OUP user menu

The Relationships of Major Lineages within Peromyscine Rodents: A Molecular Phylogenetic Hypothesis and Systematic Reappraisal

Jacqueline R. Miller , Mark D. Engstrom
DOI: http://dx.doi.org/10.1644/07-MAMM-A-195.1 1279-1295 First published online: 1 October 2008

Abstract

Peromyscine mice of the cricetid subfamily Neotominae comprise the most common and speciose assemblage of North American rodents. The composition and phylogenetic relationships within this group have been addressed using multiple character systems but remain unresolved, particularly the branching pattern of deeply rooted lineages. The relationships and status of several taxa formerly included within Peromyscus also are controversial. We present a new phylogenetic hypothesis based on combined mitochondrial and nuclear sequences using cytochrome b, interphotoreceptor retinoid-binding protein, and growth hormone receptor gene sequences, and include all major peromyscine lineages. Bayesian and parsimony approaches recover a congruent, well-resolved phylogeny, wherein Peromyscus forms a monophyletic assemblage only if a number of putative genera are included within it. Onychomys rather than Reithrodontomys is the sister taxon to Peromyscus (within which are included Habromys, Osgoodomys, Megadontomys, Podomys, and Neotomodon). Reithrodontomys in turn is the sister taxon to Isthmomys, and these 2 genera are the nearest outlier to Onychomys plus an expanded concept of Peromyscus, consistent with the recent recognition of this entire group as the tribe Reithrodontomyini. The Baiomyini, Ochrotomyini, and Neotoma (Neotomini) appear as successive outgroups to this clade: Based on these results, we present a systematic review of higher-level relationships among peromyscine rodents.

Key words
  • Baiomyini
  • classification
  • Neotominae
  • Ochrotomyini
  • Peromyscini
  • Peromyscus
  • phylogeny
  • Reithrodontomyini
  • rodents
  • systematics

The rodent suborder Myomorpha, typical “rats and mice,” includes 30% of all living mammalian species, and muroid rodents comprise the majority of myomorphs. Understanding the phylogenetic relationships within muroids has been problematic, irrespective of the taxonomic level investigated. Nowhere is this clearer than in the historical debate regarding the systematics of the Cricetidae, in particular the North and Central American Neotominae.

Distributed broadly across North and Central America, Neotominae comprises a diverse assemblage of 16 genera and approximately 120 species of rats and mice (Musser and Carleton 2005). Their ecological, behavioral, genetic, and taxonomic diversity, coupled with their ubiquitous distribution and commonality, have made the Neotominae ideal subjects for comparative studies, particularly the assemblage of mice referred to both formally (Hershkovitz 1969; Reeder et al. 2006) and informally (Carleton 1980:140) as the Peromyscini. Attempts to recover the phylogeny of peromyscines have employed morphological characters (Carleton 1980; Hooper 1957, 1958, 1959, 1960; Hooper and Musser 1964a, 1964b; Linzey and Layne 1969), banded chromosomes (Rogers et al. 1984; Stangl and Baker 1984), allozymes (Patton et al. 1981; Rogers et al. 2005), mitochondrial DNA (Bradley et al. 2004, 2007; Engel et al. 1998), and nuclear DNA (Reeder and Bradley 2004). Despite this rigorous effort, elucidating the relationships within this group has proven difficult. Combining mitochondrial and nuclear DNA sequences offers another approach to resolve branching patterns (e.g., Reeder et al. 2006), but taxonomic coverage to date has been limited.

Systematic history.—Despite intense interest, classification within Neotominae has not always been consistent, and the monophyly and composition of lineages remain unresolved.

For example, the number of recognized genera, tribes, and subfamilies varies among authors and definitions of different groups often overlap (Ellerman 1941; Glydenstolpe 1932; Hershkovitz 1966a, 1966b; McKenna and Bell 1998; Musser and Carleton 1993, 2005; Simpson 1945; Vorontsov 1959), impacting greatly on the concept of peromyscine mice or the broad definition of the nominate genus Peromyscus. The review by Musser and Carleton (2005) of classification within the newly designated Neotominae represents a framework from which to test hypothesized species relationships (Bradley et al. 2004, 2007; Carleton 1980; Hershkovitz 1962; Hooper and Musser 1964a, 1964b; McKenna and Bell 1998; Musser and Carleton 1993; Osgood 1909; Reeder and Bradley 2004; Reeder et al. 2006; Rogers et al. 2005; Simpson 1945; Vorontsov 1959). Novelties in that classification (Musser and Carleton 2005) include designation of the Tylomyinae as a subfamily, the formal description of new tribes (Ochrotomyini and Baiomyini), and redefinition of the tribe Reithrodontomyini. The original concept of the Reithrodontomyini (Vorontsov 1959) was modified to exclude Ochrotomys (now in its own tribe) and expanded to include Onychomys, in addition to Peromyscus and allied genera, and Reithrodontomys (Musser and Carleton 2005; see also Carleton 1980; Engel et al. 1998; Hooper and Musser 1964a, 1964b; Reeder et al. 2006).

Carleton (1980) proposed that Reithrodontomys was the sister-group to a restricted subset of Peromyscus (the former subgenera Peromyscus and Haplomylomys). As a result, he recognized several of the apparently basal subgenera as full genera (e.g., Habromys, Isthmomys, Megadontomys, Osgoodomys, and Podomys). Although Musser and Carleton (2005) retained these morphologically divergent taxa as genera, the pivotal sister-group relationship of Reithrodontomys to Peromyscus + Haplomylomys was not articulated in their reinterpreted concept of the Reithrodontomyini, and was explicitly rejected by Bradley et al. (2004), (2007).

Our objective is to examine phylogenetic relationships among species-groups, subgenera, genera, and putative tribes within peromyscines, using 3 molecular markers representing both the mitochondrial and nuclear genomes: cytochrome b (Cytb), interphotoreceptor retinoid-binding protein (IRBP), and growth hormone receptor (GHR). We use the classification of Musser and Carleton (2005) as a platform to test hypothesized relationships among peromyscines, and we follow Carleton (1980), (1989) and Musser and Carleton (2005) by initially defining Peromyscus to include only taxa formerly included in the subgenera Peromyscus and Haplomylomys. Habromys, Isthmomys, Megadontomys, Neotomodon, Osgoodomys, and Podomys, which are included in Peromyscus by some authors, are listed herein as separate genera. We sometimes refer to this group (minus Isthmomys) as “Peromyscus and allied genera” to facilitate discussion. Likewise our concept of “peromyscine” is congruent with Carleton's listing of taxa under the (informal) heading Peromyscini (1980:15), and includes Peromyscus, Habromys, Isthmomys, Megadontomys, Neotomodon, Onychomys, Osgoodomys, Podomys, and Reithrodontomys, as well as Baiomys, Ochrotomys, and Scotinomys (see Table 1).

View this table:
Table 1

The classification of species and genera allied with Peromyscus, under the respective arrangements of Hooper and Musser (1964a, 1964b; see Carleton 1980:10),a Carleton (1980), Musser and Carleton (2005), and Reeder et al. (2006).

Hooper and Musser, (1964a, 1964b)aCarleton (1980)b informal designationMusser and Carleton (2005)c formal classificationReeder et al. (2006)d
Neotomine–PeromyscinesNeotomine–PeromyscinesNeotominaeNeotomine–Peromyscines
NeotominiNeotomineNeotominiNeotomini
NeotomaHodomysHodomysHodomys
OtotylomysNelsoniaNelsoniaNelsonia*
TylomysNeotomaNeotomaNeotoma
XenomysXenomysXenomysXenomys
PeromysciniBaiomyineBaiomyiniBaiomyini
BaiomysBaiomysBaiomysBaiomys
NelsoniaScotinomysScotinomysScotinomys
NeotomodonPeromyscineOchrotomyiniPeromyscini
OchrotomysHabromysOchrotomysHabromys*
OnychomysIsthmomysReithrodontomyiniIsthmomys*
Peromyscus*MegadontomysHabromysMegadontomys*
ReithrodontomysNeotomodonIsthmomysNeotomodon
ScotinomysOchrotomysMegadontomysOnychomys
OnychomysNeotomodonOsgoodomys
OsgoodomysOnychomysPeromyscus
PeromyscusOsgoodomysPodomys*
PodomysPeromyscusReithrodontomys
ReithrodontomysPodomysOchrotomyini
TylomyineReithrodontomysOchrotomys
NyctomysTylomyini
OtonyctomysNyctomys*
OtotylomysOtonyctomys*
TylomysOtotylomys Tylomys
  • a In Hooper and Musser (1964a, 1964b) Neotoma included Hodomys as a subgenus, and Peromyscus included Habromys, Isthmomys, Megadontomys, Osgoodomys, and Podomys as subgenera.

  • b Derived from his concluding synopsis (Carleton, 1980) wherein he suggested informal use of tribes (Neotomine, Baiomyine, Peromyscine, Tylomyine) to indicate major subdivisions, pending further phylogenetic analyses. Nyctomys and Otonyctomys were presumed to be included within Tylomyines.

  • c Musser and Carleton (2005) assigned Nyctomys, Otonyctomys, Ototylomys, and Tylomys to their own subfamily, the Tylomyinae, distinct from the Neotominae.

  • d Reeder et al. (2006) considered their concept of Peromyscini as synonymous with that of the Reithrodontomyini of Musser and Carleton (2005).

  • * Not included in their analyses, but added by precedent or convention.

Materials and Methods

Taxonomic sampling and DNA extraction.—Peromyscine phylogeny is characterized by both recent and archaic diversification events. We selected 1 mitochondrial (Cytb) and 2 nuclear markers (IRBP and GHR) for their utility in resolving relationships at varying taxonomic levels, ranging from species within genera to genera and higher order assemblages within Cricetidae (Adkins et al. 2001; Arellano 1999; Bonvicino and Martins Moreira 2001; Bradley et al. 2004, 2007; Conroy and Cook 1999; DeBry and Sagel 2001; D'Elía 2003; Engel et al. 1998; Jansa and Weksler 2004; Serizawa et al. 2000; Smith and Patton 1999; Steppan et al. 2004; Tiemann-Boege et al. 2000; see also Page and Holmes 1998). We isolated total genomic DNA from either liver or kidney by proteinase-K digestion and phenol-chloroform extraction (Sambrook et al. 1989), purifying and concentrating an aliquot of each stock DNA by ethanol precipitation to 24 µl from a 96-µl volume. This concentrated DNA was used undiluted for the amplification of nuclear markers.

We sequenced 99 specimens, 46 species in all, representing all recognized tribes within Neotominae (Musser and Carleton 2005). All major peromyscine taxa at the level of subgenera and above are included, as well as all species-groups therein, with the exception of the P. truei group of Peromyscus. Neotomini is nominally represented by Neotoma micropus. We employed 2 species representing 2 genera of the Sigmodontinae as outgroups to neotomines, and thus outgroups to peromyscines (Appendix I; see D'Elía 2003; Jansa and Weksler 2004; Smith and Patton 1999; Steppan et al. 2004). We also included 1 species of the Tylomyinae, defined as a separate tribe by Musser and Carleton (2005).

Polymerase chain reaction amplification and sequencing.— Primers for each gene were obtained from the literature (Adkins et al. 2001; Irwin et al. 1991; Wade 1999) or designed from published or preliminary sequences or both (Table 2). The polymerase chain reaction was used to amplify a target 1,056–base pair (bp) fragment of Cytb DNA, 921-bp fragment of IRBP (exon 1), and an 852-bp fragment of GHR (exon 10). A 25-pl polymerase chain reaction master mix contained 1.0–2.0 µl of template, 2.5 µl of Erika Hagelberg buffer (Appendix II), 1.0 µl of forward primer, 1.0 µl of reverse primer, 1.0 µl deoxynucleoside triphosphates (5 mM), 0.8 units (0.2 µl) of Taq polymerase (Qiagen, Inc., Valencia, California; Fisher Scientific, Pittsburgh, Pennsylvania), 0–0.5 µl of MgCl (20 mM), and 17.5–18.3 µl of double-distilled H2O. Genomic DNA was diluted in sterile distilled water at 5:95 to 15:85 concentrations for Cytb amplification.

View this table:
Table 2

Tailed primer oligonucleotide sequences used to amplify cytochrome b (Cytb), exon 1 of interphotoreceptor retinoid-binding protein (IRBP), and exon 10 of growth hormone receptor (GHR). Tails are M13 forward and reverse (underlined).

CytbForward
CytB4F5′-CACGACGTTGTAAAACGACCTCCSACGCAGAYAAAATYCCRTTC-3′
CytB5F5′-CACGACGTTGTAAAACGACATAGCAACAGCATTYATAGGYTA-3′
CytB1F5′-CACGACGTTGTAAAACGACACCCTAACCCG ATTCTTYGCATTC-3′
JML147315′-CACGACGTTGTAAAACGACTGATATGAAAAAYCATCGTTGTAYTTC-3′
BITF (Wade 1999)a5′-CACGACGTTGTAAAACGACCCATCCAACATCTCAGCATG(ATGAAA)-3′
L1472BTF (Kocher et al. 1989)5′-CACGACGTTGTAAAACGACGAGATCTGAAAAACCATCGTGT-3′
Reverse
JMH35′-GGATAACAATTTCACACAGGCCGGTTTACAAGACCAAGTAAT-3′
CytB1R5′-GGATAACAATTTCACACAGGGAATGCRAAGAATCGGGTTAGGGT-3′
CytB4R5′-GGATAACAATTTCACACAGGGGRATTTTGTCTGCGTSGGAG-3′
CytB6R5′-GGATAACAATTTCACACAGGTGGYTTAATRTGTGCTGGAGTRT-3′
JMH65′-GGATAACAATTTCACACAGGCCTGCAATGGGYATTAGGATGATRATA-3′
MOUSE-TR (Wade 1999)5′-GGATAACAATTTCACACAGGTTCTTCATTTYTGGTTTACAAGACCA-3′
IRBP
Forward
IR3IRBP5′-CACGACGTTGTAAAACGACACCAACCTYACCCGAGAGAACTG-3′
IR4IRBP5′-CACGACGTTGTAA AACGACGCAAYGTGGGGCTACCTRCG-3′
Reverse
IR5IRBP5′-GGATAACAATTTCACACAGGTCAAARCGCADGTAGCCCAC-3′
IR6IRBP5′-GGATAACAATTTCACACAGGTTRCCYGGCAGCACNGAYACCT-3′
7RIRBP5′-GGATAACAATTTCACACAGGAGYCTTCTGGGGGRTCRYTAGGC-3′
GHRForward
GHREXON10 (Adkins et al. 2001)′-CACGACGTTGTAAAACGACGGRAARTTRGAGGAGGTGAACACMATCTT-3′
GHR7′-CACGACGTTGTAAAACGACCTTCWAC AWKG ATG ACTCCTGG-3′
GHR8′-CACGACGTTGTAAAACGACGARGGRTCAGACACRGACAGACT-3′
Reverse
GHREND (Adkins et al. 2001)′-GGATAACAATTTCACACAGGCTACTGCATGATTTTGTTCAGTTGGTCTGTGCTCAC-3′
GHR9′-GGATAACAATTTCACACAGGCACATAGCCACACGAGGAWAG-3′
GHR 10′-GGATAACAATTTCACACAGGTCTCAGGCAAAYGCAAAGCAG-3′
  • a Indicates a later revision to BITF by the addition of 6 base pairs.

For nuclear genes, a 2-step touchdown produced optimal results. Thermal profiles included an initial 3-min denaturing at 94°C followed by: 40 s at 94°C, 40 s at 55–60°C, 1 min 45 s at 72°C (15–17 cycles); and 40 s at 94°C, 40 s at 52–56°C, 1 min 45 s at 72°C (21–26 cycles). A 5-min final extension at 72°C was tailed to either protocol. For amplification of Cytb we used either this profile, or a more general profile with a constant annealing temperature of 50°C for a total of 36 cycles.

Polymerase chain reaction products were migrated across a 1.5% agarose gel. Bands were extracted and purified using a Qiagen QIAQUICK gel extraction kit, or by filter-tip centrifugation (Dean and Greenwald 1995). We removed unincorporated reagents and small fragments by alcohol precipitation. Sequencing was done on either a LICOR 4200 bidirectional automated sequence analyzer (LI-COR Biosciences, Lincoln, Nebraska) or an ABI 3000 Genetic Analyzer (Applied Biosystenis, Foster City, California), employing protocols recommended by the manufacturers. Sequences were aligned against known, published gene sequences using Sequencher 4.0 (Gene Codes Corp., Ann Arbor, Michigan) and concatenated into a single “supermatrix” (e.g., Gatesy et al. 2002), partitioned by marker domain. All sequences are accessioned in GenBank (Appendix I).

Phylogenetic analyses.—We inspected plots of transitional–transversional substitution against genetic distance using DAMBE (Xia 2000; Xia and Xie 2001), noting evidence of 3rd-position saturation in Cytb. Some authorities advocate removal of 3rd-position sequence where saturation is suspected (e.g., Ericson and Johansson 2003; Huchon et al. 2002). However, synonymous substitutions at 3rd positions are known to be phylogenetically informative, even when sequences are moderately saturated (Poux and Douzery 2004; Yoder et al. 1996; Yoder and Yang 2000). This is especially true where taxonomic units are closely related (Björklund 1999; Hästad and Björklund 1998; Pereira et al. 2002). We therefore include all codon positions in our analyses.

We evaluated variability in rates of nucleotide substitution versus an assumption of clock-like molecular change among lineages using a likelihood ratio test, assuming a chi-square distribution with n(taxa) − 2 degrees of freedom (Huelsenbeck and Bull 1996; Pereira et al. 2002; Posada 2003). We further verified rate variability among lineages and among genes by inspecting uncorrected genetic distances among taxa sequences and markers. Evidence of rate heterogeneity abrogates the value of performing incongruence tests (for instance, incongruence length distance tests) because rate differences alone will cause them to fail (Barker and Lutzoni 2002; Darlu and Lecointre 2002; Dolphin et al. 2000; Pybus 2006; Smith et al. 2004; Yoder et al. 2001). Furthermore, employing partitioned Bayesian inference of phylogeny circumvents problems of model selection when combining data by allowing independent models to be employed for each independent domain in concatenated sequences (Nylander et al. 2004). In parsimony, phylogenetic signal can be additive when employing homogeneous (e.g., unweighted) reconstructions, even when genes that differ by evolutionary process are combined (Sullivan 1996). Nonetheless, we inspected the topologies of the independent gene markers, assessing patterns of agreement–disagreement, considering branch lengths and nodal support where conflicts occurred. We considered these comparisons in the interpretation of the concatenated sequence analyses.

We used Bayesian inference (BI) and maximum parsimony (MP) to estimate phylogenetic relationships. Trees under MP were rooted with Oryzomys and Sigmodon sequences, representing Sigmodontinae (a subfamily, like Neotominae, within Cricetidae). BI only allows designations of a single outgroup taxon and we selected Sigmodon hispidus. We reviewed the overall biochemical characteristics of each gene initially, using the software programs MEGA (Kumar et al. 1993) and PAUP* version 4.0M0 (Swofford 2001).

Each gene differed with regard to compositional values, as well as gamma shape parameters and proportions of invariant sites (Table 3). For analyses employing BI, the model and parameters best representing each gene were determined using MODELTEST 3.7 (Posada and Crandall 1998), employing the Akaike information criterion (Akaike 1974; Posada and Buckley 2004). This yielded a general time reversible model (GTR+I+r) for Cytb and IRBP, and the TIM+I+r model for GHR (see Gu et al. 1995; Rodriguez et al. 1990; Yang 1993, 1994a, 1994b). TIM+I+r is a special case of the GTR+I+r model, under which its assumptions are nested (see Hillis et al. 1996; Posada and Crandall 2001). Although models are similar, parameterization varies. As such, we partitioned the data by gene, with all parameters of the GTR+I+Γ model unlinked except topology, such that base frequencies, gamma shape, the proportion of invariant sites, transition-transversion ratio, and so on, were estimated independently for each partition (e.g., see Geuten et al. 2004). We ran the partitioned model with topology unlinked concurrently, to inspect individual gene trees.

View this table:
Table 3

Sequence characterization and variability for combined data and data partitioned by loci domain. Cytb = cytochrome-b gene, GHR = exon 10 of growth hormone receptor gene, IRBP = fragment of exon 1 of interphotoreceptor retinoid-binding protein gene. R denotes the ratio of transitional to transversional change. Matrix characterization obtained using the programs MEGA (site characteristics), and PAUP (base frequencies). Model of evolution based on MODELTEST (3.7) using the aAkaike information criterion and by blikelihood ratio test.

LociTotal aligned sitesConserved sitesVariable sitesInformative sitesR total (1st, 2nd, 3rd positions)Base frequencies (all nucleotides/informative sites only)Model of evolution (PI,Γ)
Cytb1,0565205365051.7A = 30.6, C = 26.7, G = 13.2, T = 29.5GTR+I+Γa

(0.471, 0.961)
(2.6, 3.2, 1.4)A = 37.2, C = 33.3, G = 06.0, T = 23.5GTR+I+Γb

(0.471, 0.961)
GHR8525503022601.9A = 28.5, C = 28.0, G = 22.4, T = 21.1TIM+I+Γa

(0.293, 0.801)
(0.7, 2.2, 2.6)A = 25.7, C = 27.9, G = 24.3, T = 22.1HKY+I+Γb

(0.294, 0.803)
IRBP9215903312642.7A = 20.4, C = 28.8, G = 30.8, T = 20.0GTR+I+Γa
(1.4, 3.8, 3.0)A = 19.4, C = 34.4, G = 30.3, T = 15.9(0.305, 0.754) TrNeF+I+Γb

(0.325, 0.778)
Total2,8291,6621,1671,0291.8A = 26.7, C = 27.8, G = 21.7, T = 23.9GTR+I+Γa

(0.373, 0.420)
(1.9, 2.9, 1.7)A = 29.9, C = 32.2, G = 16.7, T = 21.2GTR+I+Γb

(0.373, 0.420)

Sampling under BI used a Metropolis-coupled Markov chain Monte Carlo method (Geyer 1991; Larget and Simon 1999; Li 1996; Mau et al. 1999; Yang and Rannala 1997), involving 10 × 106 generations in 2 parallel runs of 4 chains (2 cold and 6 hot), with trees sampled every 1,000 generations, implemented by MRBAYES 3.1 (Altekar et al. 2004; Huelsenbeck and Ronquist 2001; Ronquist and Huelsenbeck 2003; see also Metropolis et al. 1953). We determined stationarity by inspecting log-likelihood plots, and discarded the first 250 trees as burn-in. We also ran several independent Markov chain Monte Carlo runs to verify convergence following the recommendations of Huelsenbeck et al. (2002).

We performed unweighted MP analysis using PAUP* 4.0b 10 (Swofford 2001), employing a heuristic search, characterized by tree-bisection-reconnection branch-swapping and random sequence addition with 100 repetitions. Nodal support was calculated using 1,000 bootstrap pseudoreplicates with heuristic searches at each replication (50 random addition sequence replicates). Uninformative characters were removed.

Results

Rate heterogeneity and tests of the molecular clock.— Substitution rates differed among the 3 genes, assessed via pairwise, uncorrected distance matrices (not shown). In all cases, Cytb exhibits the most nucleotide variation (mean p-distance = 0.141 ± 0.008) relative to either nuclear marker in pairwise analyses (mean nuclear p-distances: IRBP = 0.033 ± 0.003; GHR = 0.037 ± 0.003) or the overall data matrix (0.072 ± 0.003).

We constrained the Bayesian tree using the partitioned, best-fitting model to be clocklike, and derived the corresponding likelihood score. This was compared to the likelihood score for the tree unconstrained by the assumption of a molecular clock using a likelihood ratio test (Huelsenbeck and Bull 1996; Posada 2003; see also Pereira et al. 2002). The test confirmed rate variation both among individual genetic loci and suggested that >1 lineages within the data set had different rates of substitution (concatenated supermatrix: 2 delta = 789.788, χ2 critical value = 145.78, d.f. = 97, P < 0.001). Independent likelihood ratio tests for each gene under the GTR+I+Γ model indicate a similar pattern of nonclocklike molecular evolution (Cytb, χ2= 455.035, d.f. = 97, P < 0.001; GHR, χ2 = 170.687, d.f. = 97, P < 0.001; IRBP, χ2 = 192.349, d.f. = 97, P < 0.001).

Molecular composition and patterns of substitution.— Unexpected rates or patterns of substitution can, in some cases, be attributed to nuclear pseudogene homologs. This is particularly true when dealing with mitochondrial genes that, in some rodent taxa, are known to have copies in the nuclear genome (e.g., DeWoody et al. 1999; Mirol et al. 2000; Smith et al. 1992; Zullo et al. 1991). The observed rates of substitution at the 3rd codon positions match the expected frequency (3rd position > 1st position > 2nd position) for mammalian mitochondrial genes (Irwin et al. 1991; Kimura 1983; Yang 1994a). Sequence lengths were as anticipated for the primers we used and the sequences generated were not characterized by unexpected stop codons, or fragmentation.

A chi-square goodness-of-fit test using all sites showed base composition to be homogenous across taxa in all genes (%2 = 107.1434, d.f. = 294, P = 1.000), as well as for each gene individually. However, the proportional composition of individual genes varied significantly (Table 3). Base composition also is homogeneous across taxa for either nuclear marker when only variable sites are considered, but varies significantly in Cytb (IRBP, χ2 = 148.341, P = 1.000; GHR, χ2 = 93.066, P = 1.000; Cytb, χ2 = 517.411, P < 0.001). The most dissimilar taxa represent outlying lineages (Sigmodon, Oryzomys, and Ototylomys) and, less so, lineages that have traditionally proven difficult to place phylogenetically, for example Isthmomys and Neotomodon.

We performed analyses, both for the supermatrix and for Cytb, using a functional outgroup approach (sequentially as Neotoma followed by Ochrotomys, and excluding outlying lineages), employed a purine-pyrimidine (RY) coding approach to 3rd positions in Cytb, and also performed analyses with Cytb removed (not shown) in order to assess what effect compositional heterogeneity as well as 3rd-position saturation could have on interpreting topology. The elimination of the outgroup taxa, employment of purine–pyrimidine coding, or exclusion of Cytb weakened overall nodal support but resulted in highly similar topologies.

Individually, Cytb also recovers an unusually long branch length for Neotoma, which, under this gene tree, falls among Peromyscus and its allied taxa. However, this position is poorly supported. To determine if this independent gene tree skews the overall topology, we analyzed the data with Neotoma removed from the sequence matrix. Analysis of Cytb excluding Neotoma recovered an otherwise identical topology to that with Neotoma included. Removal of Neotoma similarly has no qualitative effect on relationships within the supermatrix topology. However, the elimination of Neotoma weakens overall support. We employed a Kishino-Hasegawa test (Kishino and Hasegawa 1989) between the Cytb tree (1) and the same tree with Neotoma constrained to its position in the supermatrix analysis (2), the latter proving to be a better fit (log-likelihood [lnL] (1) −25,535.321 versus lnL (2) −25,486.140; P < 0.0001). However, we recognize that the statistical merits of such a test are inconclusive when the trees are not determined a priori. However, tree length under MP is also reduced when the position of Neotoma is likewise constrained.

Bayesian inference.—Our data plateaued by 1 million generations and showed no additional steps in the subsequent 9 million generations. Score variance of the average standard deviation of split frequencies stabilized after 7 million generations, with a standard deviation of 0.00285 at 10 million generations. Analysis yields a highly resolved tree with strong support for almost all deep branches, as well as for the majority of species-groups recovered. Log-likelihood scores were −28,509.47 for run 1, −28,509.66 for run 2, and −28,509.56 overall (arithmetic mean). Eight principle nodes are identified above the species level (Fig. 1, Roman numerals).

Fig. 1

Topology recovered from the combined Bayesian analysis of interphotoreceptor retinoid-binding protein (IRBP), cytochrome b (Cytb), and growth hormone receptor (GHR) genes, under a GTR+I+G model partitioned by gene, 10 million generations with 2 × 4 chains. Best marginal likelihood scores: run 1 = −29,509.47, run 2 = ‒29,809.66, combined = ‒29,809.56 (arithmetic means). Posterior probabilities subtend nodes. Asterisks denote probabilities of 1.00. Roman numerals correspond to higher-order branching events with strong posterior probability support. Arabic numerals in gray above nodes correspond to species-group associations. Genera: B. = Baiomys, H. = Habromys, O. = Onychomys, P. = Peromyscus, R. = Reithrodontomys, and S. = Scotinomys. Subgenera: A. = Aporodon, H. = Haplomylomys, P. = Peromyscus, and R. = Reithrodontomys. Geographically distinct species localities in parentheses: CAR = La Carpentera, Costa Rica; ESC = Escazu, Costa Rica; GUA = Baja Verapaz, Guatemala; MV = Monte Verde, Costa Rica; NICA = Isla Ometepe, Nicaragua; POAS = Volcan Poas, Costa Rica; and CAM = Campeche, Mexico.

Clade I (Neotominae, sensu Musser and Carleton 2005) appears monophyletic in our analysis, but additional taxa within the Neotomini are required to confirm this result. Clade II comprises the traditional peromyscine rodents, including the Ochrotomyini (Ochrotomys nuttalli), Baiomyini (Baiomys + Scotinomys), and the remaining peromyscine taxa (sensu Carleton 1980; see also Reeder et al. 2006). Clade III represents the Baiomyini plus the remaining peromyscine taxa. Clade IV includes Reithrodontomys + Isthmomys, sister to the remaining peromyscine taxa (i.e., exclusive of Ochrotomys, Baiomys, and Scotinomys), constituting the Reithrodontomyini of Musser and Carleton (2005). Clade V comprises the sister-groups Reithrodontomys + Isthmomys. Clade VI represents Onychomys + the remaining peromyscine taxa. Clade VE represents the genus Reithrodontomys. Lastly, clade VIII comprises Peromyscus and allied taxa, including Habromys, Haplomylomys, Megadontomys, Neotomodon, Osgoodomys, and Podomys, but excluding Isthmomys. Each of these 8 clades is recovered by the parsimony analysis and, with the exception the placement of Ochrotomys, is supported by strong posterior probabilities.

Ten additional distinct lineages or species-groups among these clades are strongly supported, with posterior probability values of 1.00 (Fig. 1). Within clade VII these include the monophyletic subgenera Reithrodontomys and Aporodon. Within . clade VIII these include a cohesive maniculatus species-group within Peromyscus (represented by P. maniculatus, P. melanotis, and P. polionotus); a clearly defined association between Peromyscus leucopus and the P. maniculatus group; the 2 species of Haplomylomys (P. californicus + P. eremicus); Haplomylomys + P. crinitus, in association with the P. leucopus + P. maniculatus group; P. melanophrys, P. mayensis, plus P. mexicanus + P. nudipes (in association with Megadontomys); Neotomodon + Podomys; a close association among P. aztecus, P. boylii, and P. levipes; and an association of Habromys with this latter group. A clade containing only the subgenera Peromyscus and Haplomylomys, sister to Reithrodontomys and exclusive of the remaining peromyscine mice (as hypothesized by Carleton [1980], [1989]) was not recovered.

Parsimony.—Unweighted MP analysis recovered 12 equally parsimonious trees of 5,130 steps, based on 1,029 parsimony-informative sites (Table 3). Consensus of these 12 trees is well resolved (strict consensus; Fig. 2). Most of the deeper nodes (I–VIII) identified under BI are recovered with moderate to strong bootstrap support. These events include strong bootstrap support for a monophyletic Neotominae clade (94%, clade I), moderate support for the placement of Ochrotomys as sister to all other peromyscines (78%, clade II), and weak support for the inclusion of the baiomyine taxa (Baiomys + Scotinomys) within peromyscines exclusive of Ochrotomys (65%). Likewise, there is strong bootstrap support (95%) for a monophyletic reithrodontomyine assemblage in accord with Musser and Carleton (2005), within which is evidence for 2 distinct clades (clade V and clade VI) in agreement with the results of BI. Bootstrap support for the Isthmomys + Reithrodontomys clade is moderate (75%), but there is no bootstrap support (51%) for the association of Onychomys with the assemblage of Peromyscus and allied genera (clade VI), although that grouping was consistently recovered, and was well supported under BI.

Fig. 2

Strict consensus representing the 12 equally parsimonious trees obtained from the parsimony analysis of unweighted interphotoreceptor retinoid-binding protein (IRBP), growth hormone receptor (GHR), and cytochrome b (Cytb) gene concatenated sequences. Tree Length = 5,130, Consistency Index = 0.2975, Rescaled Consistency Index = 0.2243. Bootstrap values appear above branches. Asterisks indicate bootstrap support of 100% for terminal branch sequences. Roman numerals indicate deep clade events suggested in text. Genera: B = Baiomys, H = Habromys, O = Onychomys, P = Peromyscus, R = Reithrodontomys, and S = Scotinomys. Subgenera: A = Aporodon, H = Haplomylomys, P = Peromyscus, and R = Reithrodontomys. Geographically distinct species localities in parentheses: CAR = La Carpentera, Costa Rica; ESC = Escazu, Costa Rica; GUA = Baja Verapaz, Guatemala; MV = Monte Verde, Costa Rica; NICA = Isla Ometepe, Nicaragua; POAS = Volcan Poas, Costa Rica; and CAM = Campeche, Mexico. Majority-rule concensus renders a sister relationship between the clades comprising (Habromys + allies) and (Neotomodon + Podomys and an expanded mexicanus group, including P. mayensis and P. melanophrys).

Parsimony recovered all of the 10 additional species associations noted in the Bayesian analysis. The subgenera Reithrodontomys and Aporodon are monophyletic each with bootstrap supports of 99%. Likewise, there is support under MP for the larger grouping of Peromyscus leucopus with P. melanotis + P. polionotus + P. maniculatus. P. californicus + P. eremicus form a monophyletic Haplomylomys, associated in turn with P. crinitus and Osgoodomys, although without bootstrap support. Haplomylomys + allies are further associated with the P. leucopus + P. maniculatus assemblage under majority rule consensus (not shown), although this larger grouping is not resolved under strict consensus. The mexicanus species-group is resolved, with outliers P. mayensis and P. melanophrys (but not Megadontomys). Peromyscus boylii, P. aztecus, and P. levipes form a group that is sister to Habromys.

There is a close relationship between Podomys and Neotomodon, congruent with the BI topology. However, under MP, these 2 taxa are allied with the mexicanus species-group and its allies rather than with the boylii species-group + Habromys.

The position of the Podomys + Neotomodon clade is equivocal under both scenarios (bootstrap < 50%, posterior probability 0.69), although consensus of the 12 MP trees by majority rule recovers Podomys + Neotomodon in a larger association, congruent with the BI analysis in membership. We consider these associations (mexicanus + allied species, Habromys +’ boylii group, and Podomys + Neotomodon) to form an unresolved polytomy.

Congruence.—There is considerable congruence between the topologies recovered under MP and under BI including the recovery of deep nodes as well as species associations. Although Ochrotomyini is monotypic and we have represented Neotomini with a single species, both taxa also fall consistently outside of Baiomyini and Reithrodontomyini in accord with their recognition as distinct tribes (Musser and Carleton 2005). There are some places where the Bayesian and parsimony trees differ (Figs. 1 and 2). For example, in the tenuirostris group of Reithrodontomys (represented by R. creper and R. microdon), the placement of R. (A.) microdon is equivocal (Figs. 1 and 2). Likewise the exact relationship of the Neotomodon–Podomys clade within Peromyscus varies between analyses, as does the placement of Megadontomys.

Independent gene trees demonstrate some noncongruent relationships, the majority of which have negligible support. These variants likely represent polytomies that reflect the reduced number of characters in each single gene, particularly the conserved nuclear markers for which there are few derived characters distinguishing the terminal branches. Between nuclear trees the position of Onychomys, for example, falls either inside or sister to Peromyscus and allied taxa. Cytb differs from the nuclear genes in its poor midtree resolution, and in the placement of Neotoma, which has an unusually long branch length for its position in this gene tree. We evaluated all levels of noncongruence by 1st assessing the number of synapomorphies and thus degree of resolution, and collapsing unsupported nodes into polytomies under individual gene trees. We also examined phylogenies generated after employing several manipulations (see above). Noncongruence between the nuclear markers and Cytb is principally due to 3rd-position saturation and compositional bias in Cytb. Congruence improved when the inconsistency was explored with a variety of methods (functional outgroup analysis, removal of 3rd positions, and purine–pyrimidine coding of nucleotides), but at the expense of reducing nodal support within the supermatrix tree. Thus, we are confident that branching sequence conflict in the position of Neotoma reflects its long branch length relative to the other taxa, and the spurious presence of nonhomologous characters supporting deep branch nodes in the rapidly evolving Cytb gene (see Galtier et al. 2006). Analysis of combined data partitions appears to have a stabilizing effect, improving clade supports overall.

Novel taxa.—Our sequences for R. gracilis obtained from Monte Verde (MV, Figs. 1 and 2) fall among those of samples of R. mexicanus (ESC, Figs. 1 and 2) obtained from the vicinity of San Jose, Costa Rica, in both the supermatrix analysis and in analyses of individual loci, and not with typical R. gracilis from the Yucatan Peninsula (CAM, Figs. 1 and 2). Uncorrected genetic distances (not shown) between Monte Verde samples and R. mexicanus for all 3 loci are shorter than those for other closely related species pairs in Reithrodontomys. Also in Aporodon, 1 lineage (ROM114291 and ROM116805) from Volcan Poas (POAS, Figs. 1 and 2; Alajuela, Costa Rica) and 1 from La Carpentera (CAR, Figs. 1 and 2; Cartago, Costa Rica; ROM116835) appear as outliers to the known R. mexicanus species-group. The Poas taxon is morphologically distinct from R. mexicanus from the same locality, based on coloration, dentition, foot size, and relative tail length, and likely represents a new species (to be described elsewhere). Likewise, ROM116835 differs morphologically from R. mexicanus caught sympatrically at La Carpentera and at neighboring localities, and demonstrates significant sequence divergence in all 3 genetic markers. This specimen might indicate the presence of cryptic species within the R. mexicanus species complex.

Discussion

In the phylogenetic analysis we recovered a highly resolved, monophyletic Neotominae consisting of 4 clades corresponding to the tribal level classification of Musser and Carleton (2005). In relative branching sequence, these represent the separation of Neotoma (Neotomini), Ochrotomyini, Baiomyini, and Reithrodontomyini (Figs. 1 and 2). Neotoma appears as an outlier to peromyscines; however, additional representatives of Neotomini (i.e., Hodomys, Xenomys, Nelsonia, and Neotoma) are required to substantiate that result.

Ochrotomyini

The placement of golden mice, Ochrotomys, has long been problematic (e.g., see Osgood [1909], who recognized it as a monotypic subgenus of Peromyscus). Using dental and phallic morphology, Hooper (1957), (1958) found Ochrotomys to be distinct from Peromyscus, and elevated it to generic rank. Independent lines of evidence collectively suggest both taxonomic distinction (Patton and Hsu 1967) and an early divergence (Carleton 1973, 1980; Engel et al. 1998; Engstrom and Bickham 1982; Hooper and Musser 1964a, 1964b). Examination of chromosomal banding data indicates that Ochrotomys is highly derived and it appears as an outlier to Reithrodontomys, Baiomys, Onychomys, and Peromyscus (Engstrom and Bickham 1982), with few homologies shared between Ochrotomys and the remaining peromyscine rodents. Examination of our data places Ochrotomys as the basal lineage within peromyscines (Figs. 1 and 2), sister to the larger assemblage comprising Baiomyini and Reithrodontomyini (Musser and Carleton 2005). This arrangement has been recovered elsewhere with mitochondrial and nuclear markers (e.g., Bradley et al. 2007; Engel et al. 1998; Reeder et al. 2006).

Baiomyini

Baiomys and Scotinomys.—A sister-group relationship between Scotinomys and Baiomys has been proposed based on numerous lines of evidence (Arata 1964; Carleton 1980; D'Elía 2003; Engel et al. 1998; Hooper 1960; Hooper and Musser 1964a, 1964b; Miller and Engstrom 2007; Reeder and Bradley 2004; Reeder et al. 2006). Musser and Carleton (2005) formally recognized the 2 genera as constituting the tribe Baiomyini. However, this interpretation has been questioned elsewhere (Packard 1960; Rogers and Heske 1984) and Onychomys has sometimes been associated with this group (Rogers et al. 2005; see also discussion in Carleton 1980). In agreement with Musser and Carleton (2005) we recovered a strongly supported, monophyletic Baiomys + Scotinomys clade to the exclusion of Onychomys or any other taxon (Bradley et al. 2004, 2007; Engel et al. 1998; Reeder and Bradley 2004; Reeder et al. 2006). Consistent with Carleton (1973), (1980) the baiomyines are distinct, and sister to the Reithrodontomyini. Emerging data on vocal signaling and “singing” provide other unique synapomorphic characters for the baiomyine clade (Miller and Engstrom 2007).

Reithrodontomyini

Reithrodontomys, Isthmomys, Onychomys, and Peromyscus and allied genera.—Vorontsov (1959) proposed the Reithrodontomyini to include Peromyscus (and its allies) and Reithrodontomys, to the exclusion of Onychomys, the latter of which he segregated as a monotypic tribe. Bradley et al. (2004), (2007) placed Onychomys with Neotoma in the Neotomini, and regarded it as the sister lineage to the Peromyscini. However, Onychomys appeared as the sister-group to Peromyscus and allied taxa in both Reeder and Bradley (2004) and Reeder et al. (2006). Likewise, Musser and Carleton (2005) defined the Reithrodontomyini as comprising Onychomys, Reithrodontomys, Isthmomys, Peromyscus, Megadontomys, Osgoodomys, Podomys, Neotomodon, and Habromys (but see Reeder et al. [2006], who instead refer to this group as Peromyscini). Our results support the composition of this group; however, we recognize a 2nd, major division within the Reithrodontomyini, separating an Isthmomys + Reithrodontomys clade from Onychomys + Peromyscus (and allied taxa).

Reithrodontomys + Isthmomys.—A sister-group relationship between Reithrodontomys and Isthmomys has been found previously in phylogenetic analyses of electrophoretic data (Patton et al. 1981; Rogers et al. 2005) and recently based on Cytb sequences (Bradley et al. 2007). However, the association of Isthmomys and Reithrodontomys is surprising given the gross disparity in body size and morphology between them (Isthmomys is 10 times heavier than the largest ReithrodontomysReid 1997). Moreover, Isthmomys lacks grooved incisors, a key synapomorphy defining Reithrodontomys. Stangl and Baker (1984) likewise distinguished Isthmomys as a subgenus under Peromyscus based on chromosome banding data, noting that Isthmomys fell outside a putative Peromyscus + Onychomys assemblage.

Among peromyscines, Reithrodontomys is 2nd only to Peromyscus in species richness. Two subgenera, Aporodon and Reithrodontomys, are recognized (Hooper 1952; Howell 1914), although Hooper (1952) regarded R. fulvescens and R. hirsutus as intermediate between them, based on dental morphology (see also Arellano 1999). In our data set, the subgenera within Reithrodontomys are clearly distinct and monophyletic, with R. fulvescens included in the subgenus Reithrodontomys. Within our sample of Aporodon are diminutive harvest mice from Monte Verde (Puntarenas, Costa Rica) identified as R. gracilis (Reid and Langtimm 1993), which represented a substantial southern extension of its distribution. In our analysis, however, R. gracilis haplotypes from Monte Verde (MV, Figs. 1 and 2) fall among samples of R. mexicanus and not with typical R. gracilis from the Yucatan Peninsula (CAM, Figs. 1 and 2). We therefore assign the sample from Monte Verde to R. mexicanus and we doubt that R. gracilis occurs in Costa Rica. Remaining specimens assigned herein to R. mexicanus (ESC, GUA, and NICA, Figs. 1 and 2) do not form a monophyletic group and thus, based on our limited sampling of R. mexicanus, the species appears to be paraphyletic.

Peromyscus + allied taxa.—Peromyscus, as defined by Musser and Carleton (2005), is not monophyletic (Avise et al. 1979; Linzey and Lane 1969; Rogers et al. 1984; Stangl and Baker 1984). Osgood (1909) broadly conceived of Peromyscus as including species, subgenera, and genera subsequently viewed as constituting the Peromyscini (Carleton 1980, see also Carleton 1989; Engel et al. 1998; Hershkovitz 1962, 1966a, 1966b, 1972; Reeder et al. 2006), with the majority of species in the genus subsumed under the subgenera Peromyscus and Haplomylomys. Six apomorphic subgenera were subsequently removed from Peromyscus and ranked as genera (Habromys, Isthmomys, Megadontomys, Osgoodomys, Peromyscus, and Podomys), in addition to Neotomodon (Carleton 1980, 1989; Musser and Carleton 2005; see also Linzey and Layne 1969). Carleton (1980) elevated many of these subgenera based on his conclusion that Reithrodontomys was the sister-group to the subgenera Haplomylomys + Peromyscus, and that the other subgenera were outliers to this clade (see also Hooper and Musser 1964a, 1964b). Nonetheless, other authors have placed all or nearly all of these taxa within Peromyscus (Bradley et al. 2007; Rogers et al. 1984, 2005; Stangl and Baker 1984; Williams et al. 1985; Yates et al. 1979), a broad concept for the genus similar to that presented by Hooper and Musser (1964b) and often referred to as “Peromyscus (sensu lato).”

Examination of our data indicates that Onychomys is the sister-group to most genera or subgenera formerly included in Peromyscus and that this larger group is distinct from Reithrodontomys + Isthmomys. Given that Reithrodontomys is not the sister taxon to a restricted subset of deer mice comprising the subgenera Peromyscus + Haplomylomys (see also Bradley et al. 2004, 2007; Engel et al. 1998; Reeder and Bradley 2004; Reeder et al. 2006; Rogers et al. 2005), the original rationale for the generic recognition of Habromys, Megadontomys, Podomys, Neotomodon, and Osgoodomys (Carleton 1980; see also Carleton 1989) is falsified. Instead, Habromys, Neotomodon, Osgoodomys, Podomys, and, in most analyses, Megadontomys fall within and are associated with individual lineages of the subgenera Peromyscus and Haplomylomys (i.e., Peromyscus as defined by Musser and Carleton [2005]).

Podomys, Neotomodon, and Habromys.Podomys is allied with Neotomodon in our analyses, and the 2 are associated with either the Habromys + boylii + aztecus + levipes assemblage (Bayesian analysis), or the mayensis + melanophrys + mexicanus assemblage (parsimony). Thus, both genera always nest within Peromyscus, a result consistent with earlier observations based on chromosomal data (Yates et al. 1979; see also Greenbaum et al. 1994). However, their exact position within the genus varies. Habromys, likewise, is sometimes viewed as closely related to Podomys and Neotomodon (Carleton 1980, 1989; Hooper 1968). Engel et al. (1998) placed Habromys as sister to a mexicanus + boylii species-group clade, but Rogers et al. (2005) found Habromys associated with Osgoodomys based on allozyme data. Examination of our data consistently places Habromys as sister to a boylii + aztecus clade; and this group is allied with Neotomodon and Podomys in the Bayesian analysis, but not in parsimony. Like Neotomodon and Podomys, Habromys is positioned within the species-groups constituting the subgenus Peromyscus.

Megadontomys.—Megadontomys traditionally has been associated with Isthmomys. The 2 taxa were once considered congeneric, and Isthmomys flavidus and I. pirrensis were originally described as species of Megadontomys (Bangs 1902; Goldman 1912, 1920). A sister relationship between the 2 genera was proposed by Carleton (1980), but has not been recovered by additional lines of evidence (Linzey and Layne 1969; Rogers et al. 1984, 2005; Stangl and Baker 1984; see also Carleton 1989). Megadontomys is not closely related to Isthmomys in our analyses, but instead is either associated with an expanded P. mexicanus assemblage, or lies as sister to Peromyscus and allied genera (excluding Isthmomys). Although its exact phylogenetic position relative to Peromyscus and allied genera is not resolved in our analyses, inclusion within an expanded concept of Peromyscus (excluding Isthmomys) is consistent with alternative branching sequences.

Osgoodomys.—Based on phallic morphology, Hooper and Musser (1964a) found P. (Osgoodomys) bander anus to be distinct from other Peromyscus and recognized it as a monotypic subgenus. It was later elevated to generic rank (Carleton 1980; Linzey and Layne 1969). In several other studies, however, Osgoodomys was found to be derived from within Peromyscus; within a basal polytomy in Rogers et al. (1984) and Stangl and Baker (1984); associated with Habromys, P. boylii, and P. eremicus in Rogers et al. (2005); or as the sister-group to Haplomylomys in Bradley et al. (2004) and Engel et al. (1998). Our MP analyses indicated a close association of Osgoodomys with members of the subgenus Haplomylomys, a relationship also reported by Bradley et al. (2004) and Engel et al. (1998). In our BI analysis P. crinitus (not sampled by Bradley et al. [2004] or Engel et al. [1998]) was associated with this group, whereas Osgoodomys was removed. As was true of Neotomodon, Podomys, and Habromys, Osgoodomys is clearly derived from within the restricted cluster of taxa (subgenera Peromyscus and Haplomylomys) comprising the genus Peromyscus as defined by Musser and Carleton (2005).

Species-Groups

Peromyscus + Haplomylomys.—Many species-groups we recovered conform to those identified elsewhere in the literature for Peromyscus (Bradley et al. 2004, 2007; Carleton 1980; Engel et al. 1998; Musser and Carleton 1993, 2005; Sullivan et al. 1997). Although our taxonomic sampling within Peromyscus is limited, we also found evidence of more broadly defined groupings, lending credence to expanded concepts of some of the more traditional species-groups. Within Peromyscus these represent: a close relationship among the aztecus + boylii + levipes species-groups (conforming to a former view of the P. boylii group, after Osgood [1909] and Hooper [1968]); this complex allied to Habromys; the grouping of P. californicus + P. eremicus (Haplomylomys); association of P. mayensis with the mexicanus and melanophrys species-groups; and monophyly of the leucopus + maniculatus species-groups, a sister-group relationship that has been demonstrated elsewhere (Bradley et al. 2004; Engel et al. 1998; Rogers and Engstrom 1992; Rogers et al. 2005). In our analyses, the leucopus and maniculatus species-groups are distinct from other members of the subgenus Peromyscus and, when included within that taxon, render it paraphyletic. We anticipate that additional taxon sampling will result in removal of the leucopus + maniculatus species-groups to a separate subgenus or genus, and possibly other species-groups as well.

A Taxonomic Reappraisal

Within the Neotominae, the traditional concept of peromyscine rodents is corroborated by a monophyletic group of Ochrotomyini + Baiomyini + Reithrodontomyini. Their distinction as tribes also supports aspects of the deep phylogenetic structure within Neotominae proposed by Musser and Carleton (2005). Within Reithrodontomyini, there is a pronounced dichotomy between the clade comprising Isthmomys + Reithrodontomys and that of Onychomys + Peromyscus. We recommend that the dichotomy be formally recognized by restricting the tribe Reithrodontomyini to include Reithrodontomys Isthmomys, and reintroducing the tribe Peromyscini to include Onychomys and Peromyscus and its allied genera.

There has been recent agreement from many lines of evidence that a concept of Peromyscus restricted to the subgenera Peromyscus and Haplomylomys (Carleton 1980, 1989) is paraphyletic (e.g., Bradley et al. 2004, 2007; Engel et al. 1998; Reeder and Bradley 2004; Reeder et al. 2006; Rogers et al. 1984, 2005; Stangl and Baker 1984; Tiemann-Boege et al. 2000). Construction of a monophyletic Peromyscus requires substantial modification: either incorporation of a number of putative genera that do not consistently appear as outgroups to the subgenera Peromyscus and Haplomylomys, or further fragmentation of the genus. At present, we recommend that Peromyscus be redefined to include Megadontomys, Osgoodomys, Neotomodon, Podomys, and Habromys, but to exclude Isthmomys (as represented by I. pirrensis). This classification renders Peromyscus monophyletic, recognizing that this generic concept encompasses a comparatively broad range of morphological and genetic diversity. The alternative of maintaining the 5 taxa at the level of genus requires that Peromyscus be broken into numerous generic shards comprising current genera and subgenera and species-groups, including the 5 aforementioned taxa and at least separate maniculatus + leucopus, melanophrys + mexicanus, Haplomylomys + P. crinitus, and boylii + aztecus + levipes groups. Our sampling of the subgenera Peromyscus and Haplomylomys is not exhaustive, and inclusion of additional taxa undoubtedly will reveal other distinct groups. Although a broadly defined Peromyscus is more in keeping with the “traditional” concept of Peromyscus (e.g., Hooper 1968; Hooper and Musser 1964b), the construct is novel in that Neotomodon is included in the genus whereas Isthmomys is not. We anticipate that examination of additional data may ultimately find this concept of Peromyscus inflated, such that disintegration of the genus into 10 or more well-defined genera will be the most viable option.

As we near the 100th anniversary of Osgood's (1909) revision of peromyscine rodents, it is gratifying to witness the coalescence of numerous lines of evidence, from morphology to molecular biology, which will ultimately result in a resolved phylogeny for the most ubiquitous group of North American mammals. To achieve that end, we look forward to future syntheses, which include a more comprehensive set of taxa in Peromyscus and Reithrodontomys; additional molecular, morphological, and behavioral data; and synthetic analyses of multiple data sets.

Acknowledgments

We thank R. D. Bradley (Texas Tech University, Lubbock, Texas), D. S. Rogers (Brigham Young University, Provo, Utah), and R. C. Dowler (Angelo State University, San Angelo, Texas), as well as P. Waddell and the Peromyscus Genetic Stock Center for providing tissue samples for this research. We also thank D. S. Rogers, S. J. Steppan, W. D. Kilburn, and A. J. Baker for comments on earlier stages of this manuscript; R. M. Adkins, S. L. Pereira, and O. Haddrath for technical guidance; and R. M. Adkins for additional laboratory support. This study was funded by the Royal Ontario Museum (MDE), Connaught Foundation open fellowship (JRM), the Ontario Graduate Scholarship Program (JRM), and the University of Toronto Doctoral Completion Grant (JRM).

Appendix I

Specimens examined and utilized for data analyses.—For each specimen is given the collection locality, the museum accession or collector number, and the GenBank accession numbers (EF) for interphotoreceptor retinoid-binding protein (IRBP), cytochrome b (Cytb), and growth hormone receptor (GHR) genes, respectively. Abbreviations for museum specimens are as follows: Royal Ontario Museum (ROM), Angelo State Natural History Collection (ASNHC), Universidad Nacional Autónoma de México (CNMA), Brigham Young University (BYU), Carnegie Museum (CMNH), University of South Carolina Peromyscus Genetic Stock Center (USC-PGSC), Museum of Texas Tech University (TTU), and Angelo State University (ASK: field number). Locality acronyms (CAM, CAR, ESC, GUA, MV, NICA, and POAS) are defined in the captions for Figs. 1 and 2.

Baiomys musculus.—MEXICO: Chiapas; Frontera Comolapa (ROM97641: EF989834, EF989933, EF989735); MEXICO: Morelos; Réserva de la Biosfera Sierra De La Hualtla (ROM117128: EF989835, EF989934, EF989736; ROM117133: EF989836, EF989935, EF989737).

Baiomys taylori.—United States: Texas; Brown County; Camp Bowie (ROM114886: EF989838, EF989937, EF989739; ROM114884: EF989837, EF989936, EF989738); United States: Texas (ASNHC11056: EF989839, EF989938, EF989740).

Habromys ixtlani.—MEXICO: Oaxaca; Llano de la Flores, km 132 Tuxtepec–Oaxaca (CNMA29849: EF989842, EF989941, EF989832).

Habromys lepturus.—MEXICO: Oaxaca; Cerro Zempoaltepetl, 3 km E of Santa Maria Yacochi (CNMA29970: EF989841, EF989940, EF989742; CNMA29972: EF989840, EF989939, EF989741).

Habromys lophurus.—GUATEMALA: Huehuetenango; Santa Eulalia, 16 km NW of Santa Eulalia (ROM98341: EF989858, EF989943, EF989744); GUATEMALA: Huehuetenango; Santa Eulalia, 12 km NW of Santa Eulalia (ROM98299: EF989843, EF989942, EF989743; ROM98342: EF989845, EF989944, EF989745).

Isthmomys pirrensis—PANAMA: Darien Province; Mt. Pirri, summit (ROM116308: EF989846, EF989945, EF989746; ROM116307: EF989848, EF989947, EF989748; ROM116309: EF989847, EF989946, EF989747).

Megadontomys thomasi.—MEXICO: Oaxaca; La Esperanza, 11 km SW of La Esperanza (CNMA29186: EF989849, EF989948, EF989749; CNMA29188: EF989850, EF989949, EF989750).

Neotoma micropus.—United States: Texas; Winkler County, Winkler County Road 406 (ROM114902: EF989853, EF989952, EF989753; ROM114903: EF989854, EF989953, EF989754).

Neotomodon alstoni.—MEXICO: Distrito Federale; 3 km S of Paires (ASNHC 1595: EF989851, EF989950, EF989751; ASNHC 1596: EF989852, EF989951, EF989752).

Ochrotomys nuttalli.—United States: Florida; 9.2 miles NE of Panama City (ROM113008: EF989862, EF989961, EF989761; ROM113007: EF989863, EF989962, EF989762).

Onychomys arenicola.—United States: Texas; Presidio County, Hip O Ranch, 5 miles W of Marfa (ROM114904: EF989855, EF989954, EF989755; ROM114894: EF989856, EF989955, EF989833).

Onychomys leucogaster.—United States: Texas; Port Isabel, 11 miles N of Port Isabel (ASNHC4348: EF989859, EF989958, EF989758); United States: Texas; Crane County, Dune Oil Field and University–Waddell Oil Field (ROM114892: EF989860, EF989959, EF989759).

Onychomys torridus.—United States: Arizona; Pima County, 4.5 miles S and 5.5 miles E Continental (ROM11491: EF989868, EF989967, EF989767); United States: New Mexico; Rodeo, 1.6 miles N of Rodeo (ASNHC4066: EF989861, EF989960, EF989760).

Oryzomys couesi.—MEXICO: Tamualipas; 4 km W of La Carbonera, 46.5 km ESE of San Fernando (ROM96122: EF989867, EF989966, EF989766); United States: Texas; Cameron County, Port Isobel (ASK0840: EF989866, EF989965, EF989765).

Osgoodomys banderanus.—MEXICO: Michoacán, La Mira, 8 km N of La Mira (ASNHC2664: EF989857, EF989956, EF989756); MEXICO: Colima; 3 km SE of Colima (ASNHC2634: EF989858, EF989957, EF989757).

Ototylomys phyllotis.—MEXICO: Campeche; 44 km S of Constitution, 70 km E of Escarcega (ROM95675: EF989864, EF989963, EF989763; ROM95676: EF989865, EF989964, EF989764).

Peromyscus aztecus.—EL SALVADOR: Santa Ana; Parque Nacional Montecristo, Los Planes (ROM101490: EF989870, EF989969, EF989769; ROM101489: EF989869, EF989968, EF989768).

Peromyscus boylii.—MEXICO: Colima; Hacienda San Antonio (ASNHC3449: EF989871, EF989970, EF989770; CMNH103724: EF989872, EF989971, EF989771).

Peromyscus californicus.—University of South Carolina (stock origin—Peromyscus Genetic Stock Center [USC-PGSC IS 1590: EF989873, EF989972, EF989772]).

Peromyscus crinitus.—United States: Utah; Uintah County, Bitter Creek Canyon (BYU16629: EF989874, EF989973, EF989773; BYU16630: EF989875, EF989974, EF989774).

Peromyscus eremicus.—MEXICO: Sonora; 22 km S (by road) Hermosillo (BYU17952: EF989876, EF989975, EF989775); United States: Utah; Washington County, Fort Pierce, Fort Pierce Wash (BYU18684: EF989877, EF989976, EF989776).

Peromyscus leucopus.—CANADA: Ontario; Toronto (ROM101861: EF989880, EF989979, EF989779); MEXICO: Quintana Roo; San Miguel, 20.3 km SE of San Miguel (CMNH92801: EF989881, EF989980, EF989780).

Peromyscus levipes.—MEXICO: Chiapas; Cerro Tzontehuiz (ROM97624: EF989882, EF989981, EF989781); GUATEMALA: Huehuetenango; Santa Eulalia, 10 km NW of Santa Eulalia (ROM98294: EF989883, EF989982, EF989782).

Peromyscus maniculatus.—United States: Michigan; Ann Arbor (stock origin), University of South Carolina—Peromyscus Genetic Stock Center (USC-PGSC BW 27740: EF989887, EF989986, EF989786); CANADA: Ontario; Kwataboahegan, Kwataboahegan River (ROM98941: EF989884, EF989983, EF989783; ROM98946: EF989885, EF989984, EF989784); CANADA: Quebec; near Tadoussac (ROM114367: EF989886, EF989985, EF989785).

Peromyscus mayensis.—GUATEMALA: Huehuetenango; Santa Eulalia, 16 km NW of Santa Eulalia (ROM98360: EF989888, EF989987, EF989787; ROM98339: EF989889, EF989988, EF989788).

Peromyscus melanophrys.—MEXICO: Zacatecas (stock origin), University of South Carolina—Peromyscus Genetic Stock Center (USC-PGSC XZ 1073: EF989890, EF989989, EF989789).

Peromyscus melanotis.—University of South Carolina—Peromyscus Genetic Stock Center, stock origin (USC-PGSC 25: EF989891, EF989990, EF989790).

Peromyscus mexicanus.—COSTA RICA: Puntarenas; Monte Verde Biological Station (ROM113250: EF989895, EF989994, EF989794); COSTA RICA: Cartago; Capellades (ROM113188: EF989892, EF989991, EF989791); COSTA RICA: Guanecaste; Volcan Santa Maria (ROM113237: EF989894, EF989993, EF989793).

Peromyscus nudipes.—COSTA RICA: Cartago; Cerro de la Muerte, San Gerardo del Dota (ROM113216: EF989893, EF989992, EF989792).

Peromyscus polionotus.—United States: Florida; Ocala National Forest (stock origin), University of South Carolina—Peromyscus Genetic Stock Center (USC-PGSC PO 11033: EF989896, EF989995, EF989795).

Podomys floridanus.—United States: Florida; Columbia County, O'Leno State Park (TTU97866: EF989878, EF989977, EF989777; TTU97867: EF989879, EF989978, EF989778).

Reithrodontomys brevirostris.—COSTA RICA: Alajuela; Parque Nacionale Juan Castro Blanco, 10 km E of Sucre (ROM116804: EF989918, EF990017, EF989817).

Reithrodontomys creper.—COSTA RICA: Cartago; Parque Nacional Volcan Irazu (ROM116798: EF989900, EF989999, EF98979; ROM116799: EF989899, EF989998, EF989799); COSTA RICA: Cartago; Rio Birris, 12 km N of Porter (ROM97321: EF989898, EF989997, EF989797; ROM97311: EF989897, EF989996, EF989796).

Reithrodontomys darienensis.—PANAMA: Danen Province, Mt. Pirri, summit (ROM116311: EF989916, EF990015, EF989815).

Reithrodontomys fulvescens.—MEXICO: Nayarit; Santiago, 4.8 km N of Santiago (ASNHC3465: EF989901, EF990000, EF989800); MEXICO: Tamaulipas; Soto La Marina, 5 km W of Soto La Marina (ROM96109: EF989902, EF990001, EF989801); United States: Texas; Brown County, Camp Bowie (ROM114900: EF989903, EF990002, EF989802; ROM114901: EF989904, EF990003, EF989803).

Reithrodontomys gracilis (CAM).—MEXICO: Campeche; Champoton, 52 km SW of Champoton (ROM95890: EF989905, EF990004, EF989804).

Reithrodontomys gracilis (MV).—COSTA RICA: Puntarenas; Monte Verde, and Monte Verde Biological Station (ROM97308: EF989906, EF990005, EF989805; ROM116845: EF989908, EF990007, EF989807).

Reithrodontomys megalotis.—MEXICO: Distrito Federal; Paires, 3 km S of Paires (ASNHC2133: EF989909, EF990008, EF989808; ASNHC2136: EF989910, EF990009, EF989809).

Reithrodontomys mexicanus (ESC, NICA, GUA).—COSTA RICA: San Jose; San Antonio de Escazu, foothills near Pico Blanco (ROM116841: EF989912, EF990011, EF989811; ROM116844: EF989907, EF990006, EF989806); NICARAGUA, Ometepe Island, Cerro Madera (ROM116824: EF989921, EF990020, EF989820; ROM116839: EF989917, EF990016, EF989816); GUATEMALA:, Verapaz; Purulha, 5 km E of Purulha (ROM98468: EF989911, EF990010, EF989810).

Reithrodontomys microdon.—GUATEMALA: Huehuetenango, Santa Eulalia, 12 km NW of Santa Eulalia (ROM 98300: EF989914, EF990013, EF989813); GUATEMALA: Huehuetenango, Santa Eulalia, 16 km NW of Santa Eulalia (ROM 98382: EF989915, EF990014, EF989814).

Reithrodontomys sp. (CAR).—COSTA RICA: Cartago; Cerro de la Carpentera, Iztaru Scout Camp (ROM116835: EF989913, EF990012, EF989812).

Reithrodontomys sp. (POAS).—COSTA RICA: Alajuela; Parque Nacionale Volcan Poas (ROM114291: EF989919, EF990018, EF989818; ROM116805: EF989920, EF990019, EF989819).

Reithrodontomys spectabilis.—MEXICO: Quintana Roo, San Miguel, 30 km SE of San Miguel (ASNHC2139: EF989922, EF990021, EF989821); MEXICO: Quintana Roo, Isla Cozumel, 1.5 km N of El Cedral (ROM97733: EF989923, EF990022, EF989822).

Reithrodontomys sumichrasti.—GUATEMALA: Huehuetenango, Santa Eulalia, 10 km SW of Santa Eulalia (ROM98383: EF989924, EF990023, EF989823); GUATEMALA: Chimaltenango, Santa Apolonia, 15 km SW of Santa Apolonia (ROM98384: EF989925, EF990024, EF989824).

Scotinomys teguina.—EL SALVADOR: Santa Ana; Parque Nacionale Montecristo, Los Planes (ROM101507: EF989928, EF990027, EF989827); GUATEMALA: Baja Verapaz; 5 km E of Purulha (ROM98466: EF989929, EF990028, EF989828); COSTA RICA: Cartago; La Carpentera, Iztaru Scout Camp (ROM116802: EF989930, EF990029, EF989829).

Scotinomys xerampelinus.—COSTA RICA: Cartago; Parque Nacionale Volcan Irazu (ROM97311: EF989931, EF990030, EF989830; ROM116813: EF989932, EF990031, EF989831).

Sigmodon hispidus.—MEXICO: Campeche; La Valeta (ROM95214: EF989926, EF990025, EF989825; ROM95215: EF989927, EF990026, EF989826).

Appendix II

Formula for Erika Hagelberg polymerase chain reaction buffer.

  • 195 µl of double-distilled H2O

  • 100 µl of 1 M Tris HCl (pH 8.3)

  • 500 µl of 1 M KCl 25 µl of 1 M MgCl2

  • 80 µl of bovine serum albumin (1.6 mg/ml)

  • 100 µl of 1% gelatin

  • Total volume = 1,000 µl

Footnotes

  • Associate Editor was Carey W. Krajewski.

Literature Cited

View Abstract