Marmosa mexicana, Merriam, 1897

Flores, D. A., 2009, Phylogenetic Analyses Of Postcranial Skeletal Morphology In Didelphid Marsupials, Bulletin of the American Museum of Natural History 2009 (320), pp. 1-81 : 41-68

der, Marmosa mexicana , M. murina , M. rubra , and Micoureus .

TIBIA AND FIBULA

Character 97: Sesamoids in the articular area between tibia, fibula, and astragalus: (0) absent; (1) one sesamoid present. This character is modified from Horovitz and Sánchez-Villagra (2003: ch. 86), since in the sample there are no taxa with two sesamoids. One sesamoid in the area between tibia, fibula, and astragalus is present in Glironia , Metachirus , Philander , Didelphis virginiana , Tlacuatzin , Thylamys , Monodelphis , Marmosops incanus , M. parvidens , M. pinheiroi , and Cryptonanus unduaviensis . Polymorphism is exhibited only in Chironectes (coded {01}). No information is currently available about this character for Philander mcilhennyi , Marmosa rubra , Lutreolina , Thylamys pusillus , Lestodelphys , and Gracilinanus microtarsus (coded ‘‘?’’).

Character 98: Tibia length relative to femur length: (0) tibia shorter than femur; (1) tibia longer than or equal to femur. This character was described by Horovitz and Sánchez-Villagra (2003: ch. 89). Glironia is autapomorphic in this character, as the femur is longer than the tibia. Maynard Smith and Savage (1955) found similar proportions only in large mammals, such as Rhinoceros , Mastodon , and horses. In marsupials, this characteristic was also evidenced in some Australasian taxa with a diversity of habits, such as Phalanger , Pseudochirops , Phascolarctos , and Vombatus ( Horovitz and Sánchez-Villagra, 2003) , as well as the fossil Mayulestes ( Muizon, 1998) . In the remaining American marsupials, the tibia is longer than the femur (see table 3 in Hershkovitz, 1999).

Character 99: Tibia shape: (0) sigmoidshaped; (1) sigmoid shape present but not so marked ( fig. 28 ). Although the sigmoid shape of the tibia is the most common condition in didelphids, some taxa exhibit a notable sigmoid shape. This morphology was interpreted as plesiomorphic and is not restricted to didelphids ( Szalay and Sargis, 2001). The diaphysis starts in a sigmoid curvature at the level of the insertion of hamstring muscles, which suggests the possibility that this shape appears due to the pull of these extensors of the leg ( Argot, 2002). On the other hand, Lanyon (1980) demonstrated the biomechanical advantages of the curved shape on load transmission. The sigmoid shape of the tibia is also related to the asymmetrical condition of the femoral condyles in the knee joint. The lateral displacement of the load line is linked to the tibia shape and its function on load transmission ( Szalay and Sargis, 2001). In the sample, Chironectes , Tlacuatzin , and Gracilinanus microtarsus exhibit a remarkably sigmoid-shaped tibia.

Character 100: Tibial tuberosity developed: (0) absent; (1) present. The tendon of the m. quadriceps, a powerful extensor of the knee, inserts directly on the tibial tuberosity, as the patella is absent in didelphids. In some taxa with well-developed arboreal habits (e.g., Caluromys , Caluromysiops , Micoureus ), the tibial tuberosity is neither very evident or anteriorly expanded (personal obs.; Muizon and Argot, 2003). Only the terrestrial Metachirus exhibits this structure as notably developed, which is coherent with a more stabilized knee joint, useful for the saltatorial mode of locomotion. In slow-climbing didelphids, a more mobile and less stabilized knee joint is necessary, because of the range of hindlimb movements during arboreal displacement.

Character 101: Tibia, development of the posterior crest for the insertion of m. flexor digitorum tibialis: (0) absent; (1) present ( fig. 29 ). The crest is present in Hyladelphys , Marmosa mexicana , Marmosops noctivagus , M. impavidus , M. incanus , Lestodelphys , Gracilinanus , and Cryptonanus .

Character 102: Head of fibula notably developed craniocaudally: (0) absent; (1) present. An anteroposteriorly expanded head of the fibula is associated with the development of the area of insertion of the m. peroneus longus, which is involved in flexion of the tarsus ( Evans, 1993). Consequently, this is linked to arboreal habits, as the mentioned muscle inserts on the proximal portion of the first metatarsal ( Muizon and Argot, 2003; Argot, 2003a) and is associated with the opposability of the hallux. The head of the fibula is notably developed anteroposteriorly in Caluromys , Caluromysiops , Chironectes , Tlacuatzin , Marmosa , Marmosops , Thylamys (except T. macrurus ), Lestodelphys , Micoureus ,, Gracilinanus , and Cryptonanus .

CARPUS AND METACARPUS

Character 103: Lunate: (0) small (contacting only with scaphoid and cuneiform); (1) relatively large (contacting with scaphoid, cuneiform, magnum, and unciform) ( fig. 30 ). This character is modified from Horovitz and Sánchez-Villagra (2003: ch. 62) since in this sample there are no taxa with the lunate absent or fused with other elements. In most didelphids, the lunate is relatively large and in contact with other elements. However, this bone is notably smaller in Philander mcilhennyi , Tlacuatzin , Thylamys pusillus , Marmosa murina , Marmosops , Cryptonanus unduaviensis , and Gracilinanus agilis . No information is currently available about this character for Micoureus regina , M. paraguayanus , Lestodelphys , and Gracilinanus microtarsus (coded ‘‘?’’).

Character 104: Prepollex: (0) absent; (1) present. This character was described by Horovitz and Sánchez-Villagra (2003: ch. 63). Because the prepollex is the smallest element on the wrist, its loss is common during the cleaning process. The prepollex is present in most groups analyzed herein, except for Metachirus , Chironectes , and Thylamys pusillus , where this element seems to be absent. No information is currently available about this character for Lutreolina , Tlacuatzin , Micoureus paraguayanus , Lestodelphys , Marmosops impavidus , M. noctivagus , Cryptonanus unduaviensis , and Gracilinanus microtarsus (coded ‘‘?’’).

Character 105: Distolateral process of scaphoid separating lunate from magnum dorsally: (0) absent; (1) present ( fig. 30 ). This character was described by Horovitz and Sánchez-Villagra (2003: ch. 64). This process of the scaphoid is present in most groups analyzed, except for Glironia and Cryptonanus unduaviensis . No information is currently available about this character for Lutreolina , Tlacuatzin , Monodelphis brevicaudata , M. adusta , Micoureus paraguayanus , Lestodelphys , and Gracilinanus microtarsus (coded ‘‘?’’).

TARSUS

Character 106: Astragalus, angle between medial and lateral facets for tibia: (0) intermediate, between 90 ° and 180 °; (1) 180 °. This character is modified from Horovitz and Sánchez-Villagra (2003: ch. 94), since in the current sample there are no taxa with a 90 ° angle, and the character is treated as binary. According to Jenkins and McClearn (1984) and Szalay (1982, 1994), the medial and lateral astragalotibial facets form a broad and almost flat plane in didelphids. However, I observed a continuous variation in the angle formed by the facets, and all conditions were met in a single interval (90 ° –180 °). In terrestrial forms, as Metachirus or Monodelphis , the facets are better delimited, forming a sharper angle (personal obs.; Szalay, 1994). On the other hand, the angle between medial and lateral facets for the tibia was 180 ° only in some large opossums such as Philander , Lutreolina , Didelphis , and Chironectes . No information is currently available about this character for Monodelphis adusta , Micoureus paraguayanus , and Lestodelphys (coded ‘‘?’’).

Character 107: Astragalus, dimensions of astragalonavicular facet in distal view: (0) transversely wider; (1) dorsoventrally wider. This character was described by Horovitz and Sánchez-Villagra (2003: ch. 97). In the taxa analyzed here, only Metachirus exhibits the astragalonavicular facet dorsoventrally wider. No information is currently available about this character for Monodelphis adusta , Micoureus paraguayanus , and Lestodelphys (coded ‘‘?’’).

Character 108: Astragalus, ridge between medial and lateral astragalotibial facets: (0) absent; (1) present ( fig. 31 ). This character was described by Horovitz and Sánchez-Villagra (2003: ch. 107). The ridge between medial and lateral astragalotibial facets is present in large opossums, such as Glironia , Caluromys , Caluromysiops , Didelphis , Metachirus , Lutreolina , and Chironectes . Contrary to Horovitz and Sánchez-Villagra (2003), I coded 0 for Monodelphis , since I did not observe a ridge in this area of the astragalus. No information is currently available about this character for Monodelphis adusta , Micoureus paraguayanus , and Lestodelphys (coded ‘‘?’’).

Character 109: Astragalus, ridge between lateral astragalotibial facet and astragalofibular facet: (0) absent; (1) present ( fig. 31 ). This character was described by Horovitz and Sánchez-Villagra (2003: ch. 108). Similar to the anterior character, the ridge is present only in large opossums. No information is currently available about this character for Monodelphis adusta , Micoureus paraguayanus , and Lestodelphys (coded ‘‘?’’).

Character 110: Astragalus, contact between astragalonavicular and sustentacular facets: (0) absent; (1) present. This character was described by Horovitz and Sánchez-Villagra (2003: ch. 110). I found evidence of contact between both facets in most taxa, except Glironia , Caluromysiops , Metachirus , Chironectes ( fig. 31 ), and Thylamys pallidior . Individual variation is observed in Didelphis virginiana and D. marsupialis (coded {01}). No information is currently available about this character for Monodelphis adusta , Micoureus paraguayanus , and Lestodelphys (coded ‘‘?’’).

Character 111: Astragalus, continuous lower ankle joint pattern: (0) absent ( fig. 31 ); (1) present. This pattern results from the contact between posterior calcaneoastragalar and sustentacular facets, which is in relation to the absence of the sulcus astragali. I found evidence of the absence of this pattern in most didelphid groups analyzed here, except for Gracilinanus microtarsus and Hyladelphys , where a continuous lower ankle joint pattern is observed. Opposite to this, individual variation is observed in Marmosa mexicana . No information is currently available about this character for Monodelphis adusta , Micoureus paraguayanus , and Lestodelphys (coded ‘‘?’’).

Character 112: Astragalus, astragalonavicular facet vertically oriented and distal calcaneocuboid facet deep: (0) absent; (1) present. Most didelphid groups analyzed herein exhibit the astragalonavicular facet transversely oriented, except for the terrestrial Metachirus , where the astragalonavicular facet is almost vertically oriented, which is accompanied by an increase in depth of the distal calcaneocuboid facet of the calcaneus. Although the orientation of the astragalonavicular facet of Monodelphis seems to be somewhat vertical (as was also observed by Szalay, 1994: 191), its position reflects the condition exhibited by most of didelphids groups (i.e., condition 0). The particular vertical orientation of the astragalonavicular facet showed by Metachirus (and partially by Monodelphis ) suggests the increased functional importance of flexion-extension rather pronation-supination of the hindfoot. Polymorphism is evidenced in Philander opossum (coded {01}). No information is currently available about this character for Monodelphis adusta , Micoureus paraguayanus , and Lestodelphys (coded ‘‘?’’).

Character 113: Calcaneus, development of peroneal process of calcaneus: (0) small; (1) well developed ( fig. 32 ). Most of the didelphid group analyzed herein exhibits the peroneal process well developed, except for Metachirus and Monodelphis , where this process is smaller (see Szalay, 1994: figs. 8– 12 ). No information is currently available about this character for Monodelphis adusta , Micoureus paraguayanus , and Lestodelphys (coded ‘‘?’’).

Character 114: Calacaneus, calcaneal sustentaculum position: (0) subterminal; (1) on anterior tip ( fig. 32 ). The calcaneal sustentaculum is placed on the anterior tip of the calcaneum only in Chironectes . No informa-

TABLE 5

Tree Statistics from Parsimony Analyses of Different Didelphid Data Sets, Considering Polymorphism as Composite Entries (CO) and Transformation Series (TS)

tion is currently available about this character for Monodelphis adusta , Micoureus paraguayanus , and Lestodelphys (coded ‘‘?’’).

POSTCRANIAL DATA SET SUMMARY

The data set described above includes 114 postcranial characters, of which 106 (93%) are parsimony informative and 8 (7%) are autapomorphic. Sixty-six characters (58%) are binary, 47 (41%) describe ordered multistate (additive) transformations, and only 1 character (0.9%) describes unordered multistate (nonadditive) transformations (table 3). The data matrix (appendix 2) has 114 X 38 5 4332 cells, of which only 79 (2%) are scored as missing (‘‘?’’) and 44 (1%) are scored as inapplicable (‘‘–’’). The remaining 4209 matrix cells (97%) record organismal traits, with data completeness for individual terminal taxa ranging from 88 to 100% (table 4). Polymorphism was detected for 30 characters (2% of the total matrix cells), which showed intraspecific variation in the sample analyzed (table 2).

ANALYTIC RESULTS OF POSTCRANIAL CHARACTERS

A heuristic analysis of the postcranial data analyzed with CO polymorphic entries resulted in three equally most parsimonious trees (502 steps, CI 5 0.27, RI 5 0.62; table 5) whose highly resolved strict consensus is shown in figure 33 . The three caluromyines included (except the root) form a single clade in which Caluromys is recovered as a monophyletic group. Unlike results from the previous analysis of a different nonmolecular data set analyzed by Jansa and Voss (2005: fig. 1C ), the deep branch topology in the didelphine group is well resolved in the consensus topology, and some already recognized groups are recovered. Hyladelphys appears in basal and intermediate positions between caluromyines and didelphines, which is consistent with hypotheses attained by previous nonmolecular data, IRBP sequences, and combined analyses (table 6). Successively, Marmosa robinsoni and Cryptonanus unduaviensis appear as sister taxa of the remaining didelphines, which form a well-resolved topology. As in previous morphological and molecular analyses, Thylamys is recovered as a monophyletic group, arranged in the sequence ( T. pallidior - T. venustus ( T. macrurus ( T. pusillus ))). On the other hand, all species of Marmosops included in this analysis ( pinheiroi , parvidens , noctivagus , impavidus , and incanus ) form a monophyletic group (node M), which have been recovered

TABLE 6

Different Data Sets Supporting an Intermediate Position of Hyladephys between Caluromyinae and Didelphinae .

Postcranial

DMP-1

RAG-1

Jansa and Voss, 2005 (combined) b

c

Gruber et al., 2007 (combined)

a Morphological data set including postcranial evidence described here, and the nonmolecular characters defined by Voss and Jansa (2003).

b

Data set

combining nonmolecular characters, IRBP, and DMP-1 sequences (postcranial excluded).

c Data set combining nonmolecular characters, IRBP, DMP-1, and RAG-1 sequences (postcranial excluded).

d Data set combining nonmolecular characters, IRBP, DMP-1, and RAG-1 sequences (postcranial excluded, RAG-1 third position eliminated).

e

Data set combining nonmolecular characters, IRBP, DMP-1, and RAG-1 sequences (postcranial included).

f Data set combining nonmolecular characters, IRBP, and DMP-1 (postcranial included, RAG-1 excluded).

in previous results with other kinds of evidence. In this group, Marmosops impavidus and M. noctivagus are sister species, as well as M. parvidens and M. pinheiroi . Lastly, Lestodelphys is placed as sister of Marmosops . Up in the tree, Tlacuatzin appears as sister of the group Gracilinanus agilis - G. microtarsus , and Micoureus is recovered as a monophyletic group in the sequence ( M. demerarae - M. paraguayanus ) M. regina )). The three remaining species of marmosa ( M. mexicana , M. murina , and M. rubra ) are successively arranged. Finally, the species of Monodelphis , resolved in the sequence ( M. brevicaudata ( M. adusta + M. theresa )) are recovered as sister group of the large opossums (node O). Although in general the large opossums ( Didelphis , Philander , Metachirus , Chironectes , and Lutreolina ; node G) form an unresolved clade, Didelphis is recovered as a monophyletic group in the sequence ( Didelphis virginiana ( D. marsupialis + D. albiventris ), as well as Philander in the sequence ( P. mcilhennyi ( P. opossum - P. frenatus )). Both genera form a monophyletic group with Lutreolina (node E).

The heuristic search of the postcranial data set with polymorphic entries analyzed with TS resulted in seven most parsimonious trees of 540 steps (CI 5 0.25, RI 5 0.59). The strict consensus ( fig. 34 ) resulted in a less resolved topology than CO analysis, where Hyladelphys appears again in an intermediate position. Under this criterion, the species of Marmosa , Gracilinanus , Cryptonanus unduaviensis , and Tlacuatzin are arranged in a basal polytomy. The monophyletic Monodelphis is resolved in the sequence ( M. adusta (M. brevicaudata-M. theresa ), with Thylamys macrurus in basal position, forming a trichotomy with T. pusillus and the T. pallidior-T. venustus group. The last clade is the more speciose one and includes the monophyletic Marmosops (node M; arranged in the same sequence as in the CO analysis), Lestodelphys , and the large opossums (node G). Unexpectedly, the position of the Patagonian Lestodelphys is not close to some group of +

jackknife frequencies (cutoff value 5 50%). Heavy lines denote branches with a decay index of $4. Outgroup taxa are indicated with asterisks, and alphabetic labels indicate didelphine clades discussed in the text.

mouse opossums, but as sister of large opossums. In the last group, the topology is almost similar to the one obtained in CO analysis, but Philander is paraphyletic and Metachirus and Chironectes are sister groups.

NONMOLECULAR EVIDENCE ON DIDELPHID PHYLOGENY: THE INCLUSION OF POSTCRANIAL CHARACTERS

In table 3 are summarized and compared basic statistics for the nonomolecular data set published by Jansa and Voss (2005), the postcranial data presented herein, and the combination of both morphological data sets. Note that both nonmolecular data sets are proportionally similar regarding some basic statistics, that is, as percentages of missing and inapplicable data, parsimony informative characters, and autapomorphies. However, the percentage of ordered multistate characters is higher in the postcranial data set presented here. Combining both nonmolecular data sets analyzed with CO polymorphic entries, I recovered 120 most parsimonious trees (723 steps, CI 5 0.33, RI 5 0.68) whose strict consensus is shown in figure 35. In this topology, the caluromyines ( Caluromys + Caluromysiops ) form a monophyletic group, with both species of Caluromys as sister taxa. The intermediate position of Hyladelphys between caluromyines and didelphines appears here again (table 6). Up in the tree, the topology appears as a deep dichotomy, which includes the remaining didelphids arranged basically in two monophyletic clades. In the first group, Tlacuatzin is sister taxon of the group Gracilinanus- Cryptonanus-Chacodelphys-Thylamys-Lestodelphys (node B). In this clade, Gracilinanus and Cryptonanus are respectively recovered as monphyletic groups, and Lestodelphys is nested in the paraphyletic Thylamys (node A) in an unresolved trichotomy with T. venustus and T. pallidior . T. pusillus and T. macrurus are successively basal to this group. The second clade is formed by a basal polytomy including Micoureus regina , Marmosa murina , M. lepida , M. mexicana , and M. robinsoni ; the group resolved as the sequence Marmosa rubra ( Micoureus paraguayanus - M. demerarae )); and the group conformed by the monophyletic Marmosops (node M) and the clade formed by the monophyletic Monodelphis and the large opossums as sister groups (node O). The species of Marmosops are split in two sister clades, one formed by ( Marmosops incanus ( M. impavidus - M. noctivagus )), and another including M. parvidensM. pinheiroi . The four species of Monodelphis considered in the morphological analysis are clustered in a well-supported monophyletic clade where M. brevicaudata and M. emiliae form a pair and M. adusta and M. theresa complete a trichotomy. The clade formed by the large opossums (node G) includes the monophyletic Didelphis , Philan- der (sister of the monotypic Lutreolina , node E), and Chironectes and Metachirus successively arranged in increasingly basal positions (nodes F and G, respectively). The three species of Didelphis form an unresolved polytomy, and the species of Philander are resolved as ( P. mcilhennyi (P. opossum-P. frenatus )).

A heuristic search of all morphological data sets with polymorphic entries analyzed as TS resulted in six most parsimonious trees of 926 steps (CI 5 0.29, RI 5 0.65). The strict consensus ( fig. 36) results in a barely less resolved topology than the CO analysis (table 5). Deep in the tree, Hyladelphys is placed in its typical intermediate position, and Marmosa lepida , M. rubra , and M. robinsoni are successively arranged basal to the remaining didelphids, which are split in two groups: the clade conformed by the sister species Micoureus paraguayanus - M. demerarae , and a polytomy including the remaining didelphine species. Seven natural groups can be recognized in this unresolved polytomy. Three out of seven monophyletic groups consist of pairs of species: Gracilinanus agilis-G. microtarsus , G. aceramarcae-G. emiliae , and Cryptonanus unduaviensis-C. chacoensis . The four remaining natural groups include the complex Thylamys-Lestodelphys (node A), the monophyletic Marmosops (node M) and Monodelphis , and the large opossums (node G) in the same topology as in the CO analysis. However, in the TS analysis Monodelphis adusta and M. theresa are sister species, Philander is recovered as an unresolved trichotomy, and Didelphis is recovered in the arrangenment ( Didelphis virginiana (D. albiventris-D. marsupialis ).

COMBINED ANALYSIS

Parsimony analysis combining the new postcranial evidence with the nonmolecular characters previously defined, plus IRBP, DMP-1, and RAG-1 sequences analyzed with CO polymorphic entries, resulted in two most parsimonious trees (6203 steps, CI 5 0.61, RI 5 0.80). The strict consensus topology ( fig. 37) resembles, in some positions, the didelphid relationships obtained in the combined evidence by Jansa and Voss (2005: fig. 1D ) and Gruber et al. (2007: fig. 2 ), although the deep branch topology differs remarkably. The relationships within the outgroup are similar to the one obtained with nonmolecular evidence (i.e., Caluromys and Caluromysiops forming a monophyletic group, and both species of Caluromys as sister taxa). Hyladelphys kept its intermediate position between caluromyines and didelphines. From this point of the tree, two traditionally recognized groups in the didelphine subfamily are recovered as monophyletic groups: the 2 n 5 22 large opossums (node F), and the mouse opossums (node L, although including Metachirus ). Among the large opossums, the two polytypic living genera ( Didelphis and Philander ) are recovered as monophyletic and sister groups (node D). Lutreolina and Chironectes are successively arranged in increasingly basal positions (nodes E and F, respectively). The remaining taxa are split into three diverse clades. The first one contains the monophyletic Monodelphis , resolved in the sequence ( M. emiliae ( Monodelphis theresa (M. adusta-M. brevicaudata ))), and Chacodelphys as sister taxa. More nested in the tree, the complex Micoureus-Marmosa (node I) is resolved on the monophyletic Micoureus in the sequence ( Micoureus regina (M. demerarae-M. paraguayanus )), and Marmosa as paraphyletic. Marmosa lepida and M. murina are successively basal to Micoureus . On the other hand, this group is sister of the group ( Marmosa rubra (M. mexicana-M. robinsoni )). The last clade (node C) includes the paraphyletic Marmosops , Metachirus , and the complex Thylamys-Lestodelphys (node A), sister of Gracilinanus-Cryptonanus group (node B). Although Marmosops is recovered as monophyletic in other analyses (e.g., nonmolecular and previous combined analyses), the complete evidence presented here recovered the species in two different clades. Three species are resolved in the grouping ( Marmosops incanus (M. impavidus-M. noctivagus ), and the sister taxa M. parvidens and M. pinheiroi are clustered with Metachirus . The species of Thylamys are clustered in a monophyletic group as an unresolved polytomy, although T. venustus is placed basal in relation to the remaining species of the genus. As in other results, Lestodelphys is sister of Thylamys . Both species of Cryptonanus ( chacoensis and unduaviensis ) are recovered as sister taxa, clustered with the monophyletic Gracilinanus (node N), which shows the sequence ( Gracilinanus emiliae ( G. aceramarcae (G. agilis-G. microtarsus ))).

A heuristic search of combined data sets with polymorphic entries analyzed as TS resulted in four most parsimonious trees (6464 steps, CI 5 0.59, RI 5 0.79). The strict consensus ( fig. 38) under this parameter is notably less resolved than the topology obtained in CO analysis (table 5), and some differences can be detected. In this scheme, the position of the 2 n 5 22 large opossums (node F), which under CO treatment is a sister group of the mouse opossums ( fig. 37), is clustered with most of the mouse opossums in a polytomy where the nodes B and C (observed in CO analysis) are not recovered ( figs. 37, 38). Another minor difference with CO analysis is the inverted position of Monodelphis theresa and M. emiliae .

NODAL SUPPORT

Bremer support values obtained from the postcranial evidence are, in general terms, low. For instance, in the CO analysis ( fig. 33 ), 21 nodes (68% of resolved nodes) collapse in trees that are one step longer, 4 additional nodes (13%) collapse in trees that are two steps longer, whereas 2 more nodes (6%) collapse in trees that are three steps longer. Only four nodes (13% of the total) have a decay index $4 (table 5). Although the TS analysis has less resolved ingroup nodes in the consensus tree ( fig. 34 ; table 5), the amount of well-supported nodes is higher (table 5). Five nodes (28%) collapse in trees that are one step longer, five additional nodes (28%) collapse in trees two steps longer, and two nodes (11%) collapse in trees three steps longer. Lastly, the remaining six nodes (14%) have a decay index $4 (table 5). Resampling values were also low: in CO analysis 22 nodes (71%) have jackknife values below 50%, and 6 nodes (19%) have jackknife values between 50% and 85%, while the remaining 3 nodes (9%) have jackknife values higher than 85%. In the TS postcranial analysis 11 nodes (61%) have jackknife values below 50%, 4 nodes (22%) have jackknife values between 50% and 85%, while only 2 nodes (11%) have jackknife values higher than 85%.

Nodal support values for morphology of the total tree are slightly higher. In the CO analysis ( fig. 35), 17 nodes (59%) collapse in trees that are one step longer, 2 additional nodes (7%) collapse in trees two steps longer, 1 node (3%) collapses in trees that are three steps longer, and 9 nodes (31%) have a decay index $4 (table 5). The consensus tree of the TS analysis ( fig. 36) is less resolved than the CO analysis, although in the TS analysis there is one more well-supported node than for the CO results. Twelve nodes (44%) collapse in trees that are one step longer, four additional nodes (15%) collapse in trees two steps longer, and only one node (4%) collapses in trees three steps longer. The remaining 10 nodes (37%) have a decay index $4 (table 5). Similarly, resampling values are slightly higher in morphology of the total tree: in the CO analysis 16 nodes (55%) have jackknife values below 50%, 8 nodes (27%) have jackknife values between 50% and 85%, while the remaining 4 nodes (14%) have jackknife values higher than 85%. In the TS morphology/total analysis, 15 nodes (55%) have jackknife values below 50%, 8 nodes (29%) have jackknife values between 50% and 85%, and the remaining 4 nodes (15%) have jackknife values higher than 85%.

Nodal support values for the combined CO analysis indicate that most of 38 resolved ingroup nodes in the consensus tree ( fig. 37) are moderately well supported. Only eight nodes (21%) collapse in trees one step longer, one additional node (3%) collapses in trees two steps longer, and three more nodes (8%) collapse in trees three steps longer. The remaining 26 nodes (57%) have a decay index $4 (table 5). The consensus tree of the TS analysis ( fig. 38) shows less resolved ingroup nodes than for the CO analysis. Only three nodes (10%) collapse in trees one step longer, two additional nodes (6%) collapse in trees two steps longer, and six more nodes (19%) collapse in trees three steps longer. The remaining 20 nodes (64%) have a decay index $4 (Table 5). Similarly, resampling values were relatively high both in the CO and TS combined analyses. In the CO analysis 10 nodes (26%) have jackknife values below 50%, 6 nodes (16%) have jackknife values between 50% and 85%, and the remaining 22 nodes (58%) have jackknife values higher than 85%. On the other hand, in the TS combined analysis 16 nodes (52%) have jackknife values below 50%, 8 nodes (26%) have jackknife values between 50% and 85%, and the remaining 7 nodes (23%) have jackknife values higher than 85%.

DISCUSSION

EFFECTS OF DIFFERENT CODINGS OF POLYMORPHIC DATA

The kind of treatment of polymorphic characters may have a significant impact on phylogenetic analyses. Different methods for dealing with polymorphism may lead to very different estimations of phylogeny, even when relationships are strongly supported by one or more methods ( Wiens, 1999). The abundance and impact of polymorphic characters are especially clear for closely related species, but the application of different criteria for analyzing polymorphic data may affect higher level relationships as well. In this sense, although different codings of polymorphic postcranial characters in didelphids produced topologies in general not contradictory, some discrepancies were evident. The principal difference was the loss of resolution of the TS analysis compared to the CO analysis, with the consistency index and retention index being slightly higher in the CO analysis (table 5). In the postcranial results, the nodes weakly supported in the CO analysis collapsed in the TS topology ( figs. 33 , 34 ). However, more resolution is expected for the additional phylogenetic information with TS coding (Mabee and Humpries, 1993). Different codings caused little impact on the nodal support, since in both topologies the number of nodes with high decay index ($4) was similar in all analyses, except in the combined analysis considering RAG-1 (table 5), where the CO coding had more well-supported nodes. An inverted bias on nodal support and resolution was observed in a combined analysis in Oryzomyini rodents performed by Weksler (2006). In the case of didelphids, despite the differences in resolution in the combined total evidence, most relationships obtained in the TS analysis are not contradicted by the CO analysis, except for two punctual cases: the inverted position of Monodelphis emiliae and M. adusta , and the basal position of the Marmosa-Micoureus group (node I) in TS analysis (being instead sister of node C in the CO analysis, figs. 37, 38).

In the total morphology analysis, the effects of different codings on polymorphic data were similar with respect to the postcranial-only data set: the topology obtained with the CO coding analysis is notably better resolved than the TS coding analysis ( figs. 35, 36), although the values of Bremer support are in general rather similar (table 5). Contrary to the congruence observed in the topology obtained by both kinds of coding in the potscranial-only analysis, the low resolved topology observed with the TS coding in the total morphology analysis differs considerably in some positions with regard to the CO coding analysis. In the TS coding total morphology analysis, the change of position of Marmosa lepida and M. rubra is unexpected for different reasons. In the first case, all postcranial characters are missing, and the remaining morphological characters do not show polymorphic entries (see Voss and Jansa, 2003: appendix 5). In Marmosa rubra , although 95% of the postcranial characters were scored (table 4), there were no polymorphisms since the sample consisted of only one specimen (see appendix 1), and the remaining morphological characters do not show any polymorphic entries (see Voss and Jansa, 2003: appendix 5). Lastly, the change observed in Marmosa robinsoni is perhaps a consequence of the high polymorphism present in postcranial morphology (table 2; appendix 2). Other differences can be noted between both kinds of coding, such as the monophyletic condition of Gracilinanus in CO coding total morphology analysis, the changing position of Thylamys macrurus and T. pusillus , and the relationship among the species of the monophyletic Didelphis and Philander . Despite the differences in deep branch topology, the well-supported clades in CO coding total morphology analysis were all recovered in TS coding morphology-total analysis as well (i.e., the monophyly of Monodelphis , Philan- der, Marmosops , and Didelphis , the relationship of large opossums [nodes E, F, and G], and the relationship of Thylamys-Lestodelphys [node A]). Similarly to postcranial-only analysis, in the case of total morphology, the TS coding analysis seems not to contribute to the retention of more phylogenetic information.

As described in the results, the topologies obtained including the genetic evidence (i.e., combined analysis) are in general highly resolved and better supported than the morphology-only analyses. Despite the fact that the topologies from both kind of codings of polymorphic data are significantly congruent in some aspects, the clustering of the mouse opossums and the better resolution applying CO coding analysis ( fig. 37) are interesting. The mouse opossums are the most speciose group in the didelphid living radiation, and the genera currently recognized (sensu Gardner, 2005) are not always recovered as natural groups in the cladistic context. Contrasting with the large opossums, which were considered as a monophyletic group based in a diverse array of previous evidence, the mouse opossums were partially supported only in a morphological framework (e.g. Creighton, 1984; Reig. et al., 1987; Goin, 1995; Flores, 2003). Here, the addition of postcranial evidence to the previous nonmolecular and genetic evidence causes the moderately supported clustering of the mouse opossums in a group just in CO coding analysis (although Metachirus is nested in the group when RAG-1 is included; fig. 37; appendix 3). However, the group is paraphyletic in all previous nonmolecular and molecular (IRBP, DMP-1, RAG-1) analyses ( Kirsch and Palma, 1995; Patton et al., 1996; Jansa and Voss, 2000, 2005; Voss and Jansa, 2003; Voss et al., 2005; Jansa et al., 2006; Gruber et al., 2007), and even in all remaining analyses of this report. In this sense, applying different criteria for the treatment of morphological polymorphic data, the relationships and monophyly of the mouse opossums are strongly affected.

EFFECT OF THE INCLUSION OF POSTCRANIAL CHARACTERS IN PREVIOUS NONMOLECULAR

AND COMBINED HYPOTHESES

NONMOLECULAR HYPOTHESES: Comparing the topology of nonmolecular evidence illustrated by Jansa and Voss (2005: fig. 1C ), the inclusion of the postcranial data set causes considerable changes and better resolution in topology and support values ( figs. 35, 36; tables 5–7). As mentioned above, the intermediate position of Hyladelphys is consistent in postcranial-only and combined analyses (table 6). The trichotomy conformed by the monophyletic Didelphis and Philander , as well as the monotypic Lutreolina (node E), is kept both in CO and TS analyses of total morphology evidence, as well as the monophyly of Monodelphis and its sister relationship with large opossums (node O in the CO analysis, fig. 35), the relationship of Lestodelphys-Thylamys (node A), and the monophyly of Marmosops (node M), Gracilinanus , and Cryptonanus (although the TS coding analysis produces the rupture of

TABLE 7 Nodes Recovered under Different Analyses and Values of Absolute Bremer Support and Jackknife Frequencies a

Node 1 2 3 4 5 6 7 8

A 5/61 4/60 4/83 3/89 4/83 5/86

B 1/,50 1/,50 1/,50 2/,50

C 1/,50 b 1/,50 2/,50

D.7/99.7/100 4/73 4/78

E 1/,50 3/,50 6/,50.7/100.7/99.7/97.7/99

F.7/66 6/84.7/100.7/100.7/100.7/100

G.7/98.7/98 6/83 7/99 5/99 7/99

H 2/,50

I 7/100 6/100 7/99.7/94

J 1/,50 2/,50

L 1/,50 b 4/75 c

M 4/50 4/50 5/71 5/67 5/97 6/98

N 2/60 1/50 1/,50 2/,50

O 1/,50 1/,50

a Letters indicate the nodes labeled as in figures 33–40 . Each column corresponds to different analyses described and compared in the text. 1, Postcranial data, CO coding for polymorphic entries; 2, postcranial data, TS coding; 3, all nonmolecular data set (including postcranium), CO coding; 4, all nonmolecular dataset (including postcranium), TS coding; 5, combined data (including postcranium), CO coding; 6, combined data (including postcranium), TS coding; 7, combined data (including postcranium), RAG-1 eliminated, CO coding; 8, combined data (including postcranium), RAG-1 eliminated, TS coding.

b Including Metachirus .

c Excluding Metachirus .

the monophyly of Gracilinanus , fig. 36). However, some alterations in the topology can be detected by including the postcranial characters. In the CO coding analysis ( fig. 35), both species of Cryptonanus ( chacoensis and unduaviensis ) appear as a monophyletic group sister to the clade Chacodelphys -node A ( Thylamys - Lestodelphys ), and the species of Gracilinanus are also clustered as monophyletic basal in node B. The monophyly of node B in the CO coding analysis ( fig. 35) is recovered by including the postcranial evidence. In this sense, consideration of the postcranial characters on the morphological evidence previously defined ( Jansa and Voss, 2005) is consistent with the genetic evidence, since clustering of the monophyletic genera Gracilinanus and Cryptonanus with node A was obtained using only genetic and combined evidence ( Jansa and Voss, 2005; Gruber et al., 2007). The position of Metachirus is also altered when the postcranial evidence is included, since this monotypic taxon is basal to the large 2 n 5 22 opossums in the total morphology analysis (node G in figs. 35, 36), whereas it is located as the sister taxon of the group consisting of 2 n 5 22 opossums- Monodelphis in the morphological analysis excluding the postcranial characters (see Jansa and Voss, 2005: fig. 1C ; table 7). Although Metachirus shows a particular mode of locomotion and some postcranial autapomorphies (see the character descriptions and appendix 2), its close relationship with the 2 n 5 22 large opossums (node G in figs. 35, 36) is well supported by postcranial morphology (table 7). Similarly, the inclusion of postcranial evidence notably affects the position of the recently redescribed Tlacuatzin canescens . This taxon appears in a basal polytomy together with some species of Marmosa and Micoureus in the morphological evidence of Jansa and Voss (2005: fig. 1C ). A similar position is obtained when the TS coding criterion is applied for polymorphic characters in the total morphology analysis ( fig. 36), but when the postcranial data set is included in the CO coding analysis, this taxon is placed as sister to node B, although with low support ( fig. 35).

The relationship Thylamys-Lestodelphys (node A) and the paraphyly of Thylamys are also kept when postcranial evidence is considered (table 7), but the position of this clade in the total morphology consensus tree is remarkably different. In the topology obtained by Jansa and Voss (2005: fig. 1C ) this clade is located as sister of the Monodelphis -large opossums group (node O), whereas when including the postcranial evidence this group appears as forming part of node B in the CO coding analysis ( fig. 35). The monophyly of Marmosops (node M) is also kept, but its position is different when the postcranial characters are considered. When omitting postcranial evidence, its position is basal in the clade that includes the large opossums, Monodelphis (node O), and the complex Thylamys-Lestodelphys- Chacodelphys ( Jansa and Voss, 2005: fig. 1C ). Nonetheless, in the CO coding total morphology analysis this monophyletic genus appears as sister to node O (node G and Monodelphis , fig. 35).

Although the parsimony analysis considering only the postcranial data set resulted in a well-resolved strict consensus but with lower consistency and retention indices compared to the total morphology analysis (i.e., the postcranial characters defined here and the 71 nonmolecular characters defined by Voss and Jansa, 2003; see tables 5, 7; figs. 33 , 35), several relationships (which are recovered in both analyses separately) are kept with the inclusion of postcranial evidence (tables 6, 7): the intermediate position of Hyladelphys between caluromyines and didelphines is recovered in all morphological analyses, as is the monophyly of Monodelphis , Marmosops (node M), and the large opossums ( Didelphis , Philander , Lutreolina , Chironectes , and Metachirus [node G]). Several nodes recovered by including genetic evidence ( Jansa and Voss, 2005; Jansa et al., 2006; Gruber et al., 2007) were not found based on morphological evidence omitting postcranial characters. As mentioned above, the inclusion of postcranial evidence also produces better resolution. Nodes B and G ( fig. 35) specifically are also recognized in a more inclusive morphological data set, with the postcranial characters concatenated to the previously defined nonmolecular evidence (table 7).

HYPOTHESES BASED ON COMBINED EVIDENCE: The inclusion of postcranial characters in the combined data set performed by Jansa and Voss (2005), Jansa et al. (2006), and Gruber et al. (2007) causes some interesting alterations in the resulting hypotheses (table 7). The general topology ( figs. 37, 38) shows several congruences with diverse aspects of molecular (IRBP, DMP-1) and nonmolecular evidence (table 8). The recent inclusion of RAG-1 sequences ( Gruber et al., 2007: figs. 1 , 2 ) resulted in the well-supported clustering of distantly related clades based on profuse evidence: clade B (Thylamys-Cryptonanus-Gracilinanus) as sister of clade I (Marmosa-Micoureus complex). However, the addition of postcranial evidence to the supermatrix analyzed by Gruber et al. (2007) did not recover this apparently spurious clade ( figs. 37, 38). The homoplasy caused by the convergence in CG content on the third position in RAG-1 sequences ( Gruber et al., 2007) is hidden by the effect of the phylogenetic information coming from the postcranial evidence, independent of the treatment applied to polymorphic characters.

Another example of the influence of postcranial data on previously combined data sets is the position of Hyladelphys (table 6). As mentioned before, this monotypic and recently recognized genus was located in a basal position, intermediate to didelphines and caluromyines, based on profuse molecular and morphological support ( Jansa and Voss, 2005). The recent inclusion of RAG-1 sequences alters the typical phylogenetic position of this taxon, being sister to the already mentioned and apparently spurious clade (B + I in Gruber et al., 2007: fig. 2 ), although with low support values. The inclusion of postcranial characters in the combined data set (even including the RAG-1 sequences and its third positions) replaces the typical intermediate position of Hyladephys ( figs. 37, 38) between caluromyines and didelphines.

Although the combined evidence incorporating the postcranial characters shows some relationships congruent with earlier evidence (tables 7, 8), the position of the cursosaltatorial Metachirus and the recently described Tlacuatzin are highly affected, as was also demonstrated in total morphology analyses. In previous molecular and combined analyses (table 8), Metachirus was within the clade conformed by the large opossums ( Didelphis , Philander , Lutreolina , and Chironectes ; node G), usually in a basal position (see Gruber et al., 2007; Jansa and Voss, 2005; Jansa et al., 2006). Nonetheless, the addition of postcranial characters to the combined evidence relates Metachirus with Marmosops parvidens-M. pinheiroi , nested in the speciose clade labeled C in figures 37 and 38 (table 7). The inclusion of Metachirus in that group is unlikely in view of the huge amount of phylogenetic information (even postcranial) relating this species to other large opossums. The inclusion of postcranial characters in previously combined evidence (considering also RAG-1 sequences, see appendix 3) apparently affects the presumably true phylogenetic position of Metachirus .

Similarly, consideration of postcranial morphology in combined analyses notably alters the position of Tlacuatzin canescens , which is sister of Monodelphis in all recent combined evidence (see Jansa and Voss, 2005: fig. 1D ; Jansa et al., 2006: fig. 5 ; Gruber et al., 2007: fig. 2 ). By adding the new data set, Tlacuatzin became sister to the Marmosa-Micoureus complex (node I), a relationship not found in previous analyses. This topology is highly interesting because T. canescens was traditionally included in the nonmonophyletic genus Marmosa . However, even when the decay index of this clade is high in the TS and CO coding combined analyses, the jackknife values in both analyses are notably lower ( figs. 37, 38).

As described in the results, the topology obtained by including genetic evidence in the CO coding analysis is highly resolved and better supported than the morphology-only analyses (tables 7, 8). However, even if the topologies from both kinds of coding of polymorphic characters are considerably congruent in some aspects ( figs. 37, 38; table 7), the clustering of the mouse opossums recovered with the CO coding analysis is remarkable (node L; table 7; although Metachirus is nested in this clade). Here, the addition of postcranial evidence to previous nonmolecular and genetic evidence causes the clustering of the mouse opossums only in the CO coding analysis, although with most of the support values being low ( fig. 37; table 7), which were also paraphyletic in all previous nonmolecular and molecular (table 8) analyses (e.g. Voss and Jansa, 2003; Jansa and Voss, 2000, 2005; Voss et al., 2005; Jansa et al., 2006; Kirsch and Palma, 1995; Patton et al., 1996), and even in all remaining analyses in this report (table 7).

As discussed above, the addition of postcranial characters recovered some relationships already supported by previous combined analyses (tables 7, 8), but the resulting topologies are contradictory in some positions. Despite the mentioned incongruences in the positions of Metachirus and Tlacuatzin , the node labeled O, which includes Monodelphis as sister to the large 2 n 5 22 opossums, is recovered when considering the postcranial evidence, but it has not been obseved before in any previous molecular or combined analyses (table 8), except in two analyses in this report (table 7), and in the partial morphological evidence from Jansa and Voss (2005: fig 1C ). However, this node is recovered with low support ( fig. 38; table 7). Another clear difference concerns the already recognized nodes J and H (Voss and Jansa, 2005: fig. 1D ; Gruber et al., 2007: fig. 6A–C ; Jansa et al., 2006: fig. 4B ; table 8), which are well-supported sister groups in previous combined evidence, although this relationship is broken by inclusion of the RAG-1 sequence ( figs. 37, 38; see Gruber et al., 2007: fig. 2 ).

In view of the high convergence of CG content on the third positions, a more congruent topology was obtained when the third positions of RAG-1 were experimentally eliminated (table 8: Gruber et al., 2007). When omitting the RAG-1 sequence from the combined data set presented here, the relationships obtained are highly congruent with previous evidence recently published, recovering almost all nodes already recognized, even in the TS coding analysis ( figs. 39, 40; tables 7, 8). Independent of deep branch differences of the partitioned evidence (depending on the CO or TS treatment of polymorphic characters), Metachirus recovers its traditional position as sister of the remaining 2 n 5 22 large opossums (node G), and Tlacuatzin is relocated as sister to Monodelphis . The topology obtained with the TS partitioned analysis ( fig. 40) is basically similar to the one obtained in the combined analysis by Jansa and Voss (2005: fig. 1D ) and some combined topologies obtained by Gruber et al. (2007: fig. 6A–C ) and Jansa et al. (2006: fig. 5 ). Most nodes already documented are recovered: nodes H, J, C, G, B, and I (table 7). However, as in the complete combined data set, the application of different criteria for treatment of polymorphic data in the partitioned analysis strongly affects the relationships and phyletic condition of the mouse opossums, since in the CO coding analysis node H is not recovered, and the mouse opossums (node L) are monophyletic although moderately supported ( fig. 39; table 7). In this sense, the inclusion of postcranial evidence concatenat- ed to the previous combined data set has notable influence on deep branch topology, depending of the mode of coding polymorphic characters. The partitioned TS combined analysis (excluding RAG-1; fig. 40) recovers all of the topologies already observed (table 7), which indicates that the inclusion of postcranial characters does not contradict the relationships obtained with profuse previous combined evidence (table 8).

THE POSTCRANIAL ANATOMY AS EVIDENCE OF DIDELPHID RELATIONSHIPS AND POSTCRANIAL

SYNAPOMORPHIES IN DIDELPHINAE

The study of diversity of the cranioskeletal system is one of the most critical areas of research for the understanding of various aspects of behavior and ecological morphology, particularly locomotion and feeding habits in marsupials ( Szalay, 1994). Even if the most common conception points toward the close evolutionary relationship between craniodental anatomy and feeding demands, the movements linked to locomotion, posture, and other behavioral patterns are particularly dependent on the musculoskeletal system. Moreover, the skeletal morphology of the most abundantly represented Neogene forms, or extant marsupials, has not been adequately studied from the perspective of evolutionary morphology ( Szalay, 1994). Because the skeletal structure is highly correlated with posture, habits, and locomotion, several patterns (both in the axial and appendicular skeleton) have been associated with an apparent functionality. Several recent papers (e.g. Argot 2001, 2002, 2003a, 2003b, 2004a, 2004b; Szalay, 1994; Sears, 2004; Szalay and Sargis, 2001; Weisbecker and Sánchez-Villagra, 2006) have revised the postcranial morphological patterns in marsupials and their associated forms-functions on the metatherian postcranium.

The didelphid relationships have been examined with different kinds of data (molecular and morphological), but the application of postcranial characters has never been considered as evidence of didelphid phylogeny in a cladistic frame on a denser taxon sampling. Several postcranial characters defined by Horovitz and Sánchez-Villagra (2003) are highly variable within the didelphid taxa included in this report, although the postcranial unambiguous synapomorphies proposed for Didelphidae in the cited work are also evidenced in all didelphid taxonomic samples considered herein.

Some postcranial topologies recovered in this study ( figs. 33 , 34 ) are clearly congruent and noncontroversial with clades already recognized based on other evidence (morphological and molecular; tables 6–8). The postcranial morphology has showed inherent phylogenetic information in recovering some traditionally recognized relationships and monophyletic groups (tables 7, 8; appendix 3). However, the postcranial evidence in didelphids also produces some unusual relationships, a product of the convergence of characters strongly associated with form-function patterns. In other words, several postcranial characters exhibited the same condition in taxa with analogous locomotion and/or posture patterns, which are clearly linked to specific form-function. For instance, the particular vertical orientation of the astragalonavicular facet of the astragalus (ch. 112[1]) and the depth of the distal calcaneocuboid facet of the calcaneus showed by Metachirus and Monodelphis suggest an increased functional importance of flexion-extension of the hindfoot in both terrestrials but not closely related taxa ( Szalay, 1994). This is a hint that some atypical relationships obtained are possibly caused by morphological constraints of form-function in some structures, which can hide the true relationships obtained from other kinds of evidence when the postcranial data set is analyzed separately. This homoplasy is presumably the cause of the low values of consistency index and the poor resolution of the strict consensus (mainly in the TS coding postcranial-only topology, fig. 34 ) compared to the more resolved trees obtained from the total morphology ( fig. 35) and combined evidence ( figs. 37–40; table 5).

In the topologies based on the postcranial-only data set, the monophyly of several traditionally recognized polytypic genera (sensu Gardner, 2005), such as Gracilinanus , Marmosa , and Cryptonanus , are not recovered, and some relationships had not been recovered in previous analyses ( figs. 33 , 34 ). For instance, the monophyly between the partially terrestrial Thylamys and the highly terrestrial Monodelphis is recovered in the strict consensus of the TS coding postcranial analysis ( fig. 34 ), which is supported by an array of characters clearly related to specific capacities of movements: position of the vertebra where the accessory process is differentiated from the transverse process on T7 (implying an anterior point restricting lateral flexibility; ch. 30[1]), or absence of a longitudinal groove on the lateral surface of the ulna for insertion of Mm. abductor pollicis longus and anconeus (muscles well developed in arboreal forms [ Argot, 2001]; ch. 77[0]). In all other topologies obtained here (i.e., total morphology and combined, figs. 35–40), as well as in previous hypotheses ( Jansa and Voss, 2000, 2005; Gruber et al., 2007; Voss and Jansa, 2003; Jansa et al., 2006; Reig et al., 1987; Kirsch and Palma, 1995), Monodelphis and Thylamys are clustered in clearly distant clades. Only in the pioneer nonmolecular analysis of Creighton (1984) are both taxa clustered in a monophyletic clade. Another interesting example of a functional component in postcranial characters is the unusual basal position of Lestodelphys on the large opossum clade (node G) in the TS coding postcranial evidence ( fig. 34 ). The typical position of this terrestrial mouse opossum is close to Thylamys (node A), based on profuse previous evidence ( Jansa and Voss, 2000, 2005; Voss and Jansa, 2003; Jansa et al., 2006; Flores, 2003; Creighton, 1984; Reig et al., 1987; Kirsch and Palma, 1995; tables 7, 8). However, Lestodelphys and most of the large opossums shared some character states linked to functional implications on posture and locomotion. In both Lestodelphys and most of large opossums, the spinous process on C6 is laminar (ch. 17[2]) and the femoral lesser trochanter is scarcely developed (ch. 93[0], see Metachirus in fig. 27 ).

An additional example of an unusual relationship possibly caused by constraints from functional demands is the sister relationship between the terrestrial Metachirus and the specialized swimmer Chironectes in the TS coding postcranial analysis, being closely related to the long recognized monophyletic Didelphis . The previous phylogenetic evidence does not recover Metachirus-Chironectes as sister taxa, with both being successively arranged in the clade containing the large opossums ( Reig et al., 1987; Kirsch and Palma, 1995; Jansa and Voss, 2000, 2005; Voss and Jansa, 2003; Gruber et al., 2007). In the scheme obtained from postcranial evidence ( fig. 34 ), both monotypic taxa are clustered as a monophyletic group sharing some synapomorphies associated with specific locomotion or postural patterns: ventral tubercle of atlas of triangular shape (ch. 5[3]), cranial notch of neural arch of axis wide (ch. 15[1], fig. 1 ), ilium with the distal portion barely curved laterally (ch. 90[1], fig. 26 ), prepollex absent (ch. 104[0]), and astragalus with well-developed astragalonavicular facet not contacting the sustentacular one (ch. 110[0], fig. 31 ), although in AMNH 148720 both facets are slightly in contact (see Szalay, 1994: fig. 7.12 ).

Alternatively, as stated above, the postcranial evidence supports the monophyly of some traditionally already recognized groups (tables 6–8; fig. 33 ), such as Thylamys , Micoureus , Monodelphis , Marmosops (node M), Didelphis , Philander , and the large opossums (node G), which are currently recognized by other kinds of evidence (e.g., Reig et al., 1987; Kirsch et al, 1995; Kirsch and Palma, 1995; Jansa and Voss, 2000, 2005; Jansa et al., 2006; Voss and Jansa, 2003; Gruber et al., 2007; table 8). Similarly, the intermediate position of Hyladelphys among caluromyines and didelphines is also recovered here based on postcranial evidence ( figs. 33 , 34 ; table 6), proving its basal position in the didelphid crown group and reinforcing its intermediate phylogenetic position between caluromyines and didelphines already obtained from other sorts of evidence (table 6).

Although not strongly supported, the species of Monodelphis included in this analysis are clustered in a monophyletic group supported by an array of postcranial traits coming from caudal vertebrae morphology, scapula, and some characters from the forelimb and hindlimb. In the same way, Marmosops (node M) is also recovered as monophyletic based on postcranial morphology, although the support values are not high in both TS and CO coding (table 7). This group was already recovered as monophyletic by different kinds of evidence (tables 7, 8); in this report, I add two postcranial characters (deltopectoral crest notably developed, ch. 68[1], and osseous posteroventral extension of the ischium, ch. 86[1]; fig. 23 ) that reinforce the monophyletic nature of this genus (appendix 3). Other groups of mouse opossums previously recognized, such as Thylamys and Micoureus , are also recovered under CO coding postcranial morphology alone ( fig. 33 ). However, even if both genera where recovered as well-supported natural groups in some published analyses (e.g., Jansa and Voss, 2005; Gruber et al. 2007), their monophyletic condition was proved only in a molecular or combined frame. Here, both groups are slightly supported in a morphological context (appendix 3): in Thylamys the diaphragmatic element is placed on T10 (ch. 27[0]), the longitudinal groove in the lateral surface of the ulna is absent (ch. 77[0]), and the bicipital tuberosity of the radius is scarcely marked (ch. 81[0]), whereas in Micoureus the medial relief for m. teres major on the humerus is absent (59[0]).

As mentioned above, the large opossums ( Didelphis , Lutreolina , Chironectes , Philander , and Metachirus ; node G) were widely recognized in previous papers (table 8). Some postcranial characters, principally traits coming from vertebrae, ribs, and humerus morphology, also add synapomorphies that support the monophyletic condition of this group (appendix 3). A last noncontroversial outcome is the monophyly of Didelphis and Philander , since those genera were largely recovered as natural groups based on several kinds of characters (e.g., Patton et al., 1996; Kirsch et al., 1995; Jansa and Voss, 2003; Flores, 2003; Jansa et al., 2006; Jansa and Voss, 2005; Gruber et al., 2007). The three species of Didelphis included in this report shared the especially strong and scarcely mobile articulation of cervical and thoracic vertebrae (appendix 3), which was illustrated by Coues (1869), whereas the three species of Philander included in this study are clustered by the morphology of the axis, as well as by some special patterns of the forelimb and femur (appendix 3).

CONCLUSIONS

The recent impulse and increase of knowledge on didelphid phylogeny is the result of the contribution of the inclusion of a denser taxon sampling and the consideration of varied evidence coming from both nuclear sequences (IRBP, DMP-1, and RAG-1) and morphology. Although the nonmolecular aspects frequently have resulted in low resolved topologies compared with hypotheses based on genetic evidence, they are in general agreement with genetic and combined data sets. Topologies coming from postcranial characters only are in general well resolved (mainly in CO coding analysis) and do not conflict with well-supported groups and relationships (e.g., Hyladelphys as intermediate between didelphines and caluromyines; the monophyly of large opossums; and Didelphis , Philander , Monodelphis , Thylamys , Micoureus , and Marmosops as natural genera), although some unusual clusterings are observed that result from convergences possibly caused by functional demands. However, the contribution of phylogenetic information from postcranial morpohology substantially improves the resolution of previous morphological hypotheses. Some relationships, formerly evidenced only with nuclear sequences, are now recovered with the addition of postcranial characters as well in a morphological framework. Even with the recent consideration of RAG-1 sequences, which reveals some unusual relationships, the phylogenetic information coming from postcranial morphology produces topologies in better agreement with earlier combined hypotheses. However, combined evidence considering postcranial evidence (including and excluding RAG-1 sequences) recovered the clustering of the mouse opossums in the CO polymorphic character coding. In this sense, the phylogenetic condition of the mouse opossums is still problematic when postcranial characters are considered in a combined data set, although this diverse group is not recovered in other partitioned analyses performed in this report.

The inclusion of new informative sequences and other kinds of morphological characters could provide additional support to groups recognized before or to new topologies. In this sense, the anatomical comparisons on forearm muscles in didelphids and some Australasian taxa performed by Abdala et al. (2006) added some potential new morphological synapomorphies to clades already recognized (i.e., nodes C, G, I, and D). Including a denser taxon sample in such alternative anatomical systems is an important priority in future research on didelphid phylogeny for the sake of completeness. Finally, the pending postcranial observations in taxa still not analyzed, as well as the exact condition of missing data of some skeletal traits, could also reinforce several phylogenetic topologies that are slightly supported.

Figures

Fig. 28. Chironectes minimus (AMNH 212909) and Didelphis marsupialis (AMNH 210439), right tibia in anterior view. In Chironectes the tibia has a sigmoid shape (ch. 99[0]), whereas in Didelphis it is almost straight (ch. 99[1]). Scale bars: 5 mm.

Fig. 29. Marmosops impavidus (AMNH 139226) and Micoureus regina (AMNH 148757), right tibia in mesial view. In Marmosops there is a notable crest for insertion of m. flexor digitorum tibialis (cfdt) (ch. 101[1]), which is absent in Micoureus (ch. 101[0]). Scale bars: 5 mm.

Fig. 30. Philander frenatus (MVZ 182067), dorsal aspect of distal ulna (ul), radius (ra), carpal, and metacarpal (Mc) bones. Note the lunate (lu) relatively large and in contact with other elements (ch. 103[1]). A distolateral process of the scaphoid (sca) separates the lunate and magnum (mag) (ch. 105[1]). Other abbreviations: un, unciform; cu, cuneiform; tm, trapezium; tr, trapezoid. Scale bar: 5 mm.

Fig. 31. Chironectes minimus (AMNH 148720), plantar and dorsal views of left astragalus. The ridge between the medial (atim) and lateral (atil) astragalotibial facets is present (ch. 108[1]), as well as being between the lateral atragalotibial and astragalofibular (afi) facets (ch. 109[1]). There is no contact between the astragalonavicular (an) and sustentacular (su) facets (ch. 110[0]), and the sustentacular facet is separated from the calcaneoastragalar (caa) facet by the well-developed sulcus astragali (sa) (ch. 111[0]). Other abbreviation: ampt, astragalar medial plantar tuberosity. Scale bar: 5 mm.

Fig. 32. Chironectes minimus (AMNH 148720), dorsal and plantar views of right calcaneus. The peroneal process (pp) is well developed (ch. 113[1]), and the calcaneal sustentaculum (su) is placed on the anterior tip (ch. 114[1]). Other abbreviations: caa, calcaneoastragalar facet; cacud, distal calcaneocuboid facet; cacup, proximal calcaneocuboid facet; sc, sulcus calcanei. Scale bar: 5 mm.

Fig. 8. Chironectes minimus (AMNH 264571) and Caluromys philander (AMNH 267001), fifth lumbar vertebra in anterior view. In Chironectes, the transverse process (tp) is ventrally extended beyond the vertebral body (b) (ch. 37[1]), whereas in Caluromys it is not extended beyond the vertebral body, being more laterally extended (ch. 37[0]). Other abbreviations: mp, mammillary process; sp, spinous process; vf, vertebral foramen. Scale bars: 5 mm.

Fig. 9. Hyladelphys kalinowskii (RSV 1572) and Caluromysiops irrupta (AMNH 244364), pelvis, sacrum, and posterior lumbar vertebrae in dorsal view. In Hyladelphys, just the first sacral vertebra (S1) is contacting the illium (il) (ch. 38[0]), whereas in Caluromysiops both sacral elements (S1 and S2) are contacting the illium (ch. 38[1]). Note the absence of the posterior process on the lateral sacral crest of S2 in Hyladelphys (ch. 39[0]). Other abbreviations: ac, acetabulum; is, ischium; of, obturator foramen. Scale bars: 5 mm.

Fig. 10. Monodelphis brevicaudata (AMNH 257203) and Lestodelphys halli (UWZM 22422), sacral vertebrae in dorsal view. In Monodelphis there is a posterior process (pp) on the lateral sacral crest (lsc) on S2 (ch. 39[1]), whereas in Lestodelphys it is absent (ch. 39[0]). Scale bars: 5 mm.

Fig. 11. Marmosops incanus (MVZ 182768) and Metachirus nudicaudatus (AMNH 267009), right os coxae with sacral vertebrae in lateral view. In the sacral vertebrae of Marmosops incanus the spinous process is present only in S1 (ch. 40[0]), whereas in Metachirus the process is present in both vertebrae (ch. 40[1]), and the S1 spinous process is taller than the S2 spinous process (ch. 41[1]). In Metachirus, the acetabulum (ac) is close, deep, and with the dorsal part laterally extended (ch. 83[1]), and the iliac wing (iw) forms a large blade (ch. 84[1]). Scale bars: 10 mm.

Fig. 12. Didelphis marsupialis (AMNH 210439 and Caluromys philander (AMNH 267001), sacral vertebrae in ventral view. A ventral foramen (vef) in the S1 body is present in Didelphis (ch. 42[1]), whereas in Caluromys it is absent (ch. 42[0]). Other abbreviations: psf, pelvic sacral foramen; w, wing. Scale bars: 10 mm.

Fig. 33. Strict consensus of three equally most parsimonious trees resulting from cladistic parsimony analysis of 114 postcranial characters for 38 taxa described in this report, where polymorphic data were treated as composite entries (CO) (see tables 2–4 for summary data set characteristics and tree statistics). Numbers above branches refer to absolute Bremer support values ($1). Numbers below branches refer to

Fig. 1. Didelphis marsupialis (AMNH 21439) and Metachirus nudicaudatus (AMNH 244617), axis in lateral view. The spinous process (Sp) in Metachirus is posteriorly extended (ch. 10[1]), whereas in Didelphis it is not extended beyond the neural arches (Na) (ch. 10[0]). In Didelphis, the anterior extension of the dens (de) is barely anterior to the anterior tip of the spinous process (ch. 11[0]), whereas in Metachirus the dens is cranialy extended (ch. 11[2]). The ventral tubercle (Vt) is a uniform crest in Didelphis (ch. 14[0]), whereas in Metachirus it forms two protruding lobes (ch. 14[1]). Other abbreviations: As, articular surface; Tc, transverse canal; Tp, transverse process. Scale bars: 10 mm.

Fig. 34. Strict consensus of seven equally most parsimonious trees resulting from cladistic parsimony analysis of 114 postcranial characters for 38 taxa described in this report, where polymorphic data were treated as transformation series (TS) (see tables 2 and 4 for summary data set characteristics and tree statistics). Numbers above branches refer to absolute Bremer support values ($1). Numbers below branches refer to jackknife frequencies (cutoff value 5 50%). Heavy lines denote branches with a decay index of $4. Outgroup taxa are indicated with asterisks. Alphabetic labels indicate didelphine clades discussed in the text.

Fig. 2. Marmosa rubra (FMNH 124612) and Philander opossum (AMNH 262415), anterior cervical vertebrae in ventral view. In Marmosa rubra, the transverse process (Tp) on the third cervical (C3) has one head (ch. 21[0]), but in Philander it shows an anterior head (h) (ch. 21[1]). In both species the fourth cervical (C4) has two heads (ch. 22[1]). Scale bars: 5 mm.

Fig. 5. Caluromysiops irrupta (AMNH 208101) and Caluromys philander (AMNH 267001), cervical and thoracic vertebrae in dorsal view. The T1 body in Caluromysiops is lengthened craniocaudally (ch. 26[1]), whereas this condition is not observed in Caluromys (ch. 26[0]). However, in both taxa the hemal arches in T1 are notably shorter than hemal arches in T2 (ch. 32[1]). Note the ribs (ri) flattened and craniocaudally expanded in both taxa (ch. 49[0]). Other abbreviations: Ax, axis; he, hemal arch. Scale bars: 10 mm.

Fig. 6. Metachirus nudicaudatus (AMNH 267009) and Caluromys philander (AMNH 267001), lumbar vertebrae in dorsal view. In Metachirus the mammillary process (mp) of L3 is not anteriorly extended beyond the articulation with L2 (ch. 34[0]), whereas in Caluromys it is extended (ch. 34[1]). Scale bars: 10 mm.

Fig. 4. Philander opossum (AMNH 262415) and Marmosa mexicana (ROM 99608), last cervical vertebrae and first thoracic vertebra in lateral view. In Philander, the C7 spinous process is taller than the C6 process (ch. 20[0]), whereas in Marmosa mexicana the C6 and C7 spinous processes are similar in size (ch. 20[2]). Note the C6 spinous process resembling a protuberance in both species (ch. 17[2]). Abbreviation: r, rib. Scale bars: 5 mm.

Fig. 27. Caluromys philander (AMNH 267001) and Metachirus nudicaudatus (AMNH 244617), proximal portion of right femur. In Metachirus, the greater trochanter (gt) is well developed (ch. 95[1]), whereas in Caluromys it is not very developed (ch. 95[0]). The lesser trochanter (lt) is more developed in Caluromys (ch. 93[1]) than in Metachirus (ch. 93[0]). Note in both species the protuberance between the trochanteric fossa (tf) and femoral head (h) (ch. 94[1]). Scale bars: 5 mm.

Fig. 26. Monodelphis brevicaudata (AMNH 257203) and Chironectes minimus (AMNH 212909), pelvis in dorsal view. In Chironectes the posterior part of the ischium (is) body is laterally deflected (ch. 92[1]), whereas in Monodelphis it is almost straight (ch. 92[0]). Other abbreviations: il, illium; syp, symphysis pubis. Scale bars: 10 mm.

Fig. 7. Thylamys macrurus (MSB 70700) and Metachirus nudicaudatus (AMNH 244617), lumbar vertebrae in dorsal view. An intervertebral space (ip) is evident dorsally in Thylamys (ch. 36[0]), whereas in Metachirus this space is absent (ch. 36[1]). Other abbreviations: mp, mammillary process; po, postzygapophysis; sp, spinous process; tp, transverse process. Scale bar: 5 mm.

Fig. 23. Marmosops parvidens (AMNH 267348), os coxae in lateral view showing the posteroventral extension (pex) on the pubis (ch. 86[1]). Scale bar: 5 mm.