Mammals Of The Rio Juruá And The Evolutionary And Ecological Diversification Of Amazonia Author PATTON, JAMES L. Author DA SILVA, MARIA NAZARETH F. Author MALCOLM, JAY R. text Bulletin of the American Museum of Natural History 2000 2000-01-25 2000 244 1 306 http://www.bioone.org/perlserv/?request=get-abstract&doi=10.1206%2F0003-0090(2000)244%3C0001%3AMOTRJA%3E2.0.CO%3B2 journal article 10.1206/0003-0090(2000)244<0001:MOTRJA>2.0.CO;2 0003-0090 5347311 Oryzomys Baird, 1858 Rice rats A review of the species of rice rats of the genus Oryzomys from the lowland tropical forests of South and Central America (including Amazonia, the Mata Atlântica of coastal Brazil , and the trans­Andean forests of the Pacific Coast of Ecuador and Colombia north into Central America) has been published recently by Musser et al. (1998). In this thorough and extremely important summary, these authors define, characterize, amply illustrate, and allocate all available names to 11 species, two of which they describe as new. Six of these species ( O . megacephalus , O . tatei , O . yunganus , O . macconnelli , O . nitidus , and O . emmonsae ) occur within Amazonia, and in various combinations of sympatry. In the course of their studies, Musser et al. examined most specimens of Oryzomys in our collection from the Rio Jurua´. They included these materials in both univariate and multivariate statistical summaries for each species, and they also examined patterns of morphological variation along the river among our samples that they allocated to O . megacephalus . We make no effort here to redo the more geographically extensive descriptions and analyses presented by these authors, as there is no point to duplicate the wealth of detail or exemplary illustrations. Rather, we provide sufficient information on character differences to permit the identification of each species found within the Rio Jurua´. We also summarize in a more detailed way patterns of geographic differentiation of these species within the Rio Jurua´, and provide details on population ecology and life history. Where data are available, we also place our samples in the broader context of Amazonia, summarizing comparisons between our samples from the Rio Juruá with those from elsewhere in Amazonia based on our molecular sequence data. Our results differ in a few key elements from the conclusions reached by Musser et al. (1998); we detail these differences under the appropriate sections below. PHYLOGENETIC RELATIONSHIPS AND SPECIES BOUNDARIES OF RAINFOREST ORYZOMYS Musser et al. (1998) presented our preliminary molecular sequence data in describing geographic units of many of the species of the Oryzomys ‘‘ capito ’’ complex that they characterized. These data were totally concordant with the patterns of morphological variation they described. We now have expanded this database considerably, and here present a more substantial analysis of the relationships and species boundaries hypothesized by Musser et al. (1998). We present these analyses separately for each of the two morphological groupings of Oryzomys : (1) what we term the ‘‘ megacephalus ’’ group, those taxa with, among other features, a derived cephalic arterial system ( O . laticeps , O . megacephalus , and O . yunganus ), and (2) the ‘‘ macconnelli ’’ group, those with the primitive cephalic arterial condition ( O . emmonsae , O . macconnelli , O . nitidus , and O . russatus ). THE ‘‘MEGACEPHALUS’’ GROUP: Sequences from the cytochrome­b gene are available for 37 individuals from 32 localities within Amazonia and the Mata Atlântica of coastal Brazil (fig. 96; table 34). The three species of this complex recognized by Musser at al (1998) are readily separable on the basis of sequence differences, but four strongly divergent clades, with bootstrap values of 94 or higher, are apparent (fig. 97). Two of these correspond to the ‘‘eastern’’ and ‘‘western’ clades they allocated to Oryzomys megacephalus , clades that are somewhat more divergent from each other (15.9%, on average) than either is from O . laticeps of the Mata Atlântica (12.9 and 13.4%, respectively) Oryzomys yunganus , which is sympatric with both ‘‘eastern’’ and ‘‘western’’ O . megacephalus at several localities (Musser et al., 1998), is the most divergent of the four clades, at an average of nearly 17%. Although our sampling of this species is limited, it exhibits only moderate sequence divergence averaging 5.5% among localities several thousands of kilometers distant (fig 96). This is the maximal amount of sequence divergence found within any of the four sampled species (table 35). Fig. 96. Map of localities of members of the Oryzomys megacephalus group sampled for mtDNA cytochrome­b sequences: left : O . yunganus (open circles); right : O . perenensis , solid circles; O . megacephalus , solid triangles; O . laticeps , solid squares. Locality numbers correspond to the list of provenance and voucher specimen catalog numbers provided in table 34. Distributions are based on Musser et al. (1998). Our enlarged data set presents the same topology of relationships illustrated by Musser et al. (1998; compare their fig. 6 with fig. 97). Most significantly, our broadened geographic sampling of both their ‘‘western’’ and ‘‘eastern’’ divisions of O . megacephalus continue to show that these groupings are both strongly divergent from one another with a very sharp region of transition in central Amazonia (fig. 96) and likely do not even form a monophyletic assemblage (fig. 97). Although Musser et al. (1998) could find only size differences among samples of O . megacephalus from western to eastern Amazonia, their two geographic units do differ in karyotype as well as in mtDNA sequences. All western samples that have been karyotyped (from localities 1–9, c, e, and f in fig. 96) are 2n = 52; those eastern samples (localities o, p, q, r, t, and v) are 2n = 54. The concordance of molecules, chromosomes, and body sizes suggests separate species status, despite the apparent lack of diagnosable morphological characteristics. Although a general size cline occurs among population samples across central Amazonia south of the Rio Solimões­Amazonas axis, a vast area for which no molecular or karyotypic data are available, we believe the summation of molecular, chromosomal, and size differences between the two clades identified in figure 97 and mapped in figure 96 justify species recognition. Consequently, contrary to Musser et al. (1998), we consider our samples from the Rio Juruá as well as those from throughout western Amazonia, to represent O . perenensis Allen, 1901 . With its type locality in the Departamento de Junín in eastcentral Peru´, this is apparently the earliest available name for the western Amazonian 2n = 52 taxon (Musser et al., 1998: 45). THE ‘‘MACCONNELLI’’ GROUP: We have also augmented the molecular sampling of this group of species relative to the data summarized by Musser et al. (1998), although numerous critical geographic gaps remain (fig. 98, table 36). Results of this expanded data set are concordant with the morphological analyses presented by Musser et al (1998), and thus support most hypotheses of Fig. 97. Strict consensus tree of three equal minimum­length parsimony trees for haplotypes of the mitochondrial cytochrome­b gene (801 bp; only 401 bp are available for the individual from Paraguay [Myers et al., 1995]) for members of the Oryzomys megacephalus complex of species. Length = 627 steps, CI = 0.593, RI = 0.870. Sequences of members of the Oryzomys macconnelli species complex are used as the outgroup. Bold numbers at internal nodes are bootstrap values, based on 1000 replicates percentages are mean Kimura two­parameter distances for all haplotypes below a given node. Haplotypes are identified by locality, as in the map, fig. 96, and provenance and voucher catalogue numbers are listed in table 34. species boundaries they advanced. Two exceptions, however, are evident. First, a larger number of reciprocally monophyletic and deeply divergent clades are identifiable than would be expected by the current species boundaries in the group. Specifically, samples of O . macconnelli form two strongly differentiated and geographically delimited clades, each with strong bootstrap support of 92 or 100%. These do join as a supportable monophyletic lineage, but only with a bootstrap value of 73% (fig. 99) perhaps less than would be expected if they belonged to a single taxon. Our samples from north of the Rio Solimões comprise two groups (one from east and northeast of the Rio Negro, including north of Manaus in Brazil , Venezuela , and Surinam ; the other from the Rio Jaú to the west of the Rio Negro) that differ by a substantial degree (5.7%, table 37) and that are karyotypically distinct. Specimens from Venezuela are 2n = 76, FN = 85 (Musser et al., 1998) whereas those from the Rio Jaú are 2n = 58, FN = 90 (M N. F. da Silva, unpublished data). These northern samples, however, differ from a collection of samples of O . macconnelli from south of the Rio Solimões that stretch across southern Amazonia from central Perú through the Rio Juruá to as far east as the Fig. 98. Map of localities of members of the Oryzomys macconnelli group sampled for mtDNA cytochrome­b sequences: left : O . nitidus (solid triangles) and O . legatus (open triangle); right : O . macconnelli (solid circles), O . emmonsae (solid pentagons), and O . russatus (solid squares). Localities from the Rio Juruá are numbered; other localities are given letters. Both letters and numbers correspond to the list of provenance and voucher specimen catalog numbers provided in table 36. The transition in geographic units of O . macconnelli across the Rio Solimes is indicated by the hiatus in the cross­hatched distribution of the species. Distributions are based on Musser et al. (1998). Serra dos Carajás in Estado do Pará (fig. 98). The average sequence divergence between these geographic groupings of O . macconnelli is 11.1% (table 37), a value only slightly less than that between any other pair of species in the complex (range 12.4% to 15.7%, table 37), and more than twice the variation present within any single clade. The southern mitochondrial DNA clade also has a distinct karyotype, with 2n = 64, FN = 70 reported for specimens from the Rio Juruá (see below and Musser et al., 1998). Musser et al. (1998) document clinal size variation from west to east across the range of O . macconnelli , but they lacked samples from much of the geographically intermediate area between southern Venezuela and the south bank of the Rio Solimões necessary to determine whether the size cline is smoothly graded or stepped. As with O . megacephalus and O . perenensis , detailed sampling in the central Amazon on both sides of the Rio Solimões will provide the necessary documentation of whether O . macconnelli , as hypothesized by Musser et al. (1998), represents a single species, or is composite, as is suggested by the molecular and karyotypic data we summarize here. Fig. 99. Strict consensus of two equal minimum­length parsimony trees for haplotypes of the mitochondrial cytochrome­b gene (801 bp) for members of the Oryzomys macconnelli complex of species Length = 550 steps, CI = 0.567, RI = 0.811. Sequences of members of the Oryzomys megacephalus species complex are used as the outgroup. Bold numbers at internal nodes are bootstrap values, based on 1000 replicates; percentages are mean Kimura two­parameter distances for all haplotypes below a given node. Haplotypes are identified by locality, as in the map, fig. 98, and provenance and voucher catalogue numbers are listed in table 36. The second area where the molecular sequence data suggest an alternative to the hypotheses advanced by Musser et al. (1998) is the species identification of specimens from the Departamento de Tarija in southern Bolivia . Although this issue does not involve our perspectives of Oryzomys from the Rio Jurua´, we provide the following comments both for completeness, and to suggest anoth­ er future research direction. The voucher specimens for the two haplotypes from locality p in figure 98 and table 36 were examined by Musser et al. (1998), who considered them to represent the taxon O . legatus Thomas , which they further considered a synonym for O . russatus Wagner , otherwise distributed within the Mata Atlântica of southeastern Brazil . The sequence data, however, unequivocally link these two specimens within the clade of O . nitidus haplotypes, specifically those from specimens identified as that species by Musser et al. (1998) from localities in nearby Departamento de Santa Cruz (fig. 99). Consequently, the linkage of these specimens with O . nitidus in molecular characters suggests that the morphological similarly between O . legatus and O . russatus is convergent, and not indicative of common ancestry. Certainly, these data do not support the hypothesis that legatus is a synonym of O . russatus . Given the morphological distinctness of legatus relative to samples of O . nitidus from Bolivia (Musser et al., 1998), these taxa should be considered separate species, even if their mtDNA sequences are not reciprocally monophyletic. G. G. Musser (personal commun. to J.L.P., October 27, 1997 ) wrote that they ‘‘... almost kept legatus separate as a species but could not justify it morphologically.’’ As a final point, we note our specimens from the Serra dos Carajás (locality e in the map, fig. 98, and in table 36) are of both O macconnelli and the recently described O emmonsae (Musser et al., 1998) , and thus represent the first instance of sympatry between these species. ORYZOMYS species from the Rio Juruá Four species of rice rats are present in the Rio Juruá basin: (1) Oryzomys perenensis which occurs along the entire length of the Rio Juruá and is widespread throughout western Amazonia; (2) Oryzomys yunganus morphologically similar to O . perenensis and also present throughout much of western and northern Amazonia, including the length of the Rio Jurua´; (3) Oryzomys macconnelli patchily distributed throughout much of Amazonia, as well as within the Rio Juruá basin; and (4) Oryzomys nitidus , which is limited largely to western Amazonia and the Paraná basin, and was only taken within the Headwaters Region of the Rio Jurua´. As noted above, these four are readily separable into two groups, each comprising two species, by several external and qualitative cranial features (table 38). The salient features of the skin are coloration and hair length, with both O . perenensis and O . yunganus typically dark brownish to reddish brown and with short dorsal fur. The range of color can be extensive within O . perenensis , both locally and geographically, but is much less so in O . yunganus . The pelage of O . yunganus also has a distinctive sheen that is hard to describe but evident to the expe­ rienced eye. In general, however, these two species are difficult to distinguish externally, so that identification may be quite problematical if one were only handling live animals in a longitudinal trapping program where vouchers are not prepared. Oryzomys nitidus has similar short, but yellowish or reddish brown dorsal fur, and immature individuals of this species have reddish, as opposed to grayish, fur. Finally, O . macconnelli is unique in its long and thickly luxuriant reddish brown pelage, making it one of the most readily recognizable murids in Amazonia. Both O . macconnelli and O . nitidus have the primitive cephalic arterial supply, characterized by the combination of enlarged stapedial foramen, squamosal­alisphenoid groove, and sphenofrontal foramen (pattern 1 of Voss, 1988; see Carleton and Musser, 1989). Oryzomys perenensis and O . yunganus lack the latter two features while maintaining an obvious stapedial foramen; they display a derived condition (pattern 2 of Voss, 1988; Carleton and Musser, 1989). In specimens with relatively unworn dentition, a medial fossette (­id) is present on the second molars of O . yunganus , O . macconnelli , and O . nitidus , but not in O . perenensis . Otherwise, O . perenensis and O. yunganus are quite similar morphologically, and careful examination of each specimen is required for secure identification. Additional characters, for example, which can be used to distinguish these two species include an increased degree in the development of palatal excrescences in O . perenensis , and shape of the incisive fo­ raminal septum (Gardner and Patton, 1976; Musser et al., 1998), although the differences are subtle and comparisons must be carefully matched by age. Among other characters that are useful in distinguishing between O . macconnelli and O . nitidus , respectively, are shallow versus deep zygomatic notches, narrow versus wide zygomatic plates, teardropshaped and short rather than oval and long incisive foramina, and, therefore, long versus short palates. Exhaustive descriptions accompanied by informative illustrations of each of these taxa are presented by Musser et al (1998), a work that should be examined carefully by anyone with an interest in any of these species. External and cranial dimensions of our specimens of all four species from the Rio Juruá are given in table 39. All variables exhibit strong statistical significance when comparisons are made across all taxa simultaneously (one­way ANOVA, p <0.001 for all variables except BOL and ZPL; table 40) Not surprisingly, however, individual variables exhibit combinations of significance and nonsignificance in pairwise comparisons using Fisher’s protected least significant difference, a multiple t­statistic, as a post hoc test of differences in one­way ANOVAs (table 40). The smallest species in virtually all measurements is O . yunganus , which is significantly smaller than O . perenensis in all dimensions except for IFL, AW, BOL, and ZPL. This is in contrast to other sites within western Amazonia where O . yunganus is typically larger (e.g., Quincemil, Perú [see Musser et al., 1998]). Oryzomys macconnelli is the largest species, in 15 of 24 variables, and O . nitidus is usually next largest. Thus, the combination of size dimensions coupled with coloration, fur quality, cephalic arterial pattern, and details of the second molars serve to differentiate all four species (table 38). Multivariate relationships among the four species were ascertained by discriminant analysis, using log 10 transformations of the 20 cranial variables. Standardized coefficients for each variable for the three possible discriminant axes are given in table 41. The first axis cleanly separates O . yunganus from O . perenensis (fig. 100), with taxa combining relatively short skulls with longer rostra, or the reverse, based on those characters with the highest loadings. The second axis primarily separates O . macconnelli and O . nitidus , as a pair, from the first two, while these two species become completely separated on the third axis (fig. 100). Posterior probability scores generated from the discriminant analysis allocated individual specimens nearly perfectly to their appropriate species. Three of 258 specimens of O . perenensis were misclassified as O . yunganus and two of 46 O . yunganus were misclassified as O . perenensis . All specimens of both O . macconnelli and O . nitidus were correctly classified. In the species accounts below, we present analyses of patterns of morphometric variation relative to age and sexual dimorphism and examine interpopulation differentiation along the Rio Juruá for both O . perenensis and O . yunganus . Inadequate sample sizes precluded these analyses for the other two species. We also summarize habitat and other ecological data and discuss patterns of reproduction.