This is the first installment of a revision of the didelphid marsupial genus Philander, commonly known as gray four-eyed opossums. Although abundant and widespread in lowland tropical forests from southern Mexico to northern Argentina, species of Philander are not well understood taxonomically, and the current literature includes many examples of conflicting species definitions and nomenclatural usage. Our revision is based on coalescent analyses of mitochondrial gene sequences, phylogenetic analyses of mitochondrial and nuclear genes, morphometric analyses, and firsthand examination of relevant type material. Based on these results, we provisionally recognize eight species, of which three are formally treated in this report: P. quica (Temminck, 1824), an Atlantic Forest endemic formerly known as P. frenatus (Olfers, 1818); P. canus (Osgood, 1913), a widespread species formerly treated as a synonym or subspecies of P. opossum (Linnaeus, 1758); and P. pebas, a new species endemic to Amazonia. The remaining, possibly valid, species of Philander can be allocated to two clades. The first is a cis-Andean complex that includes P. andersoni (Osgood, 1913); P. mcilhennyi Gardner and Patton, 1972; and P. opossum. The second is a trans-Andean complex that includes P. melanurus (Thomas, 1899) and P. pallidus (Allen, 1901). Among other nomenclatural acts, we designate a neotype for the long-problematic nominal taxon Didelphis superciliaris Olfers, 1818, and (in an appendix coauthored by Renate Angermann), we establish that Olfers' coeval binomen D. frenata is based on an eastern Amazonian type and is a junior synonym of P. opossum.
INTRODUCTION
Species of Philander, commonly known as “gray four-eyed opossums” or “pouched foureyed opossums” (fig. 1), occur in lowland tropical and subtropical forests from Mexico to northern Argentina.4 Closely related to other large didelphids with 22 chromosomes (tribe Didelphini: Chironectes, Didelphis, and Lutreolina), species of Philander are scansorial predators that eat a wide variety of invertebrates, small vertebrates, and fallen fruit (Charles-Dominique et al., 1981; Santori et al., 1997; Cáceres, 2004; Ceotto et al., 2009; Macedo et al., 2010). Because they are not reluctant to enter baited traps, species of Philander are abundant in museum collections, but their taxonomy has long been controversial.
Although 18 nominal taxa are currently referred to Philander (table 1), influential mid-20th century checklists (Cabrera, 1958; Hall and Kelson, 1959) recognized only a single widespread species, Philander opossum, with seven subspecies in South America and two in Central America. This hypothesis, implying reproductive continuity among populations spanning numerous zoogeographic barriers (mountains, rivers, open habitats) and many thousands of kilometers, persisted until Gardner and Patton (1972) reported sympatry between two phenotypically distinguishable species in western Amazonia. Subsequent publications have reported other examples of sympatry between two kinds of Amazonian Philander (Hutterer et al., 1995; Patton et al., 2000; Hice and Velazco, 2012), and DNA-sequencing studies have discovered deep genetic divergence—equivalent to that seen between sympatric Amazonian taxa—among allopatric forms that have long been considered synonyms or subspecies of P. opossum (see Patton and da Silva, 1997; Nunes et al., 2006; Chemisquy and Flores, 2012).
Although there is now broad consensus that multiple valid species of Philander merit recognition, authors disagree about what to call them (e.g., Patton and da Silva, 1997; Hershkovitz, 1997), and there are significant problems with the data currently available to test alternative taxonomic hypotheses. Among other difficulties, major geographic sampling gaps complicate the interpretation of phenotypic and genetic differences, morphological analyses of voucher specimens are often insufficient to support the application of names to haplotype groups, and nuclear-gene sequences are unavailable to assess whether mitochondrial haplogroups are really species. Additionally, no revision of the genus has been based on firsthand examination of relevant type material.
This is the first of several technical reports on the taxonomy of Philander. In this installment we analyze the most extensive set of mitochondrial DNA (mtDNA) sequences yet assembled for the genus. Additionally, we obtained sequence data from several nuclear markers that we use to test inferences about phylogenetic relationships previously based exclusively on mtDNA. Coalescent analyses of the mtDNA data, together with phenotypic information obtained from morphological vouchers, type material, and other specimens, support the provisional recognition of eight species, of which three are formally treated as valid in this report.
FIG. 1.
A member of the Philander melanurus complex attacking a large specimen of the venomous elapid snake Micrurus nigrocinctus (photo credit: Mario J. Gómez-Martínez). The pale supraocular spots and ashy dorsal coloration are diagnostic external traits of the genus Philander.
Materials and Methods
Source of Material: Except as noted, all specimens are preserved in the following collections: AMNH (American Museum of Natural History), New York; BMNH (Natural History Museum), London; CM (Carnegie Museum of Natural History), Pittsburgh; EBD (Estación Biológica Doñana), Seville; EBRG (Museo de la Estación Biológica de Rancho Grande), Maracay; FMNH (Field Museum), Chicago; INPA (Instituto Nacional de Pesquisas da Amazônia), Manaus; ISEM (Institut des Sciences de l'Evolution), Montpellier; KU (Biodiversity Research Center, University of Kansas), Lawrence; LSUMZ (Museum of Zoology, Louisiana State University), Baton Rouge; MACN (Museo Argentino de Ciencias Naturales “Bernardino Rivadavia”), Buenos Aires; MCZ (Museum of Comparative Zoology, Harvard University), Cambridge; MHNLS (Museo de Historia Natural La Salle), Caracas; MN (Museu Nacional), Rio de Janeiro; MPEG (Museu Paraense Emílio Goeldi), Belém; MSB (Museum of Southwestern Biology, University of New Mexico), Albuquerque; MSU (Michigan State University Museum), East Lansing; MUSM (Museo de Historia Natural de la Universidad Nacional Mayor de San Marcos), Lima; MVZ (Museum of Vertebrate Zoology, University of California), Berkeley; MZUSP (Museu de Zoologia da Universidade de São Paulo), São Paulo; NMW (Naturhisorisches Museum Wien), Vienna; RMNH (Naturalis Biodiversity Center), Leiden; ROM (Royal Ontario Museum), Toronto; TTU (Museum of Texas Tech University), Lubbock; UFES (Universidade Federal do Espírito Santo), Vitória; UFMG (Universidade Federal de Minas Gerais), Belo Horizonte; USNM (National Museum of Natural History, Smithsonian Institution), Washington D.C.; ZMB (Museum für Naturkunde), Berlin.
TABLE 1.
Nominal species-group taxa currently referred to Philander.a
FIG. 2.
Cis-Andean collection localities for sequenced specimens of Philander. Numbers are keyed to localities listed in the gazetteer (appendix 1).
Taxon Sampling and Laboratory Methods: The sequences analyzed in this report were obtained from specimens representing most of the phenotypes that have previously been considered valid taxa of Philander (e.g., by Cabrera, 1958; Hall and Kelson, 1959; Patton and da Silva, 1997; Hershkovitz, 1997). Sequenced specimens include paratypes, topotypes, and otherwise geographically referable material of andersoni, azaricus, canus, crucialis, frenatus, fuscogriseus, mcilhennyi, melantho, melanurus, mondolfii, olrogi, opossum, pallidus, and quica. To the best of our knowledge, the only nominal taxa not represented by these molecular data are deltae (currently known only from a handful of specimens in Venezuelan museums) and nigratus (for which we were unable to obtain tissue).
FIG. 3.
Trans-Andean collection localities for sequenced specimens of Philander. Numbers are keyed to localities listed in the gazetteer (appendix 1).
We extracted DNA from preserved tissue or from fragments of dried skin obtained from museum specimens as explained below. We also downloaded sequence data deposited in Gen-Bank by previous researchers, and several unpublished sequences were kindly made available to us by the Patton lab at the University of California at Berkeley. Careful checking of these data for provenance revealed that three pairs of GenBank accessions are duplicates (based on the same tissue/specimen; asterisks indicate the sequence used by us): JQ778972 and KT153576* were both obtained from MVZ 197405 (field number JLP 16968); GU112937 and U34679* were both obtained from a tissue with identifier “ORG 01,” apparently corresponding to an uncataloged specimen at the Museu Nacional (Rio de Janeiro); and JQ778971 and JF281029* were both obtained from MZUSP 29212 (field number MAM 208). The sequences we analyzed are listed in tables 2 and 3 with voucher, tissue, and GenBank identifiers. Although we tried to examine morphological voucher material for every sequence analyzed in this report, we were not entirely successful in doing so (examined voucher specimens are marked with asterisks in table 2). The collection localities of sequenced ingroup (Philander) specimens are mapped in figures 2 and 3.
Laboratory Methods: We extracted DNA from preserved tissues or dried museum specimens using methods described in Voss and Jansa (2009) and Giarla et al. (2010). To minimize risk of contamination, all extractions from museum specimens were performed in an isolated laboratory where mammalian polymerase chain reaction (PCR) products were not present. We PCR-amplified six loci for this study (CYTB, BRCA1, IRBP, OGT, SLC38, and Anon128) using the primers listed in appendix 2 and methods described in Voss and Jansa (2009), Giarla et al. (2010, 2014), Gutiérrez et al. (2010), and Pavan et al. (2014). The resulting PCR products were Sanger-sequenced on an ABI 3730xl automated sequencer. Sequences were edited and assembled in Geneious Pro ver. 7.0 ( http://www.geneious.com; Kearse et al., 2012), and length heterozygotes in the nuclear loci were resolved using Indelligent v. 1.2 (Dmitriev and Rakitov, 2008). Individual genes were aligned using the default parameters of MUSCLE (Edgar, 2004), and alignments of all protein-coding genes were examined with reference to translated amino-acid sequences.
Phylogenetic and Coalescent Analyses: We performed maximum-likelihood and Bayesian phylogenetic analyses of a cytochrome-b (CYTB) matrix that included sequences obtained from 135 specimens of Philander together with several outgroup sequences (tables 2, 3). The best-fitting nucleotide substitution model for these data was determined under the corrected Akaike Information Criterion (AICc) in jModelTest (Posada, 2008). We conducted five independent maximum-likelihood (ML) searches in GARLI 2.0 (Zwickl, 2006) and evaluated nodal support based on bootstrap analyses of 1000 pseudoreplicated datasets with the same parameters as the initial searches. Bootstrap support (BS) values were summarized on the best ML tree using Sumtrees version 3.3.1 (Sukumaran and Holder, 2010). Bayesian inference (BI) was implemented in MrBayes v3.2 (Ronquist et al., 2012) by running two independent Markov Chain Monte Carlo (MCMC) analyses for 50 million generations each, sampling every 5000 generations and including one cold chain and three heated chains. To evaluate convergence, the results of the MCMC runs were inspected in Tracer v1.6 (Rambaut et al., 2014). We discarded the first 50% of trees from each run as burnin, combined the remaining trees into a final set of 10,000 trees, and summarized all parameters in a maximum-cladecredibility tree with TreeAnnotator v1.7.2 (Drummond et al., 2012). All phylogenetic analyses (including those described in subsequent paragraphs) were implemented in the CIPRES Science Gateway (Miller et al., 2010). We estimated uncorrected genetic distances (p-distances) within and among putative species using MEGA7 (Kumar et al., 2016).
To delimit putative species from our CYTB sequence data we first constructed an ultrametric tree in BEAST v1.7.2 (Drummond et al., 2012) including only unique haplotypes across the aligned region (125 terminals; table 2); we used a lognormal relaxed-clock model, a coalescent constant-size tree prior, and relative time set with a prior on the ingroup age of one (normal distribution: mean = 1, SD = 0.01). We ran two independent MCMC analyses for 50 million generations each, sampling every 5000 generations. Procedures for assessing convergence, fractions of trees discarded as burnin, and the summarization process followed those described above for the MrBayes analysis. To estimate the threshold between interspecific and intraspecific branching processes we used the likelihood version of the General Mixed Yule Coalescent model (GMYC) as implemented in SPLITS (Pons et al., 2006). GMYC analyses were implemented on the ultrametric topology recovered by BEAST allowing only single shifts across the phylogeny (Fujisawa and Barraclough, 2013). For the purposes of this report, we recognize as putative species any CYTB lineage with strong support from either ML or BI that crosses the estimated threshold between cladogenetic and coalescent branching processes.
TABLE 3.
Ingroup and outgroup specimens sequenced for multigene phylogenetic analyses.
Lastly, we analyzed two concatenated-gene matrices that included a single representative from each putative species, the first matrix containing only the five nuclear loci, and the second containing the five nuclear loci plus cytochrome b. Both matrices were concatenated using Sequence Matrix 1.8 (Vaidya et al., 2011). We used the Bayesian information criterion and a greedy search algorithm (implemented in PartitionFinder; Lanfear et al., 2012) to identify the best partitioning scheme and substitution models. For each matrix, we partitioned proteincoding genes (BRCA1, CYTB, IRBP) by locus and codon position, whereas noncoding genes (Anon128, OGT, SLC38) were partitioned only by locus. We performed partitioned ML and BI phylogenetic analyses on each dataset following the methods and software described for the CYTB analyses described above, with the unique exception that we ran 5 million generations sampling each 500 generations on each MCMC analysis of MrBayes.
Craniodental Measurements: Craniodental measurements were taken with digital calipers as skulls were viewed under low (6–12×) magnification. Measurement values were recorded to the nearest 0.01 mm, but those reported herein are rounded to the nearest 0.1 mm. The following dimensions were measured (fig. 4):
Condylobasal length (CBL): measured from the occipital condyles to the anteriormost point of the premaxillae.
Nasal length (NL): the anteroposterior dimension of the longest intact nasal bone.
Nasal breadth (NB): measured between the triple-point sutures of the nasal, frontal, and maxillary bones on each side.
Least interorbital breadth (LIB): measured at the narrowest point across the frontals between the orbits (anterior to the postorbital processes).
Least postorbital breadth (LPB): measured at the narrowest point across the frontals between the temporal fossae (behind the postorbital processes).
Zygomatic breadth (ZB): measured at the widest point across both zygomatic arches.
Palatal length (PL): measured from the anteriormost point of the premaxillae to the postpalatine torus, including the postpalatine spine (if present).
Palatal breadth (PB): measured across the labial margins of the M4 crowns, at or near the stylar A position.
Maxillary toothrow length (MTR): measured from the anterior margin of C1 to the posterior margin of M4.
Length of molars (LM): measured from the anteriormost labial margin of M1 to the posteriormost point on M4.
Length of M1–M3 (M1–M3): measured from the anteriormost labial margin of M1 to the posteriormost point on M3.
Width of M3 (WM3): measured from the labial margin of the crown at or near the stylar A position to the lingual apex of the protocone.
Age Criteria: Unless otherwise noted below, we recorded measurements and scored qualitative morphological data from adult specimens only. Following Voss et al. (2001), a specimen was judged to be juvenile if dP3 was still in place; subadult if dP3 had been shed but P3 and/or M4 was still incompletely erupted (M4 is the last upper tooth to erupt in Philander); and adult if the entire permanent upper dentition (I1–5, C1, P1–3, M1–4) was fully erupted.
Morphometric Analyses: Adult male specimens of Philander are about 3% to 5% larger, on average, than conspecific adult females in most measured craniodental dimensions, so we tabulate descriptive sample statistics separately by sex. After the molar toothrow is fully erupted (in young adults), the measurement LM is ontogenetically invariant, so we often use LM as a univariate surrogate for size when comparing species. Estimates of central tendency and dispersion mentioned in these accounts (e.g., 12.4 ± 0.5 mm) are the sample mean plus or minus one sample standard deviation.
FIG. 4.
Dorsal and ventral cranial views and occlusal view of the maxillary dentition of Philander opossum, showing the anatomical limits of craniodental measurements defined in the text.
Because males are more numerous than females in our samples, we computed multivariate sample comparisons from adult male measurements. For the multivariate analyses reported below we deleted two measurements (MTR, M1–M3) that redundantly index variation along the same anterior-posterior dental axis as LM, we eliminated all specimens with missing values, and we log10-transformed our data. We computed generalized (Mahalanobis) distances among our samples and summarized the similarity structure of the resulting distance matrix using cluster analysis (with the Unweighted Pair-Group Method with Arithmetic Means, UPGMA).5
We extracted principal components from the variance-covariance matrix computed from log-transformed measurements for selected pairs of samples, and we inspected specimen scores in two-dimensional projections to assess sample overlap on the first several axes. On the assumption that the first eigenvector of the pooled within-group covariance matrix is an appropriate estimate of general size (growth, including ontogenetic allometry; Jolicoeur, 1963; Bookstein et al., 1985), we used Burnaby's (1996) method to obtain size and size-invariant shape factors for pairwise sample ordinations (Rohlf and Bookstein, 1987). All multivariate computations were made using NTSYS Version 2.2 (Exeter Software, Setauket, NY).
MOLECULAR RESULTS
Analyses of Cytochrome-b Sequence Data
The 135 cytochrome-b sequences we obtained from specimens of Philander ranged in length from 285 to 1149 bp (table 2), resulting in 74.6% overall nucleotide coverage for this matrix. The best-fitting nucleotide substitution model for these data was GTR+I+Γ, and the optimal topologies recovered from each of our independent analyses (ML, MrBayes, BEAST) were nearly identical, differing only with respect to weakly supported details. All three analyses provided strong support for the monophyly of Philander and for several groups that we interpret as multispecies clades (fig. 5). Nine lineages cross the GMYC species threshold, but two of these are not strongly supported by nodal statistics. For the purposes of this report, we recognize eight putative species, seven of which can be associated with available names based on phenotypic and geographic criteria.
Sister to all other putative species of Philander is a haplogroup from southeastern Brazil, for which the oldest available name is P. quica. The remaining putative species belong to a single strongly supported clade, but the two deepest nodes within this clade are not strongly supported. Among the robustly supported groups we recovered are: (1) a western Amazonian haplogroup that corresponds to P. andersoni, (2) a pair of trans-Andean haplogroups that we identify as P. melanurus and P. pallidus, (3) a pair of cis-Andean haplogroups that we identify as P. mcilhennyi and P. opossum, and (4) another pair of cis-Andean haplogroups that correspond to P. canus and a new species (P. pebas). Percent pairwise uncorrected sequence divergence among these putative species ranges from 1.8% (between P. canus and P. pebas) to 11.9% (between P. pallidus and P. quica; appendix 3).
FIG. 5.
Ultrametric tree from BEAST analysis of cytochrome-b sequences of Philander with putative species represented as cartooned terminals. Dashed vertical line indicates the threshold between Yule and coalescent processes as estimated by the likelihood implementation of the general mixed Yule coalescent model (GMYC). Bases of triangles at branch tips are proportional to the number of sequences belonging to each clade. Filled semicircles at each internal node indicate strong support from Bayesian (BEAST: PP) and maximum-likelihood (GARLI: BS) analyses of these data.
Analyses of Concatenated-gene Datasets
Our concatenated-gene alignments contained 4280 bp for the nuclear-gene (nucDNA) dataset and 5429 bp for the combined nuclear and mitochondrial dataset (CYTB+nucDNA); the nuclear sequence data in these alignments include 619 bp from Anon128, 946 bp from BRCA1, 1158 bp from IRBP, 653 bp from OGT, and 904 bp from SLC38. Each dataset was analyzed using the partitions and DNA substitution models listed in table 4. Phylogenetic analyses of both datasets yielded identical topologies with differences observed exclusively in nodal support (fig. 6). Philander quica remains the first-diverging species of the genus, but relationships among the other species are rearranged from those previously observed from our CYTB analyses. Most importantly, P. andersoni is now recovered as the sister taxon of P. mcilhennyi + P. opossum, and this trio of Amazonian endemics is resolved as the sister group of the previously described trans-Andean lineage (P. melanurus + P. pallidus). Interestingly, although these relationships are robustly supported by the combined (nuclear + mitochondrial) data, some nodes are only weakly supported by the nuclear genes analyzed separately. However, the monophyly of Philander is strongly supported by both datasets, as is the P. opossum complex (P. andersoni + P. mcilhennyi + P. opossum), the P. melanurus complex (P. melanurus + P. pallidus), and the sister-group relationship between the two last-named clades.
TABLE 4.
Optimal partitioning schemes and substitution models for two concatenated-gene datasets.
Morphometric Analyses
Generalized distances (Mahalanobis D values) computed from craniodental measurements of adult male specimens representing the putative species identified by coalescent analysis of our molecular data range from about 1.7 to 7.4 (appendix 4). Notably higher values (mostly >5.0) were obtained for comparisons of P. quica and P. canus with other congeneric taxa, whereas lower values (mostly <4.0) were obtained for comparisons between P. melanurus and P. pallidus and among members of the P. opossum complex. This similarity structure can be heuristically summarized by cluster analysis, the results of which (fig. 7) clearly illustrate the wide divergence of P. quica and P. canus from other congeneric taxa. Principal-components analyses of selected pairs of taxa (see below) suggest that generalized distance values >4.5 are associated with nonoverlapping multivariate distributions.
FIG. 6.
Result of Bayesian analysis of concatenated sequence data from cytochrome b and five nuclear loci (Anon128, BRCA1, IRBP, OGT, SLC38) from exemplar specimens of each putative species (table 3). Gray boxes provide nodal support statistics (PP/BS) from analyses of nuclear genes only.
DISCUSSION
Like many other systematists (e.g., deQueiroz, 1998, 2007), we regard species as evolutionarily independent lineages. Widely accepted evidence for the evolutionary independence of candidate species includes—but is not limited to—reciprocal mtDNA monophyly, substantial sequence divergence, phenotypic diagnosability, ecological differences, and sympatry. In our opinion, none of these criteria is necessary or sufficient for recognizing species, but as such evidence accumulates, the case for species recognition becomes more compelling.
The putative species identified by the molecular analyses reported above correspond to reciprocally monophyletic mtDNA haplogroups that exhibit levels of sequence divergence equaling or exceeding the estimated threshold value between coalescent and branching processes. However, neither GMYC nor any other delimitation method based exclusively on genetic sequence data provides an infallible guide for recognizing species (Carstens et al., 2013; Sukumaran and Knowles, 2017), so it is important to assess our molecular results for congruence with other lines of evidence. The nongenetic evidence at hand consists of morphology, geographic distributions, and ecology, which we briefly review here in advance of formal taxonomic treatment. Additionally, the following paragraphs serve to explain the synonymies implicit in our use of binomina for putative species.
FIG. 7.
Dendrogram resulting from UPGMA clustering of putative species of Philander using generalized distances computed from log-transformed craniodental measurement data (appendix 4).
Three Highly Corroborated Species
Three of the putative species identified by our interpretation of the GMYC results are strongly supported as independent evolutionary lineages by other types of evidence, and we are confident that they are valid species.
Philander quica: In addition to its wide genetic divergence from other congeneric taxa (p-distances ≥9.9%; appendix 3), this species is morphologically and biogeographically distinctive. It is smaller than all other species with the exception of P. canus—from which it can be distinguished by qualitative craniodental traits (see below)—and it is the only species of Philander that occurs in the Atlantic Forest (Mata Atlântica), a well-known center of vertebrate endemism. It is not currently known to occur sympatrically with any other congener (table 5), but its geographic range must contact that of P. canus in eastern Brazil, eastern Paraguay, and northeastern Argentina and might historically have contacted that of P. opossum in northeastern Brazil. This species was formerly widely known as P. frenatus based on erroneous information about where the type of frenatus was collected (appendix 5). Thomas's (1923) name azaricus is a synonym, but Olfers' (1818) superciliaris is not.
TABLE 5.
Geographic relationships among putative species of Philander.
Philander canus: Long considered a subspecies of P. opossum (e.g., by Cabrera, 1958; Patton and da Silva, 1997, 2008; Gardner, 2005), these phylogenetically remote taxa are conspicuously divergent in molecular and morphometric traits. Philander canus is widely distributed across several cis-Andean biomes and is known to occur sympatrically with P. andersoni, P. mcilhennyi, and P. pebas (table 5). The names crucialis, mondolfii, and olrogi are junior synonyms.
Philander pebas: This new Amazonian species is easily distinguished from all other species of Philander by dental morphology, and it additionally differs from its sister taxon, P. canus, in size and pelage pigmentation. It occurs sympatrically with P. andersoni, P. canus, and P. mcilhennyi (table 5). Whereas P. pebas apparently occurs in seasonally flooded habitats and secondary growth, sympatric congeners typically occur in upland (unflooded) primary forest.
Five Problematic Species
The remaining putative species form a clade of morphometrically similar allopatric taxa. Although pelage traits distinguish P. mcilhennyi and P. andersoni from each other and from P. opossum, morphological diagnoses of P. opossum, P. melanurus, and P. pallidus are difficult to formulate based on material examined to date. Given that P. andersoni and P. mcilhennyi are currently recognized as valid species (Patton and da Silva, 2008) and that we do not currently have any compelling evidence to suggest otherwise, there are only two nomenclatural options that merit consideration. One is to recognize a paraphyletic P. opossum with one subspecies in eastern Amazonia (P. o. opossum) and two others that only occur west of the Andes (P. o. melanurus, P. o. pallidus). To our knowledge, no other animal species shares this disjunct distribution. The second option, which we adopt below, is to treat all five putative species in this complex as provisionally valid, with the caveat that three of them are not yet certainly distinguishable except by mtDNA sequence characteristics.
Philander melanurus: This is the oldest available name for a robustly supported haplogroup that includes specimens from western Ecuador, western Colombia, and eastern Panamá. By comparison with specimens from southern Mexico and northern Central America that we refer to P. pallidus, these are darker-furred animals with a marked tendency to have shorter white tail-tips; in fact, some specimens from northwestern Ecuador and southwestern Colombia are mostly blackish and have all-dark tails. The nominal taxa fuscogriseus, grisescens, and melantho are junior synonyms.
Philander pallidus: This is the only available name for a strongly supported haplogroup that occurs in southern Mexico, Belize, and El Salvador. As noted above, examined specimens from these regions (including those sequenced for this study) are paler-furred than specimens from Panamá, western Colombia, and western Ecuador that we refer to P. melanurus, and most of them have longer white tail-tips (none has all-dark tails). Whether these phenotypes intergrade somewhere in the wide Central American hiatus from which we lack sequence data (Guatemala, Honduras, Nicaragua, and Costa Rica), and whether genetically intermediate haplogroups occur in the same region, are obvious topics for future research.
Philander andersoni: This is the only available name for a robustly supported haplogroup that occurs in northeastern Peru, eastern Ecuador, southeastern Colombia, southern Venezuela, and northwestern Brazil (north of the Amazon and west of the Rio Negro). It is known to occur sympatrically with P. canus and P. pebas in different pairwise combinations (all three species have yet to be found at the same locality). Sequenced specimens that we examined have a distinct middorsal stripe of blackish fur, pale preauricular spots, mostly black hind feet, and at least half-white tails, but other external and cranial traits are variable. Patton and da Silva (1997, 2008) listed nigratus (from southeastern Peru) as a synonym of P. andersoni, but specimens of nigratus are larger animals that (among other differences) lack a distinct middorsal blackish stripe and have only short white tail-tips.
Philander mcilhennyi: This is the only available name for a robustly supported haplogroup that is currently known to occur south of the Amazon in eastern Peru and western Brazil, where it is known to occur sympatrically with P. canus and P. pebas. Many sequenced specimens (and most other referred material) are phenotypically distinctive, with almost uniformly blackish pelage, but some sequenced specimens that we refer to P. mcilhennyi on the basis of haplogroup membership (e.g., AMNH 273055, 273089) resemble P. andersoni in pelage coloration, and the other external and craniodental characters by which Patton and da Silva (1997, 2008) diagnosed these taxa do not appear to consistently distinguish them.6 Not unreasonably, Hershkovitz (1997) ranked mcilhennyi as a subspecies of P. andersoni, but as we did not recover andersoni and mcilhennyi as sister groups, we provisionally treat both as valid species.
Philander opossum: By contrast with traditional usage, we restrict P. opossum to the large, uniformly gray form with a long white tail-tip that occurs throughout the Guianas (Guyana, Surinam, French Guiana) and the eastern part of Amazonian Brazil (Amapá, Pará, Roraima, and part of Amazonas). In terms of physical geography, Brazilian populations of this species occur east of the Rio Negro (along the north bank of the Amazon) and east of the Rio Madeira (along the south bank). As understood in this report (see Remarks under P. quica and appendix 5, below), P. opossum includes frenatus and superciliaris as junior synonyms.
TAXONOMIC ACCOUNTS
The following accounts include an emended description of the genus Philander, redescriptions of P. quica and P. canus, and a description of our new Amazonian species, P. pebas. Additionally, these accounts serve to summarize geographic distributions, comment on relevant issues of nomenclature and identification, and list the morphological specimens we examined. Our abbreviated synonymies include only original descriptions (subsequent name combinations can be found in Patton and da Silva, 2008). Qualitative morphological comparisons of P. quica, P. canus, and P. pebas are summarized in table 6, and descriptive statistics are summarized in tables 7 and 8. Morphological comparisons with other species are restricted to members of the cis-Andean P. opossum complex.
Philander Brisson, 1762
Type Species: Didelphis opossum Linnaeus, 1758, by plenary action of the International Commission on Zoological Nomenclature (ICZN, 1998).
Contents: Based on evidence summarized in this report, we tentatively recognize the following eight species as valid (synonyms in parentheses): andersoni Osgood, 1913; canus Osgood, 1913 (including crucialis Thomas, 1923; mondolfii Lew et al., 2006; and olrogi Flores et al., 2008); mcilhennyi Gardner and Patton, 1972; melanurus Thomas, 1899 (including fuscogriseus Allen, 1900; grisescens Allen, 1901; and melantho Thomas, 1923); opossum Linnaeus, 1758 (including frenatus Olfers, 1818; and superciliaris Olfers, 1818); pallidus Allen, 1901; pebas, new species (described below); and quica Temminck, 1824 (including azaricus Thomas, 1923).
In the absence of genetic information, we are currently unable to assess the validity of deltae Lew et al., 2006, and nigratus Thomas, 1923, either or both of which might also be good species.
Description:7 Combined length of adult head and body ca. 250–350 mm; adult weight ca. 280–700 g. Rhinarium with one ventrolateral groove on each side of median sulcus; dark circumocular mask present, usually continuous with dark coronal fur; pale supraocular spots present; dark midrostral stripe absent; throat gland absent. Dorsal pelage unpatterned-grayish or -blackish, or with grayish flanks and black middorsal stripe when fresh (foxing to brownish tones in old museum skins); dorsal underfur gray; dorsal guard hairs usually short (but sometimes much longer middorsally than along flanks in P. mcilhennyi); ventral fur variously pigmented (self-whitish or -buffy, gray-based buffy or cream, or entirely grayish; variable within and among species). Manus mesaxonic (dIII > dIV); manual claws about as long as fleshy apical pads of digits; dermatoglyph-bearing manual plantar pads present; central palmar epithelium smooth or sparsely tüberculate; carpal tübercles absent. Pedal digits unwebbed; dIV longer than other pedal digits; plantar surface of heel naked. Pouch present, opening anteriorly; mammae usually 2–1–2 = 5 or 3–1–3 = 7; cloaca present. Tail longer than combined length of head and body, slender and muscular (not incrassate); furred dorsally and ventrally to about the same extent at base; naked caudal integument blackish proximally and abruptly whitish distally (but a few specimens of some species have all-black tails); caudal scales in spiral series, each scale with 4–6 bristlelike hairs emerging from distal margin; ventral caudal surface modified for prehension distally, with apical pad bearing dermatoglyphs.
FIG. 8.
Dorsal, ventral, and lateral cranial views of Philander opossum (based primarily on AMNH 266387, an adult female from Paracou, French Guiana).
FIG. 9.
Collection localities of examined specimens of Philander quica, P. canus, and P. pebas. The symbol for sympatry marks localities where P. canus and P. pebas have been collected together.
TABLE 6.
Morphological and geographical comparisons among three species of Philander.
Skull in general aspect (fig. 8) smaller and less robust than that of Didelphis (which it otherwise resembles). Premaxillary rostral process absent. Nasals short, usually not extending anteriorly above I1 (exposing nasal orifice in dorsal view), and widened posteriorly near maxillary-frontal suture. Maxillary turbinals elaborately branched. Lacrimal foramina usually two on each side, exposed laterally on orbital margin or on face just anterior to orbit. Interorbital region smoothly rounded, without supraorbital beads or crests; short, blunt postorbital processes usually present in large adult specimens. Left and right frontals coossified (midfrontal suture incomplete or absent), but left and right parietals separated by persistent midparietal suture. Parietal and alisphenoid in contact on lateral braincase (no frontal-squamosal contact). Sagittal crest present, well developed on parietals and extending anteriorly onto frontals. Petrosal not laterally exposed through fenestra in squamosal-parietal suture (fenestra absent). Parietal-mastoid contact absent (interparietal narrowly contacts squamosal).
Maxillopalatine and palatine fenestrae present; maxillary fenestrae absent; posterolateral palatal foramina small, not extending anteriorly between M4 protocones; posterior palate (behind toothrows) with prominent lateral corners, the choanae constricted behind. Maxillary and alisphenoid usually separated by palatine on floor of orbit (but maxillary-alisphenoid contact occurs unilaterally or bilaterally in a few specimens). Transverse canal foramen usually present. Alisphenoid tympanic process small and uninflated, usually with broad lamina enclosing extracranial course of mandibular nerve (secondary foramen ovale present), and not contacting rostral tympanic process of petrosal. Anterior limb of ectotympanic indirectly suspended from basicranium (by malleus). Stapes usually triangular with large obturator foramen. Fenestra cochleae exposed (not concealed by rostral and caudal tympanic processes of petrosal). Paroccipital process large, erect, directed posteroventrally. Dorsal margin of foramen magnum bordered by exoccipitals only (incisura occipitalis absent).
TABLE 7.
Summary statisticsa for craniodental measurements (mm) of adult male specimens of Philander quica, P. canus, and P. pebas.
Two mental foramina present on lateral surface of each hemimandible; angular process acute and strongly inflected.
Unworn crowns of I2–I5 with much longer anterior than posterior cutting edges. Upper canine (C1) alveolus in premaxillary-maxillary suture; C1 simple, without accessory cusps. First upper premolar (P1) smaller than posterior premolars but well formed and not vestigial; third upper premolar (P3) taller than P2; P3 with posterior cutting edge only; upper milk premolar (dP3) large and molariform. Molars highly carnassialized (postmetacristae conspicuously longer than postprotocristae; relative widths M1 < M2 < M3 < M4; centrocrista only weakly inflected labially on M1–M3; ectoflexus usually distinct only on M3; anterolabial cingulum and preprotocrista discontinuous (anterior cingulum incomplete) on M3; postprotocrista with carnassial notch. Last upper tooth to erupt is M4.
Lower incisors (i1–i4) without distinct lingual cusps. Lower canine (c1) erect, acutely pointed, and simple (without a posterior accessory cusp). Second lower premolar (p2) much taller than p3; lower milk premolar (dp3) large and molariform with complete (tricuspid) trigonid. Hypoconid labially salient on m3; hypoconulid twinned with entoconid on m1–m3; entoconid much taller than hypoconulid on m1–m3.
TABLE 8.
Summary statisticsa for craniodental measurements (mm) of adult female specimens of Philander quica, P. canus, and P. pebas.
Philander quica (Temminck, 1824)
Didelphis quica Temminck, 1824: 36; type locality (fixed by lectotype selection; Hershkovitz, 1959: 342) “Sapitibi” (= Sepetiba at 22°58′ S, 43°42′ W; Paynter and Traylor, 1991), Rio de Janeiro, Brazil.
Metachirus opossum azaricus Thomas, 1923: 604; type locality “Sapucay” (= Sapucaí at 25°41′ S, 56°57′ W; Paynter, 1989), Paraguarí, Paraguay.
Type Material: Temminck (1824: 36–38) based his description of Didelphis quica on an unknown number of specimens from various museums, presumably including one or more examples collected by Johann Natterer, whose information about the species was prominently acknowledged (“Nous devons à M. Natterer la connaissance plus exacte de cette espèce qu'il a envoyée au musée impérial de Vienne . . .”). In a published catalog of Natterer's mammals, Pelzeln (1883: 110–111) listed two specimens of D. quica, male and female, collected in 1818 at “Sapitiba” (= Sepetiba) near Rio de Janeiro. Of these, Hershkovitz (1959) designated the female as lectotype, apparently sight unseen. Unfortunately, this specimen is no longer in Vienna, where only the male topotype (NMW 7687/ST 1012) can now be found. The female may have been exchanged or gifted to Temminck, whose cabinet was subsequently transferred to the Leiden museum (formerly the Rijksmuseum van Natuurlijke Historie, now the Naturalis Biodiversity Center; S. Engelberger, personal commun., 26 November 2014), but no specimen currently in Leiden can be positively identified as Hershkovitz's lectotype (S. van der Mije, 25 November 2014).
In the absence of any compelling evidence to the contrary, we accept Hershkovitz's (1959) lectotype designation as valid. Although the specimen in question appears to have been lost (or to be unidentifiable), the fixation of the type locality is sufficient for confident application of Temminck's epithet to the Atlantic Forest species of southeastern Brazil. The male topotype mentioned above (consisting of the mounted skin and extracted skull of a very old animal with much-faded pelage and teeth worn away almost to the roots) is not taxonomically informative, but other examined specimens of Philander from the Brazilian state of Rio de Janeiro (in the AMNH and ZMB; see below) exhibit all the diagnostic traits that we attribute to P. quica in the description that follows.
Distribution and Sympatry: Sequenced material and examined specimens that we assign to Philander quica are from rainforested tropical and subtropical landscapes in southeastern Brazil, northeastern Argentina (Misiones), and eastern Paraguay (fig. 9). Although P. quica is the only species of Philander known to occur throughout this biome (the Mata Atlântica of Brazilian authors), it might occur sympatrically with P. canus along its margins, where Atlantic rainforests come into contact with (or grade into) the gallery formations and semideciduous forests apparently preferred by the latter species. Additionally, the range of P. quica might contact that of P. opossum somewhere along the coastline between Bahia and Rio Grande do Norte.8 Published records of this species from the Cerrado, the Chaco, and western Amazonia (in Patton and da Silva, 1997, 2008; Hershkovitz, 1997) are based on misidentifications (see Remarks, below).
Description: Dorsal pelage short (usually <14 mm) and uniformly grayish, sometimes indistinctly darker along the midline but never with a distinctly blackish middorsal stripe (fig 10); fur of crown (between the ears) usually grizzled gray but sometimes blackish; pale preauricular spot absent or indistinct; ventral fur pale, usually self-whitish or -yellowish, sometimes with broad lateral zones of gray-based hairs on the throat and between the fore- and hind legs, but apparently always self-colored in the midline (fig. 11); pinnae pale (unpigmented) basally, but abruptly blackish distally; dorsal pelage of hind feet often indistinctly darker laterally than medially, but never distinctly blackish or with blackish markings; scaly part of tail usually onethird to slightly less than one-half white distally. Skull (fig. 12A, D) substantially smaller than those of most congeneric species (except P. canus). Nasal bones neither conspicuously elongated nor very short (about 47% of condylobasal length on average), not extending posteriorly to or between postorbital processes. Third upper premolar (P3) labial cingulum incomplete, apparently never extending to anterior base of crown; crown length of upper molar series 12.3 ± 0.4 mm (sexes combined; observed range 11.4–13.2 mm, N = 40); enameled lingual surfaces of upper molars smooth, not crenulated; pre- and postcingula consistently absent; lower molar posterior cingulids absent.
Phylogeography and Geographic Variation: Our phylogenetic analysis of 28 cytochrome-b sequences of Philander quica spanning some 10 degrees of latitude (from the Braziltoian state of Bahia in the north to the Argentinian province of Misiones in the south) provides scant evidence of phylogeographic structure (fig. 13). This lack of genetic differentiation with distance is accompanied by an absence of conspicuous geographic variation in morphology. In effect, this appears to be a genetically and phenotypically homogeneous taxon.
Comparisons: Philander quica closely resembles P. canus, which is similar in size (tables 7, 8); also has uniformly grayish dorsal pelage and self-whitish, -yellowish, or -buffy ventral fur (figs. 10, 11); and is not visually distinctive in any aspect of cranial appearance (fig. 12). Chemisquy and Flores (2012) suggested that these taxa could be distinguished by the width of the postorbital constriction (least postorbital breadth in our terminology), but the samples we measured exhibit broad overlap in this dimension (e.g., 7.8–8.8 mm in P. quica males versus 7.4–8.4 mm in P. canus males). Philander quica and P. canus also have broadly overlapping distributions in the plane of the first two principal components that we computed from craniodental measurements of both taxa (not shown), a result consistent with our impression that these species are metrically very similar. Instead, qualitative morphological comparisons are more informative.
In side-by-side comparisons, the molars of Philander quica appear to have somewhat less well-developed anterolabial cingula, narrower protocones, deeper ectoflexi (especially on M3), and longer postmetacristae than those of P. canus, but the single most useful dental trait that distinguishes these taxa is the morphology of P3. Whereas the third upper premolar of P. canus always has a complete labial cingulum that extends along the entire base of the tooth from anterior to posterior (fig. 14A), the labial cingulum of P3 is incomplete in P. quica, apparently never extending anteriorly past the middle of that tooth (fig. 14B). Unfortunately, we have not found any external trait by which these species can be reliably identified in the field.
Close comparisons between Philander quica and our new species, P. pebas, seem unnecessary given their widely separated geographic distributions (fig. 9), large genetic and morphometric distances (appendices 3, 4), and salient qualitative differences (table 6).
By contrast, Philander quica and P. opossum merit comparison because they are externally similar (with uniformly grayish dorsal fur and mostly self-colored ventral fur) and might eventually be found to occur sympatrically in eastern Brazil (see above). Although same-sex univariate comparisons (table 9) reveal some overlap in all measured craniodental dimensions, principal-components analysis indicates that these species have discrete multivariate distributions (fig. 15A). Because the axis of species discrimination is approximately perpendicular to the more or less parallel axes of within-species variation, and because these axes are oblique to PC1 and PC2, we computed size and size-independent shape factors to obtain vectors with more interpretable coefficients (fig. 15B; table 10). The latter suggest that, independent of general-size allometries, P. opossum has longer but narrower nasals, a longer palate, and much larger molars than P. quica.
Differences in nasal shape between Philander quica and P. opossum are subtle but useful for visual identification of skulls: expressed as a percentage, the ratio NB/NL is about 25% on average in P. quica versus about 22% in P. opossum. Additionally, the shorter/broader nasals of P. quica never extend posteriorly to or between the postorbital processes, whereas the longer nasals of P. opossum often (in about two-thirds of examined specimens) extend to or between the postorbital processes. Whereas P. quica always has an incomplete labial cingulum on P3 (fig. 14B), the labial cingulum of P3 is narrowly complete (ending along the entire base of the tooth when unworn) in about 28% of examined specimens of P. opossum. Externally, P. opossum has better-defined preauricular spots and more saturated (consistently buffy) underparts than P. quica (which usually has whitish or yellowish ventral fur), and P. opossum tends to have a more extensively white-tipped tail (over half the specimens we examined have tails that are about ⅔ white) than P. quica (in which most specimens have tails that are ≤ ½ white). In side-by-side comparisons, the furred basal portion of the tail is visibly longer in P. opossum than in P. quica; unfortunately, this trait is difficult to quantify due to the absence of a definite anterior landmark for relevant measurements.
Remarks: Most recent authors have used the binomen Philander frenatus for this species following Patton and da Silva (1997), but the holotype of frenatus was collected in eastern Amazonia, and we treat that name as a junior synonym of P. opossum (see appendix 5).
Hershkovitz (1997) used the trinomen Philander opossum quica for material that we refer to P. quica, P. canus, and P. pebas. He mapped the range of P. o. quica as including much of the Cerrado, Pantanal, and western Amazonia, but none of his material from Peru, Bolivia, or central Brazil corresponds to P. quica as recognized in this report. Hershkovitz (1997) regarded Didelphis myosuros Temminck, 1824, as a synonym of P. o. quica and designated a lectotype for this purpose, but his lectotype designation is invalid because Pohle (1927) had previously designated a lectotype for D. myosuros, which is currently recognized (e.g., by Gardner and Dagosto, 2008) as a subspecies of Metachirus nudicaudatus (Geoffroy, 1803).
According to Patton and da Silva (1997) this species—which they called Philander frenatus (see above)—occurs in the Brazilian state of Goiás, but they did not list any examined specimens from Goiás, and all the specimens of Philander that we examined from that state are unambiguously referable to P. canus (see below). Subsequently, Patton and da Silva (2008) mapped the range of P. “frenatus” as extending to the Chaco biome in the Argentinian province of Formosa, but the marginal record in question is based on AMNH 256980, a juvenile specimen that we reidentified as P. canus.
The status and relationships of Thomas's azaricus has long been unsettled. Whereas Cabrera (1958) and Patton and da Silva (1997) treated this nominal taxon as a valid subspecies of P. opossum, Hershkovitz (1997) assigned it to the synonymy of P. o. quica, and Patton and da Silva (2008) included it in their synonymy for P. frenatus. Recently, Chemisquy and Flores (2012) analyzed a cytochrome-b sequence from a topotype of azaricus and found that it belonged to the Atlantic Forest haplotype group that they called P. frenatus. We examined the type series of azaricus and an additional 13 topotypes for this report; all of these specimens (21 in total) are unambiguously assignable to P. quica, exhibiting the craniodental traits of that species as diagnosed above.
The identity of Didelphys superciliaris Olfers, 1818, which Patton and da Silva (1997, 2008) listed as a synonym of Philander frenatus, cannot now be determined. Both epithets were first published as nomina nuda by Illiger (1815), but their availability dates from Olfers (1818), who based his descriptions on material that he examined in Berlin (see appendix 5). However, whereas the application of frenatus can now be established based on an extant holotype, no type material of superciliaris is known to survive. Because the collections that Olfers examined in Berlin included specimens collected in both eastern Amazonia and southeastern Brazil, (Voss and Angermann, 1997; see also appendix 5), the lost type of superciliaris could have come from either place: if from the former region, the name would be a junior synonym of P. opossum; if from the latter, then it would be a senior synonym of P. quica. Because superciliaris has not, to our knowledge, been recognized as a valid taxon for almost two centuries, it would seem pointless to use this name to replace quica, the application of which is undisputed. Therefore, to fix the application of Didelphys superciliaris Olfers for the species that occurs in eastern Amazonia, we select as neotype a specimen in the American Museum of Natural History (AMNH 203348) consisting of the skin and skull of an adult male collected by personnel from the Instituto Oswaldo Cruz on 6 December 1960 near “Capim” (= São Domingos do Capim; 1°40´S, 47°47´W; Paynter and Traylor, 1991) at Km 92 on highway BR 14, Para, Brazil.
Habitats: Within the Atlantic Forest biome or ecoregion (Mata Atlântica), Philander quica seems to be eurytopic, occurring in a wide range of vegetation types including mature lowland rainforest (formerly the dominant climax vegetation of southeastern Brazil; Por, 1992), submontane forest, secondary growth, agricultural fields, and coastal restinga scrub (Cerqueira et al., 1993; Bergallo, 1994; Bonvicino et al., 1997; Passamani et al., 2000; D'Andrea et al., 2007).
Specimens Examined (N = 66): Brazil—Espírito Santo, Engenheiro Reeve (BMNH 3.9.4.110); Minas Gerais, Serra de Caparaó (AMNH 8052–8053, 61851–61853); Paraná, Parque Nacional do Iguaçu (MVZ 197401), Roca Nova (BMNH 3.7.1.108–3.7.1.110); Rio de Janeiro, Barreira (ZMB 38069, 38072, 38073, 38076, 38091), Rio de Janeiro (AMNH 133106, 133107; ZMB 38063), Sepetiba (NMW 7687/ST 1012); Rio Grande do Sul (BMNH 84.2.8.29); Santa Catarina, Hansa (BMNH 29.6.6.71), Jaraguá (NMW B2529), Teresópolis (NMW 1671–1675); São Paulo, Boracéia (MVZ 182777; USNM 460503), Fazenda Intervales (MVZ 182066, 183246, 183247), Iguape (USNM 542920), Ilha de Sebastião (MVZ 182067), Ilha do Cardoso (FMNH 141589, 141590), São Sebastião (BMNH 2.4.6.37–2.4.6.40), Ypanema (NMW 2636, 2638, 2640). Paraguay—Central, “Caroreni Viejo” (not located; ZMB 44285); Paraguarí, Sapucaí (BMNH 2.11.7.14, 2.11.7.48, 3.2.3.32–3.2.3.37 [type series of azaricus]; USNM 121412–121422, 121457, 121458), “Ipitimi” (= Ybytymí; ZMB 91277); San Pedro, Tacuatí (USNM 293133).
FIG. 10.
Dorsal pelage of cis-Andean species of Philander compared in the text. From left to right: P. quica (MVZ 183246), P. canus (LACM 10086), P. pebas (MVZ 190343, holotype), P. mcilhennyi (LSU 16393), P. andersoni (LACM 91620), P. opossum (AMNH 266996). The dorsal fur of these species is always grayish or blackish in life, but museum skins often acquire brownish tones after long storage.
FIG. 11.
Ventral pelage of cis-Andean species of Philander compared in the text. From left to right: P. quica (MVZ 183246), P. canus (LACM 10086), P. pebas (MVZ 190343, holotype), P. mcilhennyi (LSU 16393), P. andersoni (LACM 91620), P. opossum (AMNH 266996).
FIG. 12.
Dorsal and ventral views of adult male crania of Philander species formally treated in this report: A, D, P. quica (MVZ 183247); B, E, P. canus (AMNH 210413); C, F, P. pebas (MVZ 190343, holotype). All views about ×1.3.
FIG. 13.
Relationships among 28 cytochrome-b sequences of Philander quica. This subtree shows the full details of the cartooned clade labeled “quica” in figure 5.
FIG. 14.
Lateral view of P2–M1 of Philander canus (A, AMNH 210409) and P. quica (B, MVZ 182066). Whereas P3 has a complete labial cingulum that extends along the entire base of the tooth in P. canus, the labial cingulum of P3 is incomplete (extending only along the posterior part of that tooth) in P. quica.
TABLE 9.
Same-sex comparisons of summary statisticsa for craniodental measurements of Philander quica and P. opossum.
TABLE 10.
Coefficients of principal components (PC1, PC2), general size (Size), and size-adjusted group differences (Shape) for multivariate analyses of Philander quica versus P. opossuma
FIG. 15.
Projections of specimen scores on the first two principal components (A) and on factors representing general size and size-invariant shape differences (B) from analyses of craniodental measurements of Philander quica (open circles) and P. opossum (filled circles). The coefficients of these axes are provided in table 10.
Philander canus (Osgood, 1913)
Metachirus canus Osgood, 1913: 96; type locality Peru, San Martín, Moyobamba (6°03′ S, 76°58′ W; Stephens and Traylor, 1983).
Metachirus opossum crucialis Thomas, 1923: 604; type locality Bolivia, Santa Cruz, Santa Cruz de la Sierra (17°48′ S, 63°10′ W; Paynter, 1992).
Philander mondolfii Lew et al., 2006: 229; type locality Venezuela, Bolívar, Reserva Forestal de Imataca, Unidad V, between Tumeremo and Bochinche (8°00′ N, 61°30′ W).
Philander olrogi Flores et al., 2008: 17; type locality Bolivia, Santa Cruz, 7 km N Santa Rosa (17°03′ S, 63°35′ W).
Type Material: The holotype (by original designation, FMNH 19347) consists of the skin and skull of an adult male collected by W.H. Osgood and M.P. Anderson on 4 August 1912. Although the skin is well preserved, the skull is broken and incomplete (the left zygomatic arch, the left squamosal, and the left bulla are all missing).
Distribution and Sympatry: Sequenced specimens and other examined material that we refer to Philander canus have been collected in central and western Brazil, northern Argentina, Paraguay, eastern Bolivia, eastern Peru, northeastern Colombia, and Venezuela (fig. 9). Although we have not examined specimens from northeastern Peru (Loreto), eastern Ecuador, or southeastern Colombia (Caquetá, Putumayo), future collecting may eventually fill in this geographic hiatus. Philander canus occurs sympatrically with P. andersoni in southern Venezuela (at El Platanal in Amazonas state), with P. mcilhennyi in eastern Peru (e.g., at Balta, in Ucayali department) and western Brazil (at Sobral, in Acre state), and with P. pebas in eastern Peru (e.g., at Balta, in Ucayali department).
Description: Dorsal pelage short (usually <14 mm) and uniformly grayish, usually without any trace of darker middorsal pigmentation; fur of crown (between the ears) usually grizzled gray; pale preauricular spot often present; ventral fur continuously self-whitish, -cream, or -buffy, at least along the midline, but sometimes with broad lateral zones of graybased hairs; pinnae pale (unpigmented) basally, but blackish distally; dorsal pelage of hind feet usually pale, but sometimes indistinctly darker over lateral metatarsals (never distinctly marked with black); scaly part of tail usually <½ white distally but seldom <¼ white. Nasal bones short (about 46% of condylobasal length on average), never extending posteriorly to or between postorbital processes. Unworn third upper premolar (P3) apparently always with complete labial cingulum extending along entire base of tooth; crown length of upper molar series 13.0 ± 0.4 mm (sexes combined; observed range 12.2–14.1 mm, N = 99); enameled lingual surfaces of upper molars smooth, not crenulated; pre- and postcingula consistently absent; lower molar postcingulids absent.
Phylogeography and Geographic Variation: Our phylogenetic analysis of 32 sequences of Philander canus reveals no comprehensible phylogeographic structure, with haplotypes from northern populations (in Colombia and Venezuela) mixed in among those from central Brazil, Bolivia, Paraguay, and northern Argentina (fig. 16). Uncorrected sequence divergence at the cytochrome-b locus among haplotypes that we assign to P. canus is only about 0.8% despite the very wide geographic dispersion of collecting localities. Although we have not statistically tested for geographic variation in morphology among our samples, this appears to be another phenotypically rather uniform species despite modest sample differences in pelage pigmentation (some populations tending to have self-whitish or -cream underparts, whereas others have self-buffy ventral fur).
Comparisons: Morphological comparisons of Philander canus with P. quica have already been described (see above) and comparisons with P. pebas will be described subsequently (see below). It remains to compare this species with members of the P. opossum complex, which—as defined earlier in this report—includes P. opossum, P. andersoni, and P. mcilhennyi.
Philander canus is superficially similar to P. opossum, with which it has long been associated as a subspecies or synonym (e.g., by Cabrera, 1958; Patton and da Silva, 1997, 2008; Gardner, 2005; Chemisquy and Flores, 2012; Hice and Velazco, 2012). Although the geographic ranges of P. canus and P. opossum are not currently known to come into contact, it seems plausible that these species are sympatric or parapatric in the Brazilian states of Mato Grosso, Tocantins, and southern Para, where Cerrado vegetation comes into contact with southeastern Amazonian rainforest. Philander canus is substantially smaller, on average, than P. opossum in several same-sex univariate comparisons (table 11), notably in condylobasal length (CBL), nasal length (NL), least interorbital breadth (LIB), least postorbital breadth (LPB), palatal length (PL), and maxillary toothrow (MTR). Despite some overlap in observed ranges for all dimensions, measured samples of these species have nonoverlapping multivariate distributions (fig. 17), and general-size-adjusted shape coefficients indicate that nasal morphology accounts for much of the observed divergence (table 12). Visual comparisons suggest that the posterior portion of the nasals of P. canus are typically much broader than those of P. opossum, do not extend as far posteriorly, and lack the deep posterolateral notches that are often present in the latter species (fig. 18). Philander canus and P. opossum both have uniformly grayish dorsal fur, mostly pale hind feet, and self-colored ventral fur, but tail pigmentation might be useful for field identification. Whereas the scaly part of the tail is almost always <½ white in specimens of P. canus, the scaly part of the tail is typically ≥½ white in specimens of P. opossum, and this modal difference might be expected to become even more pronounced in sympatry.9
Philander canus is much smaller, on average, than either P. andersoni or P. mcilhennyi, from which it also differs in nasal shape as described and illustrated above (all members of the P. opossum complex have long, narrow nasals that are often laterally notched and often extend posteriorly to or between the postorbital processes). Large generalized distances (appendix 4) suggest that multivariate ordinations of P. canus with either P. andersoni or P. mcilhennyi would show nonoverlapping distributions, but we have not performed those analyses because these species are easy to tell apart by other characters. The dorsal pelage pigmentation of P. canus (uniformly gray; fig. 10) is quite unlike that of P. andersoni (with a distinctly blackish middorsal stripe) and P. mcilhennyi (some specimens of which are completely blackish). The self-whitish, -cream, or -buffy underparts of P. canus likewise contrast with the mostly gray-based ventral pelage of P. andersoni and the almostblackish ventral fur of P. mcilhennyi (fig. 11). Whereas the hind feet of P. canus are covered dorsally with pale fur, the hind feet of P. andersoni and P. mcilhennyi are either completely blackish or have black metatarsals and abruptly whitish digits. Lastly, the scaly part of the tail is almost always <½ white in P. canus but apparently always ≥½. white in P. andersoni and P. mcilhennyi.
Remarks: As understood herein, Philander canus includes the nominal taxa crucialis, mondolfii, and olrogi. The latter are represented in our molecular analyses by: (1) a CYTB sequence we obtained from a Bolivian specimen (AMNH 260034) that was collected near the type locality of crucialis and that resembles the holotype of crucialis in qualitative and morphometric traits; (2) two CYTB sequences, one from a Colombian specimen (KU 123943) and another from a Venezuelan specimen (KU 120245) that were part of Lew et al.'s (2006) original material of mondolfii; and (3) CYTB sequences that we obtained from two specimens (AMNH 261271, 261272) that were part of Flores et al.'s (2008) original material of olrogi. All these specimens conform to our morphological diagnosis of P. canus, so the conclusion that the nominal taxa in question are conspecific seems straightforward, but brief comments on each synonym are appropriate.
Thomas (1923) described crucialis on the basis of a single specimen, which he compared only with azaricus (= P. quica; see above). Later, with more Bolivian material at hand for comparison with topotypical specimens of Osgood's species, he (Thomas, 1928) judged crucialis and canus to be indistinguishable. We agree.
Lew et al. (2006) described mondolfii based on several dozen specimens from Colombia and Venezuela that the authors compared carefully with other species of Philander known to occur in or near those countries, including P. andersoni, P. deltae, P. “fuscogriseus” (= P. melanurus), and P. opossum, but they did not compare mondolfii with P. canus. Although we have not examined the holotype or paratypes of mondolfii—all currently inaccessible in Venezuelan museums—we did examine 16 specimens that were part of Lew et al.'s (2006) original material. These specimens (AMNH 16951, 30709, 30711–30714, 133119, 133120, 136163, 136167–136169, 139221; KU 120233, 120245, 123943) so closely resemble the type of P. canus and other referred material from eastern Peru that we could not find any phenotypic basis for retaining mondolfii even as a subspecies.
Flores et al. (2008) described olrogi on the basis of seven specimens from Bolivia and Peru. Although we have not seen the holotype, we examined the skull of a paratype (AMNH 246441) as well as several other specimens that were part of Flores et al.'s (2008) original material (AMNH 261269–261272). Despite careful study, we confess ourselves unable to consistently distinguish these specimens from material that the authors referred to P. opossum canus. Although they reported a principal-components analysis that was said to support the recognition of olrogi as a distinct taxon, canus and olrogi have overlapping distributions in their illustrated results (Flores et al., 2008: fig. 5). In the absence of compelling evidence for the phenotypic distinctness of these genetically indistinguishable nominal taxa, we interpret the allegedly diagnostic traits of olrogi to be aspects of intraspecific morphological variation within P. canus.
Cabrera (1958: 35) listed nigratus as a synonym of canus (which he ranked as a subspecies of P. opossum), but the holotype (BMNH 0.7.7.62) and other material that we refer to nigratus10 are larger animals (LM = 14.7–16.4 mm) with much darker dorsal fur, completely gray-based ventral fur, blackish feet, shorter white tail-tips (less than ¼ of the tail is unpigmented in most specimens), and an incomplete labial cingulum on P3. Although we do not know whether nigratus is a valid species, its phenotype more closely resembles those of species in the P. opossum complex than that of P. canus.
Hershkovitz (1997) used the name Philander opossum quica for many specimens that we refer to P. canus, including all the material he listed from central Brazil (Goiás, Mato Grosso) and Bolivia; among the material that he listed from eastern Peru are specimens that we refer to both P. canus and P. pebas. The “dichromatism” that he (Hershkovitz, 1997: 49) noted among specimens of “P. o. quica” from Balta (in the Peruvian department of Ucayali) is the result of sympatry rather than polymorphism: of the six specimens in question that we examined, three (LSUMZ 12006, 12008, 12009) are P. canus and the others (LSUMZ 12007, 12010, 14011) are P. pebas (see Specimens Examined for both taxa, below; a third congener, P. mcilhennyi, also occurs at Balta).
The western Amazonian specimens that Patton et al. (2000) referred to Philander opossum canus include examples of both P. canus and P. pebas. Of the five that we were able to examine— the others having been returned to Brazil—four (MVZ 190343–190346) are P. pebas; only one (MVZ 190347, from the state of Acre, near the Peruvian border) is actually P. canus. The central Amazonian specimens that Nunes et al. (2006) identified as P. canus are also examples of P. pebas.11 The only other material of P. canus that we have seen from the Brazilian Amazon (besides the MVZ specimen from Acre) is a small series collected many years ago along the lower Rio Madeira (in Amazonas state) and a single specimen from the upper Madeira (in Rondônia).
Habitats: The geographic range of Philander canus extends over a wide range of biomes or ecoregions (including the Cerrado, Chaco, Pantanal, Llanos, and Amazonia), and it is possible that the species occurs in a corresponding variety of habitats, but definite ecological information associated with collected specimens is hard to find. Other species of Philander are known to live in rainforest, so collection records from biomes dominated by savanna vegetation and/or dry forest (e.g., the Cerrado, Chaco, and Llanos) seem anomalous, but the literature on Cerrado mammals provides a few relevant observations.
In the Cerrado landscapes of eastern Bolivia, Philander canus is apparently restricted to tall evergreen gallery forests and does not seem to occur in the savannas and dry forests that cover much of the landscape (Emmons et al., 2006). In the Cerrado of central Brazil, P. “opossum” (presumably P. canus) is also said to be a gallery-forest species (Mello and Moojen, 1979; Redford and Fonseca, 1986; Alho, 2005), but in one report of a multiyear trapping study P. “opossum” was said to prefer gallery forest but to occur frequently in other local habitats, including open grassland, shrub savanna, and dry forest (Alho et al., 1986). Following Pulliam's (1988) ecological terminology, we conjecture that gallery forests are probably the source habitat for P. canus in Cerrado landscapes, whereas open formations (including dry forests) are likely to be sink habitats. In the Chaco of northern Argentina, Philander “opossum” (presumably P. canus) is also said to occur in gallery forests (Huck et al., 2017).
Information about the habitat distribution of Philander canus appears to be unavailable from trapping studies in the Pantanal and Llanos, but we suspect that it is largely restricted to gallery forests in those ecoregions as well. Nevertheless, Lew et al.'s (2006) summary of macrohabitats where Philander “mondolfii” (= P. canus) has been collected in Venezuela (including lowland and submontane rainforest, semideciduous forest, and tree savannas) suggest that it has broad ecological tolerances, at least where other sympatric congeners are not known to occur.
We have not found any published accounts of where specimens that can definitely be identified as Philander canus have been collected in Amazonia. Although the natural climax vegetation throughout this enormous ecoregion can be broadly characterized as lowland rainforest, local disturbance (e.g., from lateral migration of rivers within their meander belts) and edaphic factors can result in a surprising diversity of natural vegetation types at many Amazonian localities (Pires and Prance, 1985; Puhakka and Kalliola, 1995), and anthropogenic habitats are also scattered throughout the region. The geographic distribution of Amazonian collection localities for P. canus provides no habitat clues, because these localities are not clustered around savanna enclaves, human settlements, or other obvious landscape features. The known Amazonian range of P. canus broadly overlaps those of P. andersoni, P. mcilhennyi, and P. pebas, so it would be reasonable to suppose that competitive interactions with sympatric congeners might restrict the habitat occupancy of this species to some extent, but the information compiled for this report is entirely inadequate even for conjecture.
Specimens Examined (N = 154): Bolivia—Beni, Arruda (FMNH 114701), Camiaco (AMNH 210402), Casarabe (AMNH 261269–261272; MSB 55854), 8 km N Exaltación (AMNH 210403), Magdalena (FMNH 114714), Mamore River (AMNH 210409), 4 km SE Palacios (210410), Puerto Caballo (AMNH 210411), Puerto Siles (AMNH 210413, 210414), Rio Tijamuchi (AMNH 261273), San Joaquin (FMNH 114685, 114694, 114707); Pando, Bella Vista (MSB 57006, AMNH 262413); Santa Cruz, 7 km E aserradero Moira (EBD 8736), 6 km W Asención (MSB 55855), Ayacucho (USNM 390564), Becerra (390565), 2 km N Chapare River mouth (AMNH 210416), 2 km SE Cotoca (MSB 59887), Estancia Cachuela Esperanza (AMNH 260034, MSB 55073), Hamecas (AMNH 135887), La Laguna (MSB 55856), 3 km SE Montero (AMNH 263964, MSB 67025), Palmar (USNM 390562), San Miguel Rincón (AMNH 260037, MSB 55074, 55075), 10 km N San Ramon (AMNH 261277, 261278), Santa Cruz de la Sierra (BMNH 47.11.22.15 [holotype of crucialis]), 15 km S Santa Cruz (AMNH 263966, MSB 58517), 7 km N Santa Rosa (AMNH 246441 [paratype of olrogi]), near Warnes (USNM 390005, 390009–390012). Brazil—Acre, Sobral on Rio Juruá (MVZ 190347); Amazonas, Auara Igarapé on Rio Madeira (AMNH 91749, 91750), Borba on Rio Madeira (AMNH 91748), Lago Sampaio on Rio Madeira (AMNH 92761, 92762), “Santo Antonio de Uayara” on Rio Madeira (= Santo Antonio de Guajará; AMNH 92293); Goiás, Anápolis (AMNH 133043, 133046, 133047, 133055, 133056, 133062, 133064, 133068–133070, 133073–133075, 133082, 133084–133086, 133091–133094, 133096–133101, 133123, 133172, 133171, 133182, 133192, 133195), 24 km SE Formoso (LACM 10086–10088); Mato Grosso, Cáceres (USNM 390014), Fazenda São Luis (MVZ 197403); Mato Grosso do Sul, Corumba (USNM 390013), Passo do Lontra (MVZ 197402); Rondônia, Porto Velho (USNM 390001). Colombia—Boyacá, Río Cobaría (FMNH 92297); Meta, Finca El Capricho (KU 123943), Restrepo (AMNH 133119), Villavicencio (AMNH 136168, 136169, 139221). Paraguay—Alto Paraguay, Estancia Doña Julia (TTU 79753); Central, 17 km E Luque (MVZ 144304); Presidente Hayes, Estancia Loma Porá (TTU 80404). Peru—Cusco, Camisea (MUSM 14150), Hacienda Cadena (FMNH 66412, 68332), Quincemil (FMNH 75094–75096); Huánuco, Moyuna (MUSM 83); Madre de Dios, “Albergue Lodge Cuzco Amazónico” (= Cusco Amazónico; MVZ 157613, 165927), Boca Rio Colorado (FMNH 84247), Lago Sandoval (MVZ 157614), mouth of Rio La Torre (LSUMZ 24591), 6 km W Rio Tambopata (USNM 39002); 2.75 km E Shintuya (FMNH 169815); Pasco, San Pablo (AMNH 230034), Nevati (AMNH 230028, 230030, 230031, 254509); San Martín, Bellavista (MUSM 92), Moyobamba (FMNH 19347 [holotype of canus]), Rioja (MUSM 88); Ucayali, Balta (LSUMZ 12006, 12008, 12009), 59 km SW Pucallpa (USNM 499001, 499002), Boca Río Urubamba (AMNH 75906–75908), Lagarto (AMNH 76636), Santa Rosa (AMNH 75909). Venezuela—Amazonas, El Platanal (EBD 8954, 8956); Apure, 29 km SSW Santo Domingo (USNM 418545, 418546); Bolívar, 20 km W La Paragua (USNM 388403), Maripa (AMNH 16951), Río Yuruan (AMNH 30709–30714); Trujillo, 9.8 km NNE Motatán (KU 120233, 120245, 120246, 120251), 19 km W Valera (USNM 371322); Zulia, 60 km WNW Encontrados (USNM 418548).
FIG. 16.
Relationships among 46 cytochrome-b sequences of Philander canus and P. pebas. This subtree shows the full details of the cartooned clades labeled “canus” and “pebas” in figure 5.
TABLE 11.
Same-sex comparisons of summary statisticsa for craniodental measurements of Philander canus and P. opossum.
TABLE 12.
Coefficients of principal components (PC1, PC2), general size (Size), and size-adjusted group differences (Shape) for multivariate analyses of Philander canus versus P. opossuma
FIG. 17.
Projections of specimen scores on the first two principal components (A) and on factors representing general size and size-invariant shape differences (B) from analyses of craniodental measurements of Philander canus (open triangles) and P. opossum (filled circles). The coefficients of these axes are provided in table 12.
Philander pebas, new species
Type Material: The holotype, MVZ 190343, consists of the skin, skull, and frozen tissues of an adult male collected by J.L. Patton (original number 15395) on 1 September 1991 at Igarapé Nova Empresa, on the left bank of the Rio Juruá, Amazonas, Brazil (6°48´S, 70°44´W). A complete (1149 bp) cytochrome-b sequence that we obtained from this specimen is archived in GenBank with accession number MG491956.
Distribution and Sympatry: Sequenced specimens and other referred material of Philander pebas are from eastern Ecuador, eastern Peru, and Amazonian Brazil (fig. 9). Based on specimens we examined, P. pebas occurs sympatrically with P. andersoni in northeastern Peru (e.g., near Iquitos, in Loreto department) and with P. canus and P. mcilhennyi in southeastern Peru (e.g., at Balta in Ucayali department).
FIG. 19.
Upper molar differences between Philander pebas and P. canus (see text for explanation). A, Occlusal view of left M2–M4 of P. pebas (MVZ 190343, holotype); B, occlusal view of left M2–M4 of P. canus (AMNH 210413). Abbreviations: poc, postcingulum; prc, precingulum.
Description: Dorsal pelage very short (usually ≤12 mm) and uniformly grayish (sometimes darker middorsally than on the flanks but never with a distinct middorsal blackish stripe; fig. 10); fur of crown (between the ears) grizzled-grayish, often quite dark but apparently never clear black (at least some hairs frosted, with pale tips); pale preauricular spot absent or indistinct; ventral fur mostly gray-based (fig. 11), often self-cream or -buffy in the inguinal region but apparently never continuously self-colored along the abdominal and thoracic midline; pinnae sometimes entirely blackish but often indistinctly paler basally; hind feet often with dark metatarsals and pale digits, but not blackish or with distinctly blackish markings; scaly part of tail usually <¼ white distally. Nasal bones neither very short nor unusually elongated (about 47% of condylobasal length on average), sometimes extending posteriorly to (but apparently never between) postorbital processes. Unworn third upper premolar (P3) with complete labial cingulum; crown length of upper molar series 13.8 ± 0.5 mm (sexes combined; observed range 12.7–15.1 mm, N = 50); unworn molar enamel distinctly crenulated, especially on lingual surfaces of protocones (fig. 19A); pre- and postcingula usually present on one or more upper molars (more frequently retained on M4 than on M1–3 in older specimens with worn teeth; fig. 19A); posterior cingulids apparently always present on one or more lower molars (fig. 20A).
FIG. 20.
Lower molar differences between Philander pebas and P. canus (see text for explanation). A, Labial view of right m1–m3 of P. pebas (MVZ 190343, holotype); B, labial view of right m1–m3 of P. canus (AMNH 210413). Abbreviations: pcid, postcingulid.
Phylogeography and Geographic Variation: Some phylogeographic structure is apparent among the 14 haplotypes that we assign to Philander pebas, with partial separation of Brazilian sequences on the one hand from Peruvian and Ecuadorean sequences on the other (fig. 16), but neither haplogroup received consistently strong support in our analyses. The only phenotypic evidence of geographic variation we observed was the caudal pigmentation of the easternmost specimens (from central Amazonia), most of which have ⅓ to ½ white tails, whereas those from western Amazonia usually have tails that are ≤¼ white.
Comparisons: Philander pebas is the only species in the genus with distinctly crenulated (folded and grooved) molar enamel, a trait that is most clearly visible on unworn teeth, but which persists on the lingual surfaces of the protocones even in old adults. Additionally distinctive traits, apparently unique among didelphids, are narrow enamel shelves along the anterolingual and posterolingual bases of the protocones; we refer to these shelves as the precingulum and postcingulum, respectively.12 These shelves tend to wear away with age, but they often persist on M4 even in old adults. Another distinctive trait, only rarely observed as a polymorphism among other didelphids, is a narrow shelf along the posterolabial surface of the hypoconid; following standard tribosphenic terminology, this shelf is called the posterior cingulid or postcingulid.
Philander pebas can be distinguished from its sister species, P. canus, by additional characters. Among others, it is substantially larger than P. canus (tables 7, 8), and specimen scores on the first two principal components that we computed from craniodental measurements of both taxa illustrate nonoverlapping multivariate distributions (fig. 21A). Coefficients of general-size-invariant shape differences computed from these data suggest that P. pebas has longer but narrower nasals, wider interorbital and postorbital dimensions, and longer palates than P. canus (fig. 21B, table 13). The two species can also be reliably identified by external traits, of which ventral pelage coloration is the most consistently useful. Whereas the ventral fur of P. canus is continuously self-whitish, -cream, or -buffy from chin to groin, the ventral fur of P. pebas is extensively gray-based. Some specimens of P. pebas have self-whitish or -buffy fur on the chin, throat, and/or groin, but none of the specimens we examined has a continuous midventral streak of self-colored fur over the chest and upper abdomen. The two species also seem to be reliably identifiable by tail markings in Ecuador, Peru, and Acre (Brazil), where specimens of P. canus have tails that are at least ⅓ to almost ½ white, but where specimens of P. pebas have tails that are ≤¼ white.
Close comparisons of Philander pebas and P. quica seem unnecessary given their widely disjunct geographic distributions (fig. 9), large genetic and morphometric distances (appendices 3, 4), and salient qualitative differences (table 6).
Philander pebas differs from members of the P. opossum complex, with which it is broadly sympatric (P. andersoni, P. mcilhennyi) or potentially sympatric (P. opossum), by the unique dental traits described above and by external morphology. By comparison with P. andersoni—with its distinctly blackish middorsal stripe (fig. 10)—the dorsal fur of P. pebas is uniformly grayish, although it can be indistinctly darker (sometimes almost blackish) middorsally. Additionally, where the ranges of P. andersoni and P. pebas overlap, they can readily be distinguished by tail markings (the scaly part of the tail of P. andersoni is ≥½. white, whereas the tail of sympatric P. pebas is ≤¼ white). By comparison with P. mcilhennyi (which is sometimes almost entirely blackish), P. pebas is uniformly grayish, and these species also differ in fur length: although observed ranges narrowly overlap, the middorsal fur of P. mcilhennyi is much longer on average (16 ± 3 mm) than the middorsal fur of P. pebas (10 ± 2 mm), and the latter species never has the typically shaggy appearance of P. mcilhennyi. As in P. andersoni, the scaly portion of the tail is at least ½ white in P. mcilhennyi, whereas the tail is mostly black in P. pebas. By comparison with P. opossum (which has mostly self-buffy underparts), the ventral fur of P. pebas is extensively gray-based (and is never buffy in the specimens we examined).
TABLE 13.
Coefficients of principal components (PC1, PC2), general size (Size), and size-adjusted group differences (Shape) for multivariate analyses of Philander pebas versus P. canus.a
Remarks: Specimens that we refer to Philander pebas were among those previously identified as P. opossum quica by Hershkovitz (1997), as P. opossum canus by Patton et al. (2000), as P. opossum by Woodman et al. (1991) and Hice and Velazco (2012), and as P. canus by Nunes et al. (2006). Although we were unable to examine any of the specimens from northeastern Peru identified as P. opossum by Díaz (2014), we suspect that most of them are P. pebas.
We have not examined specimens of Philander deltae (known only from northeastern Venezuela), but Lew et al.'s (2006) description of that species includes several external traits (including brownish dorsal fur, a broad strip of “uniformly cream” ventral fur, very small and poorly defined supraorbital spots, and sparsely pigmented ears) that are quite unlike the corresponding attributes of P. pebas. Because Lew et al. (2006) did not publish measurement data for P. deltae, no morphometric comparisons with P. pebas are possible.
Habitats: All examined specimens of Philander pebas are from western and central Amazonian landscapes where the natural climax vegetation is lowland rainforest, but many localities in this region support a wide range of habitats. The floodplains of white-water rivers, in particular, typically include a mosaic of successional stages and edaphic formations (Salo et al., 1986; Puhakka and Kalliola, 1995), and they are interdigitated with floristically distinct upland forests that grow on well-drained terraces and hillsides. Although fragmentary and incomplete, available ecological information from several localities suggest that P. pebas occupies a distinctive suite of natural and anthropogenic habitats within this diverse ecological matrix.
FIG. 21.
Projections of specimen scores on the first two principal components (A) and on factors representing general size and size-invariant shape differences (B) from analyses of craniodental measurements of Philander canus (open triangles) and P. pebas (filled triangles). The coefficients of these axes are provided in table 13.
According to Patton et al. (2000), who trapped in both upland (terra firme) forest and seasonally flooded (várzea) forest along the Rio Juruá, Philander “opossum” was taken only in flooded forest, except in the headwaters region, where one specimen was trapped in upland forest. Of the 15 specimens of P. “opossum” they collected, we were able to examine only five, of which four were P. pebas and one was P. canus. All four specimens of P. pebas were taken at localities where the trapping habitat was described as várzea, by contrast with specimens of sympatric P. mcilhennyi, which the authors trapped in both terra firme and várzea habitats.
Another record of Philander pebas from seasonally flooded forest is based on the specimens of Philander “canus” analyzed by Nunes et al. (2006). These specimens, which we reidentified as P. pebas, were collected in the Mamirauá Sustainable Development Reserve, a protected area consisting entirely of várzea at the confluence of the Rio Japura and the upper Amazon (Solimões). When the floodwaters are at their highest, virtually the entire reserve is flooded and only the forest canopy is visible above the water line (de Queiroz and Peralta, 2010). During the low-water season, emergent land is covered by tall forest growing on levees, shrubby vegetation in lower areas, and a variety of other floodplain habitats (Ayres, 1995). The specimens in question were trapped in seasonally flooded forest (C. Nunes, personal commun., 17 October 2017).
In addition to seasonally flooded riparian formations, this species has also been trapped in swamps (habitats with permanently waterlogged soils). Several specimens of Philander “opossum” have been collected in the vicinity of Cusco Amazónico, an ecotourist lodge on the Río Madre de Dios in southeastern Peru (Woodman et al., 1991). Although Cusco Amazónico is located within the meander belt of the Madre de Dios, the various habitats sampled by zoological collectors at this locality were not seasonally flooded by river water (Duellman, 2005). Of the two specimens of P. pebas that we examined from Cusco Amazónico—where P. canus also occurs— only one is accompanied by definite habitat information. This specimen (KU 1441209) was collected in a Heliconia swamp; judging from information provided by Duellman (2005), the capture site is probably seasonally inundated by accumulated rainwater in the wet season.
Lastly, this species has been collected in anthropogenic habitats on well-drained soils. According to Hice and Velazco (2012), who reported on material collected in the Reserva Nacional Allpahuayo-Mishana and at the nearby Fuerte Militar Otorongo in northeastern Peru, Philander “opossum” was collected only in secondary vegetation and agricultural fields, whereas P. andersoni occupied adjacent primary forest habitats. We examined 16 of the 39 specimens of P. “opossum” reported by these authors, and all were examples of P. pebas.
Based on these scant data, we hypothesize that Philander pebas is primarily a várzea species; that is, one that typically inhabits riparian formations seasonally flooded by white-water rivers (for Amazonian flooded-forest nomenclature, see Prance, 1979). In support of this conjecture, we note that the geographic distribution of the species (fig. 9) corresponds closely to the distribution of white-water catchments in the Amazon Basin (Junk et al., 2011: fig. 1), and we boldly predict that P. pebas will eventually be found to inhabit the white-water Caquetá and Putumayo drainages of southeastern Colombia, from which we have yet to examine any material. Because várzea habitats are characterized by riverine flooding, terrestrial (nonaquatic and nonarboreal) species that inhabit such forests during the low-water season must periodically migrate to higher ground, and the ability to occupy temporary refugia may preadapt terrestrial várzea species to also utilize swampy habitats (seasonally flooded by accumulated rainwater), as well as to opportunistically invade secondary vegetation resulting from human activity on adjacent terraces and hillsides.
Etymology: After Lago Pebas, the vast Miocene lake complex (Wesselingh et al., 2001) or “mega-wetland” (Hoorn et al., 2010) that filled much of the Andean foreland basin, including almost the entire known geographic range of this morphologically distinctive species.
Specimens Examined (N = 58): Brazil—Acre, Fazenda Santa Fé (on Rio Juruá; MVZ 190345), opposite Ocidente (on Rio Juruá; MVZ 190346); Amazonas, Igarapé Nova Empresa (on Rio Juruá; MVZ 190343), Lago do Baptista (on S bank of Amazon; FMNH 51095), Sacado (on Rio Juruá; MVZ 190344), Santo Isidoro [near] Tefé (on S bank of Amazon; AMNH 78954), Parintins (“Villa Bella Imperatriz,” on S bank of Amazon; AMNH 92880, 92881, 93526–93528, 93968), Tapaua (on Rio Purus; USNM 461374). Ecuador—Orellana, 42 km S Pompeya Sur (ROM 106101, 106139). Peru—Loreto, Apayacu (AMNH 74388), Avícola San Miguel (MUSM 33590, 33592, 33593), Cabo Lopez (MUSM 33566, 33567, 33569, 33570, 33572), Carretera Iquitos-Nauta km 28.8 (MUSM 34892), Caserio Cahuide (MUSM 33564, 33574, 33576), El Paujil (MUSM 33580), El Triunfo (MUSM 33586, 33587, 33583), Iquitos (AMNH 98642), 19.7 km SW Iquitos (MUSM 33588), Mishana (MUSM 33597), Orosa (AMNH 73852), Otorongo Army Base (LACM 91621, 91622), Peña Negra (MUSM 33598), Picuro Yacu (MUSM 33594), Quistococha (FMNH 122745–122748; MUSM 33599, 33600), San Gerardo (MUSM 33602), Santo Tomas (MUSM 33603), Sarayacu (on Río Ucayali; AMNH 76448–76450); Madre de Dios, Cusco Amazónico (KU 144120, 144121; MUSM 6074); Ucayali, Balta (LSUMZ 12007, 12010, 14011), Yarinacocha (FMNH 55411).
ACKNOWLEDGMENTS
We extend our thanks and express our deep appreciation to the curators and support personnel of museums on three continents, who hosted our visits or loaned us specimens in the course of this study. Without their conscientious efforts to safeguard the collections in their care, research like this would not be possible: Roberto Portela (at the BMNH); Susan McLaren (CM); Mara Sempere and Rosa Rodríguez (EBD); Bruce Patterson and John Phelps (FMNH); Robert M. Timm (KU); Jake Esselstyn and Steve Cardiff (LSUMZ); Hopi Hoekstra and Judy Chupasko (MCZ); Joseph Cook and Jonathan Dunnum (MSB); Victor Pacheco and Judith Carrasco (MUSM); Jim Patton and Chris Conroy (MVZ); Frank Zachos (NMW); Burton Lim and Jacqui Miller (ROM); Robert Bradley and Heath Garner (TTU); Darrin Lunde and Nicole Edmison (USNM); and Christiane Funk (ZMB).
This work was funded in part by National Science Foundation grants to R.S.V. (DEB-0743039), J.F.D. (DEB-1311163), and S.A.J. (DEB-0743062). Research at the AMNH was made possible by the combined efforts of Eleanor Hoeger, Brian O'Toole, and Eileen Westwig, who processed numerous specimen loans for this project. Patricia Wynne drew figures 4, 8, 14, and 18 with her customary skill and professionalism. Craig Chesek produced the skin photos (figs. 10, 11), and Suzann Goldberg photographed the skulls and teeth (figs. 12, 19, 20). We thank Lorissa Fujishin for her careful lab work and expert databasing skills at the University of Minnesota.
We are much indebted to Simon Engelberger (in Vienna) and Steven van der Mije (in Leiden), who provided detailed information about old specimens and archival material at the NMW and RMNH, respectively. We are also grateful to M. Mónica Díaz, who kindly examined specimens for us at the CML (in Tucumán); to Silvia Pavan, who examined specimens for us at MPEG (in Belém); and to David Flores, who examined specimens for us at the MACN (in Buenos Aires). Rui Cerqueira, Amelia Chemisquy, Leonora Costa, Maria da Silva, Yuri Leite, Claudia Nunes, Jim Patton, and Marcia Revelez kindly provided geographic coordinates and other information about voucher material. Thiago Semedo sent us an otherwise unobtainable copy of Ayres (1995), for which we thank him very much indeed. Mario Gómez-Martínez generously allowed us to use his photograph of Philander melanurus attacking a coral snake (an incident described by Gómez-Martínez et al., 2008). Jim Patton and Diego Astúa provided comments on our submitted manuscript that helped us improve the final draft. Lastly, we thank Nicholas Voss for trapping several specimens of Philander for us in Belize (fig. 22), and we thank Nancy Simmons for preparing them when she had many other things to do.
REFERENCES
Appendices
APPENDIX 1
Gazetteer of Collection Localities for Sequenced Material
Below we list all the localities where sequenced specimens of Philander were collected, including those sequenced by us and others corresponding to sequences that we downloaded from Gen- Bank (table 2). Italicized place names are those of the largest political units (states, departments, or provinces) within each country. Geographic coordinates (in decimal degrees) were obtained from specimen labels, field notes, or institutional databases except as noted otherwise (in square brackets, with a cited source). The name of the taxon collected at each locality (in boldface) together with the name of the collector(s) and year of collection (in parentheses) are also provided.
ARGENTINA
1. Chaco, Parque Nacional Chaco, Presidencia de la Plaza [ca. 26.93°S, 59.77°W; Lorea et al., 2008]: canus (coll. S. Heinonen, 1995).
2. Formosa, Pilcomayo, Parque Nacional Rio Pilcomayo [ca. 25.13°S, 58.13°W; Pardiñas et al., 2004]: canus (coll. S. Heinonen, 1993).
3. Misiones, Iguazu, Río Urugua-i [ca. 25.92°S, 54.30°W; Pardiñas et al., 2003]: quica (coll. J.A. Crespo, 1949, 1951; W.H. Partridge, 1951). The Río Urugua-i, a left-bank tributary of the upper Rio Parana, is also known as the Arroyo Urugua-í (presumably to avoid confusion with the much larger Río Uruguay). It was the focus of much mid-20th-century collecting activity by various investigators, of whom those from the MACN seem to have worked at a site known as Yacú-poí (Massoia et al., 1987), about 30 km east of Puerto Libertad (= Puerto Bemberg; Baldo and Basso, 2004). Several MACN specimens from which Chemisquy and Flores (2012) obtained sequence data are assumed to have been collected here, including those labeled “Río Urugua-i (curso medio),” “Río Urugua-i (curso medio) 30 km Puerto Bemberg,” and “Arroyo Urugua-i.”
4. Misiones, San Pedro, Tobuna [26.47°S, 53.90°W; Paynter, 1995]: quica (coll. J.A. Crespo, 1952).
BELIZE
5. Orange Walk, Lamanai Outpost Lodge (17.46°N, 88.39°W): pallidus (coll. N.S. Voss and N.B. Simmons, 2012).
BOLIVIA
6. Beni, Casarabe, 230 m (14.80°S, 65.45°W): canus (coll. L.A. Ruedas, 1985).
7. Beni, Rio Tijamuchi, 240 m [14.17°S, 64.97°W; Anderson, 1997]: canus (coll. AMNH/MSB expedition, 1985).
8. Pando, Bella Vista, 170 m [11.23°S, 67.12°W; Anderson, 1997]: canus (coll. T.L. Yates, 1986).
9. Santa Cruz, Estancia Cachuela Esperanza [16.78°S, 63.23°W; Anderson, 1997]: canus (coll. N. Olds, 1984).
BRAZIL
10. Acre, Fazenda Santa Fé, left bank Rio Juruá [8.60°S, 72.85°W; Patton et al., 2000]: pebas (coll. M.N.F. da Silva, 1992).
11. Acre, Igarapé Porongaba, right bank Rio Juruá [8.67°S, 72.78°W; Patton et al., 2000]: mcilhennyi (coll. M.N.F. da Silva, 1992).
12. Acre, Sobral, left bank Rio Juruá [8.37°S, 72.82°W; Patton et al., 2000]: canus and mcilhennyi (coll. M.N.F. da Silva, 1992).
13. Amazonas, Altamira, right bank Rio Juruá (6.58°S, 68.90°W): mcilhennyi (coll. J.L. Patton, 1991).
14. Amazonas, alto Río Urucu (4.85°S, 65.27°W): mcilhennyi (coll. M.N.F. da Silva, 1989).
15. Amazonas, Estrada Picarreira, Parque Nacional do Pico da Neblina (0.61°N, 66.09°W): andersoni (coll. V.C.S. Vidigal [date unknown]).
16. Amazonas, Igarapé Nova Empresa, left bank Rio Juruá [6.80°S, 70.73°W; Patton et al., 2000]: pebas (coll. J.L. Patton, 1991).
17. Amazonas, Ilha das Oncas, left bank Rio Negro (1.82°S, 61.37°W): opossum (coll. L.P. Costa and J.L. Patton, 2000). One sequenced specimen corresponding to this locality datum (INPA 4342/LPC 164) was erroneously reported to have been collected at Lago Meduinim (also on the left bank of the Rio Negro) by Patton et al. (2000: table 16; J.L. Patton, personal commun.).
18. Amazonas, Macaco, left bank Rio Jau (2.08°S, 62.12°W): andersoni (coll. Y. Leite, 2000).
19. Amazonas, Mamirauá Reserve [3.17°S, 64.68°W; Nunes et al., 2006]: pebas (coll. C. Nunes, 1995–1997).
20. Amazonas, left bank Rio Jau above mouth (1.96°S, 61.49°W): andersoni (coll. M.N.F. da Silva, 1996).
21. Amazonas, Seringal Condor, left bank Rio Juruá [6.75°S, 70.85°W; Patton et al., 2000]: mcilhennyi (coll. J.L. Patton, 1991).
22. Bahia, Fazenda Bolandeira, 10 km S Una [15.35°S, 39.00°W; Geise, et al., 2001]: quica (coll. Y. Leite, 1996).
23. Espírito Santo, Cariacica, Reserva Biológica de Duas Bocas, alto Alegre (20.28°S, 40.51°W; Y. Leite and L. Costa, personal commun.): quica (coll. L.P. Costa, 2007).
24. Espírito Santo, Estação Biológica de Santa Lúcia [ca. 19.95°S, 40.52°W; Y. Leite, and L. Costa, personal commun.]: quica (coll. M.A. Mustrangi, 1993; Y. Leite, 2007).
25. Espírito Santo, Ibitirama, Parque Nacional Caparaó, Posto Santa Maria (20.50°S, 41.70°W; Y. Leite, and L. Costa, personal commun.): quica (coll. V. Fagundes, 2006).
26. Espírito Santo, Pancas, Mata de Pedra do Camelo (19.24°S, 40.77°W; Y. Leite and L. Costa, personal com.): quica (coll. L.P. Costa, 2006).
27. Mato Grosso do Sul, Rio Miranda, above Passo do Lontra [19.58°S, 57.01°W; MVZ collection database]: canus (coll. L.P. Costa and J.L. Patton, 1998).
28. Mato Grosso, Base de Pesquisa do Pantanal CENEP/IBAMA, 110 km SSW Poconé (17.12°S, 56.95°W): canus (coll. L.P. Costa, 1998).
29. Mato Grosso, Fazenda São Luis, 30 km N Barra do Garças [15.63°S, 52.36°W; MVZ collection database]: canus (coll. L.P. Costa, 1998).
30. Minas Gerais, Cruzeiro, 8 km NE Santa Rita de Jacutinga (22.08°S, 44.03°W): quica (coll. L.P. Costa [date unknown]).
31. Minas Gerais, Estação Biológica Mata do Sossego, Simonésia (20.13°S, 42.00°W): quica (coll. Y. Leite [date unknown]).
32. Minas Gerais, Parque Estadual do Ibitipoca, 30 km N Lima Duarte (21.70°S, 43.90°W): quica (coll. M.A. Mustrangi, 29 October 1993).
33. Minas Gerais, RPPN Belgo Mineira, João Monlevade [19.80°S, 43.17°W; MVZ collection database]: quica (collector and date unknown).
34. Pará, 52 km SSW Altamira, east bank Rio Xingu (3.65°S, 52.37°W): opossum (coll. L.H. Emmons and M.D. Carleton, 1986).
35. Pará, Belém, IPEAN-APEG [ca. 1.45°S, 48.48°W; Paynter and Traylor, 1991]: opossum ([collector unknown] 1971).
36. Pará, Floresta Nacional Tapirapé-Aquiri, Município de Marabá (5.80°S, 50.52°W): opossum (collector and date unknown).
37. Pará, Itaituba, BR165 Santarém-Cuiabá zona sul [4.92°S, 55.60°W; USNM collection database]: opossum ([collector unknown] 1976).
38. Pará, Santana do Araguaia [9.63°S, 50.14°W; Rocha et al., 2015]: canus (coll. R.G. Rocha [date unknown]).
39. Paraná, Mananciais da Serra (SANEPAR), Piraquara (25.47°S, 49.07°W): quica (coll. Y. Leite [date unknown]; N. Cáceres [date unknown]).
40. Paraná, Parque Nacional do Iguaçu (25.63°S, 54.46°W): quica (coll. L.P. Costa, 1998).
41. Rio de Janeiro, Debossan, Sítio Xitaca, Nova Friburgo (22.28°S, 42.53°W): quica (L. Geise [date unknown]).
42. Rio de Janeiro, Guapimirim, Garrafão, (22.45°S, 43.00°W; R. Cerqueira, personal commun.): quica (coll. L. Geise, 1991).
43. Rio de Janeiro, Parque Nacional de Itatiaia, Município de Itatiaia (22.38°S, 44.63°W): quica (coll. M.A. Mustrangi, 1993).
44. Rio de Janeiro, Restinga de Maricá, Barra de Maricá, Maricá (22.88°S, 42.83°W): quica (coll. M.C. Lara, 1992).
45. São Paulo, Fazenda da Toca, Ilha de São Sebastião, Ilhabela, 150 m (23.82°S, 45.35°W): quica (coll. M.A. Mustrangi, 1992).
46. São Paulo, Fazenda Intervales, Base do Carmo, 5.5 km S Capao Bonito (24.33°S, 48.42°W): quica (coll. M.A. Mustrangi, 1992).
47. São Paulo, Praia do Félix, Ubatuba (23.38°S, 44.97°W): quica (coll. M.A. Mustrangi [date unknown]).
48. São Paulo, Serra do Japí, 7 km W Jundiai (23.23°S, 46.95°W): quica (coll. M.A. Mustrangi, 1993).
49. Tocantins, Lagoa da Confusao [10.87°S, 49.70°W; Rocha et al., 2015]: canus (coll. R.G. Rocha, [date unknown]).
50. Tocantins, near Pium, including sublocalities “Centro de Pesquisa Canguçu” (9.98°S, 50.03°W), “Parque Estadual do Cantão” (9.96°S, 50.12°W), and “N Pium” [9.47°S, 50.04°W; Rocha et al., 2015]: canus (coll. R.G. Rocha, 2007).
COLOMBIA
51. Caldas, Victoria, Vereda Canan, sitio La Esperanza (5.32°N, 74.93°W): melanurus (coll. J.F. Diaz- Nieto, 2006).
52. Meta, Finca El Capricho, 38 km E Villavicencio, 300 m [4.15°N, 73.29°W; KU collection database]: canus (coll. J.A.W. Kirsch, 1969).
53. Putumayo, 17 km N Puerto Asis [ca.0.67°N, 76.50°W; map estimate]: andersoni (coll. J.A.W. Kirsch, 1969).
54. Valle, 28 km NE Buenaventura, 150 m [3.60°N, 76.87°W; KU collection database]: melanurus (coll. J.A.W. Kirsch, 1969).
ECUADOR
55. Los Ríos, Lima Pareja, 4 km SW Puebloviejo (1.55°S, 79.47°W): melanurus (coll. R.G. McLean, 1975).
56. Orellana, Onkone Gare, 38 km S Pompeya Sur (0.65°S, 76.45°W): andersoni (coll. ROM expedition, 1995).
57. Orellana, Parque Nacional Yasuni, 42 km S, 1 km E Pompeya Sur [0.68°S, 76.43°W; Gregorin et al., 2006]: pebas (coll. ROM expedition, 1996).
58. Orellana, Parque Nacional Yasumi, 18 km S Pompeya Sur [ca. 0.60°S, 76.61°W; map estimate]: andersoni (coll. ROM expedition, 1996).
EL SALVADOR
59. La Paz, Zacatecoluca, Hacienda Escuintla [13.50°N, 88.87°W; TTU collection database]: pallidus (coll. J.G. Owen, 1990).
GUYANA
61. Barima-Waini, Waikerebi (7.52°N, 59.38°W): opossum (coll. B.K. Lim and S.M. Woodward, 1991).
62. Potaro-Siparuni, 30 km NE Surama (4.33°N, 58.85°W): opossum (coll. ROM expedition, 1990).
63. Potaro-Siparuni, Iwokrama Forest, Sand Stone (4.38°N, 58.92°W): opossum (coll. ROM expedition, 1999).
64. Upper Takutu-Upper Essequibo, Chodikar River, 55 km SW Gunn's Strip (1.37°N, 58.77°W): opossum (coll. ROM expedition, 1996).
MEXICO
65. Campeche, 11 km by road S Candelaria [18.09°N, 91.07°W; ASNHC collection database]: pallidus (coll. M.D. Engstrom, 1984).
66. Campeche, 3.7 km SE Chekubul [18.80°N, 90.98°W; ROM collection database]: pallidus (coll. M.D. Engstrom and R.C. Dowler, 1989).
67. Campeche, El Remate, 14 km W Tancuche [20.51°N, 90.38°W; ASNHC collection database]: pallidus (coll. R. Dowler and B. Lim, 1990).
68. Chiapas, 19 km N Palenque [17.66°N, 92.00°W; ASNHC collection database]: pallidus (coll. M.D. Engstrom, 1984).
69. Quintana Roo, 1 km W Puerto Morelos [20.85°N, 86.90°W; ROM collection database]: pallidus (coll. M.D. Engstrom, 1990).
70. Tabasco, 5 km N Jonuta [18.13°N, 92.12°W; ROM collection database]: pallidus (coll. M.D. Engstrom and R.C. Dowler, 1989).
71. Tabasco, 27 km S and 14 km E El Triunfo [17.68°N, 91.04°W; ASNHC collection database]: pallidus (coll. K.L. Curran, 1986).
PANAMA
72. Bocas del Toro, Isla Bastimentos, Old Point [9.17°N, 82.05°W; Siegel and Olson, 2008): melanurus (J. Jacobs, 1987).
73. Bocas del Toro, Peninsula Valiente, Punta Alegre [9.09°N, 81.54°W; Siegel and Olson, 2008]: melanurus (coll. F.M. Greenwell, 1990).
74. Panamá, Parque Nacional Altos de Campana, 850 m (8.68°N, 79.93°W): melanurus (coll. ROM expedition, 1995).
PARAGUAY
75. Alto Paraguay, Bahia Negra, Tres Gigantes [20.08°S, 58.16°W; de la Sancha and D'Elía, 2015]: canus (collector and date unknown).
76. Ñeembucú, Estancia Santa Teresa, ca. 2 km S Puesto Anastacio (26.57°S, 58.14°W): canus (coll. G. D'Elía, 1999).
77. Paraguarí, Sapucay (= Sapucaí at 25.67°S, 56.92°W; Paynter, 1989): quica (coll. anonymous MACN personnel, 1933).
78. Presidente Hayes, Estancia Loma Porá [23.52°S, 57.52°W; de la Sancha and D'Elía, 2015]: canus (collector and date unknown).
PERU
79. Amazonas, vicinity of Huampami, Rio Cenepa [4.47°S, 78.17°W; Patton et al., 1982]: andersoni (coll. J.L. Patton, 1977).
80. Loreto, 21 km S Iquitos, Otorongo Army Base [3.95°S, 73.37°W; C.L. Hice, personal commun.]: pebas (coll. C.L. Hice, 1998).
81. Loreto, 25 km S Iquitos, Estación Biológica Allpahuayo [3.97°S, 73.42°W; Hice et al., 2004]: andersoni and pebas (coll. C.L. Hice, 1998).
82. Loreto, Nuevo San Juan, right bank Río Gálvez [5.25°S, 73.17°W; Voss and Fleck, 2011]: mcilhennyi (coll. R.S. Voss, 1998; D.W. Fleck, 1999).
83. Loreto, San Jacinto [2.32°S, 75.87°W; Duellman and Mendelson, 1995]: andersoni (coll. N. Woodman, 1993).
84. Loreto, Yurimaguas [5.90°S, 76.08°W; Stephens and Traylor, 1983): andersoni (coll. M.P. Anderson, 1912).
85. Madre de Dios, Reserva Cuzco Amazónico [12.55°S, 69.05°W; Duellman and Koechlin, 1991]: pebas (coll. N. Woodman, 1989).
SURINAM
86. Para, Zanderij (5.45°N, 55.20°W): opossum (coll. L. Roberts, 1980).
87. Sipaliwini, Bakhuis, Transect 13 (4.55°N, 57.06°W): opossum (coll. ROM expedition, 2006).
88. Suriname, Plantation Clevia, 8 km NE Paramaribo [5.87°N, 55.13°W; CM database]: opossum (coll. S.L. Williams, 1981).
VENEZUELA
89. Amazonas, Belen, left bank Rio Cunucunuma, 150 m [3.65°N, 65.77°W; Voss and Emmons, 1996]: andersoni (coll. USNM expedition, 1967).
90. Bolívar, 8 km S and 5 km E El Manteco [7.32°N, 62.47°W; Ochoa and Ibáñez, 1985]: canus (coll. L.W. Robbins, 1981).
91. Trujillo, 9.8 km NNE Motatán, 230 m [9.47°N, 70.56°W; KU collection database]: canus (coll. J.A.W. Kirsch, 1969).
APPENDIX 4
Analyzed Craniodental Samples of Adult Male Philander with Generalized (Mahalanobis) Distances
APPENDIX 5
On the Type Locality of Didelphys frenata Olfers, 1818
Robert S. Voss and Renate Angermann13
Although the gray four-eyed opossum of southeastern Brazil is now commonly known as Philander frenatus following Patton and da Silva (1997), this usage is not consistent with available information about where the holotype was collected. The epithet was first made available by Olfers (1818: 204), but Olfers attributed the name to Karl Illiger, first director of the newly founded zoological museum in Berlin, who had previously used frenata as a nomen nudum (Illiger, 1815). Olfers is known to have studied the mammal collection of the Berlin museum in 1816 or early 1817 (Voss and Angermann, 1997), where he must have seen Illiger's original material.14 Olfers (1818) treated frenata as a variety of Didelphys opossum and stated only that it occurred in South America.
As explained elsewhere (Voss and Angermann, 1997), the nucleus of the mammal collection of the Berlin Zoological Museum (ZMB) at its inception in 1810 was a large series of Brazilian specimens donated by Johann Centurius von Hoffmannsegg, a wealthy patron of German science. As far as known, all of Hoffmannsegg's Brazilian mammals were collected either by Friedrich Wilhelm Sieber, who worked in eastern Amazonia from 1803 to 1812, or by Francisco Agostinho Gomes, who lived at Bahia (now Salvador) and sent specimens to Hoffmannsegg from 1801 to 1807. The earliest known catalog of the ZMB collection—a manuscript entitled “Catalogus mammalium et avium Musei Regii” dated 1810 and preserved in the museum's archives (Historische Bild- und Schriftgutsammlungen)—lists Didelphys frenata as a new species based on material donated by Hoffmannsegg, and a penciled specimen tally in the margin indicates that only a single specimen was present. Our inspection of the ZMB accessions register (“Eingangskatalog”) suggests that no additional material identified as D. frenata was received by the museum prior to Olfer's visit.
All the mammal specimens in the early ZMB collection were mounted in lifelike poses (with the skull inside) for exhibition in the museum's public galleries, probably between 1811 and 1814, and any original collectors' labels were probably lost or discarded at that time; such data as now accompany early 19th-century specimens in the ZMB were copied from the exhibition labels when many live-mounts were remade as conventional study skins with extracted skulls in the 1900s (Voss and Angermann, 1997). The appearance of the specimen currently cataloged as the type of Philander frenatus, ZMB_MAM 2325 (the skin and skull of an old adult female) is consistent with this scenario; its labeling records the locality as “Brasilien,” the collector as “Sieber,” and the donor as “v. Hoffmannsegg.” The fur color of this specimen is now faded almost beyond recognition, but the skull and dentition are substantially intact. Although Olfers (1818: 205) stated that he had seen young examples of Philander in Brazil, none is known to have survived as a museum specimen, and it is not clear that any served as the basis for his description of P. frenatus. In the absence of evidence to the contrary, we consider ZMB_ MAM 2325 to be the holotype by monotypy.
Because at least one ZMB specimen collected by Gomes is known to have been misattributed to Sieber (the holotype of Chaetomys subspinosus; Voss and Angermann, 1997), we examined manuscript invoices of Gomes's shipments from Bahia. These lists suggest that Gomes sent only a single marsupial specimen to Hoffmannsegg, of which only the head was preserved,15 so Gomes' opossum cannot be the skin and skull now cataloged as ZMB_MAM 2325. Therefore, the holotype of Philander frenatus must have been collected by Sieber, as it is now labeled.
The only documentary evidence of where in Brazil the holotype might have been collected is the penciled annotation “Para” (the old name for Belém, the largest city in eastern Amazonia) beneath the inked entry for ZMB_MAM 2325 in the mammal department's general catalog (initiated by Wilhelm Peters sometime after 1857; Angermann, 1989). As previously noted, Sieber is known to have collected in eastern Amazonia (Urban, 1906), and measurements of the molar dentition of the holotype fall within the range of variation observed among specimens of Philander subsequently collected in that region; for example, we obtained a value of 13.7 mm for LM from the holotype of frenatus, whereas the mean value for LM in a sample of 42 specimens from eastern Amazonia is 13.8 ± 0.4 mm. By contrast, southeastern Brazilian specimens have much smaller molars (LM = 12.3 ± 0.4 mm, N = 40), so the morphology of ZMB_MAM 2325 is also consistent with its inferred eastern Amazonian origin. Eastern Amazonian populations of Philander seem to be minimally differentiated from Guianan populations in both molecular and phenotypic traits, so we treat frenatus as a junior synonym of P. opossum.
Current usage of the epithet frenatus for the southeastern Brazilian species that we recognize as Philander quica seems to be based on Hershkovitz (1959: 343), who stated that the type was collected by “Herr Kaehne” at Bahia. According to Hershkovitz, this information was provided by Wagner (1843: 44), but Wagner simply remarked that a specimen identified as Didelphys frenata by “Lichtenstein” (= Hinrich Lichtenstein, Illiger's successor as ZMB director) was collected by Kaehne in Bahia. Because Kaehne (= Franz Kaehne, a former apothecary from Prenzlau) collected in Brazil from 1831 to 1838 (Sick, 1960), his specimen cannot have been the one seen by Olfers (1818). Hershkovitz (1997: 51) subsequently alleged that the type was “collected before 1815 by Herr Kaehne” but provided no reference for this clearly erroneous statement.
Notes
[1] 4 Both vernacular names distinguish species of Philander from superficially similar taxa referred to Metachirus, commonly known as “brown four-eyed opossums” or “pouchless four-eyed opossums.”
[2] 5 The generalized distance (D) between two groups can be interpreted as the difference between group centroids scaled in units of within-group multivariate standard deviations; it is the appropriate metric for evolutionary inference from measurement data (Lerman, 1965).
[3] 6 For example, Patton and da Silva (1997) described the middorsal fur as “ca. 10 mm long” in P. andersoni versus “ca. 18 mm in length” in P. mcilhennyi, a substantial difference. However, the middorsal fur of 21 specimens of P. andersoni that we measured from northern Peru (Amazonas) and eastern Ecuador was 13 ± 2 mm long with an observed range of 10–16 mm, whereas the middorsal fur of 21 P. mcilhennyi from south of the Amazon in Peru and Brazil was 16 ± 3 mm with an observed range of 12–22 mm.
[4] 7 After Voss and Jansa (2009: 121–123), but including corrections and supplementary observations.
[5] 8 We are told (D. Astúa, personal commun.) that this is unlikely. The southeasternmost record we have seen of Philander opossum is a specimen from Canudos, Rio Grande do Norte, Brazil, collected by F. Lima in 1920 (FMNH 24790).
[6] 9 We suspect (although there is no behavioral evidence to support our conjecture) that the black-and-white tail markings of Didelphini have some social-signaling function that might be coopted for species recognition in sympatry.
[7] 10 From the eastern Andean foothills of Junín and Ayacucho departments, Peru: BMNH 94.10.1.16, 94.10.1.17, 28.5.1.20; FMNH 65782; LSUMZ 16398, 16399; MUSM 71.
[8] 11 We are indebted to S.E. Pavan, who kindly examined the MPEG voucher specimens from Nunes et al.'s (2006) study at our request.
[9] 12 There appears to be no standard terminology for these structures, despite their essential similarity among the tribosphenic taxa that exhibit them. Simpson (1936: 5), for example, used “anterior cingulum” and “posterior cingulum,” and the postcingulum is sometimes called the “talon” (Bown and Kraus, 1979: 173).
[11] 14 Ignatz von Olfers (b. 1793), a young member of the Prussian legation to Brazil, arrived at Rio de Janeiro in the late spring or early summer of 1817. For additional information about his work, the Berlin collections he studied, and other relevant historical background (including manuscript documents mentioned in the following text), see Voss and Angermann (1997).
[12] 15 This specimen was identified as Didelphys cayopollin in a document entitled “Envoy de la Caisse Nr. 7 du 24 May 1802.” With few exceptions, the name D. cayopollin was consistently used by 19th-century authors for the species currently known as Caluromys philander (see Gardner, 2008: 10–11).
