| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
© 2005 Society of Systematic Biologists
Integration of Morphological Data Sets for Phylogenetic Analysis of Amniota: The Importance of Integumentary Characters and Increased Taxonomic Sampling
Edited by Norman MacLeod: Associate EditorDepartment of Anatomical Sciences, Stony Brook University Stony Brook, New York, 11794–8081 USA
 |
Abstract |
|---|
 TopNotes Abstract Materials and MethodsResultsDiscussionConclusionsReferences |
|---|
Several mutually exclusive hypotheses have been advanced to explain the phylogenetic position of turtles among amniotes. Traditional morphology-based analyses place turtles among extinct anapsids (reptiles with a solid skull roof), whereas more recent studies of both morphological and molecular data support an origin of turtles from within Diapsida (reptiles with a doubly fenestrated skull roof). Evaluation of these conflicting hypotheses has been hampered by nonoverlapping taxonomic samples and the exclusion of significant taxa from published analyses. Furthermore, although data from soft tissues and anatomical systems such as the integument may be particularly relevant to this problem, they are often excluded from large-scale analyses of morphological systematics. Here, conflicting hypotheses of turtle relationships are tested by (1) combining published data into a supermatrix of morphological characters to address issues of character conflict and missing data; (2) increasing taxonomic sampling by more than doubling the number of operational taxonomic units to test internal relationships within suprageneric ingroup taxa; and (3) increasing character sampling by approximately 25% by adding new data on the osteology and histology of the integument, an anatomical system that has been historically underrepresented in morphological systematics. The morphological data set assembled here represents the largest yet compiled for Amniota. Reevaluation of character data from prior studies of amniote phylogeny favors the hypothesis that turtles indeed have diapsid affinities. Addition of new ingroup taxa alone leads to a decrease in overall phylogenetic resolution, indicating that existing characters used for amniote phylogeny are insufficient to explain the evolution of more highly nested taxa. Incorporation of new data from the soft and osseous components of the integument, however, helps resolve relationships among both basal and highly nested amniote taxa. Analysis of a data set compiled from published sources and data original to this study supports monophyly of Amniota, Synapsida, Reptilia, Parareptilia, Eureptilia, Eosuchia, Diapsida, Neodiapsida, Sauria, Lepidosauria, and Archosauriformes, as well as several more highly nested divisions within the latter two clades. Turtles are here resolved as the sister taxon to a monophyletic Lepidosauria (squamates + Sphenodon), a novel phylogenetic position that nevertheless is consistent with recent molecular and morphological studies that have hypothesized diapsid affinities for this clade.
Keywords: Amniota; integument; osteoderms; phylogeny; Testudines
Received June 7, 2004; Revised August 16, 2004; Accepted December 10, 2004
Despite renewed interest in amniote phylogeny, consensus on the phylogenetic position of turtles continues to elude both morphologists and molecular systematists (Fig. 1). Based on morphological data, several conflicting hypotheses have been advanced, alternatively allying turtles with a number of wholly extinct taxa such as captorhinids (Gauthier et al., 1988a), procolophonids (Reisz and Laurin, 1991; Laurin and Reisz, 1995), and pareiasaurs (Gregory, 1946; Lee, 1993, 1997a). Other morphological studies hypothesize that turtles, despite their anapsid (imperforate) skulls, are phylogenetically nested among diapsids (reptiles with two pairs of fenestrae in the temporal region of the skull), either closely allied with Lepidosauromorpha, the taxon containing lizards, snakes, and the tuatara (Rieppel and deBraga, 1996; deBraga and Rieppel, 1997; Rieppel and Reisz, 1999; Rieppel, 2000), or with Archosauromorpha, the taxon containing birds and crocodiles (Merck, 1997). An origin of turtles from within Diapsida necessitates an evolutionary reversal whereby skull fenestrae became secondarily closed.
|
Additional support for each of these hypotheses has emerged from varying lines of molecular evidence. Studies based on mitochondrial DNA (e.g., Zardoya and Meyer, 1998; Kumazawa and Nishida, 1999) ally turtles with a monophyletic Archosauria (crocodiles + birds). In contrast, nuclear DNA evidence supports a sister-taxon relationship between turtles and crocodiles, to the exclusion of birds (Hedges and Poling, 1999; Mannen and Li, 1999; Cao et al., 2000). Analyses that have combined morphological and molecular data (e.g., Eernisse and Kluge, 1993) support placement of turtles outside Diapsida when multiple anapsid fossil taxa are included, but within Diapsida when fossils are excluded from the analysis (Lee, 2001).
The multitude of incongruent hypotheses proposed to explain the phylogenetic position of turtles among amniotes poses an explicit challenge to molecular systematists and comparative anatomists alike to resolve these conflicts. Hundreds of morphological characters have been evaluated separately in prior studies of amniote phylogeny (e.g., Gauthier et al., 1988a, 1988b; Laurin and Reisz, 1995; deBraga and Rieppel, 1997), yet no analysis to date has compiled and integrated all of this anatomical data into a supermatrix. Combining data sets is crucial to studies of amniote phylogeny because doing so may reveal novel phylogenetic groupings that do not emerge from separate analysis of individual data sets (e.g., Barrett et al., 1991; Kluge and Wolf, 1993; Olmstead and Sweere, 1994; Nixon and Carpenter, 1996). Within Amniota in particular, addition of morphological data from fossils has been shown to overturn hypotheses based on data from soft tissue anatomy (Gauthier et al., 1988a) or molecules, even when molecular characters greatly outnumber morphological ones (Lee, 2001).
Here I present a compilation of published morphological data and augment it with several new taxa and characters. This contribution represents the largest set of simultaneously analyzed morphological character data yet assembled for amniotes, and one of the largest ever amassed for any vertebrate group. Furthermore, it is the first study of its kind to incorporate a detailed examination of the morphology and histology of the integument (skin and its associated structures) as a major source of character data for phylogenetic analysis of amniotes. Cladistic analyses of these data are performed here in order to (1) determine the phylogenetic position of turtles among the major subdivisions of Amniota and (2) examine the effects of both increasing taxonomic sample size and adding new characters from the integument.
The Integument and Osteoderms
The amniote integument represents a highly diversified anatomical system that has repeatedly undergone modification into structurally and functionally distinct forms. Because the integument consists mainly of soft tissues, it is often poorly represented in morphological studies that include fossils. In many amniote taxa, however, secondary dermal ossifications (osteoderms) occur within the skin (e.g., Romer, 1956), representing hard tissue components of the integument with high preservation (fossilization) potential. Osteoderms occur in all major lineages of extant amniotes except birds and are common in the fossil record, in many cases providing the only direct evidence of the skin of ancient animals. Osteoderms are commonly referred to as dermal "armor," although a protective function is not indubitable (e.g., Frey, 1988).
Dermal armor has figured prominently in recent discussions of turtle origins. Lee (1997a) hypothesized that the elaborate carapace of turtles may have derived from the fusion of individual osteoderms in a pareiasaurian ancestor. Rieppel and Reisz (1999) disagreed, citing developmental evidence (Patterson, 1977; Starck, 1979; Burke, 1989) that the carapace and plastron must be derived at least partially from endoskeletal elements. Incorporating comparative data on osteoderms and turtle shell into a phylogenetic framework is crucial for understanding the evolution of the integument in Amniota, because doing so can help to elucidate homologies between integumentary derivatives (Hill, 2003, 2004).
|
Materials and Methods |
|---|
|
TopNotesAbstractMaterials and MethodsResultsDiscussionConclusionsReferences |
|---|
Three online appendices contain all the data relevant to this study and are available to the reader at http://www.systematicbiology.org. Appendix 1 contains a list of all specimens and references examined for character state assessments. Appendix 2 contains the list of 368 discrete morphological characters and descriptions of their primitive and derived states. Appendix 3 contains the actual data matrix. References hereafter to these appendices refer solely to the supplementary online information, as these items do not appear in print in this contribution.
Taxonomic Sampling
Classic phylogenetic analyses of Amniota have identified and emphasized certain large suprageneric ingroup taxa that are assumed, a priori, to be monophyletic. These taxa, which include Synapsida, Reptilia, Parareptilia, Eureptilia, Diapsida, and Testudines, purportedly represent large-scale divergences within Amniota. Increased sampling within each of these clades has been employed as a method of more thoroughly documenting character state changes in stem lineages of major amniote taxa, but has typically been limited to a single amniote clade of particular interest (e.g., Gauthier et al., 1988a; Lee, 1997a; deBraga and Rieppel, 1997).
Rosenberg and Kumar (2001) suggested that increased taxon sampling is of little importance to phylogenetic inference, and that higher-level evolutionary histories can be reconstructed with confidence by sampling just a few taxa. Zwickl and Hillis (2002) and Pollock et al. (2002), however, showed that adding taxa greatly reduces phylogenetic error in computer simulations, resulting in greater phylogenetic accuracy when compared to a known, artificial phylogeny. Furthermore, increased taxonomic sampling has the potential to break down large, aggregate taxa into several representative (exemplar) genera or species, facilitating the testing of hypotheses of internal relationships within suprageneric taxa (Prendini, 2001). Whenever possible in this study, a single species or monotypic genus was used as a terminal taxon in order to avoid a priori assumptions of monophyly within large, suprageneric clades.
Taxa examined here included 78 amniote taxa and two anamniote outgroups (Table 1). For selection of ingroup taxa, four recent morphological analyses of amniote phylogeny were integrated and their taxonomic samples compared. These included three analyses of amniote phylogeny by Gauthier et al. (1988a, 1988b) and Laurin and Reisz (1995), and one analysis of reptile phylogeny by deBraga and Rieppel (1997; with additions and corrections noted by Rieppel and Reisz, 1999). All amniote taxa represented in these analyses were included in the current study. In addition, several taxa new to this study were selected in order to increase taxon sampling within well-corroborated amniote clades. New taxa were selected from numerous published analyses including those for aetosaurs (Heckert and Lucas, 1999), ankylosaurs (Hill et al., 2003), crocodylians (Benton and Clark, 1988; Brochu, 1997), squamates (Estes et al., 1988), and pareiasaurs (Lee, 1997a). In order to reflect the maximal diversity (and therefore provide the strongest test of monophyly; e.g., Wheeler et al., 1993; Prendini, 2001) of these clades, at least two exemplars were chosen from each of these higher-level taxa.
|
Morphological Character Data
The morphological characters used in this study originated from several sources. First, a character list was generated by combining the osteological, physiological, and soft tissue anatomical characters used in four previous analyses of amniote and reptile phylogeny: (1) Gauthier et al. (1988a), (2) Gauthier et al. (1988b), (3) Laurin and Reisz (1995), and (4) deBraga and Rieppel (1997; with additions and corrections noted by Rieppel and Reisz, 1999). The compiled character set was examined carefully to identify and reduce redundant characters; i.e., those that were identical between two or more analyses. For example, Gauthier et al.'s (1988a) character 111, "caniniform maxillary tooth present (0) or absent (1)" is identical (although coded with varying polarity) to character 25 of Laurin and Reisz (1995) and to character 96 of deBraga and Rieppel (1997). The latter authors cite Laurin and Reisz (1995) as the source of the character, although no mention is made in either analysis of Gauthier et al.'s (1988a) use of the character much earlier. These three characters were subsumed into a single character (197 in this analysis, "caniniform maxillary tooth absent [0] or present [1]") because they describe exactly the same morphology. Single characters that represent a consolidation of identical characters from two or more published matrices are identified as such in the character list (Appendix 2) and each prior analysis in which the character was used is cited accordingly.
Elimination of redundant characters ensured that all characters in the analysis were logically independent, fulfilling one of the principal requirements of parsimony analysis (Kluge and Wolf, 1993). Compiling character lists from published analyses and eliminating redundant characters resulted in the identification of 297 discrete morphological characters. This character set was then scored for any taxa that did not appear in all (or any) previous analyses. Fewer than half of the taxa were common to all analyses (Table 1), and for this reason it was necessary to score most characters either from first-hand observations of specimens in museum collections, or from original published descriptions and photographs (Appendix 1). In addition, when two analyses disagreed on the coding of a character, an assessment was made based on first-hand observations as to which character state was exhibited. Thus, no one analysis was favored over another as having more reliable character coding.
In addition to morphological characters gleaned from published analyses, I introduced four new cranial characters (characters 82, 89, 175, and 177; Appendix 2), one physiological and two behavioral characters (characters 366 to 368; Appendix 2), and 64 characters describing the morphology and histology of the integument and osteoderms (characters 302 to 365; Appendix 2). Some of these characters have been modified from other published sources, whereas others are original to this study (see Appendix 2 for a complete reference to the source of each character). The integumentary and osteoderm characters were then scored for the entire set of taxa. These characters were based on direct observations of Recent and fossil osteoderms and histological sections that were viewed and photographed under ordinary and polarized light. They are listed in Appendix 2, but are also described in more detail elsewhere (e.g., Hill, 2004). In addition, digital images representing character states are available at http://www.morphobank.org.
The total combined data matrix (Appendix 3) consisted of 297 characters taken from previous published analyses, 71 new characters, and a total of 80 taxa, including 78 ingroup taxa and two outgroups. A diagrammatic representation of the data matrix is shown in Figure 2. The total amount of missing data in the matrix was 39.5%, owing largely to nonpreservation of many soft tissue structures in fossil taxa.
|
Although molecular sequence data are becoming increasingly abundant and accessible, they were not considered here because the task of compiling morphological data from published analyses and supplementing them with new taxa and characters warrants a separate report. The focus of the current study is to enhance morphological character sampling for amniote phylogenetics, providing a comprehensive morphological database to which new molecular data will ultimately be added.
Inapplicable Characters
The taxonomic diversity of Amniota is paralleled by its morphological diversity, and indeed there is no "generalized" amniote. For this reason, studies that sample broadly across Amniota, as done here, will inevitably encounter obstacles related to the inapplicability of certain characters. For example, the 15 characters describing dentition in this analysis cannot be scored for turtles, which lack marginal teeth entirely. Similarly, characters of osteoderm morphology original to this study are inapplicable to those taxa that lack osteoderms entirely.
Several methods have been proposed for dealing with inapplicable characters. Composite coding, advocated by Maddison (1993), effectively eliminates inapplicable cells within a data matrix by incorporating inapplicable states into a more inclusive multistate character. This practice can introduce spurious taxonomic groupings, however, when the search algorithm interprets shared inapplicability of a character as a synapomorphy (Strong and Lipscomb, 1999). An alternative is non-additive binary coding (Pleijel, 1995), which breaks down complex and potentially inapplicable characters into several discrete characters that exist in two states only. This method may also group taxa based on nonhomologous absence of evidence, rather than on shared derived character states (Strong and Lipscomb, 1999). A third possibility is reductive coding (Wilkinson, 1995), which codes inapplicable characters as question marks or dashes (missing data) instead of numerical character states. Strong and Lipscomb (1999) and Lee and Bryant (1999) noted that each method of coding inapplicable characters had serious disadvantages, and called for new computer algorithms to help distinguish between inapplicable characters and missing data. Strong and Lipscomb (1999: 370) favored reductive coding overall, however, because it resulted in trees that "best reflect the information content" of their observations.
Based on the foregoing observations, reductive coding (Wilkinson, 1995; Strong and Lipscomb, 1999) was employed in the current analysis. When reevaluating published characters used in this study, certain characters were re-coded reductively, and all instances of a third character state "inapplicable" were removed. The new characters of integumentary and osteoderm morphology were coded reductively from the outset, and hence, all characters of osteoderm morphology were scored as question marks for those taxa not possessing osteoderms.
Phylogenetic Analysis
The matrix was compiled using Nexus Data Editor 0.5.0 (Page, 2001). Because of the large number of taxa used in this analysis, exact-solution (i.e., exhaustive or branch-and-bound) searches were impossible (Schuh, 2000). Parsimony analyses were therefore conducted using the heuristic search algorithm in NONA 2.0 (Goloboff, 1993; hereafter NONA) with the following parameters: maximum trees kept (hold) = 100,000; number of replications (mult*N) = 1000; starting trees per rep (hold/) = 100; random seed = time; search strategy = multiple TBR + TBR (mult*max); unconstrained search. Results obtained from NONA were checked using PAUP* 4.0b10 (Swofford, 2001; hereafter PAUP) with the following settings in effect: uninformative characters excluded; addition sequence random; tree bisection-reconnection (TBR) branch swapping; 1000 replications; "multrees" option in effect; steepest descent option not in effect; zero-length branches collapsed to yield polytomies; trees unrooted; and multistate taxa interpreted as polymorphisms. Seymouriidae and Diadectomorpha were regarded as outgroups. All characters were weighted equally, and reversals and convergences were considered equally probable transformations. Character coding was unordered (nonadditive; Fitch, 1971) to give equal cost to any possible transformation between two states. Character state optimization was carried out using both accelerated (ACCTRAN) and delayed (DELTRAN) transformation algorithms in PAUP (Swofford, 2001). Optimizations obtained from PAUP were checked against those calculated by the comparable "fast optimization" and "slow optimization" functions in WINCLADA (Nixon, 2000).
The data were evaluated in three separate analyses. First, the data compiled from previously published studies were analyzed. This analysis incorporated 37 taxa and 297 characters (241 of which were parsimony-informative) taken from the analyses of Gauthier et al. (1988b), Laurin and Reisz (1995), and deBraga and Rieppel (1997), with redundant (i.e., logically nonindependent) characters merged. The second analysis also included these 37 taxa and 297 characters, with the addition of 43 taxa original to this study. These taxa were added to achieve genus-level (and in many cases species-level; see Table 1) sampling and provide tests of phylogenetic hypotheses at low taxonomic levels. The additional 43 taxa were scored for all 297 characters assembled from the literature. The third and final analysis incorporated all 80 taxa, the 297 characters from the literature and an additional 71 characters original to this study. These characters describe the morphology and histological structure of the integument and osteoderms (64 characters), cranial morphology (4 characters), and physiology or behavior (3 characters). All characters are described in detail in Appendix 2.
Measures of support included the bootstrap (Felsenstein, 1985) and branch support (Bremer, 1994) measures. Bootstrap values were calculated using NONA, with 100 iterations of 1000 search replicates each. Support for each branch in the phylogeny was calculated by creating constraint trees containing only the branch of interest, and then using heuristic searches of 10 replicates each in PAUP to find the length of the shortest tree inconsistent with the constraint. The difference between this tree length and that of the most parsimonious unconstrained tree describes the branch support (Bremer, 1994). The relative robustness of alternative phylogenetic hypotheses was evaluated using a similar method. Reverse constraint trees (Bremer, 1994) were constructed to represent alternative hypotheses of turtle relationships, and heuristic searches of 10 replicates each were used to find the shortest tree consistent with the constraint. To further test alternative phylogenetic hypotheses, the treatment of certain controversial characters was changed a posteriori, and the data were reanalyzed. Characters identified by Lee (2001) as being incorrectly scored by Rieppel and Reisz (1999) were re-coded, and characters considered to represent morphoclines (Laurin and Reisz, 1995; Lee, 1997a) were ordered.
|
Results |
|---|
|
TopNotesAbstractMaterials and MethodsResultsDiscussionConclusionsReferences |
|---|
Combined Published Data Sets
Combining the published data sets of Gauthier et al. (1988a, 1988b), Laurin and Reisz (1995) and deBraga and Rieppel (1997) and removing redundant characters resulted in a data matrix consisting of 37 taxa and 297 characters. When analyzed using NONA, 30 most parsimonious trees were found (TL = 969; CI = 34; RI = 63). The strict consensus of these trees (Fig. 3) supports the monophyly of Amniota and several highly nested amniote clades, but lacks resolution at intermediate nodes, obscuring relationships among major groups. Synapsida, Reptilia, Eosuchia (diapsids exclusive of araeoscelidians), and Parareptilia are all supported, but relationships within and among these taxa remain unresolved. Mammals are nested within Synapsida. The content of Parareptilia is that of deBraga and Rieppel (1997), albeit more poorly resolved. Furthermore, Captorhinidae, Paleothyris, and Araeoscelidia, taxa traditionally considered basal eureptiles, form a polytomy at the base of the reptile radiation, precluding the corroboration of both eureptilian monophyly and parareptilian relationships. Mesosaurs are here resolved as part of the ingroup Amniota, but their relationships with synapsids and reptiles are ambiguous.
Within Eosuchia, the monophyly of Sauria is supported, as is that of Lepidosauria (squamates + Sphenodon). Lepidosauria here represents the sister taxon to an unresolved clade containing turtles and sauropterygians. Archosauriforms constitute the sister taxon to the (Lepidosauria (Testudines + Sauropterygia)) clade. Several taxa traditionally considered to be stem lineages of either Archosauromorpha or Lepidosauria (e.g., Kuehneosauridae, Choristodera, and Prolacertiformes) are resolved as saurians, but their relationships to more highly nested taxa remain equivocal.
|
(1); 31, (1) Addition of Taxa
The addition of 43 ingroup taxa had a dramatic effect on the resulting tree topology. The preset maximum of 100,000 equally parsimonious solutions was retrieved (TL = 1516; CI = 23; RI = 73), the strict consensus of which (Fig. 4) showed very little resolution. A monophyletic Ankylosauria was recovered, including a traditional sister-taxon relationship between two ankylosaurids, Euoplocephalusand Pinacosaurus, to the exclusion of one nodosaurid, Sauropelta (Coombs, 1978; Coombs and Marya
ska, 1990). A polychotomous assemblage just outside Ankylosauria included the remaining thyreophorans (armored ornithischian dinosaurs) in the analysis: Scelidosaurus, Scutellosaurus, and Stegosaurus. The remainder of the ingroup was completely unresolved in the strict consensus.
|
Addition of Integumentary Data
The final analysis incorporated all taxa and characters collected from the literature, with the addition of 43 ingroup taxa and 71 new characters mainly describing the osteology and soft tissue anatomy of the integument. Addition of this new character data had a remarkable effect on tree topology. Parsimony analysis of the entire data set (published data + new integumentary data) recovered four most parsimonious trees, each requiring 1738 evolutionary steps (CI = 25; RI = 73). The strict consensus of these trees (Fig. 5) supports the monophyly of the following major clades: Amniota, Synapsida, Reptilia, Parareptilia, Eureptilia, Romeriida, Diapsida, Eosuchia, Neodiapsida, Sauria, Lepidosauria, Testudines, and Archosauria. In addition, several well-recognized, highly nested clades are supported within each of these major taxa. The historically problematic mesosaurs are here found to be the sister taxon to all other amniotes, suggesting a polyphyletic Sauropsida sensu Laurin and Reisz (1995). Also supported by this analysis are a sister-taxon relationship between Lepidosauria (squamates and Sphenodon) and Testudines. The results of the final analysis represent the hypothesis of amniote relationships based on the greatest amount of available data examined for this study.
|
Character-State Optimization
Character-state optimization was used to identify synapomorphies and determine the ancestral condition for particular clades. The strongest support for nodes within a cladogram comes from unambiguous synapomorphies; i.e., those that support a certain node regardless of whether homoplasy is attributed to parallelism or reversals. Each of the major nodes in the final analysis is supported by at least four unambiguous synapomorphies (Table 2) and numerous synapomorphies favored by either fast or slow optimization alone.
|
The monophyly of the ingroup, Amniota, is supported by 10 unambiguous synapomorphies. Six of these are characters of the skull and jaws, and four are postcranial characters. Primitively, amniote skulls possess a frontal bone that contributes to the orbital margin, absence of dermal sculpturing on the cranial bones, an occipital flange of the squamosal, a rounded occipital condyle, a distinct retroarticular process, and absence of labyrinthodont infolding of tooth dentine. The ancestral amniote postcranium possessed an anterodorsally sloping axis, two coracoid ossification centers, a cleithrum that does not cap the scapula anterodorsally, and a discrete astragalus. These characters have been traditionally recognized as amniote synapomorphies (e.g., Gauthier et al., 1988a, 1988b), a hypothesis corroborated by the present study. None of the new characters of integumentary and osteoderm morphology unambiguously supports Amniota.
Synapsids are recovered here as a monophyletic taxon including mammals, supported by five unambiguous cranial and postcranial synapomorphies. The monophyly of Reptilia is supported by seven characters of the skull, and, in the postcranial skeleton, the presence of only a single pedal centrale. A monophyletic Sauropsida (sensu Laurin and Reisz, 1995) is not recovered, as Mesosauridae here constitutes the sister taxon to all other amniotes, and not merely to Reptilia.
The final analysis supports monophyly of Parareptilia sensu deBraga and Rieppel (1997), including Millerettidae as the sister taxon to a monophyletic Ankyramorpha (deBraga and Reisz, 1996). Ankyramorpha in turn consists of two clades, Lanthanosuchoidea (including Acleistorhinus, Lanthanosuchidae, and Macroleter) and Procolophonia (including procolophonids and pareiasaurs). Monophyly of Parareptilia is supported by seven unambiguous synapomorphies of the skull and postcranial skeleton.
The analysis also supports eureptilian monophyly on the basis of five cranial synapomorphies, diapsid monophyly on the basis of six characters of the skull and postcrania, and saurian monophyly on the basis of ten cranial and postcranial characters. Placement of Testudines as the sister taxon to a monophyletic Lepidosauria is justified on the basis of 16 unambiguous synapomorphies of the skull, postcranial skeleton, osteoderms, soft tissue anatomy, physiology, and behavior. It is noteworthy that the position of turtles has the highest number of unambiguous synapomorphies of all the nodes mentioned, and that support for this position comes from several anatomically distinct regions, including the integument.
Bootstrap analysis (Felsenstein, 1985) of the data set indicated a wide range of support values, some groups being robustly supported; others only weakly so (Fig. 5). The monophyly of the ingroup Amniota was supported in 82% of bootstrap replicates. Clades supported by 100% of bootstrap replicates included armadillos (Dasypodidae) and their extinct relatives the glyptodonts (Glyptodontidae), Cordyliformes (represented here by the lizard genera Cordylus and Zonosaurus), Velosauria (advanced pareiasaurs), Parasuchia (phytosaurs), Placodontia, and Testudines. The most robustly supported node consistent with a diapsid origin of turtles is the node supporting a (Kuehneosauridae (Sauropterygia (Archosauria (Lepidosauria + Testudines)))) clade (node k; Fig. 5), which was supported in 58% of bootstrap replicates. Clades that exhibited high bootstrap values typically also had strong branch supports.
Reverse constraint trees were used to examine the additional number of evolutionary steps required to move turtles to selected positions inconsistent with the most parsimonious cladogram. These constraints provided a test of the alternative phylogenetic positions that have been suggested for turtles (Fig. 1). Constraining the analysis so that turtles are the sister taxon of sauropterygians (deBraga and Rieppel, 1997; Rieppel and Reisz, 1999) incurs four additional steps. Allying turtles with archosauromorphs (e.g., Merck, 1997; Zardoya and Meyer, 1998, 2001; Kumazawa and Nishida, 1999) incurs six additional steps. Pruning turtles from the cladogram and grafting them into Parareptilia requires numerous additional steps: 13 are required to ally turtles with procolophonids (Laurin and Reisz, 1995); 21 are incurred if turtles are placed as the sister taxon to a monophyletic Pareiasauria (Lee, 1993), and 26 are incurred if turtles are allied with Anthodon within Pareiasauria (Lee, 1997a). If turtles are resolved as the sister taxon to captorhinids (Gauthier et al., 1988a, 1988b), 54 additional steps are required, and if they are placed as the sister taxon to all other amniotes (Gaffney, 1980), 57 additional steps are required.
|
Discussion |
|---|
|
TopNotesAbstractMaterials and MethodsResultsDiscussionConclusionsReferences |
|---|
The foregoing investigation supports several traditional clades within Amniota, but also introduces certain novel phylogenetic groupings. Examining only the character data amassed from the published analyses of Gauthier et al. (Gauthier et al., 1988a, 1988b), Laurin and Reisz (1995), and deBraga and Rieppel (1997) reveals a consensus on the monophyly of Amniota, Synapsida, Reptilia, and Parareptilia (Fig. 3). This result is not surprising, as these are some of the clades upon which these four analyses explicitly agree. More remarkable is the phylogenetic placement of turtles based upon characters compiled solely from published sources: nested among sauropterygians, to the exclusion of lepidosaurs. This result is consistent with the analyses of deBraga and Rieppel (1997) and Rieppel and Reisz (1999), who recovered a sister-taxon relationship between turtles and sauropterygians, indicating that turtles, despite the anapsid condition of their skulls, are members of Diapsida.
The previous analyses that form the foundation of the current study have differed to no small degree on the point of turtle relationships. Gauthier et al. (Gauthier et al., 1988a, 1988b) suggested that turtles were the sister taxon to captorhinids, and that together these two taxa were the sister taxon to Romeriida, which included extant diapsids. Reisz and Laurin (1991) and Laurin and Reisz (1995) suggested instead that turtles were closely related to procolophonids and, as such, were parareptiles. Lee (1993, 1997a) also suggested that turtles were parareptiles, closely related to pareiasaurs. The cladistic analysis of deBraga and Rieppel (1997) was the first of its kind to suggest turtles to be closely allied with sauropterygians and, as such, to be lepidosauromorphan diapsids. Combining data from several published sources supports the latter hypothesis of a testudine origin from within Diapsida.
Scoring an additional 43 taxa for the characters derived from the literature introduced a tremendous amount of homoplasy, obscuring any phylogenetic signal almost entirely. The outcome of roughly doubling the number of taxa for an existing set of characters was a dramatic decrease in the phylogenetic resolution of the resulting trees. Characters traditionally used for amniote-level phylogenetics are therefore clearly insufficient to resolve relationships among more highly nested amniote taxa.
The final analysis included the same taxonomic and character sample from the second analysis, with the important distinction that 71 new characters were added, primarily describing the morphology of osteoderms and integumentary histology. The main result of this analysis was an increase in overall phylogenetic resolution at all hierarchical levels throughout the tree. New data from the integument revealed relationships between terminal taxa that had been obscured in the first two analyses, recovering monophyly of numerous well-recognized diapsid taxa, such as Anguidae, Scincidae, Crocodylomorpha, Ornithodira, Ornithischia, Ankylosauria, and Thyreophora. This higher resolution provides an opportunity to test hypotheses of relationships within these highly nested amniote taxa, each of which is extremely derived relative to anamniote and early amniote morphology.
Addition of integumentary data helps resolve relationships not only among the newly added taxa, but at more fundamental nodes in the phylogeny, as well. The addition of new data can be interpreted as having a stabilizing effect, allowing major relationships among amniote groups to be revealed despite a high degree of homoplasy. This effect has led to an overall increase in information content relative to analyses based solely on data available in the literature. For example, eureptilian monophyly and the position of mesosaurs, which were ambiguous based on the data set compiled from the literature, are, with the addition of integumentary characters, well-supported by numerous unambiguous synapomorphies (Table 2).
The final analysis supports phylogenetic placement of turtles within Diapsida, a hypothesis that has only recently been advanced in the context of modern cladistic analyses (e.g., deBraga and Rieppel, 1997; Merck, 1997; Hedges and Poling, 1999). Measures of support in this intermediate part of the tree are weak, however; diapsid monophyly is supported by fewer than 50% of bootstrap replicates, and only a single additional step is required to collapse the node supporting Diapsida. Inclusion of turtles within Diapsida is supported by 58% of bootstrap replicates and a branch support of one step. In contrast, testudine monophyly is highly corroborated, being supported by 100% of bootstrap replicates, and a branch support of five steps.
The results presented here differ fundamentally from the traditionally held view that turtles arose from one of the groups of Permian or Triassic anapsids: captorhinids (Gauthier et al., 1988b; Gaffney, 1990), procolophonids (Reisz and Laurin, 1991; Laurin and Reisz, 1995), or pareiasaurs (Lee, 1993, 1997a). As the idea of turtles as derived diapsids gains momentum (e.g., Rieppel and deBraga, 1996; Rieppel, 2000), new debates have arisen among advocates of this hypothesis as to where exactly among diapsids turtles belong (e.g., deBraga and Rieppel, 1997; Merck, 1997; Zardoya and Meyer, 1998, 2001; Hedges and Poling, 1999; Mannen and Li, 1999; Kumazawa and Nishida, 1999; Rieppel and Reisz, 1999; Cao et al., 2000). To date, no molecular or morphological studies have supported a sister-taxon relationship between turtles and lepidosaurs, as shown here. As deBraga and Rieppel (1997) noted, their analysis found that the closest living relatives of turtles were lepidosaurs, but extensive fossil stem lineages separated each of these crown groups. The analysis presented here therefore represents a new interpretation of amniote phylogeny, i.e., a direct sister-taxon relationship between lepidosaurs and turtles.
One potential explanation for this hitherto unexplored hypothesis is the inclusion of soft tissue, physiological, histological, and behavioral characters along with traditional osteological morphology. The study of Gauthier et al. (1988b) was the last major investigation of amniote phylogeny that incorporated new soft tissue characters, most of these culled from the previous studies of Gardiner (1982) and Løvtrup (1985). Subsequent analyses of amniote phylogeny have relied entirely on either osteology of the skull and skeleton (e.g., Laurin and Reisz, 1995; deBraga and Rieppel, 1997) or molecular sequence data (e.g., Zardoya and Meyer, 2001, and references therein). The current analysis therefore represents a step toward a more integrated approach to amniote phylogenetics, incorporating data from bony morphology and other sources. This analysis provides the long overdue foundation of morphological character data for use in future total evidence analyses.
The idea of turtles as diapsids that have secondarily lost their temporal fenestrae is not implausible in the overall context of reptilian evolution (Zardoya and Meyer, 2001). There exist numerous instances of secondary closure of temporal fenestrae within other reptilian clades, such as Araeoscelidia, Acleistorhinidae, Crocodylomorpha, and Dinosauria (e.g., Coombs, 1978; Reisz et al., 1984; deBraga and Reisz, 1996; deBraga and Rieppel, 1997; Brochu, 1999). Furthermore, as Carroll (1988) suggested, the highly derived turtle postcranium may reflect a predisposition toward extreme modifications of the skull, as well. Independent molecular studies have also helped substantiate the claim that turtles are highly derived members of crown-group Diapsida (e.g., Zardoya and Meyer, 1998; Kumazawa and Nishida, 1999; Hedges and Poling, 1999; Mannen and Li, 1999; Cao et al., 2000). Nevertheless, placement of turtles among diapsids has met with considerable criticism (e.g., Lee, 1997b, 2001).
If turtles are accepted as derived members of Diapsida, certain characters previously used to ally them with extinct anapsid taxa must be interpreted as instances of parallelism in clades that are only distantly related. For example, a slender, imperforate stapes is found in turtles and procolophonids, and has been resolved as a synapomorphy uniting these two clades (Testudinomorpha; Laurin and Reisz, 1995). Here, increased sampling within Eureptilia, which was restricted in the study of Laurin and Reisz (1995), reveals this character to be a synapomorphy of Neodiapsida, as well. The presence of a slender stapes is thought to be related to the elaboration of a tympanic middle ear, an adaptation for hearing in air rather than in water, and is thought to have arisen several times within tetrapods (Laurin and Anderson, 2004). The current phylogeny suggests that the slender stapes and associated tympanic middle ear arose independently in turtles and procolophonids. Similarly, characters of the postcranial anatomy used to link turtles with procolophonids (Laurin and Reisz, 1995), such as overlapping metapodials and the absence of a cleithrum, can also be found within highly nested diapsid clades, indicating that procolophonids may have independently acquired these modifications. Increased taxonomic sampling, then, allows the emergence of phylogenetic signals that otherwise remain obscured when considering Eureptilia as a single terminal taxon.
A novel outcome of this analysis is the phylogenetic placement of mesosaurs at the base of the amniote radiation. Gauthier et al. (1988b) tentatively included mesosaurs with parareptiles, a hypothesis later supported by Modesto (1999, 2004). Laurin and Reisz (1995) suggested instead that Mesosauridae is the sister taxon of Reptilia, i.e., all nonsynapsid amniotes. Unfortunately, the otherwise comprehensive studies of reptile phylogeny performed by deBraga and Rieppel (1997) and Rieppel and Reisz (1999) did not include mesosaurs in the taxonomic sample, and so could not test their phylogenetic relationships. The current analysis suggests a phylogenetic position for mesosaurs that has not been previously hypothesized. Mesosauridae is here resolved as the sister taxon to a clade containing Synapsida and Reptilia, i.e., all other amniotes. This grouping is supported by eight unambiguous synapomorphies of the skull, vertebral column, and shoulder girdle. Mesosaurs are particularly enigmatic because their highly specialized adaptations for a fully marine lifestyle seem to have appeared very early in amniote evolution, by the Early Permian (Reisz, 1997; Modesto, 1999).
Another result unique to this study is the recovery of a clade of four problematic taxa usually considered basal archosauromorphs (e.g., Merck, 1997; Brochu, 2001): Rhynchosauria, Trilophosaurus, Prolacertiformes, and Choristodera. This new clade is here resolved as the sister taxon to all other saurians, and its monophyly is fairly robustly supported by 71% of bootstrap replicates and a branch support of three steps.
Alternative Hypotheses of Turtle Relationships
Imposing reverse constraints on the position of turtles reveals that the hypothesis presented here is considerably more parsimonious than certain familiar alternatives (Fig. 1). Allying turtles with diapsid groups other than Lepidosauria incurs four to six additional steps. Removing turtles from Diapsida and allying them with fossil anapsid groups incurs at least 13 additional steps, or as many as 57. These results suggest that, although a sister taxon relationship between turtles and lepidosaurs is not extremely well supported, a position of turtles among diapsids is considerably more parsimonious than one among parareptiles or captorhinids.
Lee (2001) identified eight characters in the analysis of Rieppel and Reisz (1999), which, when re-coded, yielded a dramatically different tree, with turtles as the sister taxon to pareiasaurs. These characters (characters 65, 82, 103, 120, 121, 127, 140, and 152 of Rieppel and Reisz, 1999, and Lee, 2001) are all represented in this study (Appendix 2), but for clarity I will refer to them in the following section by their original numbers. Because of the controversy surrounding these particular characters, I examined the effects of using alternative codings.
When all eight characters are coded according to Rieppel and Reisz (1999), a diapsid position of turtles is unsurprisingly supported. Here, however, I employ several of the changes in character coding recommended by Lee (2001). I agree with Lee (2001) that the ventral braincase tubera of turtles and pareiasaurs should be considered homologous (character 65 of Lee, 2001), that pareiasaurs possess keeled cervical vertebrae, two coracoid ossifications and a coracoid foramen completely enclosed by the coracoid (characters 103, 120 and 121 of Lee, 2001, respectively), and that pareiasaurs lack a femoral fourth trochanter (character 140 of Lee, 2001). Scoring these five characters according to Lee (2001) recovers the same tree topology shown in Figure 5, and turtles are therefore recovered as diapsids, and not as parareptiles.
Recoding the remaining three characters is a separate issue that relates to polymorphism in higher taxa. Rieppel and Reisz (1999) coded the position of the craniomandibular joint, presence of an ectepicondylar foramen, and presence of the first distal tarsal as polymorphic in turtles (characters 82, 127, and 152 in Lee [2001], respectively). Lee (2001) ignored the polymorphism in turtles, instead scoring what he regarded as primitive character states, based on observations of the primitive turtles Proganochelys, Kayentachelys, and Australochelys. The current study addresses this issue differently: instead of scoring Testudines as a single terminal taxon, I incorporated six turtle species as exemplars to reflect morphological diversity within this large clade. Scoring at the lowest possible taxonomic level eliminates the need to code polymorphisms (e.g., Rieppel and Reisz, 1999) or to hypothesize the primitive condition for aggregate ingroup taxa (e.g., Lee, 2001), and is another advantage of increased taxonomic sampling. In the large data set analyzed here, coding all eight controversial characters according to Lee's (2001) recommendations still recovers a diapsid position for turtles.
The Importance of Integumentary Characters for Amniote Phylogeny Reconstruction
Laurin and Reisz (1995) and Lee (1997a) interpreted the presence of dorsal dermal ossifications of any kind (trunk osteoderms or carapacial elements) as a potential synapomorphy linking turtles with parareptiles. These studies, however, omitted other major groups of amniotes that possess osteoderms, and in many cases erected a unique character state for the extremely derived condition of turtles. Lee (1997a) also ordered many osteoderm characters, making transformations to more heavily armored states less costly in his analysis. In contrast, Rieppel and Reisz (1999) suggested that the carapace and plastron of turtles are developmentally unique (e.g., Starck, 1979; Burke, 1989), and could not be considered homologous with the osteoderms of any amniote group.
The differences in primary homology statements between the analyses of Lee (1997a) and Rieppel and Reisz (1999) represent distinct philosophies regarding how different types of data should be handled in phylogeny reconstruction. A thorough discussion of these distinctions is beyond the scope of this paper. In order to minimize assumptions about the evolutionary history of the turtle carapace, I have taken a relatively conservative approach, leaving all characters unordered and scoring many characters of osteoderms as inapplicable if they could not be unequivocally assessed in the intricate carapace and plastron of turtles. Nevertheless, ordering certain "morphoclines" identified by Laurin and Reisz (1995) and Lee (1997a) had no effect on the resulting tree topology, and ordering all 368 characters resulted only in slight rearrangements within Diapsida, which still included a close alliance of turtles and lepidosauromorphs.
Although addition of multiple characters of integumentary morphology and histology aids in revealing patterns of relationships among major groups of amniotes, none of the new characters is explicitly resolved as a synapomorphy of these deep nodes in the phylogeny (Table 2). Instead, the new integumentary characters emerge as synapomorphies of more highly nested taxa, illuminating patterns of evolution within less inclusive clades of amniotes.
The foregoing analysis reveals that dorsal trunk osteoderms may have arisen independently at least five times over the course of amniote evolution—within Synapsida in cingulates, within Parareptilia in pareiasaurs, within Lepidosauria in squamates, within Archosauria, and within Sauropterygia in placodonts. This means that although they are morphologically similar, the dorsal osteoderms of such diverse taxa as armadillos, lizards, and dinosaurs are not homologous with one another in a phylogenetic sense.
Optimization of integumentary character states further facilitates reconstruction of soft tissue and other unpreserved conditions in fossil taxa. For example, extant lepidosaurs (squamates and Sphenodon) possess two epidermal generations, the outer of which is shed (sloughed) as a whole (e.g., Alibardi and Maderson, 2003). In contrast, extant archosaurs continuously lose small fragments of keratinized epidermis. Character state optimization reveals that the ancestral condition for lepidosaurs is the presence of epidermal generations, providing support for the hypothesis that extinct lizards such as the Glyptosaurinae shed their skins as modern lizards do. Optimization at more basal nodes indicates that the ancestral diapsid unambiguously lacked epidermal generations. This absence persisted in the lineages leading to modern turtles and archosaurs, and shedding of skin as a whole is therefore most parsimoniously interpreted as a lepidosaurian synapomorphy.
Another instance in which the new phylogeny becomes important in interpretation of fossils involves the historically confounding arrangement of bony dorsal plates in Stegosaurus. Three major hypotheses have been advanced to explain the arrangement of these structures: (1) they comprised a single, median row (e.g., Marsh, 1891); (2) they comprised a staggered, asymmetrical row (e.g., Gilmore, 1914); or (3) they comprised paired, paramedian rows (e.g., Lull, 1910) of bony plates. Because no two osteoderms of Stegosaurus are the exact same size and shape (Galton, 1990), the former two hypotheses are preferred in life restorations of Stegosaurus(e.g., Czerkas, 1987). No archosaur sampled in this study possessed asymmetrical or median rows of plates, and other stegosaurs such as Huayangosaurus, Tuojiangosaurus, and Kentrosaurus are known to possess paired plates (Galton, 1990). Stegosaurus is nested well within Archosauria, the ancestor of which possessed paired, paramedian rows of plates that persisted in the crurotarsan and ornithodiran lineages. The most parsimonious reconstruction of Stegosaurus plates is, therefore, a system of paired, paramedian rows, and a staggered arrangement unique among archosaurs should not be inferred on the basis of negative evidence.
This study shows that addition of data beyond the point at which a phylogenetic signal is seemingly lost can recover new phylogenetic information and resolve polytomies that persisted because of restrictive taxon and character sampling. Future studies of the phylogenetic relationships of amniotes should continue to increase sampling in three important ways. First, taxonomic sampling should be expanded rootward to encompass more extant and fossil tetrapods. For example, certain frogs, as well as dissorophid amphibians, which possessed bony armor plates (e.g., deMar, 1966; Rubial and Shoemaker, 1984) represent anamniote taxa that can be scored for both skeletal and integumentary characters. Inclusion of these taxa in future analyses may help elucidate patterns of evolution in the integument of tetrapods in general. Second, increased sampling at the genus level and below will continue to provide tests for hypotheses of relationships within increasingly specific clades. Instead of scoring a single terminal taxon Heloderma, as done here, future analyses may focus upon increasing the number of species and individuals sampled within a given genus. This will help to document the intrageneric variability present for a given taxon, by sampling, say, 10 specimens each of H. horridum and H. suspectum. Finally, addition of new character data from novel sources holds the potential to revise our understanding of amniote phylogeny. Other historically neglected anatomical systems may further test the hypotheses advanced here, especially if they can be observed in the fossil record as well as in extant organisms. For example, the ossified tendons of epaxial musculature found in certain dinosaurs (e.g., Weishampel and Horner, 1990; Molnar, 1996) represent a potential source of information about the muscular system in extinct reptiles. In addition to morphological data, molecular data both from published sources and contemporary sequencing efforts will complement future total evidence analyses.
|
Conclusions |
|---|
|
TopNotesAbstractMaterials and MethodsResultsDiscussionConclusionsReferences |
|---|
This study illustrates the important effects of increasing both taxonomic and character sampling and highlights the significance of the integument as a source of meaningful morphological character data. By incrementally adding data to existing morphological matrices, I show that traditional anatomical characters used for amniote systematics are wholly insufficient to resolve relationships among more highly nested taxa. Incorporating data from the integument and osteoderms, however, can help resolve these relationships, revealing phylogenetic signals that are otherwise obscured by incomplete taxonomic or character sampling. Moreover, addition of data beyond the point at which a phylogenetic signal (i.e., the position of turtles among diapsids) seems lost can recover that signal again, further underscoring its stability.
The novel hypothesis that turtles represent the sister taxon to Lepidosauria emerges from the foregoing phylogenetic analyses, but bootstrap and branch support measures indicate that it is relatively weakly supported. The more general hypothesis that turtles represent derived diapsids that have secondarily achieved the anapsid condition is far more robustly supported, as indicated by analyses using constraint trees. In addition, ordering certain characters (Laurin and Reisz, 1995; Lee, 1997a) or re-coding controversial characters (Lee, 2001) does not ally turtles with any group of fossil anapsids. The concept of turtles as diapsids, although heterodox, is becoming increasingly corroborated by both molecular and morphological studies. This indicates that many "primitive" or "parareptilian" characters of turtles may actually represent instances of evolutionary reversal or parallelism. Parallelism is particularly evident with regard to the acquisition of dorsal osteoderms, which have arisen independently in amniotes at least five times. The convergent evolution of osteoderms in taxa that are both phylogenetically and ecologically disparate underscores the versatility of the integument and suggests that the ossification of its connective tissues may serve a variety of functions.
|
|
Acknowledgements |
|---|
For access to specimens and collections, I thank E. Daeschler, R. Elsey, A. Herrel, P. Holroyd, S. G. Lucas, C. Mehling, M. A. Norell, O. Rieppel, C. Schaff, and N. Soule. For helpful discussions and comments on the manuscript, I thank M. T. Carrano, C. A. Forster, J. Gatesy, D. W. Krause, M. A. O'Leary, and O. Rieppel. I also thank M. S. Y. Lee and an anonymous reviewer for their constructive reviews. This research was supported by the Doris and Samuel P. Welles fund (UCMP), the Jurassic Foundation, Field Museum Visiting Scholar Grant (FMNH), Theodore Roosevelt Memorial Grant (AMNH), Gabor Inke Graduate Fellowship (Stony Brook University) and National Science Foundation grant DEB-0206533.
|
Notes |
|---|
|
TopNotesAbstractMaterials and MethodsResultsDiscussionConclusionsReferences |
|---|
Current Address: Department of Anatomy, New York College of Osteopathic Medicine, Northern Boulevard, Old Westbury, New York, 11568, USA; E-mail: rhill01{at}nyit.edu
|
References |
|---|
|
TopNotesAbstractMaterials and MethodsResultsDiscussionConclusionsReferences |
|---|
-
Alibardi L., Maderson P. F. A. Observations on the histochemistry and ultrastructure of the epidermis of the tuatara, Sphenodon punctatus (Sphenodontida, Lepidosauria, Reptilia): A contribution to an understanding of the lepidosaurian epidermal generation and the evolutionary origin of the squamate shedding complex. J. Morphol. (2003) 256:111–133.[CrossRef][Web of Science][Medline]
Barrett M., Donoghue M. J., Sober E. Against consensus. Sys. Biol. (1991) 40:486–493.
Benton M. J., Clark J. M. Archosaur phylogeny and the relationships of the Crocodylia. In: The phylogeny and classification of the tetrapods—Benton M. J., ed. (1988) Oxford, UK: Clarendon Press. Pages 295–338.
Bremer K. Branch support and tree stability. Cladistics (1994) 10:295–304.[CrossRef][Web of Science]
Brochu C. A. Morphology, fossils, divergence timing, and the phylogenetic relationships of Gavialis. Syst. Biol. (1997) 46:479–522.
Brochu C. A. Phylogenetics, taxonomy, and historical biogeography of Alligatoroidea. J. Vertebr. Paleontol. (1999) 19(2, suppl.):9–100.
Brochu C. A. Progress and future directions in archosaur phylogenetics. J. Paleontol. (2001) 75:1185–1201.
Burke A. C. Development of the turtle carapace: Implications for the evolution of a novel bauplan. J. Morphol. (1989) 199:363–378.[CrossRef][Web of Science]
Cao Y., Sorenson M. D., Kumazawa Y., Mindell D. P., Hasegawa M. Phylogenetic position of turtles among amniotes: Evidence from mitochondrial and nuclear genes. Gene (2000) 259:139–148.[CrossRef][Web of Science][Medline]
Carroll R. L. Vertebrate paleontology and evolution (1988) New York: W. H. Freeman and Company.
Coombs W. P. Jr. The families of the Ornithischian dinosaur order Ankylosauria. Palaeontology (1978) 21:143–170.
Coombs W. P. Jr., Maryaska T. Ankylosauria. In: The Dinosauria—Weishampel D. B., Dodson P., Osmolska H., eds. (1990) Berkeley, California: University of California Press. Pages 456–483.
Czerkas S. A. A reevaluation of the plate arrangement on Stegosaurus stenops. In: Dinosaurs past and present—Czerkas S. J., Olson E. C., eds. (1987) Volume II. California: Los Angeles County Natural History Museum. Pages 82–99.
deBraga M., Reisz R. R. The Early Permian Reptile Acleistorhinus pteroticus and its phylogenetic position. J. Vertebr. Paleontol. (1996) 16:384–395.
deBraga M., Rieppel O. Reptile phylogeny and the interrelationships of turtles. Zool. J. Linn. Soci. (1997) 120:281–354.[CrossRef]
deMar R. E. The phylogenetic and functional implications of the armor of the Dissorophidae. Fieldiana Geol. (1966) 16:55–88.
Eernisse D. J., Kluge A. G. Taxonomic congruence versus total evidence, and amniote phylogeny inferred from fossils, molecules, and morphology. Mol. Biol. Evolution (1993) 10:1170–1195.[Abstract]
Estes R., de Queiroz K., Gauthier J. Phylogenetic relationships within Squamata. In: Phylogenetic relationships of the lizard families—Estes R., Pregill G., eds. (1988) Stanford, California: Stanford University Press. Pages 119–281.
Felsenstein J. Confidence limits on phylogenies: An approach using the bootstrap. Evolution (1985) 39:783–791.[CrossRef][Web of Science]
Fitch W. M. Toward defining the course of evolution: Minimum change for a specified tree topology. Syst. Zool. (1971) 20:406–416.[Abstract]
Frey E. Das Tragsystem der Krokodile—eine biomechanische und phylogenetische Analyse. Stuttgarter Beitrage zur Naturkunde Ser. A (1988) 426:1–60.
Gaffney E. S. Phylogenetic relationships of the major groups of amniotes. In: The terrestrial environment and the origin of land vertebrates—Panchen A. L., ed. (1980) London: Academic Press. Pages 593–610.
Gaffney E. S. The comparative osteology of the Triassic turtle Proganochelys. Bull. Am. Mus. Nat. Hist. (1990) 194:1–263.
Galton P. M. Stegosauria. In: The Dinosauria—Weishampel D. B., Dodson P., Osmolska H., eds. (1990) Berkeley, California: University of California Press. Pages 435–455.
Gardiner B. G. Tetrapod classification. Zool. J. Linn. Soc. (1982) 74:207–232.[CrossRef]
Gauthier J., Kluge A. G., Rowe T. Amniote phylogeny and the importance of fossils. Cladistics (1988a) 4:105–209.[Web of Science]
Gauthier J. A., Kluge A. G., Rowe T. The early evolution of the Amniota. In: The phylogeny and classification of the tetrapods—Benton M. J., ed. (1988b) Oxford, UK: Clarendon Press. Pages 103–153.
Gilmore C. W. Osteology of the armored Dinosauria in the United States National Museum, with special reference to the genus Stegosaurus. Bull. U.S. Mus. Nat. Hist. (1914) 89:1–136.
Goloboff P. NONA (1993) Program available at www.zmuc.dk/public/phylogeny.
Gregory W. K. Pareiasaurs versus placodonts as near ancestors to the turtles. Bull. Am. Mus. Nat. Hist. (1946) 86:275–326.
Heckert A. B., Lucas S. G. A new aetosaur (Reptilia: Archosauria) from the Upper Triassic of Texas and the phylogeny of aetosaurs. J. Vertebr. Paleontol. (1999) 19:50–68.
Hedges S. B., Poling L. A molecular phylogeny of reptiles. Science (1999) 283:998–1001.
Hill R. V. Reconstructing amniote phylogeny using morphology of the skeleton, integument, and osteoderms. J. Vertebr. Paleontol. (2003) 23:61A.
Hill R. V. Morphological and histological diversity in the integument of cingulates (Mammalia: Xenarthra). J. Vertebr. Paleontol. (2004) 24:70A.
Hill R. V., Witmer L. M., Norell M. A. A new specimen of Pinacosaurus grangeri (Dinosauria: Ornithischia) from the Upper Cretaceous of Mongolia: Ontogeny and phylogeny of ankylosaurs. Am. Mus. Novitates (2003) 3395:1–29.[CrossRef]
Kluge A. G., Wolf A. J. Cladistics: What's in a word? Cladistics (1993) 9:183–199.[CrossRef][Web of Science]
Kumazawa Y., Nishida M. Complete mitochondrial DNA sequences of the green turtle and blue-tailed mole skink: Statistical evidence for archosaurian affinity of turtles. Mol. Biol. Evol. (1999) 16:784–792.[Abstract]
Laurin M., Anderson J. S. Meaning of the name Tetrapoda in the scientific literature: An exchange. Syst. Biol. (2004) 53:68–80.
Laurin M., Reisz R. R. A reevaluation of early amniote phylogeny. Zool. J. Linn. Soc. (1995) 113:165–223.[CrossRef]
Lee D.-C., Bryant H. N. A reconsideration of the coding of inapplicable characters: Assumptions and problems. Cladistics (1999) 15:373–378.[CrossRef][Web of Science]
Lee M. S. Y. The origin of the turtle body plan: Bridging a famous morphological gap. Science (1993) 261:1716–1720.
Lee M. S. Y. Pareiasaur phylogeny and the origin of turtles. Zool. J. Linn. Soc. (1997a) 120:197–280.[CrossRef]
Lee M. S. Y. Reptile relationships turn turtle.... Nature (1997b) 389:245–246.[CrossRef][Web of Science]
Lee M. S. Y. Molecules, morphology, and the monophyly of diapsid reptiles. Contrib. Zool. (2001) 70:1–22.
Løvtrup S. On the classification of the taxon Tetrapoda. Syst. Zool. (1985) 34:463–470.
Lull R. S. The armor of Stegosaurus. Am. J. Sci. Seri. 4 (1910) 29:201–210.
Maddison W. P. Missing data versus missing characters in phylogenetic analysis. Syst. Biol. (1993) 42:576–581.
Mannen H., Li S. L. Molecular evidence for a clade of turtles. Mol. Phylogenet. Evol. (1999) 13:144–148.[CrossRef][Web of Science][Medline]
Marsh O. C. Restoration of Stegosaurus. Am. J. Sci. Seri. 3 (1891) 42:179–181.
Merck J. W. A phylogenetic analysis of the euryapsid reptiles. J. Vertebr. Paleontol. (1997) 17:65A.
Modesto S. P. Observations on the structure of the Early Permian reptile Stereosternum tumidum Cope. Palaeontol. Afric. (1999) 35:7–19.
Modesto S. P. New information on the skull of the Early Permian reptile Mesosaurus tenuidens and its phylogenetic implications. J. Vertebr. Paleontol. (2004) 24:95A.
Molnar R. E. Preliminary report on a new ankylosaur from the Early Cretaceous of Queensland, Australia. Mem. Queensland Mus. (1996) 39:653–668.
Nixon K. C. WINCLADA (beta) version 0.9.9 (2000) Ithaca, New York. Published by the author.
Nixon K. C., Carpenter J. M. On simultaneous analysis. Cladistics (1996) 12:221–241.[CrossRef][Web of Science]
Olmstead R. G., Sweere J. A. Combining data in phylogenetic systematics: An empirical approach using three molecular data sets in the Solanaceae. Syst. Biol. (1994) 43:467–481.
Page R. D. M. Nexus data editor 0.5.0 (2001) Program available at http://taxonomy.zoology.gla.ac.uk/rod/NDE/nde.html.
Patterson C. Cartilage bones, dermal bones, and membrane bones, or the exoskeleton versus the endoskeleton. In: Problems in vertebrate evolution—Andrews S. M., Miles R. S., Walker A. D., eds. (1977) London: Academic Press. Pages 77–121.
Pleijel P. On character coding for phylogenetic analysis. Cladistics (1995) 11:309–315.[CrossRef][Web of Science]
Pollock D. D., Zwickl D. J., McGuire J. A., Hillis D. M. Increased taxon sampling is advantageous for phylogenetic inference. Syst. Biol. (2002) 51:664–671.
Prendini L. Species or supraspecific taxa as terminals in cladistic analysis? Groundplans versus exemplars revisited. Syst. Biol. (2001) 50:290–300.
Reisz R. R., Berman D. S., Scott D. The anatomy and relationships of the Lower Permian reptile Araeoscelis. J. Vertebr. Paleontol. (1984) 4:57–67.
Reisz R. R., Laurin M. Owenetta and the origin of turtles. Nature (1991) 349:324–326.[CrossRef]
Rieppel O. Turtles as diapsid reptiles. Zool. Scripta (2000) 29:199–212.[CrossRef][Web of Science]
Rieppel O., deBraga M. Turtles as diapsid reptiles. Nature (1996) 384:453–455.[CrossRef]
Rieppel O., Reisz R. R. The origin and early evolution of turtles. Annu. Rev. Ecol. Syst. (1999) 30:1–22.[CrossRef][Web of Science]
Romer A. S. Osteology of the reptiles (1956) Chicago: University of Chicago Press.
Rosenberg M. S., Kumar S. Incomplete taxon sampling is not a problem for phylogenetic inference. Proc. Nati. Acad. Sci. USA (2001) 98:10751–10756.
Rubial R., Shoemaker V. Osteoderms in anurans. J. Herpetol. (1984) 18:313–328.[CrossRef]
Schuh R. T. Biological systematics: Principles and applications (2000) Ithaca, New York: Cornell University Press.
Starck D. Vergleichende Anatomie der Wirbeltiere auf evolutionsbiologischer Grundlage (1979) Volumen 2. Berlin: Springer Verlag.
Strong E. E., Lipscomb D. Character coding and inapplicable data. Cladistics (1999) 15:363–371.[CrossRef][Web of Science]
Swofford D. L. PAUP* 4.0b10. In: Phylogenetic analysis using parsimony (*and other methods) (2001) Sunderland, Massachusetts: Sinauer Associates.
Weishampel D. B., Horner J. R. Hadrosauridae. In: The Dinosauria—Weishampel D. B., Dodson P., Osmolska H., eds. (1990) Berkeley: University of California Press. Pages 534–561.
Wheeler W. C., Cartwright P., Hayashi C. Y. Arthropod phylogeny: A combined approach. Cladistics (1993) 9:1–39.[CrossRef][Web of Science]
Wilkinson M. A comparison of two methods of character construction. Cladistics (1995) 11:297–308.[CrossRef][Web of Science]
Zardoya R., Meyer A. Complete mitochondrial genome suggests diapsid affinities of turtles. Proc. Nat. Acad. Sci. USA (1998) 95:14226–14231.
Zardoya R., Meyer A. The evolutionary position of turtles revised. Naturwissenschaften (2001) 88:193–200.[CrossRef][Web of Science][Medline]
Zwickl D. J., Hillis D. M. Increased taxon sampling greatly reduces phylogenetic error. Syst. Biol. (2002) 51:588–598.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  S. R. Singer, S. L. Maki, A. D. Farmer, D. Ilut, G. D. May, S. B. Cannon, and J. J. Doyle Venturing Beyond Beans and Peas: What Can We Learn from Chamaecrista? Plant Physiology, November 1, 2009; 151(3): 1041 - 1047. [Full Text] [PDF]
|
|
|
|
 |
|
 W. G Joyce, S. G Lucas, T. M Scheyer, A. B Heckert, and A. P Hunt A thin-shelled reptile from the Late Triassic of North America and the origin of the turtle shell Proc R Soc B, February 7, 2009; 276(1656): 507 - 513. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. T. Butcher and R. W. Blob Mechanics of limb bone loading during terrestrial locomotion in river cooter turtles (Pseudemys concinna) J. Exp. Biol., April 15, 2008; 211(8): 1187 - 1202. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 R. J. Asher, J. H. Geisler, and M. R. Sanchez-Villagra Morphology, Paleontology, and Placental Mammal Phylogeny Syst Biol, April 1, 2008; 57(2): 311 - 317. [Full Text] [PDF]
|
|
|
|
|
|
S. R. Harris, D. Pisani, D. J. Gower, and M. Wilkinson Investigating Stagnation in Morphological Phylogenetics Using Consensus Data Syst Biol, February 1, 2007; 56(1): 125 - 129. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||