The phylogenetic position of the enigmatic Atlantic forest-endemic spiny mouse Abrawayaomys (Rodentia: Sigmodontinae)

Abstract

Background

The phylogenetic position of the sigmodontine genus Abrawayaomys, historically assigned to the tribe Thomasomyini or considered a sigmodontine incertae sedis, was assessed on the basis of nuclear and mitochondrial DNA sequences obtained from four individuals from different localities in the Atlantic forest of Brazil. Sequences of Abrawayaomys were analyzed in the context of broad taxonomic matrices by means of maximum-likelihood (ML) and Bayesian analyses (BA).

Results

The phylogenetic position of Abrawayaomys differed depending on the gene analyzed and the analysis performed (interphotoreceptor retinoid-binding protein (IRBP) ML: sister to Thomasomyini; IRBP BA: sister to Akodontini; cytochrome (Cyt) b ML: sister to Neotomys; and Cyt b BA: sister to Reithrodontini). With the sole exception of the BA based on Cyt b sequences, where the Abrawayaomys-Reithrodon clade had strong support, all sister-group relationships involving Abrawayaomys lacked any significant support.

Conclusions

As such, Abrawayaomys constitutes the only representative so far known of one of the main lineages of the sigmodontine radiation, differing from all other Atlantic forest sigmodontine rodents by having a unique combination of morphological character states. Therefore, in formal classifications, it should be regarded as a Sigmodontinae incertae sedis.

Background

With about 86 living genera, cricetids of the subfamily Sigmodontinae are one of the most diversified and taxonomically complex groups of mammals. Predominantly distributed in South America, sigmodontines also reach Central and North America, and one extant genus is endemic to the Galapagos Islands (D'Elía 2003a). Remarkably, new sigmodontine genera are still being erected on the basis of both revisionary museum work and from newly discovered species (e.g., Pardiñas et al. 2009a; Percequillo et al. 2011; Pine et al. 2012; Alvarado-Serrano and D'Elía 2013; see comments in D'Elía and Pardiñas 2007). Similarly, statuses of several sigmodontine taxonomic forms at the species level are unclear (e.g., Alarcón et al. 2011; Bonvicino et al. 2012).

Sigmodontine genera have been arranged into different groups, most of which have been given the formal rank of tribes (e.g., Reig 1980). In the last two decades, phylogenetic analyses using either morphological or molecular data or both were used to set the limits and contents of these groups and determine the timing of their diversification (e.g., Braun 1993; Smith and Patton 1999; Steppan 1995;D'Elía 2003b; Pacheco 2003; Weksler 2003; D'Elía et al. 2003, D'Elia et al. 2006a, b; Martínez et al. 2012; Parada et al. 2013; Salazar-Bravo et al. 2013). Those studies caused a number of major reconsiderations on the limits and contents of these groups. Currently, 12 extant genera are considered as Sigmodontinae incertae sedis (see the most recently published classification in D'Elía et al. (2007) and the modification prompted by the description of a new genus by Alvarado-Serrano and D'Elía (2013)); one of these is Abrawayaomys Cunha and Cruz 1979.

Abrawayaomys is a poorly known sylvan sigmodontine genus that stands out within the sigmodontine radiation due to its spiny pelage and unusual craniodental morphology. It is found in the Atlantic forest of Argentina and Brazil and is known from a handful of trapped specimens and a few osteological remains gathered from owl pellets (Pardiñas et al. 2009b). Two species are recognized: the type species Abrawayaomys ruschii Cunha and Cruz 1979 and the recently described Abrawayaomys chebezi (Pardiñas et al. 2009a,2009b).

Abrawayaomys displays a striking combination of morphological features that was referred by Musser and Carleton (2005, p. 1,088) in the following terms ‘Diagnostic traits seem to combine aspects of Neacomys, Oryzomys, and Akodon, and both Reig (1987) and Smith and Patton (1999) acknowledged the enigmatic affinities of Abrawayaomys as uncertain.’ Recently, Pardiñas et al. (2009b) evaluated the morphology of Abrawayaomys in detail and noted a certain resemblance to the bauplan of the tribe Akodontini but also to the thomasomyine genera Chilomys and Rhagomys. Similarities with Akodontini were regarded as convergences since Abrawayaomys was assigned to Thomasomyini given that independent sets of data placed Chilomys and Rhagomys in Thomasomyini (D'Elía et al. 2006a; Salazar-Bravo and Yates 2007) and because morphology-based phylogenetic analyses placed Abrawayaomys in this tribe (Pacheco 2003; Salazar-Bravo and Yates 2007). The lack of molecular data for Abrawayaomys has prevented an assessment of its position within the available comprehensive sigmodontine trees generated on the basis of mitochondrial and nuclear DNA sequences (e.g., Engel et al. 1998; Smith and Patton 1999; D'Elía 2003b; Weksler 2003; D'Elia et al. 2006a, b).

In recent fieldwork in the states of Rio de Janeiro (Pereira et al. 2008), Minas Gerais (Passamani et al. 2011), and São Paulo (see below) in southeastern Brazil, four specimens referred to as A. ruschii were collected. Based on mitochondrial and nuclear DNA sequences gathered from these specimens, we present the first phylogenetic analyses to test previous hypotheses concerning the placement of Abrawayaomys within the sigmodontine radiation. In addition, we provide some taxonomic judgments based on the resulting phylogeny and comments on the evolution of some morphological traits.

Methods

DNA sequences corresponding to the cytochrome (Cyt) b gene and the first exon of the nuclear interphotoreceptor retinoid-binding protein (IRBP) gene were used as evidence. We sequenced four specimens of A. ruschii deposited in the following Brazilian collections: Museu Nacional, Univ. Federal do Rio de Janeiro (MN 67557; Brazil, Rio de Janeiro, Aldeia Sapucai) (Pereira et al. 2008); Museu de Zoologia, Univ. de São Paulo (MZUSP 32319; Brazil, São Paulo, Biritiba-Mirim; BO 27; Brazil, São Paulo, Estação Ecológica de Boracéia); and Coleção de Mamíferos da Univ. Federal de Lavras (CMUFLA 906; Brazil, Minas Gerais, Caeté) (Passamani et al. 2011). Sequences were gathered following the protocol of Pardiñas et al. (2003) and D'Elía et al. (2006b). We found minor differences among Cyt b sequences (see ‘Results’ below), but in the phylogenetic analyses, we used that of specimen MN 67557 because it was the only complete one (i.e., 1,140 bp; the other being 801 bp long). We found no variation in IRBP sequences of the four specimens analyzed, and therefore, Abrawayaomys was represented by a single terminal (MN 67557) in the phylogenetic analyses. New sequences were submitted to GenBank [GenBank: JX949182 to JX949189].

To appraise the phylogenetic position of Abrawayaomys within the radiation of the Sigmodontinae, we sought to ensure that sigmodontine diversity was represented as comprehensively as possible. According to the current classification, our sampling only lacked for both matrices the incertae sedis genera Phaenomys and Wilfredomys; the akodontines Gyldenstolpia and Podoxymys; the ichthyomyines Anotomys, Chibchanomys, Neusticomys, and Ichthyomys; and the oryzomyine Mindomys. In addition, the IRBP matrix lacked the thomasomyine Chilomys. Meanwhile, the ichthyomyine Rheomys, the oryzomyine Microakodontomys, and the thomasomyine Aepeomys were also missing from the Cyt b matrix. Therefore, both the IRBP (1,181 characters) and Cyt b (1,134 characters) matrices respectively included representatives of 76 and 74 sigmodontine genera (including sequences of the recently described genus Neomicroxus). The IRBP matrix lacked a sequence for Neusticomys available in GenBank [GenBank: EU649036] because a recent inspection of it indicated that it may be in fact a composite of an ichthyomyine and an oryzomyine sequence. Although the monophyly of the Sigmodontinae is well corroborated (e.g., Engel et al. 1998; Steppan et al. 2004; Parada et al. 2013), the identity of its sister group is unclear. Therefore, we integrated the outgroup with two representatives of each of the other four main lineages that, together with the Sigmodontinae, compose the family Cricetidae: arvicolines (Arvicola and Microtus), cricetines (Cricetulus and Phodopus), neotomines (Neotoma and Scotinomys), and tylomyines (Nyctomys and Tylomys). All taxa represented in the analyses, along with the GenBank accession numbers of their DNA sequences, are listed in Table 1.

Table 1 List of taxa and the DNA sequences of which were included in the phylogenetic analyses

Alignment was done with Clustal X (Thompson et al. 1997) using default parameters for all alignment parameters. Uncorrected genetic distances (p distances) with pairwise deletions were computed using MEGA 5 (Tamura et al. 2011). Each matrix was subjected to maximum-likelihood (ML) (Felsenstein 1981) and Bayesian analyses (BA) (Rannala and Yang 1996). The ML analysis was conducted in Treefinder (Jobb et al. 2004; Jobb 2008). The best fitting models of nucleotide substitution (IRBP: TVM[Optimum, Empirical]:G[Optimum]:5; Cyt b: GTR[Optimum, Empirical]:G[Optimum]:5) (see Jobb 2008) were selected with the Akaike information criterion in Treefinder using the ‘propose model’ routine. The best tree was searched under the model of nucleotide substitutions previously selected using search algorithm 2 implemented in Treefinder version March 2011; nodal support was estimated with 1,000 bootstrap pseudoreplicates (BS). The BA was conducted using MrBayes 3.1.2 (Ronquist and Huelsenbeck 2003) by means of two independent runs with three (IRBP) or seven (Cyt b) heated and one cold Markov chains each. Considering the model selected by Treefinder and models specified in MrBayes, a model with six categories of a base substitution, a gamma-distributed rate parameter, and a proportion of invariant sites was specified for both matrices; all model parameters were estimated using MrBayes. Runs were allowed to proceed for 20 million (IRBP) and 28 million (Cyt b) generations, and trees were sampled every 1,000 (IRBP) and 2,000 (Cyt b) generations. Log-likelihood values were plotted against the generation time to check that runs converged on a stable log-likelihood value. The first 25% of sampled trees were discarded as burn-in; the remaining trees were used to compute a 50% majority rule consensus tree and obtain posterior probability (PP) estimates for each clade.

Results

Cyt b gene sequences of Abrawayaomys gathered from the three specimens collected in the states of São Paulo and Rio de Janeiro varied by 0.5% ~ 0.7%; while comparisons involving the Cyt b sequence of the specimen from Minas Gerais ranged 2.6% ~ 3.0%. As mentioned above, IRBP sequences of the four specimens analyzed were identical.

Abrawayaomys, as represented by sequences of specimen MN 67557, was highly divergent from all compared sigmodontines for both analyzed genes. For the Cyt b gene, observed pairwise values involving Abrawayaomys ranged 19.0% ~ 25.9% for comparisons with Neotomys and Lundomys. For the IRBP gene, observed pairwise values involving Abrawayaomys ranged 3.42% ~ 6.65% for comparisons with Brucepattersonius and Rheomys.

Phylogenetic analyses recovered general results congruent with those of previous IRBP- and/or Cyt b-based studies (e.g., D'Elia et al. 2006a, b; Percequillo et al. 2011; Martínez et al. 2012; Parada et al. 2013; Salazar-Bravo et al. 2013). For descriptions and discussions of these findings (e.g., a strongly supported Sigmodontinae, a strongly supported Oryzomyalia, the polyphyly of the Reithrodon group, and differences between gene trees), we refer the reader to those studies because herein we focus on the phylogenetic position of Abrawayaomys.

Depending on the gene analyzed, the phylogenetic position of Abrawayaomys varied from being sister to Akodontini (IRBP BA; PP = 0.62; Figure 1), sister to Thomasomyini (IRBP ML; BS < 50), sister to Reithrodontini (Cyt b BA; PP = 0.98; Figure 2), or sister to Neotomys (Cyt b ML; BS < 50). As noted, with the sole exception of the BA based on Cyt b sequences, where the Abrawayaomys-Reithrodon clade had strong support, all sister-group relationships involving Abrawayaomys lacked any significant support. The most inclusive and well-supported clade containing Abrawayaomys was that corresponding to Oryzomyalia (sensu Steppan et al. 2004).

Figure 1

Results of the Bayesian analysis of IRBP gene sequences. Numbers indicate posterior probability (left of the diagonal) and ML bootstrap values (right of the diagonal) of adjacent nodes. Only bootstrap values of >50% are shown. Dashes signal nodes that were not recovered in the ML topology (ln = -9,764.8433). Dots at terminals indicate genera with species possessing dorsal spines.

Figure 2

Results of the Bayesian analysis of Cyt b gene sequences. Numbers indicate posterior probability (left of the diagonal) and ML bootstrap values (right of the diagonal) of adjacent nodes. Only bootstrap values of >50% are shown. Dashes signal nodes that were not recovered in the ML topology (ln = -33,444.453).

Discussion

The diversity of sigmodontine forms has long captivated students of New World mammals, but at the same time, it has seriously defied those attempting to classify them according to their evolutionary history. Problems range from species boundaries to relationships among species and genera, and limits and contents of higher taxa (e.g., tribes). These issues have direct implications for the study of the history of the diversification of the group, which in turn arguably constitutes one of the most controversial debates in muroid systematics (Voss 1993; D'Elía 2003b).

Herein, we showed that the phylogenetic position of Abrawayaomys varies from being sister to Akodontini, Thomasomyini, Reithrodontini, or Neotomys, depending on the gene analyzed (IRBP or Cyt b) and the analysis performed (ML or BA). Given that previous studies showed discrepancies between the topologies of different sigmodontine gene trees (e.g., Feijoo et al. 2010; Teta et al. 2011), the fact that different datasets (i.e., IRBP or Cyt b matrices with slightly different taxonomic sampling) provide different relationships for Abrawayaomys is not unexpected. Importantly, with the exception of the BA based on Cyt b sequences (sister to Reithrodontini), no sister relationship involving Abrawayaomys was recovered that had good support. In light of the analytical results, Abrawayaomys cannot be placed with certainty in any more inclusive clade than that of Oryzomyalia (sensu Steppan et al. 2004).

Pacheco (2003, p. 130), in a phylogenetic analysis based on morphological characters, found the genus Abrawayaomys to be sister to Rhagomys within the Thomasomyini clade. This relationship was supported by the following character states: a broad zygomatic plate, an interorbital region convergent with the supraorbital margins squared or weakly beaded, a long jugal, the absence of mesolophids (but see below), a masseteric crest anterior to the procingulum of the first lower molar, and a deeply excavated retromolar region of the mandible. Pacheco (2003) highlighted the retromolar region condition, i.e., broad and fenestrated, as a synapomorphy of Abrawayaomys + Rhagomys. We agree with Pacheco (2003) (see Musser and Carleton 2005) in the general resemblance between Abrawayaomys and Rhagomys and to a lesser extent to the remainder of the Thomasomyini, but we suggest that this similarity is not remarkable and more important and that these shared character states are not synapomorphies of a putative Abrawayaomys and Rhagomys clade. Almost all of those character states listed by Pacheco (2003) are also present in many sigmodontines, indicating the large amount of homoplasy existing within this group. In addition, we assert that the molar morphologies of both genera are quite distinct, having only the widespread brachyodont condition in common. Rhagomys has very well-developed mesolophs/phids (cf. Luna and Patterson (2003) vs. Pacheco (2003)), procingula of the first upper molars clearly crossed by a deep anteromedian flexus, well-developed posterolophs, a slightly reduced third lower molar with respect to the second lower molar, and several other traits found among taxa displaying the dental bauplan of the pentalophodont type (sensu Hershkovitz 1962), which is clearly distinguishable from the unequivocally tetralophodont molar of Abrawayaomys (cf.; Pardiñas et al. 2009b). Additional differences between these two genera are more than trenchant, including incisive foramina and palate extensions, parapterygoid plate morphology, and carotid circulatory pattern (Table 2) (see also Pardiñas et al. 2009b, Table three). Similarly, several trenchant character states are present in Abrawayaomys and representatives of other sigmodontine tribes. The external morphology of Abrawayaomys resembles that of many akodonts (cf. Pereira et al. 2008, Figure one), although it has a moderately longer tail, at least in some individuals. A morphological description of the stomach (Finotti et al. 2003) suggests a hemiglandular-unilocular type, a widespread condition among sigmodontines (Carleton 1973). Finally, Pacheco (2003) also indicated that Abrawayaomys has a peculiar genal vibrissa (called genal vibrissa 2), which is also present in the oryzomyine Oecomys and the akodontine Kunsia, and a reduced fifth pedal digit. Similarly, karyotypic evidence sheds no conclusive light on the phylogenetic position of Abrawayaomys, as the diploid number of 2n = 58 (Pereira et al. 2008) found in A. ruschii is also present in several distantly related oryzomyine species, such as Euryoryzomys lamia (Andrades-Miranda et al. 2000), Holochilus brasiliensis (Yonenaga-Yassuda et al. 1987), Nectomys squamipes (Yonenaga-Yassuda et al. 1987), Oecomys trinitatis (Patton et al. 2000), Oligoryzomys chacoensis (Myers and Carleton 1981), and Sooretamys angouya (Andrades-Miranda et al. 2000). Taking all this evidence as a whole, Abrawayaomys cannot be placed with certainty in any clade less inclusive than the large clade Oryzomyalia (see also Voss 1993). Therefore, in light of all of the evidence at hand, in formal classifications, Abrawayaomys should be kept as an incertae sedis sigmodontine (D'Elía et al. 2007). Results of future phylogenetic analyses, especially those including sequences of two other Atlantic forest inhabitants, Phaenomys and Wilfredomys, which were also considered part of the tribe Thomasomyini (e.g., Pacheco 2003) and are now regarded as Sigmodontinae incertae sedis (e.g., D'Elía et al. 2007), may prompt changes in this classification.

Table 2 Morphological comparisons among Abrawayaomys and other members of the Sigmodontinae

The prevailing biogeographic view is that the Andes played a major role in sigmodontine diversification, in which the main sigmodontine lineages originated there and later colonized lowlands of South America (Reig 1984,1986; see also Salazar-Bravo et al. 2013). However, Abrawayaomys, an Atlantic forest endemic, is the sole living representative of one of the main sigmodontine lineages (i.e., those classified at the tribal rank in formal classifications or those genera that do not belong to any recognized tribe). A similar scenario was found for two other Atlantic forest endemics, the genera Delomys and Juliomys (Figures 1 and 2) (Voss 1993; D'Elía et al. 2006a; see also the classification in D'Elía et al. 2007). The other main sigmodontine lineages are distributed outside the Atlantic forest (e.g., Abrotrichini and Sigmodontini) or, even when present in this biome, are not endemic to it (e.g., Akodontini and Oryzomyini). As mentioned above, the Atlantic forest-endemic Phaenomys and mostly Atlantic forest-resident Wilfredomys have not been included in any molecular-based phylogenetic analysis. Until their phylogenetic position is assessed, it is unclear whether they in fact represent additional main sigmodontine lineages almost endemic to the Atlantic forest or simply constitute additional genera belonging to other already identified main lineages of the Sigmodontinae either already known from the Atlantic forest or not. Whatever this result is, the finding that at least three unrelated main sigmodontine lineages, those currently represented by Abrawayaomys, Delomys, and Juliomys, are endemic to the southern Atlantic forest supports early claims (Smith and Patton 1999; D'Elía 2003b; see also Salazar-Bravo et al. 2013) highlighting the role of the Atlantic forest in harboring sigmodontine phylogenetic diversity. Future studies should be designed to test if these lineages originated in the Atlantic forest or simply invaded it after originating elsewhere.

Spines of varying hardness and architectures are present in several rodents (Chernova and Kuznetsov 2001) and are conspicuous in some Neotropical groups, such as porcupines (Erethizontidae) and spiny rats (Echimyidae). The vast majority of sigmodontine rodents have soft fur, but a few genera and species have dorsal spines. These spines are present in both species of Abrawayaomys, both species of Scolomys, all eight species of Neacomys, but only in one of the two species of Rhagomys. Rhagomys longilingua from the Andes has spiny fur, but Rhagomys rufescens from the Atlantic forest has soft fur (Luna and Patterson 2003). Considering the phylogeny portrayed here, we concluded that these spines are the result of evolutionary convergence, and this trait evolved at least four times in sigmodontines (Figure 1). None of these four genera are sister groups, and each one of them shares a more recent common ancestor with soft-furred genera or species (in the case of Rhagomys). The functional significance of spines remains unknown. Unlike porcupine quills, spines of muroid or echimyid rodents are insufficiently rigid to provide much protection against predators, including snakes, birds, and mammals, which are known to prey heavily on many spiny species (Hoey et al. 2004). Patterson and Velazco (2008) suggested a thermoregulatory interpretation based on the geographic distribution of echimyid rodents: the spiniest members occur in tropical lowland forests, while many of the softest-haired members of the family range into high elevations or latitudes, but they also noted several exceptions. The same pattern does not occur in sigmodontine rodents, since soft-furred taxa occur at all latitudes and elevations throughout the Neotropics. Regarding spiny taxa, both Neacomys and Scolomys are lowland forest genera found at lower latitudes (Patton et al. 2000); Abrawayaomys also occurs in lowland forests, but at higher latitudes in the Atlantic forest (Pardiñas et al. 2009b); and the spiny R. longilingua is found in Andean cloud forests at 1,900 ~ 2,100 m in elevation (Luna and Patterson 2003), while the soft-furred R. rufescens occurs mostly in montane Atlantic forest at 500 ~ 1,000 m in elevation (Steiner-Souza et al. 2008). Therefore, we still lack robust hypotheses for the ecological role, if any, of spiny fur in sigmodontine rodents, but the character distribution mapped on a phylogenetic tree presented here is the first step toward understanding its evolutionary importance.

Conclusions

Phylogenetic analyses show that Abrawayaomys constitutes the single representative so far known of one of the main lineages of the radiation of Sigmodontinae. In addition, it differs from all other Atlantic forest sigmodontine rodents by having a unique combination of morphological character states. Therefore, in formal classifications, it should be regarded as a Sigmodontinae incertae sedis.

Finally, the observed variation of the four Cyt b sequences analyzed and their geographic pattern, where the one gathered from the northernmost-collected specimen (in the state of Minas Gerais) was the most divergent, are enticing to further explore variation of a larger sample of A. ruschii sequences. Such a study would clarify into which of the already known phylogeographic patterns of Atlantic forest mammals (e.g., Colombi et al. 2010; Ventura et al. 2012; Valdez and D'Elía 2013; see reviews in Martins 2011; Costa and Leite 2012) would A. ruschii fit, or if this species presents a so far undescribed pattern. Similarly, such a study would help assess the alpha diversity of Abrawayaomys (Pardiñas et al. 2009b). Now, that A. ruschii is becoming more frequent in specimen collections, such a study seems feasible.