Amphinectomys, Malygin, 1994

Weksler, M., 2006, Phylogenetic Relationships Of Oryzomine Rodents (Muroidea: Sigmodontinae): Separate And Combined Analyses Of Morphological And Molecular Data, Bulletin of the American Museum of Natural History 2006 (296), pp. 1-149 : 59-78

domys, Amphinectomys , Neacomys minutus , Nesoryzomys swarthi , Oecomys catherinae , Oligoryzomys fornesi , Ol. stramineus , and Oryzomys polius (coded ‘‘?’’).

Character 99: Gastric glandular epithelium of stomach limited to antrum, not extending beyond incisura angularis (0); or gastric glandular epithelium covers antrum and proximal portion of corpus near esophageal opening (1). Oryzomyines have the basic unilocularhemiglandular stomach pattern characteristic of sigmodontines, with a shallow incisura angularis that does not split the stomach in two major chambers, and a glandular epithelium that covers the anterior gastric chamber or antrum ( Carleton, 1973; Vorontsov, 1979). Carleton (1973) recognized two major configurations of the glandular epithelium among taxa with the unilocular-hemiglandular pattern: (1) a group with the glandular epithelium restricted to the antrum that included most sigmodontines (fig. 33A); and (2) a smaller group with glandular epithelium extending into the corpus (the posterior chamber) that included Holochilus brasiliensis and Nectomys squamipes . Subsequently, the latter condition was also described for Microryzomys minutus (see Carleton and Musser, 1989) and Pseudoryzomys (see Voss and Carleton, 1993). Among the specimens I analyzed, state 1 was observed in Holochilus , Microryzomys , Neacomys musseri , Nectomys , Oryzomys balneator , Pseudoryzomys , and Sigmodontomys aphrastus (fig. 33B). The remaining analyzed oryzomyines, as well as all of my outgroups, have restricted glandular epithelium (state 0). Information on the stomach morphology is not available for Wiedomys , Amphinectomys , Neacomys minutus , Nesoryzomys swarthi , Oecomys catherinae , Oligoryzomys flavescens , O. fornesi , O. stramineus , Oryzomys angouya , O. lamia , O. levipes , O. polius , O. russatus , and Sigmodontomys alfari (coded ‘‘?’’). This character is inapplicable for Peromyscus (which has a bilocular-discoglandular stomach structure; coded ‘‘–’’).

SUMMARY OF MORPHOLOGICAL DATASET

The final morphological dataset consists of 99 characters out of 350 originally examined features (table 6). The dataset includes 16 characters based on external morphology; 32 characters of the skull and mandible; 29 dental characters; 7 postcranial characters; and 10 characters from the phallus, accessory male reproductive glands, and digestive system (table 6). Ninety-one characters are parsimony-informative, and the remaining eight are autapomorphic for oryzomyine taxa. Sixty-four characters are binary, 25 describe simple additive transformations, 2 describe complex (i.e., based on step matrices) additive transformations, and 8 describe nonadditive (unordered) transformations. The data matrix (table 5) has 54 × 99 5 5346 cells, of which 590 (11 %)

TABLE 6 Character Information Summary

are scored as missing (‘‘?’’), 17 (0.3 %) are scored as inapplicable (‘‘–’’), and 70 (1 %) are scored as polymorphic. Data completeness for most terminal taxa range from 78 to 100 % (table 4); however, Nesoryzomys swarthi was scored for only 63 % of characters, and only 32 % were coded for Amphinectomys savamis .

PHYLOGENETIC RESULTS

MORPHOLOGY- ONLY ANALYSES: A heuristic analysis of the morphological data matrix with polymorphic entries analyzed as composites (CO) resulted in one most parsimonious tree of 522 steps (CI 5 0.25, RI 5 0.63; table 7) consistent with the assumption of ingroup (oryzomyine) monophyly (fig. 34). After rooting the tree using Nyctomys and Peromyscus , the basal tree structure depicts the sequence ( Thomasomys ( Delomys ( Wiedomys + Oryzomyini ))). Oryzomyines are split basally into two major clades, the first with 28 and the second with 21 taxa. The first clade contains the genera Handleyomys , Microryzomys , Neacomys , Oecomys , Oligoryzomys , Scolomys , Zygodontomys , and 10 of the 19 analyzed species of Oryzomys . The second clade is composed of Amphinectomys , Holochilus , Lundomys , Melanomys , Nesoryzomys , Pseudoryzomys , Nectomys , Sigmodontomys , and the remaining nine Oryzomys species. Of the eight genera with multiple representatives, only Oryzomys and Sigmodontomys are not recovered as monophyletic.

The first oryzomyine clade is divided basally into two large subclades. In the first, Scolomys and Zygodontomys form a monophyletic group (labeled A in fig. 34) that is sister to the group containing Microryzomys , Neacomys , Oligoryzomys , and Oryzomys balneator (labeled C). Neacomys and Oligoryzomys are both recovered as monophyletic genera, and the latter is the sister group to a clade formed by Microryzomys and Oryzomys balneator . The second oryzomyine clade (labeled B*) contains Handleyomys , Oecomys , and eight species of Oryzomys . Oecomys is recovered as monophyletic, with O. hammondi as its sister group. Going up the tree, a clade with the three species of the nitidus species group ( O. macconnelli , O. russatus and O. lamia ) is joined by O. yunganus and then by a clade with O. megacephalus , O. talamancae , O. albigularis , O. levipes , and Handleyomys . The latter genus is the sister group of the albigularis species group ( O. albigularis and O. levipes ).

Oryzomys polius appears as the sister group to the remaining members of a second large group (labeled D* in fig. 34). The next branch contains O. angouya , and the remaining taxa are split into two clades. The first clade contains O. subflavus as the most basal taxon, followed by (1) the alfaroi group of Oryzomys — O. chapmani , O. alfaroi and O.

TABLE 7

Summary of Phylogenetic Analyses

rostratus —of which the latter two appear as sister species; (2) the palustris group of Oryzomys ( O. palustris and O. couesi ); and (3) a clade containing Pseudoryzomys , Lundomys , and Holochilus , with the latter two as sister taxa. Another clade contains Nesoryzomys as the most basal taxon, followed by six taxa in the sequence ( Oryzomys xanthaeolus ( Amphinectomys ( Melanomys ( Sigmodontomys aphrastus ( S. alfari + Nectomys ))))).

A heuristic analysis of the morphological data with polymorphic entries analyzed as transformation series (TS) resulted in 22 most parsimonious trees of 693 steps each (CI 5 0.22, RI 5 0.62). The consensus tree (fig. 35) is the least resolved among the analyzed datasets, with only 31 of 51 (61 %) resolved nodes and major polytomies between and within the major oryzomyine clades. The TS tree differs from the CO tree in several important aspects: (1) clade A is not recovered in the consensus tree, appearing in only in a few fundamental cladograms; (2) Amphinectomys appears as the sister taxon of Handleyomys within clade B*; (3) the alfaroi group of Oryzomys is recovered within clade B*; and (4) Sigmodontomys is recovered as monophyletic. Nevertheless, many relationships are congruent between the two analyses, including: (1) monophyly of Oryzomyini ; (2) monophyly of clade C; (3) the three tetralophodont taxa Pseudoryzomys , Holochilus , and Lundomys and the palustris group of Oryzomys form a clade; (4) Oryzomys polius as basal member of clade D*; and (5) Nectomys and Sigmodontomys form a clade. Additionally, clades B* and D* are similar between the two analyses, with only minor differences in taxa content, and the monophyletic genera recovered in the CO analysis are also observed in the TS tree. Note also that all nodes at which the two analyses conflict are weakly supported in both trees.

IRBP-ONLY ANALYSIS: A heuristic analysis of the IRBP dataset resulted in four equally parsimonious trees of 667 steps (CI 5 0.55, RI 5 0.71; table 7). The consensus topology (fig. 36) is identical to that previously recovered by Weksler (2003: fig. 5) when pruned of the outgroup taxa not included here. Within oryzomyines, a basal polytomy involving Scolomys , Zygodontomys , and a clade containing all of the remaining ingroup taxa is present in the consensus tree and in all fundamental cladograms. The largest group in this trichotomy is further divided into two clades. The first (labeled B in fig. 36) contains the genera Oecomys , Handleyomys , and 9 of the 16 analyzed species of Oryzomys . The internal structure of clade B in the strict consensus topology is characterized by a five-fold polytomy resulting from two solutions found in the fundamental cladograms, which depict either ( Oryzomys talamancae ( O. macconnelli ( O. lamia + O. russatus ))) or ( Handleyomys intectus ( O. alfaroi + O. rostratus )) as the basal clade relative to the remaining taxa.

The second major group, containing the remaining 11 oryzomyine genera and 7 species of Oryzomys , is divided into two clades. The first (labeled C) contains Microryzomys , Oligoryzomys , Neacomys , and Oryzomys balneator . The second (labeled D) consists of six species of Oryzomys interspersed among eight other genera. Within this group, (1) Oryzomys polius appears as the sister group to remaining species, and (2) three pairs of genera are recovered as sister groups: Holochilus + Pseudoryzomys , Amphinectomys + Nectomys , and Melanomys and Sigmodontomys . The species of Oryzomys included in this clade are not recovered as monophyletic.

COMBINED ANALYSES: A heuristic search of the combined data analyzed with CO polymorphic entries resulted in two most parsimonious trees of 1214 steps (CI 5 0.38, RI 5 0.65; table 7). Oryzomyines are again recovered as monophyletic, and Wiedomys is again its sister group (fig. 37). Oryzomys hammondi appears as the sister group of remaining oryzomyines, followed by the clade containing Scolomys and Zygodontomys (labeled clade A in fig. 37). The basal phylogenetic structure of remaining oryzomyines is similar to the one recovered by the IRBP-only analysis, with three large clades (labeled B, C, and D) containing 17, 10, and 18 species each. Clades C and D appear as sister groups. Within clade B, the clade containing the albigularis species group ( O. albigularis and O. levipes ) plus O. talamancae is recovered as sister group of the ( O. megacephalus + O. yunganus ) clade. Going up the tree, this clade is joined by a cluster containing Handleyomys and three species of Oryzomys ( O. rostratus , O. chapmani , and O. alfaroi ) and then by the monophyletic Oecomys . Oryzomys chapmani is found as the sister species to O. rostratus ; the three species of the nitidus species group ( O. macconnelli , O. russatus , and O. lamia ) form a clade that is the sister group of the remaining species.

Within clade C, Microryzomys and Oryzomys balneator are recovered as sister groups and are joined successively by the monophyletic genera Oligoryzomys and Neacomys . Oryzomys polius appears as the sister group to remaining members in clade D, which is then split into two subclades. The first contains two species of Oryzomys ( O. palustris and O. couesi ) as the sister group to the clade containing Pseudoryzomys , Lundomys , and Holochilus , with the latter two as sister groups. The phylogenetic sequence of the second subclade starts with Oryzomys angouya , followed by O. subflavus , Nesoryzomys , and O. xanthaeolus . Nested within the subclade, ( Amphinectomys + Nectomys ) and ( Melanomys + Sigmodontomys ) are sister groups. Nectomys is recovered as monophyletic, but Sigmodontomys is not: S. aphrastus is recovered as the sister group of Melanomys , and the two are joined to S. alfari .

The polytomy present in the consensus tree (fig. 37) is the result of different arrangements within the genus Oligoryzomys in the fundamental cladograms: O. fulvescens is recovered either as the sister group of the ( O. fornesi + O. flavescens ) clade or as the sister group of the ( O. stramineus + O. nigripes ) clade.

Heuristic search of combined datasets with polymorphic entries analyzed as TS resulted in nine most parsimonious trees of 1388 steps (CI 5 0.34, RI 5 0.64). The consensus tree (fig. 38) resembles the CO combined tree, having six differences: (1) hammondi is within clade B; (2) the basal topology of major oryzomyines clades is unresolved; (3) the internal topology of clade B is unresolved; (4) Microryzomys and O. balneator do not form a monophyletic group; (5) Oligoryzomys is the basal taxon in clade C; and (6) O. angouya is the sister group of all members of clade D except O. polius . The polytomies present in the consensus tree (fig. 38) are the result of (1) different arrangements within the Sigmodontomys - Melanomys clade; (2) different ordering of xanthaeolus and Nesoryzomys ; (3) different positioning of O. hammondi within clade B, either as sister group to Oecomys or as the most basal taxon in clade B; and (4) different positioning of clade A, either as basal to all oryzomyines or as sister group to clade B.

A node-by-node description of all groups recovered in the combined analysis with CO coding for polymorphisms (fig. 37) is presented in appendix 3.

REDUCED ANALYSIS: Heuristic search of the reduced dataset (i.e., excluding the five oryzomyines without molecular sequences) resulted in eight most parsimonious trees of 1174 steps (CI 5 0.40, RI 5 0.65; table 7). The recovered trees and the consensus cladogram (fig. 39) are similar to those retrieved by the CO combined analysis, except for three changes: (1) in clade B, Oryzomys albigularis is recovered as the most basal taxon, while the nitidus group is recovered as sister group to O. talamancae (i.e., O. albigularis and the nitidus group swap places); (2) in clade C, O. balneator and Microryzomys are not recovered as sister taxa, and instead are placed in a polytomy with Oligoryzomys and Neacomys ; and (3) in clade D, Oryzomys angouya is recovered in a polytomy also containing the two major subclades of clade D.

NODAL SUPPORT

Nodal support values for the morphological trees are generally low. For instance, in the CO analysis (fig. 34), 34 nodes (67 % of resolved nodes) collapse in trees that are one step longer, 6 additional nodes (12 %) collapse in trees that are two steps longer, and 4 more nodes (8 %) collapse in trees that are three steps longer. Only 7 nodes (14 % of the total) have a decay index greater than three. Resampling values were also low: 28 nodes (55 %) have jackknife values below 50 %, and 17 nodes (33 %) have jackknife values between 50 and 85 %, while the remaining 6 nodes (12 % of the total) have jackknife values higher than 85 %.

Nodal support values for the IRBP tree (fig. 36) are considerably higher. Seventeen nodes (47 %) collapse in trees that are one step longer, 2 additional nodes (6 %) collapse in trees that are two steps longer, and 2 more nodes (6 %) collapse in trees that are three steps longer. Fifteen nodes (42 % of the total) have a decay index greater than three. Jackknife values show higher support: 2 nodes (6 %) have jackknife values below 50 %, and 15 nodes (42 %) have jackknife values between 50 and 85 %, while the remaining 19 nodes (53 % of the total) have jackknife values higher than 85 %.

Nodal support values for the combined CO analysis indicate that most of the 50 resolved nodes in the consensus tree (fig. 37) are at least moderately well supported. Only 16 nodes (32 %) collapse in trees that are one step longer, 5 additional nodes (10 %) collapse in trees that are two steps longer, and 4 more nodes (8 %) collapse in trees that are three steps longer. The remaining 25 nodes (50 % of the total) have a decay index greater than 3. Jackknife resampling suggests a similar pattern of support. Eleven nodes (22 %) have jackknife values below 50 %, and 20 nodes (40 %) have jackknife values between 50 and 85 %, while the remaining 19 nodes (38 % of the total) have jackknife values higher than 85 %. Nodal support is slightly higher in the TS analysis (fig. 38): 8 nodes (18 %) collapse in trees that are one step longer, 8 additional nodes (18 %) collapse in trees that are two steps longer, and 6 more nodes (13 %) collapse in trees that are three steps longer. The remaining 23 nodes (51 % of the total) have a decay index greater than three. In terms of resampling values, only 5 nodes (11 %) have jackknife values below 50 %, and 20 nodes (44 %) have jackknife values between 50 and 85 %, while the remaining 20 nodes (44 % of the total) have jackknife values higher than 85 %.

The effect of taxon removal in the reduced analysis is visible in nodal support values: 7 nodes (18 % of the total) have jackknife values increased by at least 10 %, 3 of them more then 20 %. Twenty-three nodes (61 %) have minimal changes in jackknife values (differences between 23 % and +3 %), while 4 nodes (11 %) decreased at least 4 %. The maximum increase in jackknife values was 39 % (from 58 to 97 %; clade B), while the maximum decrease was 5 % (from 87 to 82 %; Amphinectomys + Nectomys + Melanomys + Sigmodontomys ).

DISCUSSION

EFFECTS OF DIFFERENT CODINGS OF POLYMORPHIC DATA

The different codings for polymorphic characters produced similar topologies, espe- cially in the combined trees, as expected by the low number of polymorphic cells in the matrix (1 %). The major difference was the lack of resolution of the consensus tree derived from the morphology-only TS analysis (fig. 35). Different codings had little impact on nodal support (fig. 40), with TS coding providing slightly higher values for most nodes, as expected from the retention of more phylogenetic information in this kind of coding ( Mabee and Humphries, 1993).

In the morphology-only analyses, both codings placed some taxa in unexpected positions. In the CO analysis, the alfaroi - chapmani - rostratus clade is found nested within clade D*, while the TS analysis placed Amphinectomys within clade B*. All of these unexpected arrangements received low nodal support and none was recovered in the combined (TS or CO) or IRBP analyses. The major difference in the combined analyses using different polymorphism coding was the position of O. hammondi , recovered either as the sister group of all oryzomyines (CO coding) or as a member of clade B* (TS coding). The latter placement was recovered in the morphology-only analyses with both CO and TS coding.

The remaining discussion is based on the consensus tree of the combined analysis with polymorphic data treated as composites (fig. 37). The differences in topology recovered by the two treatments of polymorphic data do not affect any of the interpretations presented below, except as noted.

EFFECTS OF MISSING DATA ON TOPOLOGY AND SUPPORT

Neither Amphinectomys , which could be scored for only one-third of all morphological characters, nor the five taxa without IRBP sequences behaved as wildcard taxa ( Nixon and Wheeler, 1992; Kearney, 2002) in the combined analysis. All had secured positions, not floating on the fundamental cladograms as the result of alternative optimizations of question marks, leading to a well-resolved strict consensus tree. Removal of the five taxa without IRBP sequences in the reduced analysis, however, affected the tree topology, as it reduced the resolution of two major clades (C and D) and changed the structure of another (clade B). Nevertheless, almost all of these changes involved clades with low nodal support.

Removal of the taxa with missing data, however, had a marked effect in nodal support for several oryzomyine lineages. Foremost among these changes were the percentile increases for resampling support of the major oryzomyine lineages; for example, the jackknife support of clade A increased from 53 to 74 %; clade B, from 58 to 96 %; clade D, from 82 to 94 %; clade (B + C + D), from 49 to 78 %; and clade (C + D), from 80 to 92 % (fig. 40). Support for clade C did not change significantly (jackknife from 74 to 77 %), as this is the only major oryzomyine clade without missing data. Other significant increases were observed in nodes closer to the tips in which terminal taxa were removed; for example, Nectomys + Amphinectomys ( N. apicalis removed), jackknife support increase from 75 to 89 %; Melanomys + Sigmodontomys ( S. aphrastus removed), from 66 to 84 %.

Among the six taxa with large amounts of missing data, three were recovered in noncontroversial positions within oryzomyines. Nectomys apicalis was found as the sister group to Nectomys squamipes ; Oryzomys levipes was found together with O. albigularis , conforming to the expectation of the albigularis species group ( Patton et al., 1990; Musser and Carleton, 1993; Percequillo, 2003); and O. chapmani clustered with O. rostratus and O. alfaroi , compatible with the presumably close association of these species ( Goldman, 1918; Musser and Carleton, 1993). In contrast, Amphinectomys , Sigmodontomys aphrastus , and Oryzomys hammondi have distinct and sometimes unexpect- ed placements in the different analyses. Amphinectomys appears as the sister group to Nectomys in the combined analysis, but in the morphology-only tree it is placed as the sister group to the clade Melanomys + Sigmodontomys + Nectomys (CO analysis) or within clade B (TS analysis). Sigmodontomys aphrastus is not found as the sister species to Sigmodontomys alfari in either CO morphology-only or combined analyses, but is in the TS analysis of morphological data. Instead, the two species form a paraphyletic sequence relative to Nectomys in the mor- phology-only analysis and to Melanomys in the combined analysis. Finally, Oryzomys hammondi is placed as the most basal oryzomyine in the CO combined tree, and as the sister group to Oecomys in the morphology-only and TS combined analyses.

DATASET COMPARISON

The IRBP and morphological datasets produced a similar number of equally most parsimonious trees and number of resolved nodes in their separated analyses despite differences in the number of informative characters (204 vs. 91 for IRBP and morphology, respectively). Furthermore, both datasets contributed equally to the structure of the combined tree, with each data partition having 30 nodes recovered in their separate analyses and in the combined consensus cladogram. On the other hand, the IRBP dataset was much less homoplasious than was the morphology dataset, as measured by ensemble CI and RI values. This pattern contradicts the expectation of higher homoplasy for the dataset with more characters ( Sanderson and Donoghue, 1989; Sanderson, 1991) and indicates that the molecular dataset had more phylogenetic signal than did the morphological partition, despite similar levels of phylogenetic resolution. This is also reflected in the higher values of nodal support recovered for the nodes in the IRBP analysis and in the resolution of the most conflicting hypothesis between the datasets favoring the IRBP solution in the combined tree.

The resulting trees from the morphological and IRBP separate analyses exhibit topological differences that imply phylogenetic conflict. Nevertheless, node-by-node comparisons of nodal support values reveal only one case of conflicting relationships between wellsupported clades in the separate analyses, or ‘‘hard’’ incongruence—the incompatibility between well-defined patterns of morphological versus molecular synapomorphies ( Voss and Jansa, 2003). The sister group of Holochilus is Pseudoryzomys in the IRBPonly analysis and Lundomys in the morphology-only analysis, both solutions with jackknife values. 95 % and a decay index. 6. The combined tree favors the morphological solution because there are 15 morphological synapomorphies for Holochilus + Lundomys hypothesis versus 7 IRBP synapomorphies for Holochilus + Pseudoryzomys . The lack of molecular synapomorphies for Lundomys and Holochilus might be due to differential evolutionary rates between molecules and morphology. Additional data on other molecular markers, as well as denser sampling, are necessary for the resolution of this conflict. All remaining cases of conflict between morphology and IRBP clades involved weak supported nodes from one or both datasets. No other clade that was moderately or highly supported (with a decay index.2 and/or jackknife support values.70 %) in either analysis was incongruent with any equivalently supported node in the other.

Major differences of higher-level relationships between separate analyses are (1) Zygodontomys and Scolomys as advanced oryzomyines in the morphology tree and as the most basal oryzomyines in the IRBP tree; (2) the recovery of the ( O. alfaroi + O. rostratus + O. chapmani ) clade within clade D* in the morphology tree (but not in the TS analysis) and within clade B in the IRBP tree; and (3) the closer relationship between clades B* and C in the morphology tree and between clades C and D in the IRBP tree. In each case, the nodal support for the morphology solutions was low, while the support for IRBP solutions was high, and in each case the combined tree favored the IRBP resolution.

Although the higher-level topology of the ingroup was resolved toward the IRBP solution, most of the conflicts observed within the major oryzomyines clades were resolved toward morphological hypotheses. Thus, the sister group relationship between the ( Microryzomys + Oryzomys balneator ) clade and Oligoryzomys is also observed in the morphology-only tree, instead of the most parsimonious IRBP arrangement of Neacomys as the sister group to ( Microryzomys + Oryzomys balneator ) clade (but observed also in the combined TS analysis). Likewise, the placement of the Oryzomys palustris species group closer to the ( Pseudoryzomys + Holochilus + Lundomys ) clade, instead of to the ( Nectomys + Melanomys + Sigmodontomys + Nesoryzomys + Amphinectomys ) clade, was the favored morphological solution. Finally, 38 % of the nodes present in the combined tree were also recovered in both separated analyses of IRBP and morphology, and 30 of 45 nodes (66 %) 11 have both morphological and molecular synapomorphies, indicating that there is a considerable amount of phylogenetic agreement between the datasets, which the recovery of similar major oryzomyine clades (B, C, and D) indicates.

ORYZOMYINE SYNAPOMORPHIES

Voss and Carleton (1993) proposed five oryzomyine synapomorphies: (1) presence of a pectoral pair of mammae; (2) long palate with prominent posterolateral pits; (3) absence of alisphenoid strut; (4) absence of posterior suspensory process of the squamosal attached to tegmen tympani; (5) and absence of gall bladder. Steppan (1995) recovered four additional oryzomyine synapomorphies: (1) nasals extending posterior to lacrimal; (2) 12 thoracic vertebrae; (3) absence of hemal arches; and (4) fewer than 36 caudal vertebrae (character not included in the present study). Surprisingly, the unconnected tegmen tympani was the only one of these traits also recovered here as an unambiguous oryzomyine synapomorphy. The absence of the gall bladder is recovered only in DELTRAN optimization because the condition of the gall bladder is unknown in Wiedomys .

Absence of the posterior suspensory process of squamosal connected to the tegmen tympani was also the only oryzomyine synapomorphy recovered in both morphology-only and combined analyses, and it was the only synapomorphy recovered as both unreversed and unique. Reithrodon is the only sigmodontine taxon outside the Oryzomyini with unconnected tegmen tympani ( Voss and Carleton, 1993; Steppan, 1995; Pacheco, 2003). This similarity, however, is due to convergence, as Reithrodon is not the sister

11 Among the 50 recovered nodes in the combined consensus cladogram, 5 nodes are directly connected to taxa without IRBP data, and thus cannot have molecular synapomorphies. group of oryzomyines ( Steppan, 1995; Smith and Patton, 1999; D’Elía, 2003; Weksler, 2003). Four other unambiguous oryzomyine synapomorphies were recovered in the combined analysis: absence of ungual tufts on D1; long incisive foramina passing M1, undivided anterocone on M1, and medial enamel bridge connection between paracone and protocone on M1. Nevertheless, they were reversed several times within oryzomyines. In addition, they occur in several groups outside oryzomyines (see Steppan [1995] and Pacheco [2003] for distribution of some of these characters in non-oryzomyine taxa).

Reconstruction of ancestral states for the proposed oryzomyine synapomorphies in the combined tree indicates that transformation occurred prior to or after the branch leading to oryzomyines. The extended posterior terminus of the nasal appeared as a synapomorphy within oryzomyines, while five other traits were shared with non-oryzomyines: absence of alisphenoid strut and presence of pectoral mammae are shared with Wiedomys and Delomys ; and long palate with complex posterolateral pits, presence of 12 thoracic vertebrae, and presence of hemal arches are shared with Wiedomys . Nevertheless, these recovered transformation patterns may be misleading because of the rarefied sampling of non-oryzomyine taxa and of the uncertainty for the oryzomyine sister group. For instance, Wiedomys is also recovered as the sister group to oryzomyines in Steppan (1995), who first suggested that some of the proposed oryzomyine synapomorphies could actually be synapomorphies for the clade uniting oryzomyines and Wiedomys . Howev- er, in a previous IRBP analysis with denser sigmodontine taxonomic sampling ( Weksler, 2003), Wiedomys was not recovered as the sister group to oryzomyines. Consequently, the characteristics shared by Wiedomys and oryzomyines would appear as homoplasic apomorphies for each taxon.

Additional analyses are necessary for the confident designation of the other oryzomyine synapomorphies, but for at least one of the traits proposed by Voss and Carleton (1993), the presence of pectoral mammae, recent evidence indicates that it is not an oryzomyine synapomorphy, but rather a ple- siomorphy for sigmodontines. The addition of the pectoral pair was interpreted by Voss (1993) and Voss and Carleton (1993) as a derived condition within sigmodontines because the outgroups used for establishing character polarity, peromyscines and nyctomyines, do not have the pectoral pair. Nevertheless, pectoral mammae are present in almost all other muroid subfamilies ( Arvy, 1974). The reconstruction of patterns of character change in a recent comprehensive muroid phylogeny ( Jansa and Weksler, 2004) shows the presence of pectoral mammae as the ancestral condition for the sigmodontines (fig. 41). Thus, the absence of the pectoral pair in several sigmodontines lineages, such as ichthyomyines, thomasomyines, and abrothrichines, should be interpreted as apomorphic. Within oryzomyines, character reconstruction patterns indicated that the absence of the pectoral pair in Handleyomys and Scolomys was caused by two independent losses, as in both morphology-only and combined analyses these two genera were never recovered as sister taxa.

ORYZOMYINE RELATIONSHIPS

Most of the phylogenetic results of the combined analysis were either in agreement with previous phylogenies or contradicted earlier results that are weakly supported. The combined cladogram contained many strongly supported clades previously recovered in the IRBP-only analysis of Weksler (2003). Foremost are recognitions of the three higher-level major clades with most oryzomyines (clades B, C, and D), their interrelationships, and the basal position of Scolomys and Zygodontomys in the tribe. Notwithstanding, the combined analysis displayed several new hypotheses of intergeneric relationships, chiefly the union of Scolomys and Zygodontomys in a monophyletic group, and the internal topology of clades B, C, and D. The phylogenetic results also indicated that many groupings of oryzomyine taxa,

+

Patton and da Silva (1995), Goodman et al. (1999), Nowak (1999), Carleton and Goodman (2000), Lecompte et al. (2002), and Pacheco (2003).

previously recognized in formal classifications as genera ( Zygodontomys , Oligoryzomys , Oecomys , Nesoryzomys , Holochilus , Neacomys , Nectomys ) or informally as species groups within Oryzomys ( albigularis , nitidus , palustris ), are monophyletic. Finally, the results corroborate the polyphyly of Oryzomys and suggest that Sigmodontomys is paraphyletic.

Scolomys and Zygodontomys were securely placed as basal oryzomyines in the IRBPonly and combined CO analysis. In the morphology-only and combined TS analyses the two genera were recovered as sister taxa with moderate support in the latter analysis, which increased in the reduced analysis. The recovery of this clade and its position at a basal branch within oryzomyines were unexpected. Scolomys and Zygodontomys are two of the most distinctive clades of oryzomyines, and they are ecologically and morphologically dissimilar from one another. Whereas species of Scolomys are strictly forest-dwelling ( Patton and da Silva, 1995; Gómez-Laverde et al., 2004), species of Zygodontomys are highly specialized for savannas and other open vegetation formations (Voss, 1991). The IRBP divergence of both lineages was also high, suggesting an early split within oryzomyine evolution. Although the clade lacks IRBP synapomorphies, analyses using fasterevolving mitochondrial genes have also recovered it ( Garcia, 1999). This suggests an early cladogenetic event between the Scolomys and Zygodontomys lineages after the appearance of the Scolomys - Zygodontomys ancestor.

The other three large clades (B, C, and D) recovered in the combined analyses received support from the previous IRBP analysis ( Weksler, 2003). Nevertheless, relationships within the major clades are still not wellsupported, especially among the lineages in clade B. Further analyses are needed for the corroboration of the present results. With the delimitation of these clades being well secured here, separate analyses of each clade with denser taxon sampling and using varied sources of data, such as morphology and nuclear and mitochondrial genes, will allow resolution of their interrelationships.

THE ‘‘ ORYZOMYS ’’ PROBLEM

The present analysis provides compelling justification for the current generic recognition of several taxa formerly included as subgenera of Oryzomys (e.g., Microryzomys , Melanomys , Oligoryzomys , Oecomys , Sigmodontomys , and Nesoryzomys ). The results also present convincing evidence for the polyphyly of Oryzomys in its currently strict sense (i.e., Musser and Carleton, 1993), a result also congruent with previous phylogenetic analyses ( Baker et al., 1983; Patton and Hafner, 1983; Dickerman and Yates, 1995; Myers et al., 1995; Patton and da Silva, 1995; Steppan, 1995; Weksler, 1996; Percequillo, 1998; Bonvicino and Moreira, 2001; Bonvicino and Moreira, 2001; Andrade and Bonvicino, 2003; Weksler, 2003). None of the cladograms recovered from morphological, molecular, or combined data analyzed here retrieved any clade resembling Oryzomys in its currently recognized form.

Dispersion of the different Oryzomys clades over the tree was extensive, and a new taxonomic classification is obviously needed for the genus. Because the required modifications involve procedures beyond the scope of the present study, such as the designation of type species, listing of valid species (and synonyms) referred to each new genera, morphological diagnoses, and comparisons with closely related clades, the new genera are being described elsewhere (Weksler, Percequillo, and Voss, in prep.) Below, I discuss possible taxonomic arrangements.

The new classification should be compatible with the recovered phylogeny and preferably cause the least possible change in the current oryzomyine nomenclature. In addition, the new arrangement should try to recognize distinctive and diagnosable clades. The preferred arrangement is restriction of Oryzomys to the palustris group (table 2) and erection of new genera for remaining species groups or isolated species. Thus, 11 new genera would be created (fig. 42), each encompassing one ( nitidus , albigularis , talamancae , subflavus , xanthaeolus ) or multiple ( melanotis + alfaroi + chapmani ; yunganus + megacephalus ) species groups, or species recovered as independent from all other Oryzomys ( angouya , polius , balneator , hammondi ).

Other nomenclatural arrangements might provide fewer nomenclatural changes, at least in terms of creation of new names (fig. 42). For example, some species or groups of species could be absorbed into sister genera, such as the inclusion of balneator in Microryzomys , or of melanotis , chapmani , and alfaroi groups in Handleyomys . These changes, however, would otherwise change the taxonomic composition of these two genera that have detailed, unambiguous, and distinctive diagnoses ( Carleton and Musser, 1989; Voss et al., 2002). Another possible change in the proposed nomenclatural arrangement is the inclusion of albigularis and talamancae in a single genus (fig. 42). Nevertheless, the sister group relationship of these two species groups is not well secured, as demonstrated by different placements of talamancae in individual analyses of morphology and molecular data, and of the low nodal support in the combined tree. Clearly, other taxonomic arrangements are also possible, such as delimiting each major oryzomyine clade (especially B, C, and D) or each big clade within these major groups as new genera. Nevertheless, the inclusion of a variety of morphotypes and distinctive evolutionary and ecological variants would render such huge genera less useful to research of adaptation, biogeography, faunal diversification, and other topics, resulting in a classification with less heuristic value ( Wheeler, 2004).

ORYZOMYINE EVOLUTION

Oryzomyini is the most diverse tribe within the sigmodontine radiation, and this diversity is reflected in morphological and ecological variation observed among the taxa analyzed in this study. The scant available published information on oryzomyine ecology and natural history suggests that most oryzomyines are medium-sized, unspecialized, forest-dwelling, omnivorous rats, with nocturnal and cursorial habits (e.g., Flemming, 1970, 1971; Wolfe, 1982; Ernest, 1986; Janos et al., 1995; Musser et al., 1998; Nowak, 1999; Patton et al., 2000; Voss et al., 2001; Guabloche et al., 2002; Voss et al., 2002). Nevertheless, there are several conspicuous anatomical and ecological deviations from

TABLE 8

Size of Oryzomyines as Measured by Head-and-Body Length (HBL)

this generalized bauplan, which I interpret below in light of the recovered phylogeny.

SIZE: Of the 31 genera, species groups, or isolated oryzomyine lineages recovered in the analysis, 16 have species with the adult range of head-and-body length (HBL) between 90 and 175 mm (considered medium-sized; table 8). Two additional groups have HBL between 100 and 200 mm (considered largemedium) and two others have HBL between 80 and 145 (small-medium). Five groups are composed of small taxa, with HBL between 60 and 110 mm, while six groups are composed of large taxa, with HBL between 150 and 260 mm (some Nectomys and Holochilus can be as small as 125 mm, but are probably young adults). Thus 71 of 116 oryzomyines (61 %) are either medium, smallmedium, or large-medium in size, 29 (25 %) are small-sized, and 16 (14 %) are large-sized.

Inspection of body size distribution among the major oryzomyine lineages (fig. 43) reveals that all members of clade B are medium- or small-medium-sized, all members of clade C are small-sized, and all members of clade D are medium, large-medium, or large in size. O. hammondi is a large rat, while clade A includes both medium and small taxa. Reconstruction of size transformation in the recovered phylogeny (fig. 44) shows that the putative primitive condition for oryzomyines is that of a medium-sized rat. There are two transformations into the small class ( Scolomys and clade B) and four transformations into the large class ( hammondi , Lundomys +