae” amounts to members of Myliobatidae that are not part of Mobulinae, Rhinoptera, or Aetobatus (i.e., extinct and extant species of Myliobatis, Apocopodon, Weissobatis, and Igdabatis) supporting Nelson's hypothesis of paraphyly.

Examining dentitions comparatively, in the context of whole-body specimens, provides our only means of testing the phylogeny of the total clade (Shimada 1997). The present study is an example of how fragmentary or incomplete specimens, here dentitions, can possess an important suite of characters and states for phylogenetic analysis. This new information allows us to generate testable hypotheses of evolutionary relationships and morphological transformations among closely related taxa.