OUP user menu

A Phylogenetic Assessment of Pocket Gophers (Geomys): Evidence from Nuclear and Mitochondrial Genes

Ryan R. Chambers, Philip D. Sudman, Robert D. Bradley
DOI: http://dx.doi.org/10.1644/08-MAMM-A-180R1.1 537-547 First published online: 2 June 2009

Abstract

Phylogenetic relationships within Geomys historically have been difficult to assess using morphometric and chromosomal data. DNA sequences from the nuclear-encoded interphotoreceptor retinoid-binding protein gene (Rbp3) and mitochondrial 12S ribosomal RNA (12S rRNA) gene were used to examine the relationships within Geomys. In addition, sequence data from Rbp3 and mitochondrial 12S rRNA gene regions were combined with DNA sequence data from the mitochondrial cytochrome-b gene (Cytb) reported in a previous study. Results from phylogenetic analyses support previously established relationships in the recognition of 4 species groups (breviceps, bursarius, personatus, and pinetis) and a minimum of 12 species. Additionally, results agree with previous studies in considering the elevation of G. pinetis mobilensis and G. breviceps sagittalis to species-level status and in reevaluating the taxonomic status of 2 additional subspecies (G. personatus davisi and G. p. maritimus).

Key words
  • cytochrome b
  • Geomys
  • interphotoreceptor retinoid-binding protein
  • phylogenetics
  • pocket gophers
  • 12S ribosomal RNA

Members of Geomys occur from southern Canada throughout the Great Plains of the United States and into northeastern Mexico, and along the coastal prairies of the southeastern United States (Fig. 1). Despite being studied for more than 100 years, phylogenetic relationships remain problematic. Early studies involving morphological data struggled to provide well-supported, stable classifications. Hall (1981) recognized 8 species (arenarius, bursarius, colonus, cumber-landius, fontanelus, personatus, pinetis, and tropicalis) and 35 subspecies proposed in earlier works (Alvarez 1963; Baker 1950; Baker and Genoways 1975; Baker and Glass 1951; Bangs 1898; Davis 1938, 1940; Hall 1932; Hooper 1940; Jackson 1957; Komarek and Spencer 1931; McLaughlin 1958; Russell 1968; Sherman 1940, 1944; Swenk 1939, 1940; Villa-R. and Hall 1947). Williams and Genoways (1980) recommended that 3 of the recognized species (colonus, cumber-landius, and fontanelus) be synonomized with G. pinetis, but this recommendation came too late to be included by Hall (1981).

Fig. 1

Map depicting the approximate distributions of pocket gophers within the genus Geomys. Dots indicate sampling localities included in this study.

As chromosomal and allozymic data became available, results from numerous studies (Baker et al. 1989; Block and Zimmerman 1991; Bradley et al. 1991; Dowler 1989; Heaney and Timm 1983, 1985; Qumsiyeh et al. 1988; Tucker and Schmidly 1981) supported recognition of 10 species: arenarius, attwateri, breviceps, bursarius, knoxjonesi, lutescens, personatus, pinetis, texensis, and tropicalis. However, Patton (1993) took a more conservative approach, recognizing only 5 species (arenarius, bursarius, personatus, pinetis, and tropicalis) and excluded the other taxa pending a more thorough review.

Results from more recent studies of the genus using advanced chromosomal and molecular techniques (Baker et al. 2003; Burt and Dowler 1999; Jolley et al. 2000; Smolen and Bickham 1994, 1995; Sulentich et al. 1991) led to the recognition of 11 species: arenarius, attwateri, breviceps, bursarius, knoxjonesi, lutescens, personatus, pinetis, strecken, texensis, and tropicalis. In the most recent synopsis of the Geomyidae, Patton (2005) recognized only 9 species, having excluded lutescens and strecken pending further evaluation. The most complete study of Geomys to date (Sudman et al. 2006) examined DNA sequences from the mitochondrial cytochrome-b gene (Cytb) to assess the phylogenetic relationships within the genus. Four species groups were defined (breviceps, bursarius, personatus, and pinetis) with a minimum of 11 species (listed above). In addition, Sudman et al. (2006) suggested that 5 additional taxa (G. bursarius jugossicularis, G. personatus davisi, G. p. maritimus, G. breviceps sagittalis, and G. pinetis mobilensis) may warrant species-level status. Genoways et al. (2008) examined chromosomal, nuclear DNA sequences, and mitochondrial restriction site data in a study of 2 hybrid zones in Nebraska involving G. bursarius and G. lutescens. Genoways et al. (2008) agreed with Sudman et al. (2006) in that G. bursarius jugossicularis should be recognized as a species and that a former subspecies of G. lutescens (G. l. halli) should be placed under jugossicularis. Herein, we follow the taxonomic arrangement of Sudman et al. (2006), as supplemented by Genoways et al. (2008), and examined the 12 species (arenarius, attwateri, breviceps, bursarius, jugossicularis, knoxjonesi, lutescens, personatus, pinetis, strecken, texensis, and tropicalis) currently recognized within Geomys.

Although the current hypothesis by Sudman et al. (2006) is informative, Cytb has a high rate of nucleotide substitution, possibly affecting resolution at basal nodes (Hafner et al. 2007; Moritz et al. 1987). One of the most common methods used to counter this characteristic of Cytb is to incorporate sequence data from a more slowly evolving nuclear gene into analyses. By doing so, potential gene-tree biases associated with mitochondrial DNA (mtDNA) can be mitigated. These include biased base composition, lineage sorting of ancestral polymorphisms, and homoplasy (Avise 1994; Moore 1995; Prychitko and Moore 2000; Zhang and Hewitt 1996). However, a more slowly evolving nuclear gene may provide incomplete resolution if branch lengths are short (Hafner et al. 2007; Lanyon 1988). Given that extant species of Geomys presumably diverged in the relatively narrow time interval between 2.5 and 5.7 million years ago (Jolley et al. 2000), a phylogeny constructed from multiple unlinked loci (e.g., nuclear and mitochondrial genes) that evolve at different rates would assist in resolving relationships with a higher degree of confidence and avoid single-gene biases.

This study had 3 goals. First, we expanded upon the studies of Jolley et al. (2000) and Sudman et al. (2006) by increasing the taxonomic sampling of 12S ribosomal RNA (12S rRNA) gene sequences and adding sequence data from the nuclear-encoded interphotoreceptor retinoid-binding protein gene (Rbp3) for the same samples and localities in Sudman et al. (2006). Second, we assessed congruence among the 12S rRNA, Rbp3, and Cytb data sets. Third, we combined and analyzed the 3 data sets to determine phylogenetic relationships within the genus.

Exon I of Rbp3 has been used to study mammalian phylogeny at a variety of taxonomic levels (DeBry and Sagel 2001; Jansa and Weksler 2004; Serizawa et al. 1999; Springer et al. 1997; Stanhope et al. 1992, 1996, 1998; Weksler 2003) and for this study, it will provide a phylogeny reconstruction independent of that obtained with DNA sequence data from a mitochondrial gene. The mitochondrial 12S rRNA gene has an intermediate rate of evolution in that it is not as conservative as exons in nuclear genes, yet changes more slowly than other mitochondrial genes that code for proteins (Ferris et al. 1983; Pesole et al. 1999). This gene has been shown to be a useful marker in studies of rodents at various taxonomic levels (Allard and Honeycutt 1992; Delpero et al. 2001; Hafner et al. 2007; Kuznetsova et al. 2002; Ledje and Arnason 1996; Montgelard et al. 2002; Nedbal et al. 1994, 1996; Olson et al. 2005). Jolley et al. (2000) examined phylogenetic relationships among pocket gophers (Geomys) using the 12S rRNA gene. The 12S rRNA phylogeny was well supported and provided a testable hypothesis; however, the taxon sampling scheme fell short of the number of taxa examined by Sudman et al. (2006). For this study, DNA sequences (12S rRNA) were obtained for 7 taxa not examined by Jolley et al. (2000) and combined with the existing dataset to be analyzed independently of the Rbp3 data set.

Materials and Methods

Taxonomic sampling.—Twenty-four specimens representing the 12 recognized species, 9 subspecies, and 3 outgroup taxa (Appendix I) were examined. Nine of the 24 specimens sequenced for Rbp3 and 7 specimens sequenced for the 12S rRNA gene were the exact samples included in Sudman et al. (2006). Most other specimens included were from the same or nearby collecting localities as those examined by Sudman et al. (2006).

Based on results from previous studies (Hafner and Nadler 1990; Hafner and Page 1995; Jolley et al. 2000; Sudman et al. 2006), 3 closely related genera were selected as outgroup taxa (Cr-atogeomys, Pappogeomys, and Thomomys). An Rbp3 sequence for Thomomys was obtained from GenBank (accession number AF297277). 12S rRNA sequences for 17 of 24 taxa were available from Jolley et al. (2000; GenBank accession numbers AF084288AF084307). The remaining Rbp3 and 12S rRNA sequences were generated in this study. All Cytb sequences were available from Sudman et al. (2006) and obtained from GenBank (accession numbers AY393935, AY393936, AY393939, AY393940, AY393944, AY393945, AY393947-AY393952, AY393958, AY393960, AY393961, AY393963, AY393964, AY393966AY393968, and AY393971) with the exception of accession number FJ210793, which was generated herein.

Data collection.—Genomic DNA was isolated from approximately 0.1 g of frozen liver or muscle tissue using the DNEasy kit (Qiagen, Valencia, California). Approximately 1,300 base pairs (bp) at the 5′ end of exon 1 of the single-copy Rbp3 was amplified by polymerase chain reaction (Saiki et al. 1988) using primers A and B (Stanhope et al. 1992). Thermal profiles, optimized for different taxa, were adapted from the following standard profile: initial denaturation at 95°C for 10 min, 35 cycles of denaturation at 95°C for 25 s, annealing at 58°C for 20 s, and extension at 72°C for 60 s, with a final extension at 72°C for 10 min (Jansa and Weksler 2004). Approximately 870 bp comprising the 12S rRNA gene were amplified by polymerase chain reaction using primers L82 and H900 (Nedbal et al. 1996). The thermal profile consisted of an initial denaturation at 95°C for 2 min, 30–35 cycles of denaturation at 95°C for 30 s, annealing at.56-58°C for 40 s, and extension at 72°C for 50 s, with a final extension at 72°C for 8 min. In addition, 1,120 bp of Cytb were obtained following methods and primers outlined in Sudman et al. (2006).

Polymerase chain reaction products were purified using the QIAquick PCR purification kit (Qiagen). Amplified products were sequenced with an ABI 3100-Avant automated sequencer and ABI Prism Big Dye version 3.1 terminator technology (PE Applied Biosystems, Foster City, California). Primers used to cycle sequence Rbp3 included Geo395R (5′ -GGCCGCTGGT-GCAGTGTCGGAGA-3′), 125F, Geo609F (CCCTTCCAA-CACCACCATGAGATCTGG), Geol405R (GGGGACCC-CACACC), Geo958R (GCATGGCCAGAGCCTTCTCC), B, D, and E2 (DeBry and Sagel 2001; Stanhope et al. 1992; Weksler 2003). Primers beginning with “Geo” were modified from the original citation (Stanhope et al. 1992). For 12S rRNA cycle sequencing reactions, the same primers were used as in polymerase chain reaction amplification as well as L309 and H626 (Nedbal et al. 1996) and for Cytb sequencing primers were those of Sudman et al. (2006). Cycle sequencing reactions were purified using isopropanol cleanup protocols. Sequences were aligned and proofed using Sequencher 4.0 software (Gene Codes, Ann Arbor, Michigan) and chromato-grams were examined to verify all base changes and to inspect sequences for heterozygous sites, which were coded following the International Union of Biochemistry polymorphic code. MEGA 3.0 (Kumar et al. 2004) and MacClade 4.0 software (Maddison and Maddison 2000) were used to check for stop codons and the presence of pseudogenes. DNA sequences for Rbp3 and 12S rRNA were deposited in GenBank and accession numbers are listed in Appendix I.

Data analysis.—Data analyses were conducted in 2 phases. First, independent analyses were conducted including sequence data from Rbp3 and the 12S rRNA gene, respectively. Nodes that differed between trees were not considered conflicting unless both nodes had a bootstrap support value greater than 75% (Helbig et al. 2005). Second, sequences from Rbp3 and the 12S rRNA gene were combined with Cytb sequences from Sudman et al. (2006) and herein and analyzed together. Rbp3, 12S rRNA, and Cytb gene regions are approximately 1,305, 870, and 1,140 bp, respectively, resulting in a total sequence length of 3,315 bp.

The partition homogeneity test or incongruence-length-difference test (Farris et al. 1994; Mickevich and Farris 1981) was used to examine heterogeneity in the combined data partitions. Previous studies (Bull et al. 1993) indicated that combining heterogeneous data might negatively affect phy-logeny estimates (see Reeder et al. [2006] for a summary of recent literature). However, other studies (Adkins et al. 2001; Flynn and Nedbal 1998; Pereira et al. 2002) indicated otherwise and Cunningham (1997) determined that P-values > 0.01 (obtained from the incongruence-length-difference test) demonstrated that combined data did not differ from individual partitions. The P-value (P = 0.980) obtained herein justified combining individual data sets (Rbp3, 12S rRNA, and Cytb) as described above.

For parsimony analysis, nucleotide positions were treated as equally weighted, unordered, discrete characters with possible states A, C, G, T, and heterozygous sites designated following the International Union of Biochemistry code. Uninformative characters were excluded and optimal trees were estimated using the heuristic search method with tree bisection-reconnection branch swapping and stepwise addition sequence options in the software package PAUP* version 4.0b 10 (Swofford 2002). Robustness and nodal support of topologies were assessed using heuristic bootstrapping (Felsenstein 1985) with 1,000 iterations.

For the Bayesian analysis, Rbp3 and Cytb sequences were partitioned by codon and 12S rRNA sequences were unpartitioned. The analyses were performed in MRBAYES version 3.1.2 (Huelsenbeck and Ronquist 2001; Ronquist and Huelsenbeck 2003) and the GTR+I+G model of evolution (determined from the MODELTEST program—Posada and Crandall 1998) with the following options: 4 Markov chains (1 cold, 3 heated), 10 million generations, and sample frequency every 1,000th generation. The first 10,000 trees were discarded as burn-in. A consensus tree (50% majority rule) was constructed from the remaining trees using PAUP* (Swofford 2002). Furthermore, 3 additional Bayesian analyses were performed as above, except with 1 million generations, sampling every 100 generations, and burn-in set at 1,000 trees. The 3 trees generated were examined for stability and compared to the longer run.

Pairwise genetic distances were estimated using the Kimura 2-parameter model of evolution (Kimura 1980). The Kimura 2-parameter model was selected so that genetic distances from the 2 nuclear genes could be compared to those published for Cytb (Sudman et al. 2006). Average genetic distances were calculated for individuals within and between clades and values used to measure levels of genetic divergence between taxonomic groups. Values for Cytb sequences were obtained, in most cases, from Sudman et al. (2006).

Results

The DNA sequences from the nuclear-encoded Rbp3 and the mitochondrial 12S rRNA gene and Cytb were obtained for 21 ingroup taxa and 3 outgroup taxa (Appendix I). Because of similarities between topologies obtained by parsimony and Bayesian methods and independent analyses of each gene region, only the tree from the combined Bayesian analysis is shown and emphasized.

Rbp3, 12S rRNA, and Cytb combined analysis.—The Rbp3 and 12S rRNA gene regions were combined with Cytb data from Sudman et al. (2006) and herein to generate a sequence with a length of 3,315 bp. Bayesian analysis of the 3 combined gene regions produced a topology (Fig. 2) containing 4 major clades, which were more highly resolved and strongly supported than any other analysis. Specifically, the 1st clade included the bursarius species group with 2 subclades (A and B). Subclade A contained samples representing G. arenarius and G. knoxjonesi and subclade B was divided into 2 minor clades representing samples of G. bursarius-like specimens (1 contained G. bursarius major,G. b. majusculus, G. jugossicularis jugossicularis, G.j. halli, and G. lutescens lutescens; the 2nd contained G. texensis texensis and G. t. bakeri). The 2nd clade represented the personatus species group and contained 2 subclades (C and D). Subclade C contained samples representing G. strecken and G. attwateri, although this clade received low nodal support values, whereas subclade D contained samples representing G. personatus-like specimens (G. personatus personatus, G. p. davisi, G. p. megopotamus, and G. p. maritimus) and G. tropicalis. The 3rd and 4th clades and represented the breviceps and pinetis species groups, respectively.

Fig. 2

Phylogenetic tree obtained from a Bayesian analysis of the 3 combined gene regions (interphotoreceptor retinoid-binding protein, 12S ribosomal RNA, and cytochrome b) with clade posterior probabilities (≥95 = support [shown as an asterisk]) above and bootstrap support values (only support values ≥65 are shown) below branches. Capital letters (A-D) refer to subclades discussed in the text.

Parsimony analysis used 519 informative characters to generate 1 most-parsimonious tree (length = 1,689 steps, consistency index [CI] = 0.4387, and retention index [RI] = 0.5613). There was strong bootstrap support (= 95) to resolve clades representing the bursarius, personatus, and breviceps species groups. Also, there was moderate support for clades containing members of the bursarius and personatus groups (bootstrap = 82). The bursarius group was weakly supported (bootstrap = 66), but there was support (bootstrap = 94) for a clade containing specimens that represent G. arenarius and G. knoxjonesi. Within the bursarius group, the remaining taxa formed 2 clades: the 1st clade contained bursarius, lutescens, and jugosiccularis and the 2nd clade contained the 2 specimens of G. texensis. These 2 clades were supported and nearly identical to the combined analysis. The personatus group was strongly supported (bootstrap = 95) and there was support for the clade containing G. strecken, and the placement of G. attwateri was unresolved relative to the clade containing all other specimens representing G. personatus as well as G. tropicalis.

Rbp3 sequences.—A 1,305-bp region located in exon I of Rbp3 contained the following nucleotide frequencies (estimated from the data): A = 20.1%, C = 29.5%, G = 30.4%, and T = 20.1%. The average transition/trans version ratio (estimated from the data) was 2.2 for all characters. The percent heterozygosity was 0.012% (16 of 1,305 sites).

Bayesian analysis of the 1,305-bp region located in exon I of Rbp3 (not shown) produced a similar but less-resolved topology compared to that obtained in the combined analysis (Fig. 2). A clade containing all but 2 members (attwateri and lutescens) of the bursarius species group was recovered, although most relationships within the clade were unresolved. The personatus (except strecken), breviceps, and pinetis species groups were recovered and similar in topology compared to that obtained in the combined analysis.

The parsimony analysis of Rbp3 sequences used 43 informative characters to generate 51 equally most-parsimonious trees (length = 59 steps, CI = 0.8305, andRI = 0.8810). Although an unresolved polytomy was formed between clades containing specimens that represented the bursarius, personatus, and breviceps species groups, a bootstrap consensus tree depicted relationships similar to that in the Bayesian analysis. Support was obtained for the 2 clades containing specimens of G. breviceps and G. pinetis, respectively, with G. pinetis representing an early offshoot from the remaining members of the genus. Also, there was moderate support (bootstrap = 89 and 86, respectively) for 2 clades representing the personatus species group (minus strecken). However, all other relationships, including taxa representing the bursarius species group, as well as relationships at terminal nodes, were unresolved with low to no support.

Kimura (1980) 2-parameter distance values for Rbp3 sequences averaged 0.60% between species. Values for comparisons of currently recognized subspecies ranged from 0% (G. personatus personatus versus G. p. megapotamus and G. jugossicularis jugossicularis versus G. j. halli) to 0.24% (2 samples of G. b. sagittalis), whereas values for comparisons of currently recognized species ranged from 0.08% (G. tropicalis versus G. personatus davisi) to 1.5% (G. pinetis versus G. bursarius and G. texensis).

12S rRNA sequences.—An 849-bp region of the mitochondrial 12S rRNA gene for 24 specimens contained the following nucleotide frequencies (estimated from the data): A = 36.9%, C = 20.4%, G = 19.0%, and T = 23.7%. The average transition/trans version ratio (estimated from the data) was 3.5 for all characters.

Bayesian analysis utilizing sequence data from the 12S rRNA gene produced a slightly different topology (not shown) than that generated in the combined analysis. First, there was no support for the clade including all ingroup taxa, in contrast to all other analyses. Second, there was support for the clade containing all members of the personatus species group, but this clade contained G. attwateri. Third, although there was more support and resolution within the bursarius clade, G. knoxjonesi and G. arenarius were not included. Overall, support was lacking at basal and intermediate nodes.

The parsimony analysis of 849 bp of the 12S rRNA gene used 109 informative characters to generate 8 equally most-parsimonious trees (length = 277 steps, CI = 0.5235, and RI = 0.6675). A bootstrap consensus tree depicted a similar topology to the tree obtained in the Bayesian analysis, the main difference involving less support at intermediate nodes.

Kimura (1980) 2-parameter distance values for 12S rRNA sequences averaged 3.67% between species. Values for subspecies ranged from 0.20% (G. personatus maritimus versus G. p. davisi) to 3.0% (G. pinetis pinetis versus G. p. mobilensis), whereas values for currently recognized species ranged from 0.60% (G. tropicalis versus G. personatus davisi and G. p. maritimus) to 8.1% (G. strecken versus G. pinetis mobilensis).

Discussion

The topology obtained in the Bayesian analysis of the 3 combined gene regions (Fig. 2) represents the phylogenetic hypothesis used to discuss taxonomic and systematic scenarios. Four clades were recovered containing taxa that represented the species groups (bursarius, personatus, breviceps, and pinetis) described in Sudman et al. (2006). Below, each species group is described in detail and taxonomic suggestions (if pertinent) are discussed.

Geomys bursarius species group.—Results from this study support previous hypotheses in the recognition of 6 species (arenarius, knoxjonesi, jugossicularis, lutescens, bursarius, and texensis) within this group. G. lutescens, not currently recognized in the most recent review of geomyids (Patton 2005), is placed within subclade B as the sister taxon to G. jugossicularis. It has been suggested that species-level status was warranted in studies that date to the revision of Merriam (1895). Sudman et al. (2006) reported an average genetic distance estimated from Cytb sequence data of 8.8% between G. lutescens and G. bursarius and of 8.1% between G. lutescens and G. (bursarius)jugossicularis (see below). These values fall within the range (5-11%) for which Baker and Bradley (2006) suggested further review, but exceed those for other recognized species of Geomys. In addition, G. lutescens possesses a karyotype (2n = 72, FN = 86−98 [Hart 1978]) distinct from that of G. bursarius (2n = 70−72, FN = 68−74 [Hart 1978]) and G. jugossicularis (lutescens)halli (2n = 70, FN = 72 [Genoways et al. 2008]) and no apparent gene flow between taxa (Sudman et al. 1987). Combined, these data further support G. lutescens as a monotypic species.

Results from Sudman et al. (2006) suggested that specimens currently named G. bursarius jugossicularis (subclade B) warranted specific status. Further review of these taxa was suggested based on genetic distance data and on the presence of a distinct karyotype. In addition, specimens recognized as G. lutescens halli, based on chromosomal and DNA sequence data, were more closely related to G. jugossicularis and appeared to represent a subspecies of G. jugossicularis based upon the oldest available name. Furthermore, Genoways et al. (2008) incorporated comparisons of mitochondrial restriction sites, nuclear sequences, and chromosomal morphology, in an examination of 2 hybrid zones in Nebraska. Taxa examined included G. bursarius majusculus, G. b. jugossicularis, G. lutescens lutescens, and G. l. halli. Our results lend support to the hypotheses presented by Genoways et al. (2008) that 3 species are present in this region (G. bursarius, G. jugossicularis, and G. lutescens).

Finally, many subspecies of G. bursarius were not included in this study. These taxa occur primarily in the central United States and represent a wealth of future research focused on understanding the process of speciation within the genus. For example, specimens representing G. bursarius major and G. b. missouriensis were reported by Sudman et al. (2006) as differing by an average genetic distance of 5.8% (Cytb sequences), a value in the range (5-11%) suggested by Baker and Bradley (2006) as warranting further examination.

Extensive sampling of these taxa is needed to further resolve relationships within the bursarius group.

Geomys personatus species group.—Results presented here support the hypothesis of Sudman et al. (2006) in the recognition of 4 species (G. attwateri, G. strecken, G. personatus, and G. tropicalis) within this group. G. strecken, not recognized by Patton (2005), is located in an unsupported clade (subclade C) along with G. attwateri. G. strecken has an average genetic distance of 11.9% (Cytb sequences) to G. attwateri and of 11.7% to the remaining G. personatus-like specimens (Sudman et al. 2006). This is larger than the average values observed among other species (e.g., G. arenarius versus G. knoxjonesi = 10.5%). Additionally, chromosomal (Smolen and Bickham 1995), rDNA (Davis 1986), and mtDNA (Jolley et al. 2000) differences as summarized by Sudman et al. (2006) lend support to the recognition of G. strecken as a species.

The relationships among taxa representing specimens of G. personatus, as well as G. tropicalis (subclade D), are more complicated. Placement of G. tropicalis in the personatus species group based on DNA sequence data results in a paraphyletic arrangement. Although Sudman et al. (2006) reported low levels of genetic differentiation within this clade (5.3%), G. tropicalis possesses a completely metacentric karyotype (2n = 38, FN = 72 [Davis et al. 1971; Qumsiyeh et al. 1988]), differentiating it from other members of the personatus group. This taxon also is geographically isolated in a small region in southeastern Tamaulipas, Mexico. Additionally, Sudman et al. (2006) suggested further review of G. personatus davisi and G. p. maritimus to determine if they warrant species-level status. Similar to previously reported relationships, G. p. davisi is sister to G. tropicalis based on DNA sequence data, yet possesses a karyotype more similar to that of G. p. personatus. In contrast to results presented by Sudman et al. (2006), G. p. maritimus was not the most basal member of the group. Herein, it was located in a supported clade that contained G. p. davisi and G. tropicalis, whereas the clade containing G. p. personatus and G. p. megopotamus represents the basal member of the personatus group (although this may be the result of a single sample of each individual being examined).

Although there is not sufficient evidence to suggest formal taxonomic changes at this time, 4 possible scenarios exist within the personatus clade. First, G. tropicalis could be relegated to subspecific status under G. personatus, resulting in a monophyletic grouping of personatus-like specimens. However, Selander et al. (1975) hypothesized that the processes of karyotypic and genie evolution have proceeded independently in pocket gophers and that the reduction in chromosome number in G. tropicalis was not accompanied by a similar degree of allelic substitution at structural gene loci. Other researchers (Davis et al. 1971; Grant 1971; Patton 1970, 1972) argued that this is a means of adapting to local environmental conditions. As such, a 2nd scenario, therefore, is that 2 species could be recognized that would include G. personatus (G. p. personatus and G. p. megopotamus) and G. tropicalis (G. t. tropicalis, G. t. maritimus, and G. t. davisi), respectively, following the oldest available name, although this is in conflict with the topology shown in Sudman et al. (2006). Third, the recognition of G. personatus (G. p. personatus and G. p. megopotamus),G. maritimus, and G. tropicalis (G. t. tropicalis and G. t. davisi) as distinct species within this clade is supported by Cytb sequence data alone, as depicted in Sudman et al. (2006), but examination of chromosomal data argues against synonymizing G. p. davisi with G. tropicalis. Fourth, all 4 taxa could represent separate species—G. personatus (G. p. personatus and G. p. megopotamus),G. maritimus, G. tropicalis, and G. davisi. Given these complex issues, it is clear that more research is needed within this group. Inclusion of the remaining subspecies of G. personatus (fallax and fuscus), extensive population sampling, and use of multiple samples from each taxon, as suggested by Sudman et al. (2006), are required to better resolve specific relationships.

Geomys breviceps and G. pinetis species groups.—In the independent and combined analyses described here, results support previous findings within the clades containing representatives of G. breviceps and G. pinetis. Genetic distance values from Cytb sequences obtained from Sudman et al. (2006) and herein support the presence of at least 2 phylogroups within each group. Within the breviceps group, samples of G. b. breviceps and G. b. sagittalis differed by an average value of 9.3% (Sudman et al. 2006), exceeding levels seen among other currently recognized species (e.g., G. bursarius differs from G. lutescens by 8.8%). In this study, 2 additional samples of G. b. sagittalis were included, producing a broad range of genetic divergence (4.1-9.3%) in samples from Arkansas, Louisiana, and Texas. Although too few samples are included herein to resolve the question, individuals from southern Louisiana are the most divergent of any in pairwise comparisons. Currently, there are studies underway to further examine the relationships within this group (P. D. Sudman, pers. comm). Within the pinetis group, G. p. pinetis and G. p. mobilensis differed by an average value of 8.10% (Cytb sequences—Sudman et al. 2006), which was similar to levels seen among other species. In addition, Sudman et al. (2006) noted that G. p. mobilensis hosts a different species of mallophagan louse (Geomydoecus mobilensisWilliams and Genoways 1980) and differed in both mtDNA restriction patterns and at 2 allozyme loci (Kennedy 1988) from other forms of G. pinetis, which occur east of the Apalachicola River (including G. p. pinetis and G. p. austrinus). Given the combination of these observations, mobilensis may warrant specific status. Both groups are in need of more extensive population sampling to include all currently recognized subspecies, but this may prove difficult within the pinetis group because several subspecies (colonus, cumberlandius, and fontanelus) are listed as “threatened” or “vulnerable” (Kirkland 1996) in Alabama, Florida, and Georgia.

Results presented herein generally support previous hypotheses in the recognition of 4 species groups within Geomys (bursarius, personatus, breviceps, and pinetis). At least 12 species (arenarius, attwateri, breviceps, bursarius, jugossicularis, knoxjonesi, lutescens, personatus, pinetis, strecken, texensis, and tropicalis) and possibly 14 (sagittalis and mobilensis) should be recognized. Further research should focus on expanded sampling of taxa within each species group. Also, given the limited morphological divergence among taxa (Mauk et al. 1999), future taxonomic implications should be based on genetic data. For example, if the Bateson-Dobzhansky-Muller model of speciation is the primary means of genetic isolation for many mammalian species, as suggested by Baker and Bradley (2006), then genetic isolation often is not accompanied by morphological evolution or even reproductive isolation. Consequently, multiple genetic data sets and evidence from host-specific parasites may be needed to resolve taxonomic questions.

Acknowledgments

We thank the following museums and curators for providing tissue samples: Natural Science Research Laboratory at the Museum of Texas Tech University (R. J. Baker), Louisiana State University Museum of Natural Science (M. S. Hafner), Texas Cooperative Wildlife Collection at Texas A&M University (J. W. Bickham), and University of Nebraska State Museum of Natural History (H. H. Genoways). We thank L. C. Bradley, J. D. Hanson, D. D. Henson, R. N. Piatt, S. B. Westerman, and C. W. Thompson for commenting on earlier versions of this manuscript. We thank N. Ordóñez-Garza for translating the resumen.

Appendix I

Specimens examined.—The specimens examined in this study are listed below by museum acronym (Hafner et al. 1997) or individual collector number. Gene type, museum number or collector identification, and GenBank accession numbers are provided in parentheses. GenBank accession numbers from previous studies are listed in the “Materials and Methods” section. Abbreviations for identification numbers are as follows: Natural Science Research Laboratory, Museum of Texas Tech University (TTU); Museum of Natural Science, Louisiana State University (LSUMZ); Texas Cooperative Wildlife Collection, Texas A&M University (AK); University of Nebraska State Museum of Natural History (UNSM); and Scott K. Davis (SKD). All localities are in the United States unless otherwise specified.

Geomys arenarius.—Texas; El Paso County, 1.6 km S, 0.4 km W Fabens (Rbp3: TTU69208, EU551796).

Geomys attwateri.—Texas; Calhoun County, Guadalupe Delta Wildlife Management Area (Rbp3: TTU75223, EU551794).

Geomys breviceps sagittalis.—Texas; Wood County, 5.6 km SE Quitman (Rbp3: TTU69299, EU551782; Cytb: TTU69299, FJ210793).

Geomys breviceps sagittalis.—Louisiana; Vernon Parish, 3.2 km S, 4.8 km W Rosepine (Rbp3: LSUMZ30723, EU551783; 12S rRNA: LSUMZ30723, EU551799).

Geomys bursarius major.—Texas; Lubbock County, 7.2 km S, 6.4 km E Lubbock (Rbp3: TTU69304, EU551780).

Geomys bursarius majusculus.—Nebraska; Madison County, 2.7 km W Meadow Grove (Rbp3: TTU76067, EU333407).

Geomys knoxjonesi.—Texas; Andrews County, 28.8 km S, 24 km E Andrews (Rbp3: TTU69233, EU551795).

Geomys jugossicularis halli.—Nebraska; Lincoln County, 2.6 km N, 0.8 km E Sutherland (Rbp3: TTU76073, EU333414).

Geomys jugossicularis jugossicularis.—Colorado; Fremont County, 4.8 km S, 6.4 km E Canon City (Rbp3: LSUMZ29284, EU551781; 12S rRNA: LSUMZ29284, EU551800).

Geomys lutescens lutescens.—Nebraska; Lincoln County, 7.2 km N, 4.8 km E Sutherland (Rbp3: UNSM20858, EU333411).

Geomys personatus davisi.—Texas; Zapata County, 4 km N, 6.4 km E San Ygnacio (Rbp3: AK5362, EU551791; 12S rRNA: AK5362, EU551801).

Geomys personatus maritimus.—Texas; Nueces County, Flour Bluff, Graham Rd (Rbp3: SKD176, EU551788; 12S rRNA: SKD176, EU551802).

Geomys personatus megapotamus.—Texas; Jim Hogg County, 0.8 km N Hebbronville (Rbp3: TTU69242, EU551786).

Geomys personatus personatus.—MEXICO: Tamaulipas; La Car-bonera (Rbp3: TTU104950, EU551787).

Geomys pinetis mobilensis.—Florida; Santa Rosa County, 1.3 km N Rte 90 on Rte 87 (Rbp3: LSUMZ29340, EU551789; 12S rRNA: LSUMZ29340, EU551803).

Geomys pinetis pinetis.—Florida; Volusia County, 8 km SE Deland (Rbp3: TTU40793, EU551790).

Geomys strecken.—Texas; Dimmit County, 0.8 km W Carrizo Springs (Rbp3: TTU69251, EU551793; Rbp3: TTU69295, EU551792; 12S rRNA: TTU69295, EU551804).

Geomys texensis bakeri.—Texas; Uvalde County, 20.8 km S Sabinal (Rbp3: TTU69254, EU551785).

Geomys texensis texensis.—Texas; Mason County, 4 km N, 9.6 km E Mason (Rbp3: TTU69277, EU551784).

Geomys tropicalis.—MEXICO: Tamaulipas; 3.5 km S Altamira (Rbp3: TTU44886, EU551797).

Cratogeomys castanops.—Texas; Terry County, 4.2 km W Brown-field (Rbp3: TTU69307, EU551778).

Pappogeomys bulled.—MEXICO: Jalisco; 5 km S Chamela, adjacent to N side of Chamela Station (Rbp3: TTU45109, EU551779; 12S rRNA: TTU45109, EU551798).

Footnotes

  • Associate Editor was Carey W. Krajewski.

Literature Cited

View Abstract