Phylogenetic analyses of cytochrome-b sequence data from 13 of the 15 currently recognized species of New World porcupines were used to test competing taxonomic hypotheses and to explore scenarios of morphological evolution and biogeography. Consistent with previous studies, the monophyly of Erethizontidae (Chaetomys Erethizon Coendou) and the monophyly of Erethizontinae (Erethizon Coendou) were both strongly supported. However, cytochrome- b sequence data provide no support for the reciprocal monophyly of Coendou, “Echinoprocta,” and “Sphiggurus” as those taxa were previously recognized by authors. All of the erethizontid species recognized in recent revisionary work and represented by multiple sequences in this study were recovered as monophyletic groups. Maximum-likelihood (ML) analyses of these data recovered the following phylogeny for 11 species of Coendou: ((melanurus (ichillus (pruinosus vestitus))) ((spinosus (bicolor nycthemera)) (prehensilis (mexicanus (quichua rufescens))))). Ancestral-state reconstructions based on the ML topology suggest that several morphological characters emphasized in past erethizontid classifications (size, nasofrontal sinus inflation, and long fur) have evolved homoplasiously Maximum-likelihood inference of geographic range evolution suggests that the last common ancestor of living erethizontids was a cis-Andean species, and that most subsequent cladogenesis was also cis-Andean; however, at least two trans-Andean dispersal events are plausibly indicated, as well as two separate invasions of Andean landscapes. Among the most remarkable results of this study are almost-identical sequences of Coendou prehensilis from localities spanning 27° of latitude and 25° of longitude; we speculate that a trophic-niche shift might have allowed rapid range expansion of this species, which accounts for almost all known cases of geographic range overlap and sympatry in the genus Coendou.
INTRODUCTION
Recent New World porcupines (Erethizontidae) are arboreal caviomorph rodents that defend themselves with dangerous quills and eat leaves, bark, fruit, and immature seeds (Charles-Dominique et al., 1981; Roze, 1989; Chiarello et al., 1997; Emmons, 1997; Cherubini et al., 2003; Passamani, 2010). With the exception of one boreal North American species, erethizontids occur in tropical and subtropical habitats from southern Mexico to northern Argentina. Most Neotropical porcupines inhabit lowland rainforest, but some also occur in dry (deciduous) forests, and others are found in montane (“cloud”) forests to at least 3500 m above sea level. Solitary, nocturnal, and silent, most erethizontids are seldom observed even where they are locally common, and several species are known from just a few museum specimens.
The current taxonomy (summarized by Voss, 2011) recognizes 15 erethizontid species in three genera (table 1). Two genera each include only a single living species: Chaetomys (containing only C. subspinosus, the Brazilian bristle-spined porcupine) and Erethizon (containing only E. dorsatum, the North American porcupine). The remaining 13 Recent species belong to Coendou (commonly known as prehensile-tailed porcupines), but some authors have recognized additional subgenera or genera within this group. Woods and Kilpatrick (2005), for example, referred several long-furred species of Coendou to the genus Sphiggurus, and the short-tailed species C. rufescens has long been referred to the monotypic genus Echinoprocta. Because these alternative usages were not based on any defensible hypothesis of reciprocal monophyly, Alberico et al. (1999) and Voss (2011) treated Sphiggurus F. Cuvier, 1823, and Echinoprocta Gray, 1865, as junior synonyms of Coendou Lacépède, 1799.
To date, erethizontid systematics has been almost entirely based on morphological data, only three DNA-sequencing studies having included multiple erethizontid exemplars (Bonvicino et al., 2002; Vilela et al., 2009; Leite et al., 2011). Taking synonymies and corrected identifications (Voss, 2011) into account, those studies suggest that (1) Chaetomys is the sister taxon of other living New World porcupines, (2) Erethizon is the sister taxon of Coendou, and (3) Coendou spinosus and C. melanurus are more closely related to one another than either is to C. prehensilis. However, most species of Coendou have not been sequenced for any gene, so existing molecular phylogenies are largely uninformative about erethizontid biogeography and morphological evolution.
This article reports the first phylogenetic analyses of New World porcupines based on a taxonomically dense molecular dataset; of the 15 currently recognized Recent species, only two (Coendou insidiosus and C. roosmalenorum) are not represented in this study. For several species with noteworthy geographic variation and/or problematic synonyms, we included multiple samples to test species monophyly. Because preserved tissue samples are unavailable for several species, we analyzed sequence data from the mitochondrial gene encoding cytochrome b, which can often be amplified from old museum skin-and-skull preparations. Our results corroborate some previous hypotheses about porcupine relationships, challenge others, and provide novel insights about porcupine morphological evolution and biogeography.
TABLE 1.
Classification and Distribution of Recent Erethizontidsa
MATERIALS AND METHODS
VOUCHER SPECIMENS: Morphological voucher material for erethizontid DNA sequences analyzed in this report (tables 2 and 3) is preserved in the following institutional collections: American Museum of Natural History, New York (AMNH); Angelo State Natural History Collection, San Angelo (ASNHC); Estación Biológica de Rancho Grande, Maracay (EBRG); Field Museum of Natural History, Chicago (FMNH); Instituto Nacional de Pesquisas da Amazônia, Manaus (INPA); Biodiversity Research Center, University of Kansas, Lawrence (KU); Los Angeles County Museum, Los Angeles (LACM); Museu de Ciências Naturais da Universidade Luterana do Brasil, Canoas (MCNU); Muséum National d'Histoire Naturelle, Paris (MNHN); Museo de Historia Natural La Salle, Caracas (MHNLS); Museu Nacional, Rio de Janeiro (MNRJ); Museo de Historia Natural de la Universidad Nacional Mayor de San Marcos, Lima (MUSM); Museum of Vertebrate Zoology, University of California at Berkeley, Berkeley (MVZ); Museu de Zoologia da Universidade do São Paulo, São Paulo (MZUSP); Museum of Texas Tech University, Lubbock (TTU); Universidade Federal de Minas Gerais, Belo Horizonte (UFMG); Universidade Federal da Paraíba, João Pessoa (UFPB); University of Michigan Museum of Zoology, Ann Arbor (UMMZ); National Museum of Natural History, Washington, DC (USNM). The catalog numbers of specimens preserved in the mammal collection of the Manso hydroelectric dam (in Cuiabá) are prefixed with “Manso” (after Bonvicino et al., 2002).
TABLE 2.
Erethizontid Specimens Sequenced for This Report
TABLE 3.
Erethizontid Sequences Downloaded from GenBank
DNA EXTRACTION, AMPLIFICATION, AND SEQUENCING: We extracted DNA from fresh tissues (heart, liver, or kidney) that had been preserved in ethanol or from dried material (skin fragments and/or quills) harvested from museum specimens. All DNA extractions were performed using a Qiagen DNA Minikit (Qiagen, Inc.). Ethanol-preserved tissue samples were extracted according to kit instructions, but dried skin fragments and quills required additional processing to yield uncontaminated DNA. All extractions from dried museum material were conducted in a UV-sterilized hood in a laboratory that had never been used to amplify mammalian DNA sequences. Skin samples were washed in a series of ethanol and water washes with overnight soaks (described in Giarla et al., 2010). Quills were similarly washed and soaked, but the time of each soak was shortened to 1 hour (rather than overnight). Washed skin samples were digested according to kit instructions, except that skin was allowed to sit in the lysis buffer + proteinase K mixture for two days with an extra 30 µL of proteinase K added on the second day. Quills were subjected to a similar two-day lysis, and 30 µL of 100 mg/mL dithiothreitol (DTT) was added to the digestion each day.
FIG. 1.
Collection localities of sequenced specimens of Neotropical erethizontids (Chaetomys and Coendou). See gazetteer (appendix 1) for geographic coordinates and other information.
For most tissue samples, polymerase chain reaction (PCR) amplification of the entire mitochondrial gene encoding cytochrome b was done in two overlapping fragments of ∼700 bp using primers MVZ05 paired with PorcCytb676R and either PorcCytb548F or PorcCytb565F paired with UMMZ04 (table 4, fig. 2). In cases where amplification of these two large fragments failed, the gene was amplified in four ∼300–400 bp fragments using the following primer pairs: MVZ05/PorcCytb404R, PorcCytb323F/PorcCytb676R, PorcCytb565F/PorcCytb827R, and PorcCytb761F/UMMZ04. If PCR amplifications yielded sufficient product for sequencing, we sequenced this product directly; otherwise, we performed a second round of PCR using the same primers and 1 µL of the amplification product as template. All amplifications of tissuederived DNA were done as 12.5 µL reactions using GoTaq Hot Start Polymerase (Promega Corp.) with 0.5 µL of each primer, 2.5 µL of reaction buffer, 1.0 µL of 25 mM MgCl2, 0.25 µL of dNTP mix, 0.065 µL of polymerase, 1.0 µL of template (either DNA or PCR product), and 6.685 µL of dH2O.
TABLE 4.
Primers Used to Amplify and Sequence the Cytochrome b Gene
For museum skins and quills, cytochrome b was amplified in overlapping fragments of ∼150–300 bp using various combinations of primers. In most cases, this initial amplification did not yield enough product for sequencing, so we performed a second round of PCR amplification as above. We used Platinum Taq (Life Technologies Corp.) for amplification of DNA derived from these degraded samples. Each reaction comprised 0.25 µL of each primer, 1.25 mL of reaction buffer, 0.375 µL of 50 mM MgCl2, 0.25µL of dNTP mix, 0.05 µL of polymerase, 1.0 or 2.0 µL of template, and enough dH2O to make a 12.5 µL reaction.
All PCR reactions using DNA as template were performed using a four-step touchdown protocol. Cycling started at an annealing temperature 2° C above the average primer-annealing temperature and dropped by 2° increments for each step; the first three steps were run for 5 cycles, the last was run for 20 cycles. Any reamplification reactions were performed for 25 cycles using a single annealing temperature. Before sequencing, we used exonuclease I and shrimp alkaline phosphotase (Hanke and Wink, 1994) to remove primers and unincorporated nucleotides from the PCR product. Amplification products were sequenced in both directions on an ABI 3730 sequencer using amplification primers and dye-terminator chemistry (BigDye ver. 3.1 Cycle Sequencing Kit, Life Technologies Corp.). Sequences were edited and assembled using Sequencher ver. 4.7 (GeneCodes, Inc.) and were imported into Geneious ver. 5.6 (Drummond et al., 2011) to prepare files for analysis. All sequences newly obtained for this study have been deposited in GenBank (with accession numbers KC463857—KC463889).
FIG. 2.
Names and locations of primers (table 4) used to amplify and sequence cytochrome b for this study.
PHYLOGENETIC ANALYSIS AND CHARACTER OPTIMIZATION: We downloaded cytochrome-b sequences representing 14 non-erethizontid hystricognath genera from GenBank to use as outgroups in our phylogenetic analyses: Hystrix (GenBank accession FJ472565), Bathyergus (AF012241), Cryptomys (AF012233), Thryonomys (AJ301644), Capromys (AF422915), Cavia (AY382790), Ctenomys (AF370680), Dasyprocta (AF437784), Galea (GU067494), Microcavia (GU067490), Myocastor (AF422919), Myoprocta (AF437781), Octodon (AF007058), and Proechimys (AJ251400). Sequences were aligned using the MUSCLE algorithm (Edgar et al., 2004) as implemented in Genious ver. 5.6 (Drummond et al., 2011).
We analyzed our aligned sequence data using maximum parsimony (MP) and maximum likelihood (ML) methods. To reduce computational time for the MP analysis, we first deleted 16 of 19 nearly identical sequences (differing by <1%) of C. prehensilis; we chose the two individuals with the longest sequences (AMNH 262274 and USNM 528360, each 1140 bp long) as well as the specimen from Pernambuco (MN 73383; 801bp) to represent the genetic diversity of this clade. We analyzed this abbreviated (43-terminal) dataset using 20 replicates of random-taxon addition heuristic tree searches with TBR branch swapping as implemented by PAUP* ver. 4.0b10 (Swofford, 2003). To estimate nodal support under the MP criterion, we performed tree searches on 500 bootstrap replicates using 3 replicates of random-taxon addition and TBR branch swapping for each bootstrap replicate. For ML analysis, we used the full (59-terminal) dataset (including all 19 C. prehensilis sequences). We first determined the best-fitting model of sequence evolution using MrModelTest ver. 2.3 (Nylander, 2004). We specified the resulting model in two independent runs of maximum-likelihood tree searching using GARLi 2.0 (Zwickl, 2006), with two search replicates for each run and all other parameters set as defaults. To obtain estimates of nodal support, we used RaXML ver. 7.2.8 as implemented on the RAxML BlackBox web server (Stamatakis et al., 2008) to perform 500 replicates of rapid bootstrapping.
To obtain a time-scaled ultrametric phylogeny we employed Bayesian searches under a relaxed molecular clock model as implemented in BEAST ver. 1.7.2 (Drummond and Rambaut 2007; Drummond et al. 2006). We specified an HKY+Г+I model of nucleotide substitiution with default priors, a Yule speciation process with a prior birth rate of 0.1, and a lognormal relaxed-clock model with mean = 0.01 and standard deviation = 0.333. We set a prior constraint on the height of the node defining Caviomorpha to have a normal distribution with a mean of 34.1 MYA and standard deviation of 3.5 MYA. This prior was based on the inferred mean date of origin (and associated error estimate) for crown-group Caviomorpha from a recent multilocus phylogenetic study of hystricognaths (Upham and Patterson, 2012). We performed 10 million generations of MCMC searches, with trees saved every 1000 generations and used Tracer ver. 1.5 (Rambaut and Drummond, 2007) to examine posterior parameter estimates and other output from the MCMC run.
We reconstructed the evolution of three discrete morphological traits on our time-scaled molecular phylogeny by coding binary traits for the 13 erethizontid species included in our study. To optimize character-state evolution, we trimmed our time-scaled molecular phylogeny to the species level, and used maximum-likelihood estimation under an equal-rates model of character evolution (Pagel, 1994) as implemented in the ace() command of the R-package APE 3.0–5 (Paradis et al., 2004). We reconstructed erethizontid biogeographic history using maximum-likelihood optimization under the the Dispersal-Extinction-Cladogenesis model implemented in the Python package LaGrange (Ree and Smith, 2008). We used the same time-scaled species-level tree that we used for morphological character optimization, but coded each terminal taxon as occupying one or more of three discrete areas. Dispersal among areas was unconstrained, but lineages were not allowed to occupy more than two areas (the maximum area occupancy observed among Recent species).
RESULTS
The 45 ingroup cytochrome-b sequences we analyzed range in length from 259 to 1140 bp (tables 2, 3), and no insertion-deletion events were required to align these or the 14 outgroup sequences that we obtained from GenBank, resulting in an aligned matrix with a total of 59 terminals. For parsimony analysis we removed 16 nearly identical sequences of C. prehensilis to yield a dataset with 43 terminals. There were 500 parsimony-informative characters in this reduced matrix, and tree searches resulted in 14 minimum-length trees (length = 2892; CI = 0.342; RI = 0.558) with the strict consensus shown in figure 3. For maximum-likelihood analyses we used the complete dataset of 59 terminals. Comparisons among 24 models of nucleotide substitution using the Akaike information criterion indicated that GTR+I+Г was the best-fitting model for this dataset (table 5). Two independent searches using GARLi resulted in trees with identical topologies and likelihood scores (lnL = -12848.85; fig. 4).
SPECIES LIMITS, DISTANCE COMPARISONS, AND PHYLOGEOGRAPHY: The monophyly of all erethizontid species represented by multiple terminals in our analyses (Coendou prehensilis, C. spinosus, C. bicolor, C. quichua, C. nycthemera, and C. melanurus) was strongly supported (bootstrap > 75%) by both maximum-parsimony and maximum-likelihood analyses. Additionally, uncorrected mean intraspecific distances (table 6) for these taxa are well within the range of variation commonly observed for mammalian species, from 0.9% (in C. prehensilis) to 3.4% (in C. quichua), with an average value of 1.9%. By contrast, uncorrected interspecific distances observed in this study range from 3.5% (between C. pruinosus and C. vestitus) to 12.1% (between C. nycthemera and C. prehensilis), with an average value of 9.1%.
FIG. 3.
Strict consensus of 14 equally most-parsimonious trees for 29 unique erethizontid cytochrome-b haplotypes (only ingroup relationships are shown). Sequenced specimens of Coendou are identified by country of origin, next-largest political unit (state, department, or province), collection locality number (mapped in fig. 1), and an alphanumeric identifier (tissue, voucher, or GenBank accession number; see tables 2 and 3). Nodal support values are bootstrap percentages.
TABLE 5.
Estimated Substitution Parameters of the GTR+I+Г Model
Intraspecific haplotype variation is geographically structured in several species, notably Coendou spinosus, in which a northeast-to-southwest sequence links samples from coastal Brazil to those from Paraguay. In C. bicolor, samples from adjacent localities in northern Peru cluster together, as do samples from adjacent localities in southern Peru and northern Bolivia. In both of these examples, patterns of haplotype coalescence closely correspond to geographic expectations, but sequences of C. quichua from distant localities in Panama and western Ecuador are (anomalously) more closely related to one another than either is to a geographically intermediate sample from northern Colombia.
Remarkably, our samples of Coendou prehensilis consist of a single genetically divergent sequence from eastern Brazil (Pernambuco, locality 10) and a tight cluster of nearly identical sequences from 13 localities scattered throughout the rest of tropical South America. In this case, the tabulated mean intraspecific distance (0.9%) conceals substantial sequence heterogeneity, because HM462243 (the sequence from from Pernambuco) differs from other sequences of C. prehensilis by an average distance of 4.6%, whereas the latter differ among themselves by an average distance of just 0.5%.
PHYLOGENETIC RELATIONSHIPS: Parsimony and likelihood analyses both provide very strong support (bootstrap > 90%) for erethizontid monophyly (Chaetomys + Erethizon + Coendou) and for erethizontine monophyly (Erethizon + Coendou). Parsimony also provides very strong support for the monophyly of Coendou, but support for generic monophyly from maximum likelihood is notably weaker (83%). Because the species of Coendou (sensu Alberico et al., 1999; Voss, 2011) were formerly classified in three genera (see above), we used a likelihood-ratio test (Shimodaira and Hasegawa, 1999) to compare the topology in figure 4 with the alternative hypothesis that Coendou (sensu lato), Sphiggurus, and Echinoprocta are reciprocally monophyletic.3 The resulting test statistic suggests that our sequence data are significantly (ΔlnL = 139.91; P = 0.000) less likely under the latter hypothesis, which can be rejected on this basis.
Although parsimony and likelihood analyses recovered different patterns of relationships among species of Coendou, all of the discrepant relationships are weakly supported. For example, parsimony recovered melanurus, ichillus, and pruinosus + vestitus as separate lineages in an unresolved basal polytomy, whereas likelihood recovered the same taxa as a weakly supported group (clade C in fig. 4). By contrast, both parsimony and likelihood provide strong support for two nested sets of relationships, one (clade A) consisting of (prehensilis (mexicanus (rufescens + quichua))), and the other (clade B) consisting of (spinosus (bicolor + nycthemera)). In the maximum-likelihood topology, clades B and C are weakly resolved as sister taxa.
FIG. 4.
Maximum-likelihood phylogeny for 45 ingroup (erethizontid) terminals; outgroup taxa are not shown. Labeling conventions and nodal support statistics are the same as in figure 3. Capital letters (A, B, C) indicate unnamed clades discussed in the text.
TABLE 6.
Mean Uncorrected Cytochrome-b Distances (scaled as percent sequence divergence; below diagonal) and Mean K2P-corrected Distances (above diagonal) among Sequenced Species of Coendou a
FIG. 5.
Coendou prehensilis (INPA 2875), the type species of Coendou. This is a long-tailed species that appears completely spiny because the quills conceal its short, sparse fur.
FIG. 6.
Coendou melanurus (AMNH 266565), referred to Sphiggurus by Husson (1978) and other authors. This is a long-tailed species in which the quills are concealed beneath long, dense fur.
FIG. 7.
Coendou rufescens (FMNH 88524), previously referred to Echinoprocta by many authors. This is a shorttailed species that (like C. prehensilis) appears completely spiny because the quills conceal its short, sparse fur.
FIG. 8.
Lateral cranial views: A, Coendou prehensilis (AMNH 134064); B, C. melanurus (AMNH 266565). The inflated nasofrontal sinuses of C. prehensilis (type species of the genus Coendou) result in a strongly convex dorsal profile by contrast with the flat dorsal profile of C. melanurus (referred to Sphiggurus by some authors; see text). Both skulls are life size (×1).
DISCUSSION
The monophyly of Erethizontidae (Chaetomys + Coendou + Erethizon) is a noncontroversial result previously obtained from phylogenetic analyses of cytochrome-b sequence data by Vilela et al. (2009). The alternative hypothesis—implicit in classifications that formerly referred Chaetomys to the family Echimyidae (e.g., Miller and Gidley, 1918; Patterson and Wood, 1982)—has now been so convincingly refuted by multiple lines of evidence (Martin, 1994; Carvalho, 2000) that it no longer merits serious attention. The monophyly of Erethizontinae (Erethizon + Coendou), likewise recovered from a previous analysis of cytochrome-b sequence data (Vilela et al., 2009), is also noncontroversial.
TABLE 7.
Selected Morphological Traits of 11 Species of Coendoua
By contrast, the generic classification of the Neotropical erethizontines variously referred by authors to Coendou, Echinoprocta, and Sphiggurus is fraught with controversy. Species referred to these genera differ in size and in qualitative external and craniodental characters (figs. 5–8; table 7). Because nomenclatural aspects of Neotropical erethizontid generic taxonomy have been reviewed elsewhere (Tate, 1935; Alberico et al., 1999), we limit our discussion to evidential support for monophyletic groups in the paragraphs that follow.
GENERIC CLASSIFICATION
Our results provide no support for the continued recognition of Echinoprocta and Sphiggurus as those taxa have traditionally been understood by authors (e.g., Cabrera, 1961; Husson, 1978; Honacki et al., 1982; Woods, 1984; Woods and Kilpatrick, 2005). As noted by Ellerman (1940) and White (1970), “Echinoprocta” rufescens is craniodentally indistinguishable from ordinary Coendou. Authors who recognize Echinoprocta as a valid genus emphasize its uniquely short and allegedly nonprehensile tail, but computed ratios of tail to head-and-body length (table 7) suggest that rufescens is simply the shortest-tailed member of a clade that exhibits continuous taxonomic variation in this proportion. Additionally, most specimens of rufescens have a nearly hairless dorsal callus at the tip of the tail, suggesting that this stubby organ normally retains some prehensile ability (Alberico et al., 1999). Our phylogenetic analyses recovered rufescens as the sister taxon of C. quichua, montane populations of which are also short-tailed (Voss, 2011). Because recognizing a monotypic genus for rufescens would render Coendou nonmonophyletic, the continued recognition of Echinoprocta at any taxonomic rank is indefensible.
Modern recognition of Sphiggurus dates from Ellerman (1940), who ranked it as a subgenus of Coendou. According to Ellerman (1940: 185), members of the subgenus Coendou have “spines not mixed with hairy covering,” whereas members of the subgenus Sphiggurus have “spines mixed with and typically covered by long woolly hair.” Cabrera (1961) maintained Ellerman's subgeneric classification, but Husson (1978: 488) decided that two species of Surinamese porcupines “differ so strongly both in external and skull characters that their [generic] separation seems fully justified” and ranked Sphiggurus as a full genus. Although Husson provided detailed morphological comparisons of these two species—C. prehensilis (figs. 5, 8A) and “S.” melanurus (figs. 6, 8B)—he did not explain which of the characters that distinguish them serve to diagnose Sphiggurus from Coendou as supraspecific taxa, nor has any subsequent author who adopted Husson's generic taxonomy (e.g., Honacki et al., 1982; Woods, 1984). As noted by Handley and Pine (1992), the characters that differ between Husson's exemplar species are not consistently correlated among other Neotropical porcupines and cannot be used jointly to diagnose higher taxa. In fact, “Sphiggurus” (sensu Woods, 1984, 1993; Woods and Kilpatrick, 2005) is not diagnosable by any known morphological criterion (Voss, 2011).
Our phylogenetic results clearly indicate that Sphiggurus (as recognized by Ellerman, 1940; Cabrera, 1961; Honacki et al., 1982; Woods, 1984; Woods, 1993; Woods and Kilpatrick, 2005) is not monophyletic. Additionally, the principal morphological characters that differ between Husson's exemplar species of Sphiggurus and Coendou appear to have evolved homoplasiously in erethizontids (see below). As we have pointed out elsewhere (Voss, 2011), previous analyses of cytochrome-b sequence data alleged to support the reciprocal monophyly of Sphiggurus and Coendou (Bonvicino et al., 2002; Vilela et al., 2009) were compromised by taxonomic misidentifications of voucher specimens.
Nevertheless, it is true that prehensilis (the type species of Coendou) and spinosus (the type species of Sphiggurus) belong to different lineages (clades A and B; fig 4) that could validly be recognized as genera or subgenera. We see no purpose in doing so, however, because (1) neither clade seems to be morphologically diagnosable; (2) employing Sphiggurus in a completely different sense than that established by traditional usage would create needless confusion in the taxonomic literature with no compensatory advantage for scientific communication; and (3) a new generic name would then be needed for clade C, which remains weakly supported by available sequence data.
SPECIES LIMITS
Phylogenetic analyses of cytochrome-b sequence data provide welcome support for several species concepts resulting from recent revisionary work. In particular, our results are consistent with the synonymy of lowland Panamanian populations that Thomas (1902) described as Coendou rothschildi (represented by USNM 296308) with Ecuadorean material referable to C. quichua (represented by KMH 2218) as suggested by Voss (2011), and they refute previous concepts of C. bicolor that included quichua as a subspecies (Cabrera, 1961; Woods and Kilpatrick, 2005). Additionally, these data suggest the essential genetic continuity of long-furred porcupines from Paraguay to Rio de Janeiro, all of which should be referred to C. spinosus (as recommended by Voss, 2011; Junior and Leite, 2012), by contrast with the taxonomies suggested by Moojen (1952), Woods and Kilpatrick (2005), and others who recognized multiple species in this complex. Lastly, our results are consistent with the recognition of C. melanurus as a distinct Amazonian species (Voss and Angermann, 1997; Voss et al., 2001; Bonvicino et al., 2002), rather than as a subspecies (Cabrera, 1961) or synonym (Husson, 1978) of southeastern Brazilian taxa.
We made a particular effort to obtain sequence data from as many geographic localities as possible for Coendou prehensilis. Among the nominal taxa currently treated as synonyms of this widespread species, we analyzed sequences representing boliviensis (NK12984), centralis (AF411581, AF411582, AF411584), longicaudatus (T-1626), and sanctaemartae (USNM 281898, 281904). Additionally, we included sequence data from specimens previously misidentified as C. bicolor (see Voss, 2011) from eastern Ecuador (USNM 528360), northeastern Peru (MVZ 155200, 155201), western Brazil (MVZ 191349, U34851), and southwestern Venezuela (USNM 443409).
Although both parsimony and likelihood analyses provide strong support for the monophyly of Coendou prehensilis, it is noteworthy that the most divergent sequence we analyzed is HM462243, an 801 bp fragment that Leite et al. (2011) obtained from the recently designated neotype (MNRJ 73383). We have not examined this specimen, but Leite et al.'s exemplary description and illustrations match other material that we refer to C. prehensilis in most external and craniodental characters. However, the neotype has a smaller hind foot (76 mm) than any we have measured for this species (82–105 mm; Voss, 2011 : table 7), and it has a narrower fourth premolar (4.9 mm versus 5.3–6.4 mm). Additional material from Pernambuco would be useful to put these unusual values in statistical perspective, but such comparisons—and others that can be made based on measurements in Leite et al. (2011) and Voss (2011)—suggest that the neotype is unusually small for Brazilian specimens traditionally referred to C. prehensilis.
The oldest available name that could be used for other geographic populations currently referred to Coendou prehensilis is longicaudatus Daudin, 1802, which was based on specimens from Cayenne, French Guiana. As no type material is known to exist (Rode, 1945), it would be sensible to designate a neotype if this taxon were to be recognized in any future revision of Coendou. The French Guianan voucher represented by sequence data in our analysis (MNHN 1997.643) is an obvious candidate for neotype designation.
OTHER RELATIONSHIPS
The only clade recovered in our analyses that corresponds to a previously recognized group of Coendou species is the cluster of vestitus, pruinosus, and ichillus (in the ML tree; fig. 4), comprising part of the “Vestitus Group” of Voss and da Silva (2001). The missing member of the Vestitus Group is roosmalenorum, tissues of which were not obtained for this study. These species differ from other congeners by having bristle quills (long, wirelike, nondefensive quills) in addition to the usual defensive quills and soft fur in their dorsal pelage (Voss and da Silva, 2001: fig. 1). Interestingly, the sister taxon to the Vestitus Group in our likelihood analysis, Coendou melanurus, also has three distinct structures in its dorsal pelage: defensive quills, soft fur, and yellow-tipped guard hairs (the latter producing the streaked effect seen in the dorsal pelage of this species; fig. 6). The guard hairs of melanurus are much finer and more flexible than the bristle-quills seen in members of the Vestitus Group, but it now seems plausible that these pelage structures are homologous.
The very strongly supported sister-group relationship that we recovered between Coendou bicolor and C. nycthemera contradicts Handley and Pine's (1992) hypothesis that nycthemera—a short-furred Amazonian species—is more closely related to the long-furred C. spinosus from the Atlantic forest of southeastern Brazil. Waterhouse (1848: 418–419) suspected that bicolor and nycthemera might be synonyms, but these taxa differ in size (indexed by toothrow measurements in table 7), nasofrontal sinus inflation, and cytochrome-b sequences (ca. 6.6%), so their current status as distinct species seems adequately supported. Although C. spinosus was recovered as the sister taxon of this pair, we suspect (based on morphology) that clade B will eventually prove to consist of ((spinosus + insidiosus) (bicolor + nycthemera)) when sequence data from C. insidiosus are analyzed.
Because Coendou mexicanus, C. quichua, and C. rufescens differ conspicuously in pelage and cranial traits, their close relationship was previously unsuspected. Nevertheless, this cluster makes sense biogeographically (see below), and we are not aware of any strongly contradictory evidence. Indeed, the genetic distance between C. quichua and C. rufescens (table 6) seems remarkably small given the traditional placement of these species in different genera. The sister-group relationship of C. prehensilis to this cluster is another unanticipated result, and one that merits testing with additional sequence data in future studies.
MORPHOLOGICAL EVOLUTION
Visually conspicuous morphological differences among Recent species of Coendou include variation in size, fur length and density, quill coloration, relative tail length, and nasofrontal sinus inflation4 (Voss and Angermann, 1997; Voss and da Silva, 2001; Voss, 2011). As discussed above, Husson (1978) was so impressed by the differences he observed between Coendou prehensilis (a large, short-furred porcupine with inflated sinuses; figs. 5, 8A) and C. melanurus (a small, longfurred porcupine with uninflated sinuses; figs. 6, 8B) that he assigned these species to different genera. However, other porcupines exhibit trait combinations that make generic assignments problematic. Coendou quichua, for example, has short fur (and therefore appears completely spiny like C. prehensilis), but it is a small species with uninflated nasofrontal sinuses (like C. melanurus). Similarly, C. mexicanus has long fur (that conceals the underlying quills like C. melanurus), but it is a large species that usually has inflated nasofrontal sinuses (like C. prehensilis).
To explore historical patterns of phenotypic evolution among Recent erethizontids, we coded size, nasofrontal sinus inflation, and dorsal fur morphology for phylogenetic analysis. Lacking other suitable proxy measures of size (weight data, for example, are missing for several taxa), we scored species of Coendou as “small” if they had maxillary tooth rows < 18 mm or “large” if they had maxillary tooth rows > 19 mm (table 7). Erethizon (with maxillary tooth rows > 22 mm; Stangl et al., 1991) has teeth that are occlusally similar to those of Coendou, so this taxon can be scored for size by the same criterion. However, the laminar cheek teeth of Chaetomys are disproportionately longer and narrower than those of Coendou, so we scored this taxon based on other measurements (e.g., posterior zygomatic breadth; Voss and Angermann, 1997) and published weights (in Chiarello et al., 1997) that place it well within the size range of “small” species of Coendou.
We coded the nasofrontal sinuses of Coendou species as “inflated” or “not inflated” based on obvious external differences in cranial morphology (fig. 8). As noted elsewhere (Voss, 2011), the nasofrontal sinuses of C. mexicanus are geographically variable, but the modal sinus condition in this taxon (illustrated by Hall, 1980: fig. 501)—and the phenotype represented by material sequenced for this study—is inflated. The nasofrontal sinuses of Erethizon and Chaetomys are not inflated (for cranial illustrations, see Ellerman, 1940: figs. 43, 47).
We coded the dorsal fur of Coendou species as “long & dense” if it conceals the quills (fig. 6) or as “short & sparse” if it does not (figs. 5, 7). The length and density of the dorsal fur of Erethizon is seasonally variable (Po-Chedley and Shadle, 1955) and probably also varies with latitude and elevation, but the typical pelage morphology of this taxon more closely resembles the “long & dense” phenotype than it does the “short & sparse” condition. By contrast, the dorsal fur of Chaetomys is unambiguously short and sparse (Oliver and Santos, 1991: pl. 7; Voss and Angermann, 1997: fig. 8).
Maximum-likelihood optimizations (fig. 9) suggest that each of these characters has evolved homoplasiously in the erethizontid crown clade, but ancestral-state assignments are ambiguous at many internal nodes. This ambiguity has the unfortunate effect that different combinations of reversal and convergence are often equiprobable as alternative explanations for observed taxonomic differences. In particular, ambiguous state assignments for the last common ancestor of Recent erethizontids and for the last common ancestor of erethizontines (the two basalmost nodes of each tree in fig. 9) make it difficult to choose among alternative reconstructions of morphological evolution on the long branches that separate Chaetomys, Erethizon, and Coendou. Within Coendou, however, some probabilistic inference is possible.
SIZE AND SINUS INFLATION: Large size and nasofrontal sinus inflation are identically distributed among Recent species of Coendou, so it is not surprising that analyses of these traits yield similar patterns of ancestral-state probabilities. Based on our results, it seems likely that the last common ancestor of Coendou was a small species with uninflated nasofrontal sinuses, and that large size and inflated sinuses evolved convergently in Coendou bicolor and in clade A (prehensilis + mexicanus + rufescens + quichua). In the latter group, ancestral-state probabilities marginally favor the hypothesis that large size and inflated sinuses evolved once (in a common ancestor) with a subsequent reversal to small size and uninflated sinuses on the branch leading to C. rufescens and C. quichua. However, it is almost equally likely that the common ancestor of clade A was small and had uninflated sinuses, in which case large size and sinus inflation must have evolved convergently in C. prehensilis and C. mexicanus.
FIG. 9.
Maximum-likelihood reconstructions of ancestral phenotypes for three morphological characters of Recent erethizontids. Branch tips representing species of Coendou are labeled with corresponding epithets only. See text for character definitions and scoring criteria and table 8 for divergence-date estimates. Pie diagrams at internal nodes represent estimated probabilities of alternative states.
Although these observations suggest that large size and nasofrontal sinus inflation might be developmentally or functionally correlated, it is noteworthy that Erethizon (the largest living erethizontid) does not have inflated sinuses, nor did some large fossil erethizontids (e.g., Hypsosteiromys from the early Miocene of Argentina; Dozo et al., 2004). Convincing explanations for the evolution of mammalian cranial sinuses are elusive (Moore, 1981: 277–279), perhaps because sinuses have different functions in different clades. To the best of our knowledge, no hypothesis about the functional significance of nasofrontal sinus inflation among New World porcupines has been suggested, but we hypothesize that this trait might help protect the soft tissues of the head from attacking predators.
Functional analyses of porcupine quill erection (Chapman and Roze, 1997) suggest that these defensive spines can only be erected to an angle of about 90° from the underlying dermal surface, so fully erected quills on a flat skull can only stick straight up. Nasofrontal sinus inflation increases the cranial surface area for quill deployment and allows erected quills to point anteriorly and laterally (fig. 10), effectively protecting the eyes and nose. High in the canopy, Neotropical porcupines are perhaps safe from most nonavian predators, but they are probably more vulnerable on the ground. Swollen nasofrontal sinuses may confer an adaptive advantage for Coendou species that often descend to the ground for geophagy (Montenegro, 2004; Blake et al., 2011) or to cross canopy gaps (Montgomery and Lubin, 1978), where they might be exposed to terrestrial predators such as pumas (Chinchilla, 1997; Novack et al., 2005; Foster et al., 2010) and large boas (Cherubini et al., 2003; Duarte, 2003).
FIG. 10.
Coendou prehensilis with erected cranial quills. The inflated nasofrontal sinuses of this species provide increased surface area for quill deployment, and their convex margins allow erected quills to point anteriorly and laterally to protect adjacent soft tissues. Photographed at the Frankfurt Zoo in 2008 (courtesy of Marek Polster).
PELAGE: Reconstructions of dorsal fur evolution in Coendou are minimally constrained by inferred ancestral-state probabilities. Parsimony analyses (not shown) indicate that at least four state transformations are necessary to explain observed taxonomic variation in dorsal-fur morphology among Coendou species, but many alternative scenarios of convergence and reversal are almost equally likely and it would be pointless to enumerate all of them here. Clearly, however, dorsal-fur character-state transformations have occurred rather frequently in the genus, and it would be interesting to know why.
The reason is not immediately obvious. Although fur length and/or density seem to vary with elevation within some widespread species (e.g., Coendou mexicanus; Voss, 2011), pelage traits are not correlated with elevation across species, as one might expect if the primary function of fur were thermoregulatory. Thus, porcupines with long/dense fur include both lowland species (e.g., C. melanurus) and highland species (C. pruinosus); likewise, porcupines with short/sparse fur include lowland species (C. prehensilis) and highland species (C. rufescens). Instead, taxonomic variation in fur length and density might result from divergent selection for aposematism versus crypsis, defensive adaptations that often involve evolutionary tradeoffs (Ruxton et al., 2004).
The black-and-white banding pattern of porcupine quills is a classic example of warning coloration (dangerous spines in many other animals are similarly colored; Inbar and LevYadun, 2005; Speed and Ruxton, 2005; Caro, 2009). Openly displaying such weapons might be advantageous for deterring attacks by predators that kill by biting (e.g., felids; Emmons, 1987; Moreno et al., 2006), which must often result in a painful (or even lethal) mouthful of spines. On the other hand, hiding quills under a long, thick coat of background-matching fur might be advantageous for avoiding detection by predators that kill with armored talons (e.g., harpy eagles; Izor, 1985; Piana, 2007), against which spines might be ineffective. Unfortunately, evidence of taxonomic variation in habits that expose porcupines to different sets of predators is almost entirely anecdotal, consisting of chance observations (e.g., fig. 11) that might not represent typical behaviors.
TABLE 8.
Estimated Divergence Dates (millions of years) for Selected Erethizontid Clades
BIOGEOGRAPHY
Recent erethizontids occur in lowland habitats either on the east or the west side of the Andes, but some are highland species (Voss, 2011). To explore erethizontid biogeographic history, we scored species as “cis-Andean” if they occur in lowland habitats east of the Andes, as “trans-Andean” if they occur in lowland habitats west of the Andes, or as “Andean” if they occur primarily in Andean landscapes (including adjacent foothills and piedmonts; table 1). For the purpose of this analysis we scored Coendou bicolor and C. prehensilis as cis-Andean species, despite the fact that an isolated montane population of C. bicolor occurs on the west side of the Andes in northern Peru (fig. 1: locality 28), and that an isolated lowland population of C. prehensilis occurs in northern Colombia west of the Serranía de Perijá (fig. 1: locality 18).
Maximum-likelihood modeling of geographic range evolution based on these data provides the first phylogeny-based reconstruction of erethizontid crown-clade biogeography (fig. 12). In this scenario, the last common ancestor of living erethizontids was an Early Miocene (ca. 19.1 Mya; table 8) cis-Andean species, but the erethizontine lineage (ancestral to Erethizon and Coendou) dispersed into the transAndean lowlands and subsequently split at about 9.4 Mya, giving rise to one trans-Andean daughter species ancestral to Erethizon and a second species, ancestral to Coendou, that was widespread in both cis- and trans-Andean lowlands. The latter split at about 6.4 Mya, producing a cis-Andean daughter species ancestral to clades B (bicolor + nycthemera + spinosus) and C (vestitus + pruinosus + ichillus + melanurus), and a widespread species ancestral to clade A (prehensilis + mexicana + rufescens + quichua). Whereas clade B remained cis-Andean, two dispersal events occurred in clade C: once into the trans-Andean lowlands and another, subsequently, into Andean landscapes. A basal split in clade A at about 4.4 Mya resulted in a cis-Andean species (prehensilis) and a trans-Andean lineage (mexicana + rufescens + quichua), one branch of which (rufescens + quichua) subsequently dispersed into the Andes.
FIG. 11.
Camera-trap photograph of two Coendou bicolor visiting a mineral lick near the Centro de Investigación Río Los Amigos, Madre de Dios, Peru (courtesy of Dyana LaRosa). In situations like this, porcupines are probably exposed to a greater risk of predation than they would be in the canopy. Coendou prehensilis is the only other Neotropical species known to frequent mineral licks (Montenegro, 2004; Blake et al., 2011).
This reconstruction of erethizontid crown-clade historical biogeography suggests that the cis- and trans-Andean lowlands have hosted more-or-less separate porcupine faunas since the Pliocene. From a Recent perspective, the persistence of a widespread (cis- and transAndean) lineage from about 10 to 5 Mya seems implausible, but the northern Andes did not reach their currently formidable elevations until about 3 Mya (Gregory-Wodzicki, 2000), so it is not impossible that one or more paleospecies maintained reproductive continuity across these mountains when they were substantially lower. Indeed, at least three Recent species (Coendou mexicanus, C. pruinosus, and C. quichua; Voss, 2011) have elevational ranges that are known to extend from near sea level to at least 2000 m, suggesting that the Andes were not insurmountable barriers to Miocene gene flow.
GENETIC UNIFORMITY IN A WIDESPREAD SPECIES
The striking absence of phylogeographic structure and sequence divergence among samples of Coendou prehensilis representing populations that span 27° of latitude (northern Colombia to eastern Bolivia) and 25° of longitude (French Guiana to northern Peru) is among the most remarkable results of this study. Such genetic uniformity would seem to suggest either very recent range expansion or extensive contemporaneous gene flow. Both explanations imply that geographic or biotic barriers that inhibit dispersal in other porcupines are (or were) easily surmounted by C. prehensilis.
Most Neotropical porcupines are restricted to humid forests, but Coendou prehensilis also occurs in dry forests and savannas (Montgomery and Lubin, 1978; Parker et al., 1993; Emmons et al., 2006; Bruna et al., 2010) and seems to be more eurytopic than other congeneric species. Additionally, C. prehensilis may occupy a different trophic niche than other porcupines. Whereas field studies of Chaetomys subspinosus and Coendou spinosus suggest that these species are almost exclusively folivorous (Giné et al., 2010; Lima et al., 2010; Passamani, 2010), direct observations of feeding behavior by C. prehensilis suggests that this species feeds primarily on seeds from immature fruits (Charles-Dominique et al., 1981). Either or both of these ecological traits (eurytopy and granivory) might account, in part, for the broad geographic distribution of this species and its ability to coexist with other porcupines. Indeed, it is noteworthy that most species of Coendou are allopatric (possibly because they occupy the same trophic niche), whereas C. prehensilis accounts for almost all known examples of geographic range overlap and sympatry in the genus (table 9). An ecological niche shift followed by rapid range expansion is one scenario that might account for these observations.
DIRECTIONS FOR FUTURE RESEARCH
The phylogenetic analyses of this study are based on sequence data from a single mitochondrial gene, so obtaining additional data from other (preferably nuclear) loci is an obvious future priority. Also of interest would be the discovery of craniodental characters that could be used to place fossil and Recent erethizontids in a common phylogenetic framework and thereby provide internal constraints for future attempts to date key events in erethizontid phylogeny. Unfortunately, most craniodental characters known to vary among Recent erethizontids are either autapomorphies (e.g., the exceptionally deep jugal, well-developed postorbital processes, and laminar molars of Chaetomys), or they are so variable within species as to be effectively useless for phylogenetic analysis.
Another productive avenue for future research concerns erethizontid natural history, about which we remain profoundly ignorant. Although much fascinating detail is now available concerning key aspects of the behavior, physiology, and functional morphology of the North American porcupine (reviewed by Roze, 2009), available information about most Neotropical species is pitifully limited to chance observations and results from short-term field studies. Without reliable data about diet, activity patterns, dispersal ability, predation, and other relevant topics, the adaptive hypotheses suggested in this report will remain untested.
Lastly, we encourage field biologists to make greater efforts to collect porcupines. New species almost certainly remain to be discovered, and available sample sizes for some described species (e.g., Coendou ichillus and C. roosmalenorum; Voss and da Silva, 2001) are too small to support confident inference about size and other metrical traits. Because most porcupines are hard to observe even where they are locally common, advantage should be taken of rare opportunities (e.g., flooding by hydroelectric dams) to obtain large series.
TABLE 9.
Geographic Relationships among Species of Coendou a
ACKNOWLEDGMENTS
Because porcupines are infrequently collected, tissue resources for a project like this are correspondingly scarce. We are grateful to John E. Cadle, François Catzeflis, Guillermo D'Elía, Alfred Gardner, Kristofer M. Helgen, Jeremy Jacobs, James L. Patton, Robert M. Timm, and Paúl M. Velazco, among others, for permission to use the tissues they collected. Permission to take skin samples from important museum specimens for DNA extraction was generously granted by Jim Dines (LACM) and Kris Helgen (USNM). At the University of Minnesota, Lorissa Fujishin provided expert advice about amplifying and sequencing problematic DNA. Ana Carmignotto, François Catzeflis, Yuri Leite, and Roberto Vilela kindly provided information about voucher specimens. Patricia J. Wynne drew figure 8 with her customary skill and attention to anatomical detail, Craig Chesek (of the AMNH Photography Studio) photographed the skins in figures 5–7, and Marek Polster kindly allowed us to reproduce his image of Coendou prehensilis at the Frankfurt Zoo. Louise Emmons sent us the camera-trap photo of C. bicolor from the Rio Los Amigos and Dyana LaRosa generously allowed us to reproduce it here. Lastly, we thank Marcelo Weksler and an anonymous reviewer for comments that improved the final version of our manuscript.
REFERENCES
Appendices
APPENDIX 1
GAZETTEER OF COLLECTING LOCALITIES
Below we list all of the localities where sequenced erethizontid specimens were collected, including those newly sequenced for this report and others sequenced for previous molecular studies (Bonvicino et al., 2002; Vilela et al., 2009; Leite et al., 2011). Italics identify the largest administrative unit (department, province, state) within each country, and geographic coordinates are provided parenthetically with a cited source for these data. Neotropical collection localities are mapped in figure 1.
BOLIVIA
1. Beni, Río Mamoré (13°42′S, 65°19′W; Anderson, 1997).
2. Santa Cruz, 10 km N San Ramón (16°36′S, 62°42′W; Anderson, 1997). BRAZIL
3. Acre, right bank Rio Juruá, Ocidente (8°34′S, 72°48′W; Patton et al., 2000).
4. Acre, left bank Rio Juruá, Fazenda Santa Fé (8°36′S, 72°51′W; Patton et al., 2000).
5. Amazonas, left bank Rio Juruá, Eirunepé (6°38′S, 69°52′W; Patton et al., 2000). Note that Lara et al. (1996), who also sequenced MVZ 195088 (= MNFS 439), gave incorrect coordinates for this locality, placing it almost 1000 km to the east; Vilela et al. (2009: table 4) repeated their mistake.
6. Bahia, Salvador (13°00′S, 38°30′W, near sea level; Vilela et al., 2009).
7. Espírito Santo, Usina Hidrelétrica de Rosal (20°54′S, 41°42′W; Vilela et al., 2009).
8. Mato Grosso, Usina Hidrelétrica de Manso (15°36′S, 56°06′W; Bonvicino et al., 2002).
9. Pará, Ilha de Marajó (ca. 1°00′S, 49°30′W; Paynter and Traylor, 1991).
10. Pernambuco, Mata Xanguá (8°39′S, 35°10′W; Leite et al., 2011).
11. Rio de Janeiro, Rio das Ostras (22°32′S, 41°57′W; Bonvicino et al., 2002).
12. Rio de Janeiro, Sumidouro (22°03′S, 42°40′ W; Bonvicino et al., 2002).
13. Roraima, São João da Baliza (00°57′S, 59°54′W; Bonvicino et al., 2002).
14. São Paulo, Biritiba Mirim (23°36′S, 46°00′W; Vilela et al., 2009).
15. São Paulo, 20 km NW Sorocaba (23°26′S, 47°38′W; L.P. Costa, personal commun.).
COLOMBIA
16. Cauca, Quintana (in western Andes near Cerro Munchique [ca. 2°32′N, 76°57′W; DMA, 1988], not “La Quintana,” a locality said to be on the western slopes of the Central Andes; Paynter, 1997: 236)
17. Cesar, San Alberto (ca. 7°45′N, 73°23′W; Paynter, 1997).
18. Cesar, Valledupar (10°29′N, 73°15′W; Paynter, 1997).
19. Cundinamarca, San Juan de Río Seco (4°51′N, 74°38′W; Paynter, 1997).
ECUADOR
20. Cotopaxi, Otonga (0°25′S, 79°00′W; Jarrín, 2001).
21. Sucumbíos, Río Napo, Limoncocha (ca. 0°25′S, 76°38′W; Paynter, 1993).
FRENCH GUIANA
MEXICO
PANAMA
24. “Canal Zone,” Fort Kobbe (8°54′N, 79°36′W; Fairchild and Handley, 1966).
PARAGUAY
PERU
27. Amazonas, Río Cenepa, Huampami (ca. 4°27′S, 78°10′W; Patton et al., 1982).
28. Cajamarca, 2.5 km N Monte Seco (ca. 6°50′S, 79°06′W; Cadle, 1991).
29. Loreto, Iquitos (3°46′S, 73°15′W; Stephens and Traylor, 1983).
30. Loreto, Río Gálvez, Nuevo San Juan (5°15′S, 73°10′W; Voss and Fleck, 2011).
31. Madre de Dios, 15 km NE Puerto Maldonado (12°33′S, 69°03′W; Woodman et al., 1991).
32. San Martín, Área de Conservación Municipal Mishquiyacu-Rumiyacu y Almendra (6°05′S, 76°59′W; Nava et al., 2010).
VENEZUELA
33. Amazonas, Cerro Neblina Base Camp (0°50′N, 66°10′W; Gardner, 2008).
34. Bolívar, Río Caroní (7°55′N, 63°01′W; F. Catzeflis, personal commun.).
35. Táchira, Las Mesas (8°09′N, 72°10′W; Handley, 1976).
36. Zulia, Misión Tukuko (9°50′N, 72°52′W; Voss, 1991).
UNITED STATES
Notes
[1] 3 In the context of this test, Coendou (sensu lato) is equivalent to the group bicolor + nycthemera + prehensilis + quichua, Sphiggurus is equivalent to the group ichillus + melanurus + mexicanus + pruinosus + spinosus + vestitus, and Echinoprocta is equivalent to rufescens (table 1).
[2] 4 The cranial structure responsible for the prominent convexity variously described as a “high dorsal hump” (Husson, 1978: 486) or “bulbous forehead” (Handley and Pine, 1972: 242) in some Neotropical porcupines is a pneumatic sinus that invades the substance of both the nasal and frontal bones. Sectioned skulls (e.g., USNM 36927) reveal that this sinus is divided by an unossified midsagittal septum and communicates with the nasal cavity by paired ventrolateral orifices.
