|Electronic supplementary material
Phylogenetic clade definitions
From the seminal work of Benton and Clark (1988), the interrelationship within Crocodilia is still in a state of flux. This is due both to our still preliminary understanding of the phylogenetic relationships of the different clades, with many still swinging around widely under the different phylogenetic hypotheses, as well as to the often indiscriminate, subjective or controversial usage of the different available names applied to different clades, stems or nodes present in the cladograms (e.g. Martin and Benton, 2008; Brochu et al. 2009 and see below). Rigorous standardization of the phylogenetic taxonomy of crocodilians (or crocodyliforms), aiming to stabilize this nomeclatural system, began with Sereno et al. (2001), with the logical bases of this system being laid out by Sereno (2005) and the available phylogenetic definitions redefined and made available online on TaxonSearch (Sereno et al. 2005); in this study, we follow the clade definitions of Sereno (2001) and refined in Sereno et al. (2005; see also Sereno and Larsson, 2009), although some comments are given in order to clarify the usage and composition of the different clade names considered and used throughout this study: Mesoeucrocodylia, Neosuchia, Atoposauridae Notosuchia and Eusuchia.
Mesoeucrocodylia is a taxonomic subdivision within Crocodylia, erected by Whetstone & Whybrow (1983), and meant to include all taxa more derived than protosuchians. According to Benton and Clark (1988), the Mesoeucrocodylia was erected by Whetstone and Whybrow (1983) to include the traditional paraphyletic Mesosuchia as well as Eusuchia. This usage of the term is well illustrated by such recent publications as Larsson and Sues (2007), Sereno and Larsson (2009) or the synthetic crocodilian tree of Brochu et al. (2009); according to the Paleobiology Database (http://paleodb.org), Mesoeucrocodylia includes the following clades and taxa: Hsisosuchidae, Bergisuchus, Bernissartiidae, Dianchungosaurus, Libycosuchidae, Libycosuchus, Metasuchia, Neosuchia, Neuquensuchus, Pachycheilosuchus, Paralligatoridae, Pholidosauridae, Sebecosuchia, Teleosauridae, Trematochampsidae, Ziphosuchia.
The available phylogenetic definition of Mesoeucrocodylia is that found in Sereno et al. (2005; http://www.taxonsearch.org), derived from and standardizing the usage of Sereno et al. (2001), respectively a stem-based taxon corresponding to “The most inclusive clade containing Crocodylus niloticus (Laurenti 1768) but not Protosuchus richardsoni (Brown 1933)”. This usage is conformable with the original intent of the name, if Protosuchia is considered monophyletic; however, many recent analyses (including, e.g., Pol et al. 2009 or Sereno and Larsson 2009) failed to recover a monophyletic Protosuchia, the different clades traditionally considered to be protosuchians being arrayed as successively more distant outgroups to a monophyletic Mesoeucrocodylia. Implementation of this strict definition of Mesoeucrocodylia is, however, difficult to apply; and, indeed, even Sereno and Larsson (2009) fail to comply to this definition since in their fig. 43 Mesoeucrocodylia is being figured as excluding Hsisosuchus while this same genus is shown as member of the Mesoeucrocodylia in their Fig. 3. Moreover, according to the above definition (adopted explicitly in the text of Sereno and Larsson 2009, Mesoeucrocodylia, as a stem-based taxon, should include all crocodyliforms close to Crocodylus than to Protosuchus; on the other hand, in their figure 43 (and the underlying analysis), Protosuchus is not used, being replaced as outgroup with Orthosuchus, along with another taxon usually included within the Protosuchia, Zosuchus. Based on their results, as represented in fig. 43, Zosuchus should also be considered a mesoeucrocodylian, being more closely related to Crocodylus than to Orthosuchus (that stands as a proxy for Protosuchus). This assertion would surely stand against “common wisdom” concerning the composition of “Protosuchia” and respectively Mesoeucrocodylia, and is indeed not suggested by the authors, since they mark the position of the mesoeucrocodylian clade slightly higher in the cladogram. To sum up, strict adhesion of the currently proposed definition of Mesoeucrocodylia (Sereno et al., 2005) would be both counterintuitive and counterproductive if, and when, “Protosuchia” will be found to be paraphyletic, removing many taxa traditionally regarded as protosuchians into the Mesoeucrocodylia.
In order to avoid this, we adopt a relaxed view of Mesoeucrocodylia, as including all crocodylomorphs more closely related to Crocodylus niloticus than to any of the taxa traditionally viewed as protosuchian (e.g., protosuchids, gobiosuchids, Shantungosuchus); although not attempted here formally, a strict phylogenetic definition of this taxon would require only using more than one external specifier, besides Protosuchus richardsoni. We are confident that this relaxed definition of Mesoeucrocodylia represents a suitable working hypothesis, as it: 1) is in accordance with traditional views as to the composition of Protosuchia and Mesoeucrocodylia (see above); 2) appears to correspond to the meaning intended by the original phylogenetic definition of Sereno et al. (2005) as well, as shown by the tacit usage of the clade denomination in fig. 43 of Sereno and Larsson (2009); 3) seems to be representative for the mainstream taxonomic opinion of the crocodylian researchers, as shown by the agreement tree produced by Brochu et al. (2009); and 4) corresponds to the most inclusive clade that includes Crocodylus niloticus but not Protosuchus richardsoni, and that is upheld (although only marginally) in the bootstrap support of our analysis (see below). The Fruita crocodilian (Clark, 1985; Clark, 1994) included in the original CTM and retained in the present analysis is included within the Protosuchia for the moment (see bootstrap analysis below), although its membership is highly unstable and may change with further in-depth analyses.
The second clade considered here, of far greater importance for our analysis, is represented by Neosuchia, originally defined by Benton and Clark (1988) as a receptacle "for the Atoposauridae, Goniopholidae, Pholidosauridae, Dyrosauridae, Bernissartia, Shamosuchus and eusuchians" and based on a list of potential synapomorphies (Appendix 2 of Benton and Clark 1988). This list of included taxa is expanded, according to a recent search on the Paleobiology Database, to encompass “Bernissartia, Dyrosauridae, Dyrosaurus, Eusuchia, Goniopholididae, Khoratosuchus, Mahajangasuchidae, Pholidosaurus, Sarcosuchus, Stomatosuchidae, Terminonaris, Thalattosuchia”. Accordingly, the original definition of Neosuchia rests on a taxon list (jumbling together genera and higher-level taxa whose monophyly is not always verified) and a list of potential synapomorphies that may become (unless is not already) obsolete. Noteworthy is that Benton and Clark (1988) was undecided whether to include or not Thalattosuchia into the newly erected Neosuchia, an issue that is still controversial (see below).
In order to standardize and stabilize this definition, Sereno et al. (2001) redefined Neosuchia (see also Sereno et al. 2005; TaxonSearch, http://www.taxonsearch.org) as a stem-based taxon corresponding to “the most inclusive clade containing Crocodylus niloticus (Laurenti 1768) but not Notosuchus terrestris Woodward 1896”. Following from this phylogenetic definition, the identity and composition of the included taxa varies widely between the different analyses (compare, e.g.,Larsson and Sues, 2007; Sereno and Larsson, 2009; Pol et al., 2009; Pol and Gasparini, 2009). It is noteworthy that although stable, node-stem triplet – based phylogenetic definitions are recommended to achieve nomenclatural stability (Sereno 2005), several inconsistencies can still be noted between the different recent usages of the name Neosuchia such as:
1) exclusion of Thalattosuchia from Neosuchia (e.g., Sereno et al. 2009; Young and Andrade 2009), despite the “common wisdom” as synthesized by the Paleobiology Database, or, conversely, inclusion of Thalattosuchia (e.g., Ortega et al. 2000; Pol et al. 2009; Pol and Gasparini 2009) despite the fact that this membership was not intended in the original usage of the name. Controversial status of the Thalattosuchia is well illustrated by the compound consensus phylogenetic tree developed by Brochu et al. (2009), and also appears in our different analyses (see below);
2) exclusion of taxa such as the Dyrosauridae or Pholidosaurus from Neosuchia (e.g., Pol et al. 2009; see also Brochu et al. 2009) despite the original intent of the Neosuchia that explicitely include these taxa into the clade; and
3) controversial status of different taxa such as Peirosauridae or Sebecia, either included (e.g., Ortega et al 2000; Sereno and Larsson 2009) or excluded (e.g., Pol et al. 2009; Brochu et al. 2009) from Neosuchia. Although not expressed explicitly, using the phylogenetic definition of Sereno et al. (2005) would drag into the Neosuchia even taxa customarily identified as “notosuchians”, such as Araripesuchus, Libycosuchus and Baurusuchus (e.g., Rogers, 2003).
Despite these controversial aspects, we feel confident that usage of the strict phylogenetic definition of Sereno et al. (2005) is recommended, although its unambiguous status would depend on the monophyletic status of its co-ordinated stem-taxon, the Notosuchia (and, thus, the choice of Notosuchus terrestris as an internal specifier for this clade). Regardless of future developments in this respect, however, that would be easily corrected for stability by chosing more than one specifier, we follow in our discussions the currently available phylogenetic definition of Neosuchia, that of Sereno et al. (2005).
For Eusuchia, we follow the node-based definition first proposed by Brochu (1999, 2003) as the “last common ancestor of Hylaeochampsa vectiana, Gavialis gangeticus, Alligator mississippiensis, and Crocodylus niloticus and all of its descendants” and formalized by Sereno et al. (2005; TaxonSearch, http://www.taxonsearch.org) as “the least inclusive clade containing Hylaeochampsa vectiana Owen 1874 and Crocodylus niloticus (Laurenti 1768)”. It corresponds to current usage of the name (e.g., Pol et al 2009), although basal membership of the clade is in a state of flux (see the consensus tree computed by Brochu et al 2009).
Finally, Atoposauridae, first defined by Gervais (1871) is usually defined by its membership and/or list of potential synapomorphies (e.g., Clarke 1986; Buscalioni and Sanz 1988; Wu et al. 1996), although the monophyly of the family was never tested and the synapomorphies supporting the family are in part obsolete. It was redefined within a coherent phylogenetic framework by Sereno et al. (2005; TaxonSearch, http://www.taxonsearch.org) as a stem-based taxon corresponding to “the most inclusive clade containing Atoposaurus jourdani Meyer 1850 but not Peirosaurus torminni Price 1955, Araripesuchus gomesii Price 1959, Notosuchus terrestris Woodward 1896, Baurusuchus pachecoi Price 1945, Crocodylus niloticus (Laurenti 1768)”. It should be noted, nevertheless, that despite the unambiguous nature of the above definition and the fact that it is anchored on the nominotypical species Atoposaurus beaumonti (that seems the best way to promote stability), it also raises one important problem – the fact that Atoposaurus itself was not included into any recent and comprehensive survey of mesoeucrocodylian/neosuchian relationships. Accordingly, position of Atoposauridae depends actually in that of the proxies used to represent this family (usually Theriosuchus and/or Alligatorium), whose relationships to Atoposaurus were investigated only in a preliminary manner (Buscalioni and Sanz, 1988) and is far from well-supported. Nevertheless and despite some pitfalls, we choose to use the phylogenetic definition of Sereno et al. (2005) as discussed above. Moreover, as Sereno et al. (2005) also pointed out, under this definition the recently described Pachycheilosuchus trinquei from the Albian of southern USA (Rogers 2003) would also be included into the Atoposauridae, instead of representing only the sister-taxon of it.
Several phylogenetic analyses were conducted during this study, in order to cross-check the variety of hypotheses brought by the use of different matrices. Moreover, although the primary analysis was run using TNT version 1.0 (Goboloff et al. 2003), the same data was also analysed using PAUP 4.10 (Swofford 2000) to see whether the different softwares recover the same results.
Many recent studies have been devoted at understanding the phylogeny of Mesoeucrocodylia, a variety of matrices accompanying them. We selected those matrices with the maximum characters to be coded in the fragmentary Theriosuchus sympiestodon. Here, we present the detailed results of our phylogenetic investigation, first using the matrices of Andrade and Bertini (2008) and Young and Andrade (2009) containing a proportionally high content of Gondwanan taxa. Then, we investigate the phylogenetic position of the new taxon in the matrix of Pol et al. (2009), containing a balanced sample of Gondwanan and Laurasian taxa.
Theriosuchus sympiestodon was included in the character-taxon matrix (CTM) of Andrade and Bertini (2008); this choice was motivated by the fact that this CTM allowed for the most extensive coding of the fragmentary material of T. sympiestodon. Additional taxa included in the original CTM are Araripesuchus tsangatsangana (Turner 2006) and Araripesuchus buitreraensis (Pol and Apesteguia 2005); Theriosuchus pusillus was also added, coded based on personal observations and from the thesis of Clark (1986). Theriosuchus guimarotae was recoded from the literature (Schwarz and Salisbury 2005). The resulting CTM contains 26 taxa and 183 characters and was analyzed under TNT using a traditional search (1000 random seeds and 1000 replicates), with Sphenosuchia assigned as the outgroup taxon. 4 most parsimonious trees (MPTs) with an optimal length of 567 steps (consistency index CI = 0.402, retention index RI = 0.617) were obtained, the majority rule consensus tree of these being reproduced in figure 1. Results from PAUP mirror with high fidelity those from TNT. Interestingly, the topology of the majority rule tree reproduces that of the strict consensus tree, suggesting that all the nodes recorded in the majority rule consensus tree are robust. Examination of the four MPTs show that changes are restricted to the two polytomies from within the Araripesuchus + Theriosuchus clade, and concern the relative position of A. patagonicus and A. tsangatsangana, respectively. While A. patagonicus appear as the sister-taxon of A. gomesi in two of the trees, it shifts into a more basal position in the other two MPTs; similar changes are also present in the case of A. tsangtsangana, that is either a sister-taxon of T. sympiestodon, or is slightly less derived than the Romanian taxon. Bootstrap support analysis with 1000 replicates was run in order to assess the robustness of the clades identified by the analysis. The Bremer decay analysis was computed from a sample of 70000 random trees (i.e. the maximum that the computer could hold for calculation); Bremer support values (1-3 = number of steps necessary to made a node collapse) are directly mentioned on the consensus tree (Fig. 1). As expected, the analysis recovered several long-held relationships within Crocodylomorpha, such as the sister-group relationships between Sphenosuchia and Crocodylia (or Crocodyliformes), the basal position of Protosuchia within Crocodylia (or Crocodyliformes) and the sister-group relationship between Thalattosuchia and other mesoeucrocodylians (Young and Andrade 2009); robustness of these nodes are upheld by high Bremer (>3) and bootstrap (100) support values. Also well supported are: (1) the basal position of a monophyletic Notosuchia inside Mesoeucrocodylia (a monophyletic Notosuchia is still supported, although poorly, in the bootstrap analysis, but it falls into a basal polytomy with other Metasuchia); (2) the position of Eusuchia and that of the crown-group inside this, as the most derived crocodilians by both high Bremer (>3) and highest bootstrap (99-100) values; and (3) the sistergroup relationship between eusuchians and sebecids (although only marginally – 52 – in the bootstrap analysis). Despite the fact that araripesuchuds and Theriosuchus are grouped together by the phylogenetic analysis, support for this monophyletic clade is weak – the clade collapses partially and ingroup resolution is severely reduced in trees one step longer than the MPTs; the bootstrap analysis showed very limited support only for the T. guimarotae + T. pusillus clade (61) and – surprisingly – the A. gomesi and A. patagonicus (56) grouping. This weak support is probably due to both large amount of missing data and high degree of homoplasy of the characters employed (as documented by low values of the CI).
Theriosuchus sympiestodon was also included in the matrix of Young and Andrade (2009) originally designed to solve thalattosuchian relationships. Additional taxa have been added to the original CTM, such as Araripesuchus tsangatsangana (Turner 2006) and Araripesuchus buitreraensis (Pol and Apesteguia 2005), coded from the literature, as well as Theriosuchus pusillus, coded based on our direct observations, and from Owen (1879) and the thesis of Clark (1986). Theriosuchus guimarotae was also recoded based on Schwarz and Salisbury (2005) and personal information from D. Schwarz-Wings (2008–2009). The resulting CTM includes 90 taxa and 166 characters. It was analysed using TNT (Goloboff et al. 2003), with Erpetosuchus as the outgroup taxon and the following settings: traditional search, 1000 random seeds, 1000 replicates. Bootstrap analysis was run using 1000 replicates; the Bremer decay analysis was computed from a sample of 70000 random trees (maximum that computer could support). The analysis yielded 56 MPTs with an optimal length of 667 steps (CI = 0.390, RI = 0.839), the majority rule consensus tree of which is presented in figure 2. The same relationships were recovered within basal crocodylomorphs as in the first analysis, with Protosuchus and a monophyletic Sphenosuchia successively more distant relatives to a monophyletic Mesoeucrocodylia, with high (3) and moderate (2) Bremer decay indices for the respective nodes, also present in the strict consensus tree (not shown here), However, no bootstrap support was found for these nodes, except for that supporting monophyly of Sphenosuchia. Inside Mesoeucrocodylia, represented by a large basal polytomy in the strict consensus tree, resolution is significantly lower, only some clades (also present in the strict consensus tree) receiving fair to good Bremer support, such as Notosuchia, Sebecia, Susisuchus + Eusuchia, Dyrosauridae and Thalattosauria, along with the newly identified Araripesuchus + Theriosuchus group (with the exception of A. patagonicus). However, it should be noted that bootstrap support for even these clades are low to very low at best (varying between 51–57 for Eusuchia, Dyrosauridae and Notosuchia, the last without Baurusuchus), with the exception of Thalattosuchia, recovered in 96% of the replicates; this high value of support is hardly surprising, since the original CTM was meant to resolve the position and ingroup relationships of Thalattosuchia.
Despite the low statistical support, the topology of the majority rule consensus tree invites to explore one interesting relationship. Far from ranking Thalattosuchia as a basal mesoeucrocodylian radiation, sister-group to Metasuchia (as found by, e.g., Young and Andrade 2009; Sereno and Larsson 2009, and references therein), the consensus tree shows it nested high within Metasuchia, as the most derived monophyletic clade of a group including goniopholids, pholidosaurids and dyrosaurids (Pholidosauria sensu Sereno and Larsson 2009), group that stands as the sister-taxon to eusuchians. This position of Thalattosuchia is quite unexpected and is at odds with the results of many previous analyses (but see, e.g., Rogers 2003; Pol and Gasparini 2009, for comparable patterns), but also with those derived from our analysis using the same CTM with PAUP (Swofford 2000). Although several runs of this second analysis failed to recover MPTs with the same length as TNT (667), it did recover a large number (usually in excess of 12000) of marginally longer MPTs (length 668; CI = 0.389, RI = 0.839). The topology of the PAUP majority rule consensus tree replicates the phylogenetic relationships present in recent analyses (e.g., Sereno and Larsson 2009), with: (1) Thalattosuchia as sister-group of Metasuchia inside Mesoeucrocodylia; (2) monophyletic Notosuchia as a sister-group of Neouchia; (3) monophyletic Sebecia as a sister-group of Mahajangasuchus + more derived metasuchians, inside Neosuchia. Within derived neosuchians, goniopholids appear as sister-group to Eusuchia + [parahyletic pholidosaurids + Dyrosauridae]. The unusual position of Thalattosuchia, found in our TNT analysis is intriguing and invites for more in-depth investigations, but further inquire into this problem is out of the scope of the present study.
Returning to the focus of our study, T. sympiestodon groups with T. pusillus, regardless the phylogenetic analysis (TNT or PAUP) considered; together, they form the sister-group to a T. guimarotae + A. tsangatsangana clade, with A. buitreraensis their next successive outgroup. Moreover, A. patagonicus also groups with these taxa as the basalmost member of the clade in the PAUP analysis, but not in the TNT analysis, where it lies just outside the node uniting the Araripesuchus-Theriosuchus clade and Neosuchia. Similarly to the results of the first analysis, this Araripesuchus+Theriosuchus clade represents the sister-taxon of Neosuchia, and appears to be more derived than Notosuchia (only in the TNT analysis; in PAUP, it appears in a basal polytomy with Notosuchia and Neosuchia). Majority rule support is high for this clade and its ingroup relationships, with low to moderate Bremer decay values for most of the internal nodes. These decay values are similar to those found in other "well-established" clades such as dyrosaurids, notosuchians or sebecids, and suggest that the assemblage might indeed represent a monophyletic group. Nonetheless, there is no significant bootstrap support for any of these relationships, except maybe for the grouping of T. sympiestodon and T. pusillus (found under PAUP, only); but again, no bootstrap support was found to uphold previously widely recognized taxa such as goniopholids, sebecids or pholidosaurids, either. The situation is somewhat reminiscent of that resulted from the analysis of the modified Bertini and Andrade (2008) CTM, and is probably due, again to the high degree of homoplasy and large amount of missing data throughout the analyzed CTM.
Admittedly, the results of the first two analyses are not coincident in every detail, nor are these statistically supported in optimal way. Nevertheless, these are remarkable in that both recovered the same close relationship between different species of Araripesuchus (especially A. buitreraensis and A. tsangatsangana) and Theriosuchus, respectively. The results are even more remarkable in that members of this clade appear consistently in the same relative position, as basal mesoeucrocodylians, more derived than notosuchians but less derived than neosuchians.
Finally, the phylogenetic position of T. sympiestodon was investigated in the matrix of Pol et al. (2009) originally designed to study the transition from Neosuchia to Eusuchia. The original matrix is composed of 282 characters, coded for 72 taxa. We added T. sympiestodon to the original matrix, and also T. guimarotae, coded from the literature (Schwarz and Salisbury, 2005). We also recoded T. pusillus for some characters (see below) according to our observations of the holotype and referred specimens from the NHM in London. Characters considered additive by Pol et al. (2009) was maintained as such in the current analysis. Each individual analysis was made under TNT, using traditional search with 1000 random seeds and 1000 replicates, 80 trees being saved for each replicate. Support for the clades was explored using Bremer decay analysis with memory space for 70000 trees saved. Analyses in PAUP were run using the heuristic search option, with 1000 replicates; number of trees saved for each replicate was set to 50000. Bootstrap support for the recovered trees was calculated using 100 replicates (due to memory constrains), with 10000 trees saved for each replicate. Decay analysis was also conducted in order to assess the robustness of the recovered nodes, using samples of 50000 randomly chosen trees.
Running the analysis using the Pol et al. (2009) matrix, modified only to include changes in T. pusillus did not alter the pattern recovered in the original analysis. We then included T. sympiestodon and T. guimarotae in the analysis for a total of 74 taxa. Analysis of the complete CTM returned 288 MPTs with best tree length of 1091 steps (CI = 0.321; RI = 0.706), the strict consensus of which is presented in fig. 3. The Bremer decay indexes are also detailed with values ranging from 1 (poor support) to “3” or higher (better support)
Unlike the first two analyses, using the CTM of Pol et al. (2009) recovers a monophyletic Atoposauridae (including T. sympiestodon), nested inside Neosuchia; note that Pol et al. (2009) marks the position of the clade Neosuchia high within Mesoeucrocodylia (at the node uniting Atoposauridae and the successively more exclusive clade with goniopholids, shamosuchids, and eusuchians, unlike the position implied by the phylogenetic definition adopted here, that locate Neosuchia as the monophyletic sister-taxon of Notosuchia (thus peirosaurids and all more derived mesoeucrocodylians). Moreover, Theriosuchus (or Atoposauridae, in overall) do not show any special relationships with members of the Araripesuchus clade. Otherwise, the topology recovered is closely comparable in most of its details to that of Pol et al. (2009), with a paraphyletic “Protosuchia”, including a basal clade of protosuchids and successively more derived gobiosuchids, shantungosuchids and the the Fruita cocodyliform, and a monophyletic Mesoeucrocodylia. Within Mesoeucrocodylia, Hsisosuchus is the basalmost taxon, sister-group to a node-based Metasuchia comprising two stem-based taxa: Notosuchia and Neosuchia (see Sereno et al., 2005). Inside Notosuchia, the monophyletic and well-resolved araripesuchids form the sister-group of the clade containing all other taxa considered traditionally as notosuchids. The basalmost mesoeucrocodylian clade is represented by the peirosaurids, followed by atoposaurids and then by a fundamental dichotomy between the lineage leading to crown-group crocodylians (starrting from goniopholids, through Bernissartia, the Glen Rose neosuchian and shamosauchids, to eusuchians) and another large clade including, on one side, the grouping of Pholidosaurus, sarcosuchids and dyrosaurids, as sister-group to thalattosuchians. Internal detailed topology of these lineages is remarkably similar to that found by Pol et al. (2009) and will not be discussed here further.
Although largely similar to that obtained by Pol et al. (2009) in their original analysis, the topology of the strict consensus departs from this in several respects such as yielding lower resolution of ingroup relationships inside non-araripesuchid notosuchians or thalattosuchians. However, the most important change concerns the position of atoposaurids within Neosuchia; while according to the original analysis, atoposaurids rank as one of the most derived clades of non-eusuchian mesoeucrocodylians, according to our analysis atoposaurids are positioned low within Neosuchia (sensu the definition adopted here), as the second basalmost neosuchian clade, more derived only to peirosaurids. Accordingly, the monophyletic Atoposauridae represents the sister-group of a clade made up of two monophyletic groupings: one leading to dyrosaurids and thalattosuchians, and another leading to the eusuchians. Inside Atoposauridae, Alligatorium appears to be the basalmost taxon, followed by T. sympiestodon as the sister-taxon to the clade of T. pusillus and T. guimarotae.
The majority rule consensus tree differ only marginally from the strict consensus tree; slightly better resolution can be found in the immediate eusuchian outgroups, with the Glen Rose taxon being recovered as more derived than Bernissartia (unlike the original analysis of Pol et al. 2009, where these two taxa occur as sister-taxa in the strict consensus tree), as well as within derived non-araripesuchis notosuchians, where the (Chimaerasuchus + Sphagesaurus) clade appear as the sister-group of the baurusuchids (just as in the strict consensus tree of Pol et al. 2009). Majority rule support is fairly high for both nodes, with 83% in the case of the derived notosuchians, and 75% in the case of Benissartia and the Glen Rose taxon, respectively, suggesting that these relationships are also relatively well-supported.
Bremer support values are relatively low throughout the tree, most branches collapsing into polytomies in the consensus tree of trees 1 step longer than the MPTs, and only very few clades (such as Crocodylomorpha, Crocodylians, Baurusuchidae and its ingroup, Gavialidae and the small clade made up of the two Dakosaurus species) appear relatively robust (decay index 3 or higher). Remarkably, Atoposauridae ranks among the best-supported clades (decay index 2), together with the sphenosuchids, gobiosuchids, shamosuchids, peirosaurids, thalattosuchians (and most of their ingroups), sarcosuchids, dyrosaurids, as well as the goniopholids; nevertheless, previously recovered ingroup relationships collapse into a basal polytomy even in this family in trees one step larger than the MPTs.
Surprisingly, the parallel analysis of the same CTM run under PAUP was able to recover a more parsimonious arrangement of the taxa – length of the 1447, marginally better MPTSs recovered under PAUP was 1065 (CI = 0.327, RI = 0.713; tree not reproduced here); due to the search setting limitations, it is probable a larger number of MPTs exist. The large-scale topology of the strict consensus tree is comparable to that resulted from the TNT analysis of our dataset, as well as that of the original analysis of Pol et al. (2009). Compared to the TNT analysis, the PAUP analysis found a better resolution within the most derived notosuchians (i.e., more derived than Malawisuchus), recovering here a topology identical to that found by Pol et al. (2009); however, it also shows a better resolution among the basal notosuchians, identical to those recovered in our TNT analysis, but unlike the unresolved polytomy recovered by Pol et al. (2009). The majority consensus tree differs in that is shows better resolution inside Goniopholidae as well as inside the (Bernissartia + higher Neosuchia), including within Eusuchia (where gavialoids are recovered as the sister-taxon of a clade including crocodyloids and alligatoroids). The most important departure of the PAUP analysis concerns the ingroup relationships of Atoposauridae that was recovered as a monophyletic clade with a similar position as in the TNT analysis, at the base of Neosuchia, immediately more derived than the basalmost peirosaurids. However, inside Atoposauridae, the monophyly of Theriosuchus is no longer apparent, as T. pusillus is grouped with Alligatorium, while T. guimarotae with T. sympiestodon. These ingroup relationships are due, most probably to the amount of missing data for the latter two taxa, and do not indicate any particular preferential phylygenetic relationships, as demonstrated by the supports the analysis foud for the atoposaurid inter-relationships (see below).
Bootstrap and Bremer support values for these relationships show, however, that most clades identified in the PAUP analysis are weakly supported, and point to the currently highly volatile nature of these relationships in many parts of the tree. Large-scale ingroup relationships within Crocodyliformes collapse in the bootstrap analysis, with such generally accepted groupings as the protosuchids, mesoeucrocodylians, metasuchians, neosuchians or even esuchians disappearing into a huge basal polytomy; this instability is also shown in the decay analysis, as all the clades listed above collapse in trees one step longer than the MPTs. Nevertheless, more exclusive ingroup relationships present in the strict consensus tree seem to better uphold in many cases, especially within “protosuchians” (e.g., gobiosuchids), notosuchians (e.g., araripesuchids, baurusuchids, sphagesaurids), thalattosuchians, the Dyrosauridae+Sarcosuchidae+Pholidosaurus clade, and occasionally inside Eusuchia (e.g., shamosuchids, gavialoids or alligatorids). Moderate to high Bremer support values are concentrated also at some of these nodes (such as gobiosuchids, baurusuchids, peirosaurids, diverse thalattosuchians, dyrosaurids, gavialoids). Interestingly, the atoposaurid clade (but not any of its ingroup relationships) receives a low-to-fair Bremer support (1), similar to, e.g., goniopholids, while bootstrap support was found only for the Alligatorium + T. pusillus grouping, suggesting that the ingroup relationships found in the strict consensus tree are due indeed to missing data and does not necessarily reflect true phylogenetic affinities.
In addition to the survey of the complete CTM dataset, two other analyses were run in order to control for the effects of the successive addition/removal of T. guimarotae and T. sympiestodon on the ingroup relationships of Atoposauridae, but also on the position of this clade within Mesoeucrocodylia and/or Neosuchia. This approach was taken especially due to the fact that both of these taxa are fairly incomplete, and the incompleteness of the coding for Theriosuchus was considered previously by Larsson and Sues (2007) as responsible for uncertainties in the neosuchian ingroup relationships: “the decreased resolution within the clade may be a result of the large amount of missing data for Theriosuchus. Theriosuchus is the only member of that clade [Neosuchia] to drop to the most inclusive neosuchian node when an Adams consensus is calculated, and suggests this is the most erratic taxon within the most parsimonious tree set.”
These analyses showed that the position of Atoposauridae in relation to the other neosuchians is relatively robustly upheld, and does not seem to depend on the inclusion of T. sympiestodon and T. guimarotae into the analysis. Again, both TNT and PAUP analyses were run, to cross-check any possible outcomes.
In the first analysis set, exclusion of T. guimarotae has no influence on the strict consensus tree topology. In this case (analysis of the dataset with T. guimarotae removed, shortened as aDS-G), the TNT analysis also yielded 288 MPTs (tree length = 1075; CI = 0.326; RI = 0.712). Again, a monophyletic Atoposauridae (now including Alligatorium, T. pusillus and T. sympiestodon) was found to lie close to the base of Neosuchia (Fig. 4), above Peirosauridae and below the same major clade with thalattosuchians, pholidosaurs, goniopholids and eusuchians as recovered in the analysis of the complete dataset (aCDS; see above). Details of the different ingroup relationships are identical to those recovered in the aCDS, as are those of the majority rule consensus tree. Using PAUP, aDS-G yielded 2905 MPTs (tree length = 1050; CI = 0.331; RI = 0.716) whose strict consensus tree (not represented here) is, again, better resolved than that of the aCDS, in exactly the same areas (more derived non-araripesuchid notosuchians). Unlike the TNT analysis, ingroup relationships inside a monophyletic Atoposauridae are not resolved, the three taxa falling out into a basal tritomy. Results of the boostrap and decay analyses for the aDS-G again are comparable to those reported for the aCDS, except that they found no longer a support for the atoposaurids; only a clade made up of Alligatorium and T. pusillus appear moderately well supported (bootstrap support value of 71), while T. sympiestodon falls out into a basal mesoeucrocodylian polytomy. Noteworthy is that Hsisosuchus is also part of the same mesoeucrocodylian polytomy, but not the Fruita crocodyliform, suggesting that traditional viewpoint that Mesoeucrocodylia indeed includes Hsisosuchidae (see above, Phylogenetic clade definitions).
On the other hand, removing only T. sympiestodon from the analysis (aDS-S) had a cascading effect in the TNT analysis. This yielded 1568 MPTs (length = 1081; CI = 0.324; RI = 0.710) whose strict consensus tree (Fig. 5) shows a wide polytomy involving Atoposauridae, Shamosuchidae, Pholidosauria + Thalattosuchia, Eusuchia, as well as five goniopholidid taxa, inside Neosuchia; moreover, breakdown of resolution is also present inside Eusuchia and Notosuchia, the monophyly of the latter being lost. Nevertheless, the majority rule consensus tree shows much better definition, its topology being similar to that recovered in the aCDS and the aDS-G. The aDS-S using PAUP yields, again, a strict consensus tree of superior resolution (computed from the 1442 MPTs, length = 1054, CI = 0.330, RI = 0.715; not presented here), similar to that from the aCDS and the aDS-G. Bootstrap and Bremer supports for the recovered clades are also comparable to those from the previous analyses. Similar to the aDS-G, a monophyletic Atoposauridae including Alligatorium, T. sympiestodon and T. pusillus, is not supported, unlike a group including only T. pusillus and Alligatorium (bootstrap support of 81); this, again suggest that the ingroup relationships of atoposaurids, as recovered in the aCDS using PAUP, reflect more probably missing data and not true phylogenetic relationships.
To conclude, the presence of either one of T. sympiestodon and T. guimarotae seem to influence little the position of Atoposauridae as recovered in the present analysis and, although further testing the validity of such results is beyond the scope of this contribution, it seems noteworthy that the results are not entirely tight to taxon sampling or data completeness within Neosuchia. It is even more noteworthy that using any one of these taxa, or both of them, led to results that depart sharply from those of Pol et al. (2009) concerning the position of Atoposauridae, placed very basally within Neosuchia as opposed to being sister-taxon of the Bernissartia + higher Neosuchia clade. Although support for this new position is not particularly strong, still the fact that it was recovered recurrently within the different analyses conducted, lend support to the hypothesis that atoposaurids are indeed basal neosuchians. This is also concurrent with the results from our previous analyses of the Bertini and Andrade (2008) and Young and Andrade (2009) CTMs, as well as with the results of Rogers (2003). Accordingly, we are fairly confident that the relationships of the Atoposauridae we recovered (as basal neosuchians) represent a well-supported hypothesis in our current state of knowledge. However, including other atoposaurids (and especially Atoposaurus or Brillanceausuchus) into further analyses is a pre-requisite for better pinpointing the position and ingroup relationships of the atoposaurids, as well as understanding the relationships within the different mesoeucrocodylian clades.
Character codings for the matrix of Andrade and Bertini (2008)
Theriosuchus sympiestodon 110?????2???1?0??01????????01??210?1?11?0?????????????????????????????111?010???11??0???0????0001?11?1??????????????????110?????1101?10100????????100??????????????????????????????????
Theriosuchus guimarotae ?11021222?011100?03200100??0010211002010001000?00?1000????1??0010???1?1?2??1??0011100010111000?0101101?00111??1?002?????1013310011010111001??1001111001110????00?01????????????????????
Theriosuchus pusillus 11?0?102??0?11???02200?00??0011211111????011?1??0?10????0??001?00???0?1?????????11100010011101?1101101000110101????0???01013310?11?10?010??0?10000100??????????????????????????????????
Araripesuchus tsangatsangana 11002111201?0?0??021100101?010?210?010100?1111100???1110011001?1?1100?111001001???1?0????1????00?011?1?101111000002000101123310?1301?101001100??1?10001?011?00100?0???02??00?010?101010
Araripesuchus buitreraensis ?10020?????0??0???2?1??00??11??2111010?0?1?101?010??011?0?0?010?00??????????????1110?0100110000010111111011?0???????????1?0?1??????1011??????1???1?00??????????????????????????????????
Character codings for the matrix of Young and Andrade (2009)
0000??0?01 0????????1 ????2?01?1 0????????? ??02????0? ??0?10211? ?????????? ?????????? ???2?????? 0??????11? ?????????0 ??0?21??1? 1?0210???0 01??0???0? ??0????1?? ???????10? ?0????
0000?2?10? ?01?001??1 12402?0112 02?00101?? ?0021?100? ?011102110 0???0000?0 000?000110 0?12??011? ?1?0001?1? ?1???00110 ??012??2?0 1?00101?00 010?02200? ??00001?0 0??????1?1 000??0
100000?10? ?010001200 22?02?0012 01?0??0000 00020???0? ?01020111? 01???00?0? ?00??????1 0??1?[1,2]0110 ???01 12??? 0110? ?0110 10002 ?11?? 1?021 010?0 0?00? 02000 ??0?0 000?? 1???? ??1?0 ????0 0
0000?2?12? ??1??012?0 ?0??20?012 ?2???10??? ??12?0??0? ????0????1 ?????????? ?????????? ??01?????0 0??000???? ?????????1 ??001??0?? 1?02101?00 01????200? ???????1?? ???????1?0 00????
00001?01?0 121500020? 121?310212 02000?0300 00030?1011 ?001101110 102?00102? ??010?0??? 0?11020111 ?0?0012200 11?11?1111 ?00210010 1?02101000 ?1?10020?1 1?1100?11? 100????1?0 ?0?010
Character codings for the matrix of Pol, Turner and Norell (2008)
2?1?????? 2????????? ?1101????? ?0?0?11??1 01?10????[1,2] ?????????? 0??????311 ???????1?1 ?????????? ?????????? ???????00? ?????????1 1?1??1???? ???????001 1????????? ?????0??12 1?1?1?0??? ??????0?00 ???010??00 ?0?01???0? ???1???0?????1???0??????0?????????0?0?????00?????01110????????1???0?0???????0????0?
20110111120100110000110111100110011?211010101 ?11?01111000?? ???1?20211 ?410010101 0110112110 001112001 0013010002 ?0?10?1101 10001?110 0?00?0?001 10??01??0? 00??10002 101010200 0?100?0?10 001110?0? ?0?0101?01 0??01?1000 ????0000?2 ????0?0?0? ??0010??00 00?0???000 ??100?00?? 0?0?0011?1 ??0?100110 ?00
Recoded character for T. pusillus in Pol et al. (2009):
Character 43 (modified from Clark, 1994: char. 43): Primary pterygoidean palate: forms posterior half
of the choanal opening (0), or forms posterior, lateral, and part of the anterior margin of the choana (1), or completely enclose choana (2).
Changed to 43(1)
Character 164 (modified from Ortega et al., 2000: char. 19): Maxillary dental implantation: teeth in
isolated alveoli (0), or located on a dental groove (1).
Changed to 164(1)
Character 211 (Pol and Norell, 2004a: char. 180): Posterolateral end of quadratojugal: acute or
rounded, tightly overlapping the quadrate (0), or with sinusoidal ventral edge and wide and rounded
posterior edge slightly overhanging the lateral surface of the quadrate (1).
Changed to 211(?)
Character 233 (Zaher et al., 2006: char. 197): Participation of ectopterygoid in palatine bar: no
(0), or yes (1).
Changed to 233(0)
Character 277: Shallow hemispherical depression on the lacrimal and/or prefrontal anterior to the
orbital margin (not articulation facet for palpebral): absent (0), or present (1).
Changed to 277(1)
Character 278: Anterior half of palatine bar between suborbital fenestrae: lateral margins are parallel to
subparallel (0) or flared anteriorly (1).
Changed to 278(1)
Andrade MB de, Bertini RJ (2008) A new Sphagesaurus (Mesoeucrocodylia:
Notosuchia) from the Upper Cretaceous of Monte Alto City (Bauru Group, Brazil), and a revision of the Sphagesauridae. Hist Biol 20:101–136
Benton MJ & Clark JM 1998. Archosaur phylogeny and the relationships of the Crocodylia. In Benton MJ (ed.),
The Phylogeny and Classification of Tetrapods, Vol. 1, Clarendon Press, Oxford, p. 289–332.
Brochu CA 1999. Phylogenetics, taxonomy, and historical biogeography of Alligatoroidea. J Vert Paleontol 19,
Suppl. To no. 2: 1–92.
Brochu CA 2003. Phylogenetic approaches toward Crocodylian history. Ann Rev Earth Planetary Sci 31: 357–
Brochu CA, Wagner JR, Jouve S, Sumrall CD, Densmore LD (2009) A correction corrected: consensus over the meaning of Crocodylia and why it matters. Syst Biol 58: 537–543 doi:10.1093/sysbio/syp053
Clark, J.M., 1985. A new crocodylomorph from the Late Jurassic Morrison Formation of western Colorado, with
a discussion of relationships within the ‘Mesosuchia’. Unpublished MS thesis, University of California, Berkeley, 84 pp.
Clark JM (1986) Phylogenetic relationships of the crocodylomorph archosaurs. Ph.D.
dissertation, University of Chicago, Chicago, 556 pp
Clark, J.M., 1994. Patterns of evolution in Mesozoic Crocodyliformes. In: Fraser, N.C. and Sues, H.-D., Editors,
1994. In the Shadow of Dinosaurs, Cambridge University Press, New York, pp. 84–97.
Goloboff PA, Farris JS, Nixon K (2003) TNT : tree analysis using new technologies (Program and
documentation available from the authors and at http://www.zmuc.dk/public/phylogeny).
Owen R (1874–79) Monograph on the fossil reptilia of the Wealden and Purbeck Formations.
Supplement no. VI, Crocodilia (Hylaeochampsa). 7 pp. and pl. II of Suppl. V. Palaeontogr. Soc. (Monogr.), London (1874). Supplement no. IX. Crocodilia (Goniopholis, Brachydectes, Nannosuchus, Theriosuchus and Nuthetes). 19 pp., 4 pls. loc. cit. (1879).
Pol D, Apesteguía S (2005) New Araripesuchus remains from the Early Late Cretaceous (Cenomanian-Turonian) of Patagonia. Am Mus Nov 3490:1–38
Pol D, Turner A, Norell M (2009) Morphology of the Late Cretaceous crocodylomorph Shamosuchus
djadochtaensis and a discussion of neosuchian phylogeny as related to the origin of Eusuchia. Bull. Am. Mus. Nat. Hist. 324: 1-103.
Larsson HCE, Sues HD (2007) Cranial osteology and phylogenetic relationships of Hamadasuchus rebouli
(Crocodyliformes: Mesoeucrocodylia) from the Cretaceous of Morocco. Zool J Linn Soc 149:533–567.
Martin JE, Benton MJ (2008) Crown clades in vertebrate nomenclature: correcting the definition of Crocodylia.
Syst Biol 57:173–181
Rogers, J. V. I., 2003. Pachycheilosuchus trinquei, a new procoelous crocodyliform from the Lower Cretaceous
(Albian) Glen Rose Formation of Texas. Journal of Vertebrate Paleontology, 23, 128–145.
Schwarz D, Salisbury SW (2005) A new species of Theriosuchus (Atoposauridae,
Crocodylomorpha) from the Late Jurassic (Kimmeridgian) of Guimarota, Portugal. Géobios 38:779–802
Sereno PC 2005. The logical basis of phylogenetic taxonomy. Syst Biol 54: 595–619.
Sereno PC, Larsson HCE, Sidor CA, Gado B 2001. The giant crocodyliform Sarcosuchus from the Cretaceous of
Africa. Science 294: 1516–1519.
Sereno PC, McAllister S, Brusatte SL 2005. TaxonSearch: a relational database for suprageneric taxa and
phylogenetic definitions. PhyloInformatics 8: 1–21.
Sereno PC, Larsson HCE (2009) Cretaceous crocodyliforms from the Sahara. Zookeys
Swofford DL (2000) PAUP*. Phylogenetic Analysis Using Parsimony (*and Other
Methods), Version 4. Sinauer Associates, Sunderland, Massachusetts.
Turner AH (2006) Osteology and phylogeny of a new species of Araripesuchus
(Crocodyliformes : Mesoeucrocodylia) from the Late Cretaceous of Madagascar. Hist Biol 18:255–369
Whetstone KN, Whybrow PJ 1983. A “cursorial” crocodilian from the Triassic of Lesotho (Basutoland),
southern Africa. Occasional Papers of the Museum of Natural History, The University of Kansas 106: 1–37.
Young MT, Andrade MB de (2009) What is Geosaurus ? Redescription of Geosaurus
giganteus (Thalattosuchia: Metriorhynchidae) from the Upper Jurassic of Bayern, Germany. Zool J Linn Soc 157: 551–585
Fig. 1. Majority rule consensus tree presenting the relationship of Theriosuchus sympiestodon sp. nov. as coded in the matrix of Andrade and Bertini (2008). Numbers at nodes represent: bold italic – bootstrap support value, bold – Bremer decay value.
Fig. 2. Majority rule consensus tree presenting the relationship of Theriosuchus sympiestodon sp. nov. as coded in the matrix of Young and Andrade (2009). Numbers at nodes represent: upper line – 50% majority rule consensus support; lower line, bold italic – bootstrap support value, lower line, bold – Bremer decay value.
Fig. 3. Majority rule consensus tree presenting the relationships of Theriosuchus sympiestodon sp. nov. as coded in the matrix of Pol et al. (2009). Numbers at nodes represent: regular – 50% majority rule consensus support; bold – Bremer decay value.
Fig. 4. Strict consensus tree presenting the relationships of Theriosuchus sympiestodon sp. nov. and Atoposauridae as coded in the matrix of Pol et al. (2009) but with the exclusion of T. guimarotae from the analysis.
Fig. 5. Strict consensus tree presenting the position of Atoposauridae within Neosuchia, using the matrix of Pol et al. (2009). T. sympiestodon sp. nov. was not incorporated to the analysis.