Phylogeny and temporal diversification of Calomys (Rodentia, Sigmodontinae): Implications for the biogeography of an endemic genus of the open/dry biomes of South America

Molecular Phylogenetics and Evolution 42 (2007) 449–466www.elsevier.com/locate/ympev

Phylogeny and temporal diversiWcation of Calomys (Rodentia, Sigmodontinae): Implications for the biogeography of an endemic genus of the open/dry biomes of South America

Francisca C. Almeida a,b,¤, Cibele R. Bonvicino c,d, Pedro Cordeiro-Estrela e

a Institute for Comparative Genomics, American Museum of Natural History, New York, NY, USAb Department of Biology, New York University, New York, NY, USA

c Division of Genetics, Instituto Nacional de Câncer, Rio de Janeiro, RJ, Brazild Department of Tropical Medicine, Instituto Oswaldo Cruz, Rio de Janeiro, RJ, Brazil

e Département Systématique et Evolution, UMR 5202 CNRS/USM 601 Origine, Structure et Evolution de la Biodiversité, Muséum National d’Histoire Naturelle, Paris, France

Received 22 March 2006; revised 21 June 2006; accepted 3 July 2006Available online 26 July 2006

Abstract

A thorough analysis of character evolution and biogeography of a group is only possible with a comprehensive sampling of its diver-sity. The sigmodontine genus Calomys is particularly interesting for the study of neotropical biogeography as it occurs exclusively in thedry and grassland biomes: Cerrado, Caatinga, Chaco, Pampas, Venezuelan Llanos, Puna, and a diversity of dry forests biomes. AlthoughBrazil encompasses a large part of the geographic distribution of the genus and at least three endemic species, the last published phylog-eny of Calomys included only two specimens (both representing the same species) from a single locality in this country. In the presentpaper we add complete cytochrome b sequences of Brazilian karyomorphs to previously published sequences in order to provide a phylo-genetic hypothesis including most of the genus diversity. The main objectives of this study were to clarify taxonomic issues related toBrazilian karyomorphs and to study the diversiWcation processes of the genus by analyzing its biogeography in combination with clado-genesis dates estimated with a molecular clock. The phylogeny indicates that at least six diVerent species inhabit the Brazilian territory,one of them still undescribed. Date estimates indicate that two sequential basal splits, the Wrst separating Andean from lowland speciesand the second isolating species north and south of the Amazonia, took place in the Pliocene between 3 and 4 Ma. A large-bodied andspeciose clade of lowland species associated with dry forests and ecotones of the core Cerrado and Chaco areas with adjacent biomesdiversiWed in the Pleistocene. This indicates the importance of safeguarding ecotones and the Cerrado at a time where this biome is beingrapidly destroyed.© 2006 Elsevier Inc. All rights reserved.

Keywords: Calomys; Sigmodontinae; Phyllotini; Molecular phylogeny; Cytochrome b; Molecular clock; Biogeography; Cerrado; Caatinga; South America

1. Introduction

The genus Calomys Waterhouse 1837 comprises smallbody-sized sigmodontine rodents belonging to the tribePhyllotini. Species of this genus inhabit dry vegetationareas and have a wide distribution in South America,

* Corresponding author. Fax: +1 212 769 5277.E-mail address: [email protected] (F.C. Almeida).

1055-7903/$ - see front matter © 2006 Elsevier Inc. All rights reserved.doi:10.1016/j.ympev.2006.07.005

occurring from the Venezuelan Llanos to the southern limitof the Argentinean Chaco, including the diagonal of drybiomes in Brazil, that encompasses the Cerrado (complexesof savanna shrublands and grasslands) and Caatinga(shrublands and dry forest) morphoclimatic domains. Mostof the species occur south of the Amazon, except for Calo-mys hummelincki that occurs in Venezuela, Colombia, andsome nearby islands (Musser and Carleton, 2005). Calomysspecies are rather prevalent in some areas of their distribu-tion and some of them have been associated with emerging

450 F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466

diseases, such as the viral hemorrhagic fever (Mills et al.,1991, 1992, 1994; Salazar-Bravo et al., 2002; Videla et al.,1989) and the Hantavirus pulmonary syndrome (Carrollet al., 2005; Levis et al., 2004; Williams et al., 1997; Yahnkeet al., 2001).

The taxonomy of the genus is complicated by theremarkable morphological similarity shared by some of thespecies. Based exclusively on morphology, Hershkovitz(1962) recognized only four Calomys species (Calomys lepi-dus, Calomys sorellus, Calomys callosus and Calomys lau-cha), while all other previously described morphotypes wereconsidered junior synonyms or subspecies. These four spe-cies represent the main species groups currently recognized:highland–C. lepidus and C. sorellus; lowland, large-bodied–C. callosus; and lowland, small-bodied–C. laucha (Salazar-Bravo et al., 2001). Later morphological studies combinedwith karyotypic analyses provided evidence for the revali-dation of old names, most of them listed under C. lauchaand C. callosus by Hershkovitz (1962), as well as the identi-Wcation of new forms (Bonvicino et al., 2003; Bonvicinoand Almeida, 2000; Corti et al., 1987; Lisanti et al., 1976;Olds, 1988; Pearson and Patton, 1976; Pérez-Zapata et al.,1987; Reig, 1986; Vitullo et al., 1984). Karyotypic analyseshave been most useful in the taxonomy of Calomys; in con-trast with morphology, chromosome number and formhave a great variation within this genus. Continual descrip-tion of new forms and the lack of a thorough revision of thegenus including diVerent classes of characters preclude areliable count of species (Olds, 1988; Musser and Carleton,1993, 2005; Salazar-Bravo et al., 2001; Table 1).

The second edition of the checklist Mammal Species ofthe World, published in 1993, indicates the occurrence ofthree Calomys species in Brazil: C. callosus, Calomys tener,and C. laucha (Musser and Carleton, 1993). Calomys laucha(2ND 64, FND68) is reported to occur in western-centralBrazil by Musser and Carleton (1993), although recent

records are restricted to the Pampas in southernmost Brazil(Mattevi et al., 2005). “Calomys callosus” represented allthe large-bodied Calomys species collected in most Brazil-ian open vegetation areas, while C. tener was recognized asthe small-bodied species from southeastern and centralBrazil. Since the publication of that checklist, the increaseduse of cytogenetics for species identiWcation revealed thatmany diVerent karyomorphs were actually present in thesamples of large-bodied Calomys, although they sharedvery similar morphologies. Based on morphometric andkaryotypic evidences–a chromosome complement of2ND66 instead of 2ND 36 or 2ND 50 as proposed forC. callosus from Paraguay–the most common Brazilianlarge-bodied form was recognized as a distinct species andthe name Calomys expulsus was revalidated (Bonvicino andAlmeida, 2000). Another form collected in central Brazil,with a chromosome complement of 2ND 46 and FND66,was described as Calomys tocantinsi (Bonvicino et al.,2003). The new edition of the Mammal Species of theWorld published in 2005 includes both C. expulsus and C.tocantinsi (Musser and Carleton, 2005). A third kar-yomorph with 2ND36, FND66 (here referred as Calomyssp. nov.) has been collected in two localities of MinasGerais state, but remains undescribed (Geise et al., 1996;CRB and R. Gentille, personal communication). In addi-tion to these novel karyomorphs, specimens with a karyo-type ascribed to Calomys callosus (2ND 50, FND66) havebeen recently collected in the states of Mato Grosso andMato Grosso do Sul, in the vicinities of Paraguay (CRBand J. A. Oliveira, personal communication).

The monophyly of Calomys was contended by Engelet al. (1998) in a study on sigmodontine phylogeny basedon sequence data of 4 mitochondrial genes. All othermolecular evidence, however, point to Calomys as being amonophyletic genus (Salazar-Bravo et al., 2001; Smith andPatton, 1999) and the results of Engel et al. (1998) have

Table 1Species recognized by Olds (1988) in the last revision of the genus and Musser and Carleton (2005) in the last mammal species list

The table also shows the species included in the study by Salazar-Bravo et al. (2001) and in the current study. Asterisks mark species for which new molec-ular data was obtained in the current study.

Olds (1988) Musser and Carleton (2005) Salazar-Bravo et al. (2001) This study

C. sorellus C. sorellus C. sorellus C. sorellusC. lepidus C. lepidus C. lepidus C. lepidusC. musculinus C. musculinus C. musculinus C. musculinusC. hummelincki C. hummelincki C. hummelincki C. hummelickiC. murillusC. tener C. tener C. tener C. tener*

C. laucha C. laucha C. laucha C. lauchaC. bimaculatusC. callosus C. callosus C. callosus C. callosus*

C. venustus C. venustus C. venustus C. venustusC. fecundus C. fecundusCalomys sp. (from Beni) Calomys sp. (from Beni)

C. callidusC. boliviaeC. expulsus C. expulsus*

C. tocantinsi C. tocantinsi*

Calomys sp nov. (from Minas Gerais)*

F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466 451

been interpreted as a case of sample misidentiWcation (Rue-das et al., 2000). The Wrst molecular phylogeny of the genusthat included a signiWcant number of taxa was proposed bySalazar-Bravo et al. (2001), based on sequences of the mito-chondrial gene cytochrome b (Table 1). Interestingly, theirresults contradicted the previously hypothesized relation-ships for the genus based on both morphology (Olds, 1988;Steppan, 1993) and karyology (Espinosa et al., 1997). Themain features of the molecular phylogeny were (1) the exis-tence of two major monophyletic groups, one comprised bylowland forms and the other comprised by species thatinhabit mostly high to mid altitude areas, and (2) the mono-phyly of a strongly supported clade containing all the large-bodied species (Salazar-Bravo et al., 2001). Only twospecimens from Brazil were included in this study, both rep-resenting C. tener and collected in the same locality in SãoPaulo state.

Calomys presents a very interesting distribution pattern,but its biogeographic history is controversial. Reig (1986)proposed that the genus originated in the Andes with pos-terior invasion of lowland habitats, as suggested by the highdiversity of Phyllotini rodents in this cordillera. Alterna-tively, Braun (1993) proposed, based on Marshall’s hypoth-esis (Marshall, 1978), that the genus originated in thenorthern part of its range in Venezuela and diversiWed as itexpanded its distribution southward. According to Mar-shall (1978), a corridor of open vegetation appeared on theeastern slopes of the Andes at around 3.5 million years agoallowing the southward migration of savanna/grasslandsigmodontine rodents that inhabited northern South Amer-ica. Salazar-Bravo et al. (2001) suggested that the genusappeared directly south of Amazonia and at about 8.5 – 9Ma split into two branches: one that originated the high-mid altitude clade and another that gave rise to Calomyshummelincki, through long range dispersal, and the ances-tor of all other lowland species. These authors also linkedthe evolution of the genus to the spread of the C4 grasses inSouth America. No major inferences, however, have beenmade about the origins of the remaining lowland species.

Little is known about the historical biogeography of thedry habitats and grasslands of South America (Cerrado,Pampas, Chaco, Llanos, Caatinga, etc). Despite the highbiodiversity found in some of these habitats, they have beenmuch less studied than the South American rainforests.One of the possible reasons for that is the comparativelylow endemicity, especially in taxonomical levels higher thanspecies. For instance, in the Cerrado, the largest continuoussavanna in South America, many of the endemic specieshave their closest relatives in the neighboring Atlantic andAmazonian rainforests (Costa, 2003; Porzecanski and Cra-craft, 2005). The genus Calomys is one of the few sigmodon-tine genera restricted to non-humid habitats and for thisreason constitutes a remarkable subject for the study ofbiogeography. Besides, Calomys presents an interestingcombination of distribution area and phylogenetic struc-ture that allows for testing of biogeographic scenariosthrough the use of molecular clock. The existence of fossil

records for the genus from several areas of its distributionis a plus for this kind of analysis, since it provides con-straints for molecular dating (Pardiñas et al., 2002). Twomain questions are of interest. When did the two basal splitof the genus-highland – lowland and north–south ofAmazonia – date? Do they correspond to any main geo-logic or climatic event as could be thought by invoking anallopatric mode of speciation? The biogeographic historyof a genus can only be discussed with a thorough samplingof the species and the inclusion of samples from its entiregeographic range. The inclusion of species from Brazil is,therefore, essential to the understanding of Calomys bioge-ography. In fact, more than half of the distribution area ofthe genus lies within the Brazilian territory.

The main goal of this study was to infer phylogeneticrelationships among Calomys species including most of itsknown diversity (Table 1). We used complete cytochrome bsequence data, which allowed us to combine sequences pre-viously obtained by Salazar-Bravo et al. (2001) withsequences obtained for 43 specimens representing Wve dis-tinct karyomorphs collected in a wide area in Brazil. Thephylogenetic hypothesis generated in this study was used todiscuss several aspects of the evolution of Calomys. First,the inclusion of more taxa provided a better resolution ofthe species relationships, helping to clarify taxonomic issuesregarding recently identiWed Brazilian karyomorphs. Sec-ond, the combination of our large geographic and speciessamplings, GIS vegetation maps of South America, anddivergence date estimates allowed us to discuss diversiWca-tion scenarios for Calomys in the understudied open and/ordry lowland habitats of South America.

2. Materials and methods

2.1. Sampling and karyotypes

Liver tissue samples were obtained for 43 Calomys speci-mens collected in several localities in Mato Grosso, MatoGrosso do Sul, São Paulo, Minas Gerais, Bahia, Goiás,Tocantins, and Piauí states, Brazil (Appendix A). Skins andskulls were or will be deposited at the mammal collection ofthe Museu Nacional (MN-UFRJ, Rio de Janeiro, Brazil).All collected animals were karyotyped. Throughout the textwe refer to specimens with karyotype 2ND 36/FND 66 asCalomys sp. nov. Chromosome preparations were obtainedfrom bone marrow culture in RPMI 1640 with 20% fetalcalf serum, ethidium bromide (5�g/ml) and colchicine(10¡6 M) for two hours.

Sequences of non-Brazilian Calomys specimens wereobtained from the GenBank and are listed in the AppendixA. For most of these species, we included a maximum of 3sequences, with the exception of C. callosus. A larger sam-pling of this species was important for study of closelyrelated forms collected in Brazil, especially those inhabitinglocalities nearby Paraguay. Sequences obtained from threeother Phyllotini species (Eligmodontia typus, Auliscomysmicropus and Phyllotis magister), Akodon montensis, and

452 F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466

Sigmodon hispidus were used as outgroups (Smith and Pat-ton, 1999).

2.2. Molecular techniques

Genomic DNA was isolated from tissue samples pre-served in ethanol or liquid nitrogen following Smith et al.(1987) or using the DNeasy extraction kit (Qiagen). A frag-ment containing the full-length cytochrome b gene(1143 bp) was ampliWed with the primers Mus14095 andMus15398 (Anderson and Yates, 2000), using standardPCR procedures. After puriWcation, PCR products weredirectly sequenced with the same primers used in the PCRampliWcation and additional internal primers designed forthis study: Cytb.HS2 (5� -GTGGTT TAATRTGTGCTGGAGTATTG-3�), Cytb.LS2 (5�- CAGATGTGYTMGGAGACCCCGA TAA-3�) and Cytb.LS3 (5� GGCCTAACTCCGATTCAGACAA-3�). Nucleotide sequences were deter-mined using automated sequencers ABI Prism™ 377 andABI 3730XL. At least four readings (with the two externaland two inversely oriented internal primers) were obtainedfor each specimen, assuring maximum quality of the Wnalsequences.

2.3. Phylogenetic analyses

Sequences were edited and aligned manually usingSequencher 4.2 (Gene Codes Corporation). In all phyloge-netic analyses a single specimen per haplotype per localitywas included.

Maximum parsimony (MP) analyses were carried outusing PAUP* version 4.10b (SwoVord, 1993). MP treeswere obtained using heuristic search with 10 random step-wise additions and tree-bissection-reconnection (TBR)branch swapping. Bootstrap values were obtained with1000 pseudoreplicates and the same searching methodsused for the trees. Bremer support values were calculatedwith PAUP* using the commands ENFORCE and CON-VERSE that constrain searches for the shortest trees thatdo not contain a speciWed clade. The Bremer value for theclade was calculated by subtracting the length of the mostparsimonious tree from the length of the constrained tree.This procedure was done for each clade obtained in theconsensus tree.

Maximum likelihood (ML) trees were obtained with thePHYML v2.4 software (Guindon and Gascuel, 2003). Themodel of sequence evolution for the ML analysis was esti-mated using Modeltest v3.6 (Posada and Crandall, 1998).Topology robustness was assessed through nonparametricbootstrap over 1000 replicates. The choice of the model ofsequence evolution for the Bayesian inference (BI) analysiswas made using MrModeltest version 2.2 (perl script byJohan A. A. Nylander, 2004). Tree search by Bayesian infer-ence was made using MrBayes v3.0B4 (Huelsenbeck andRonquist, 2001). Multiple runs were made with defaultsearch parameters and with generation numbers up to5,000,000 and with swap frequencyD 4 and NchainsD 8.

2.3.1. Divergence dates estimatesThe molecular clock hypothesis was tested using the

likelihood ratio test (Huelsenbeck and Rannala, 1997).Molecular dating was performed using the r8s version 1.50software (Sanderson, 2003), which implements three diVer-ent methods for estimating divergence times. The Wrst(LF—Langley-Fitch method; Langley and Fitch, 1974) isthe classical molecular clock (Zuckerkandl and Pauling,1962) that enforces a tree in which branch lengths are pro-portional to time; i.e. it accounts for only one constant rateof evolution across the tree. The second is the non-paramet-ric rate smoothing method (NPRS; Sanderson, 1997),which incorporates a penalty function for branches that donot conform to the molecular clock, allowing a relaxationof this assumption. Since variation of rates across lineagesis non-independent, this constraint smoothes by a least-squares procedure rate variation from ancestral to descen-dant lineages to penalize rapid rate change. The third is thepenalized likelihood method (PL; Sanderson, 2002), whichestimates rates across branches through the maximizationof a likelihood estimate penalized by the product of asmoothing parameter and a roughness penalty. Thismethod allows the examination of a range of solutions withdiVerent smoothing levels. For this reason, it can be appliedto any kind of data: from clock-like to highly variable rates.

Since the choice of a calibration point is a priori choice ofbiogeographical scenario and the fossil record for sigm-odontines is deWcient, we decided to use a conservativeapproach allowing for a range of basal node dating.Although this approach gave us time periods instead of spe-ciWc date points, it allowed us to test hypotheses and placecladogenesis in general time periods, without a subjectiveexclusion of any possible scenario. For all three methods, wechose to Wx the basal node in the divergence between Sigm-odon and the Oryzomayalia (Steppan et al., 2004). The old-est fossil data referent to this node is a specimen ofProsigmodon oroscoi, from Mexico, dated in approximately5 million years ago (Ma). This fossil, together with theevidence of a Phyllotini fossil from the Montehermosan(4–5 Ma; Reig, 1978), suggests a minimum age for Sigm-odon-Oryzomyalia split of 5 million years. This divergence,however, has been dated by molecular clock in between 13and 12.1 Ma, as calibrated on the appearance of the Wrstmodern murine (Steppan et al., 2004). Smith and Patton(1999) obtained similar estimates with the molecular clock,ranging from 10 to 14 Ma. Since the fossil data can only givethe minimum date of a clade, we Wxed the basal node in allvalues from 5 to 13 Ma with a 0.5 Ma step between each con-secutive value. This choice, nevertheless, excludes a scenarioof an explosive radiation of sigmodontinae after a late arrivalin South America subsequent to the closure of the Panama-nian Isthmus at around 3 Ma (Simpson, 1969; Webb, 1978).

Restriction of minimum age of some nodes was carriedout to improve the accuracy of node age estimates. Mini-mum ages were set using the following fossil and geochro-nology information from Pardiñas et al. (2002) and Cioneand Tonni (2001). C. expulsus and C. tener are found in

F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466 453

Brazil since the Lujanian (0.3–0.2 Ma). Phylotinii genera arefound since the early chapadmalalan (4–3.5 Ma). In thisway, two minimum age restrictions were set, the Wrst onefor the cladogenesis of Calomys expulsus at 0.2 Ma and thesecond for the divergence between the phyllotini andakodontini tribes at 3.5 Ma.

3. Results

3.1. Parsimony analysis

Complete cytochrome b sequences (1143 bp) wereobtained for 43 specimens collected in 17 localities. A com-parison of nucleotide data of the complete dataset showed524 variable sites, 405 of which were informative for parsi-mony analysis. Maximum parsimony (MP) analysis of thecomplete dataset resulted in 83 most parsimonious treeswith 1705 steps with consistency index of 0.445 and rescal-oned consistency index of 0.363 (Fig. 1). The genus Calomysappeared as a monophyletic clade. Relationships amongspecies were mostly resolved. All the unresolved clades arewithin species and in one case involves two species, C. callo-sus and C. tocantinsi. Specimens of the same species alwaysclustered together with bootstrap values above 90%, withthe exception of the two specimens of C. venustus.

A comparison of this tree with the MP tree obtained bySalazar-Bravo et al. (2001) shows a general correspondence,but a few diVerences in species relationships. A major simi-larity between the two trees is a clade containing the low-land large-bodied species. In our analysis, this clade alsoincludes C. expulsus, C. tocantinsi, and Calomys sp. nov. asexpected based on their distribution and body-size. Theclade containing the highland species was also recovered inour analysis. Nevertheless, our tree did not support a sisterrelationship between C. tener and C. laucha. These resultsare not totally incongruent, since this clade did not receivea signiWcant bootstrap value in the MP analysis and wasnot recovered in the ML analysis done by Salazar-Bravoet al. (2001). In the large-bodied clade, our analysis recov-ered a clade including C. venustus, C. fecundus, and Calo-mys sp. from Beni (bootstrapD83%; bremerD4), while intheir results the latter two species form a clade sister to C.callosus (bootstrapD64%; bremerD5). The placement ofC. hummelincki is also contentious. Whereas Salazar-Bravoet al. (2001) recovered it as a sister group to all the otherlowland species, our analysis recovered it as the most basalspecies of the genus, although this relationship had no sta-tistical support. All the other relationships between speciesthat were included in both studies were congruent.

The most diverse group is the clade of lowland large-bodied species. Calomys expulsus appeared as the mostbasal and diVerentiated species in the group. Calomys sp.nov. is the next species to branch oV the clade. Its close rela-tionship with the remaining species of this group is highlysupported. The remaining species are divided in twogroups. One group includes C. fecundus, Calomys sp. fromBeni, and C. venustus, with evidences that the former two

are sister species. The other group is formed by specimensassigned to C. tocantinsi and C. callosus. Calomys tocantinsiwas deeply nested in this clade, although the two haplo-types representing this species clustered together with highstatistical support.

3.2. Maximum likelihood and Bayesian inference

The sequence evolution model GTR + I + G was selectedfor both the ML and the BI analyses, according to theresults of ModelTest and MrModelTest, respectively. MLand BI searches obtained the following parameters, respec-tively: base frequencies: AD0.3104/0.3163, CD0.2506/0.315, GD0.1262/0.0988, TD0.3128/0.2698; substitutionrate matrix: [A–C]D 2.606/1.033, [A–G]D6.986/6.140, [A–T]D 1.651/1.382, [C–G]D0.577/0.471, [C–T]D13.378/10.415, [G–T]D1/1; proportion of invariable sitesD 0.472/0.436; gamma distribution parameterD 1.197/1.1597. TheML tree, shown in Fig. 2 had log likelihood of 8929.525.The Bayesian analysis produced 49,016 trees, for which weobtained a majority rule consensus with a very similartopology to that obtained with ML (Fig. 2). Besides minordiVerences in intraspeciWc relationships between the ML/BIand the MP results, two main diVerences can be noticed.The Wrst is the position of C. hummelincki that appears as asister group to all other lowland species (MLbootstrapD 77) as found by Salazar-Bravo et al. (2001), butnot in our the MP analysis. The second is the monophyly ofC. tocantinsi, obtained only in the ML and BI analyses. Inboth conXicting cases, however, the results of the MP anal-ysis did not receive statistical support. An important con-gruent result of the three methods used here is thesequential branching oV of C. tener, C. laucha, and the“large-bodied” clade.

3.3. Intra- and interspeciWc variation in Brazilian species

Calomys expulsus, represented by 13 specimens from 8localities, had 8 diVerent haplotypes with all but 2 substitu-tions in degenerate sites. The C. expulsus clade was recov-ered with high support in all analyses. No obviousintraspeciWc structure can be observed, but it is interestingto note that the three haplotypes from Caetité (Bahia State)were recovered as the most basal in all the analyses, fol-lowed by a specimen from João Costa (Piauí State). Thesetwo localities are in the Caatinga ecoregion, while all theremaining samples (from 5 diVerent localities in the statesof Bahia and Goiás) were collected in the Cerrado and clus-tered in a highly supported clade with no internal resolu-tion (Figs. 1 and 2, see also map in Fig. 3).

Among our 12 specimens with a 2ND50 karyotype, 4diVerent haplotypes were obtained. Aquidauana (3 sam-ples) and Pantanal do Poconé (2 samples) had one haplo-type each. Corumbá (7 samples) had 2 haplotypes thatdiVered by only one transition in a third position; one hap-lotype was present in 6 individuals, while the other haplo-type was found in only one individual. All 4 haplotypes are

454 F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466

Fig. 1. Strict consensus of 83 most-parsimonious trees (tree length D 1705; CI D 0.445; RI D 0.363). No weighting scheme was used. Bootstrap values above50 are included above branches. Numbers below branches refer to Bremer decay indexes.

F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466 455

very similar to those obtained from the GenBank andassigned to C. callosus. Nevertheless, the GenBank sampleswere much more variable, with 12 haplotypes found in 12specimens from 8 localities. The C. callosus sequences fromthe GenBank clustered with samples from Corumbá, Aqui-dauana, and Pantanal do Poconé in all the analyses. Thisprovides the Wrst evidence that C. callosus is also found inBrazil. The 24 C. callosus samples diVered by 40 substitu-tions, 14 of them in non-degenerate sites (35%). This isquite high if compared to the percentage of substitution in

non-degenerate sites in C. tener (11%) and C. expulsus(4.4%). This result may indicate a high genetic variation inthe clade but this conclusion is highly dependable upon thequality of the sequences.

Calomys tener presented an interesting pattern of intra-speciWc geographic structure that is consistent among thediVerent analyses. A clade containing samples from SãoPaulo state was recovered in all analyses and was clearlyseparated from the remaining specimens. This clade is com-posed by 5 specimens with unique haplotypes collected in

Fig. 2. Maximum likelihood and Bayesian trees topology with ML bootstrap values above and BI posterior probabilities below braches. Branch lengthsare from the ML analysis.

456 F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466

three localities (Fig. 4). The remaining C. tener sequencesclustered in a separate clade that was recovered only in theMP analysis with low statistical support. These specimensare from several diVerent localities in the Brazilian states ofGoiás and Bahia, plus one specimen from Bolívia. The twogroups occur in ecologically distinct areas; while in SãoPaulo state C. tener occurs in the ecotone area between theAtlantic Forest and Cerrado, the remaining specimens arefrom strict Cerrado areas. The sequence alignment revealedthat the sequences obtained by Salazar-Bravo et al. (2001)for C. tener specimens had very diVerent nucleotides in the

Wrst 20 positions as compared to C. tener sequencesobtained in the present study. Since the former sequencesare clearly clustered as expected based on the geographicorigin (see below), their beginning is dubious probably as aresult of low quality sequences. In an MP analysis exclud-ing the specimens sequenced by Salazar-Bravo et al. (2001),both C. tener groups were recovered, and the bootstrap val-ues increased from 71 to 91 for the group from São Pauloand from 58 to 61 for the other group (results not shown).Bremer support was also increased for the “ecotone group”from 1 to 3. A partial cytochrome b sequence from a speci-

Fig. 3. Sample localities of large-bodied species of Calomys. Numbered localities refer to samples used in this study and listed in the Appendix A. Theremaining data points represent sample localities of karyotyped specimens obtained in the literature (see text for references).

Andean YungasArgentine ChacoArid ChacoBeni savannasBolivian montane dry forestsCaatingaCerradoChaco savannasChiquitania dry forestsCordoba montane savannasNortheast Brazil dry forestsUruguay savannas and PampasPantanalHumid Chaco

C. callidusC. callosusC. expulsusC. fecundusC. tocantinsiC. venustusCalomys sp.Calomys sp. nov.

F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466 457

men collected in the Mato Grosso do Sul state, Brazil (seeFig. 4), in a locality close to an ecotone area clustered withthe specimens from São Paulo (result not shown).

3.4. Molecular dating

The molecular clock hypothesis was rejected by the like-lihood ratio test (DD 250.54; dfD 61; p < 0.001). Althoughthe gene is not evolving at a clocklike rate, we show the esti-mates obtained with the LF method for comparisons withthe other two methods (NPRS and PL) that do not assumethe clock. Using these three methods, we dated the follow-ing nodes referred to by the letter in parentheses: diver-gence between Akodontini and Phyllotini (A), betweenPhyllotini and Calomys (B), between highland and lowlandclades (C), Amazonian split between C. hummelincki andthe rest of the lowland clade (D), divergence of C. tenerfrom C. laucha and the large body sized Calomys (E), of C.sorellus from the C. musculinus-C. lepidus clade (F), diver-gence of the “large-bodied” clade (G), divergence of C.expulsus (H), divergence of C. musculinus from C. lepidus(I), divergence of Calomys sp. nov. (J), divergence of theC.callosus-C. tocantinsi clade from the C. fecundus-C. venu-stus clade (K), divergence between C. fecundus from C.

venustus (L), and the divergence of C. callosus from C.tocantinsi (M). The range of divergence estimates obtainedwith the three methods employed here are shown in boxplots in Fig. 5. Comparisons between the results obtainedwith each method show that NPRS has the oldest estimateswhile PL has the most recent estimates (Fig. 5, see alsoWgure in the Supplementary data). LF estimates fellbetween the other two, although they are in general moresimilar to those obtained with the PL method.

Observing the box plots of the divergence date estimates(Fig. 5), three categories of cladogenetic events can be dis-criminated. First, nodes that could not be assigned to onespeciWc geological epoch, i.e. nodes A, B, H, I. Second,nodes that can be assigned to a speciWc period when thethree methods are congruent on at least three quartiles ofthe distribution i.e. nodes E, F and G assigned to the Plio-cene and nodes K, L, and M assigned to the Pleistocene.Third, nodes that can be assigned to a speciWc geologicalepoch under the second criteria upon discussion of theaccuracy of the NPRS method, i.e. nodes C and D assignedto Pliocene and node J to Pleistocene. The NPRS method inknown to overWt the data leading to rapid rate Xuctuationsin short branches (Sanderson, 2003). In fact, the diVerenceof divergence dates between the NPRS method and the FL

Fig. 4. Sample localities of C. tener. Numbers refer to the localities listed in the Appendix A. “a” indicates the locality of Maracaju, Mato Grosso do SulState, Brazil, the collection site of a sample that clusters with the “ecotone” clade. This sample was not included in the trees shown here because only a par-tial sequence of its cytochrome b gene was available.

Brazil

Bolivia

Paraguay

Flooded grasslandsTropical and subtropical moist broadleaf forestsDeserts and xeric shrub lands; temperate, tropical, andsubtropical grasslands, savannas, shrublands, and dry leafbroadleaf forests

458 F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466

and PL methods increases towards younger nodes. TheNPRS method also presented the largest conWdence inter-vals. As Sanderson (2002) has already stressed, the NPRSmethod produces local rate variances that are higher thanthose found at optimal smoothing levels, as the onesobtained with the PL method for instance. In a generalmanner, based on cross-validation scores, he concludedthat when the data depart from a constant rate PL outper-forms NPRS. For these reasons, we decided to exclude thedate estimates obtained with the NPRS method from fur-ther discussion.

4. Discussion

4.1. Phylogenetic relationships in the genus Calomys

Our results conWrmed several relationships recoveredby Salazar-Bravo et al. (2001) and resolved some inconsis-tencies between their MP and ML analyses. For instance,a sister taxa relationship between C. tener and C. laucha,obtained in their MP analysis was never recovered in ourtrees. This diVerence in the results was probably due to theabsence of C. expulsus samples in the Salazar-Bravo et al.(2001) study. In our analyses, C. laucha appears as a sistertaxon to the “large-bodied” group, which is a quite sur-prising result, since C. laucha is the smallest species of thegenus and has never been associated with the “large-bod-ied” clade. Our analyses also gave support to the cladeincluding C. fecundus, Calomys sp. (from Beni), and C.

venustus, that appeared only in the ML analysis done bySalazar-Bravo et al. (2001). Nevertheless, the two studiesagree on the monophyly of the highland and the large-bodied species groups, which were recovered in all theanalyses done so far with high statistical support. Anotherconsensus is on the derived position of the “large-bodied”clade that presents the highest diversity in species andkaryotypes.

The position of C. hummelincki was one of the main dis-crepancies between our diVerent analyses. While the MPtree shows C. hummelincki as the most basal in the genus,the ML and BI trees show this species as the sister taxon toall lowland species. The MP results, however, did notreceive statistical support. These results suggest that thesplit between the highland clade, C. hummelincki, and theremaining lowland species occurred over a relatively shortperiod of time, very early in the history of the genus. Sincethe grouping of C. hummelincki with the other lowland spe-cies is the only alternative that received statistical support(>95% posterior probability in the BI) in at least one of ouranalyses, we prefer this hypothesis to the one that showsthis species as basal to all other Calomys species. In thisway, the genus Calomys can be described as having threemain clades: the highland (C. sorellus, C. lepidus, and C.musculinus), lowland (all the remaining species), and large-bodied (C. expulsus; Calomys sp. nov., from Minas Gerais;C. fecundus; C. venustus; Calomys sp., from Beni; C. tocant-insi, and C. callosus) species groups, with the last one con-tained in the lowland clade.

Fig. 5. Box plot of divergence dates for each node for methods NPRS, PL, and LF. Boxes are divided in quartiles. Vertical lines are the limits between geo-logical epochs: solid line for the Miocene–Pliocene and dashed line Pliocene–Pleistocene boundaries, respectively.

F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466 459

4.2. Taxonomic considerations on Brazilian species

The morphological similarity of the species within thelarge-bodied Calomys species group is remarkable. Hers-hkovitz (1962) was the Wrst to propose a close relationshipbetween the species in this group, also characterized bywell-developed supra-orbital ridges. This author consideredall the diVerent large-bodied morphotypes that had beendescribed so far to belong to a single species with two sub-species: C. callosus callosus and C. callosus expulsus. Theformer included what we now recognize as C. venustus, C.fecundus, and C. callidus. Our results showed the large-bod-ied Calomys species as the most derived group within Calo-mys conWrming Hershkovitz’s (1962) postulation based onmorphological criteria. Until a few years ago all large-bod-ied Calomys specimens collected in Brazil in ecologicalstudies were identiWed as C. callosus (e.g., Alho and Pereira,1985; Lacher et al., 1989). Only recently, with the use ofcytogenetics did the existence of several diVerent morphok-aryotypes become clear. The most widespread karyotype isthe 2ND66, FND 68 assigned to Calomys expulsus (Bonvi-cino and Almeida, 2000). The other karyomorphs recentlyfound, 2ND 36/FND66; 2ND46/FND66, 2ND 50/FND66, are far less common and seem to be restricted tosmaller areas, although more collection data would be nec-essary to delimit their distributions (Fig. 3)

The morpho-karyotype 2ND 36, FND 66, herein desig-nated as Calomys sp. nov., was Wrst described by Geise et al.(1996) for specimens collected in Lagoa Santa, MinasGerais state, Brazil, but its taxonomic status was not speci-Wed by these authors. These specimens (CEG40 andCEG41) were never deposited in a museum collection andtherefore are not available for examination. Recently, spec-imens with the same karyotype have been collected inCapitão Andrade, another locality in Minas Gerais(LBCE5556, LBCE5579, LBCE5583). Although morpho-logically very similar to C. expulsus, the karyotype and thecytochrome b sequences clearly suggest that these speci-mens represent a distinct species. Calomys sp. nov. is knownfrom only these two localities and seems to diVer from C.expulsus for occurring in regions with original forest vege-tation or ecotone, while C. expulsus has only been found inthe Cerrado and Caatinga. Since Lagoa Santa is the (desig-nated) type locality of C. expulsus and C. tener, it is possiblethat Calomys sp. nov. was actually described under one ofthese names by Lund (1841). Holotype examination is cur-rently in process to clarify the taxonomy of these two mor-pho-karyotypes and the results will be published elsewhere.

The other karyomorph recently described in Brazil(2ND46/FND 66) was considered distinct from C. callosusbased on diVerences in the karyotype and morphometry,and named C. tocantinsi (Bonvicino et al., 2003). The MPtree showed a C. tocantinsi clade deeply nested within C.callosus that appeared as a paraphyletic species. This resultwas obtained only in this analysis; in the ML and BI trees,both species appeared as separate monophyletic clades. Theclose relationship between the two species, however, is evi-

dent and reXects in the low estimates of divergence time.The distributions of C. tocantinsi and C. callosus seem to bedisjunct, although additional sampling is necessary to con-Wrm it. The result of the MP analysis does not invalidate thespeciWc status of C. tocantinsi since its recent split couldhave prevented complete lineage sorting of cytochrome bhaplotypes.

Among C. tener, specimens from Campinas, Pedreira,and Tupi Paulista (São Paulo State, Brazil) clustered in aclade apart from the specimens from Jaborandi, Mimoso,Corumbá (Brazil), and Santa Cruz de la Roca (Bolivia).The latter clade, however, was recovered only in the MPanalyses and did not have statistical support. P-Distanceestimates also suggest the existence of two groups sincewithin-group p-distance estimates (0.001–0.008) did notoverlap with between-group estimates (0.010–0.013). This isan interesting result since the two groups are distributed indiVerent ecological regions, the Wrst in an ecotone region ofCerrado and Atlantic Forest, and the second in strict Cer-rado areas. Although this may suggest an incipient splittingin C. tener, more samples are necessary to rule out samplinggap as a cause of the observed pattern. Another hypothesisis that the ecotone clade represents a recent expansion ofthe species in an originally forested area.

With the recent conWrmation of the presence of C. lau-cha (2ND 64, FND68) in the Pampas of southern Brazil,we can count a total of 6 species of Calomys in this country.Recently, a specimen with a karyotype of 2ND 48/FND66was collected in a patch of Cerrado in the Amazon Forestin the Brazilian State of Rondonia (Mattevi et al., 2005).Although this karyotype is identical to the one attributed toC. callidus, we doubt that the Rondônia specimen can beascribed to this species since their distributions are disjunct.Further data are necessary to clarify the identity of thatspecimen, which could be a new species, a chromosomalvariety of C. tocantinsi or C. callosus, or even Calomys sp.(from Beni) due to their geographical proximity.

4.3. Karyotype evolution

Although similar sized Calomys species are barely distin-guishable in terms of morphology, the karyotypes can bereliably used to discriminate among these species. The onlyexceptions are C. callosus and Calomys sp. from Beni,which share the same karyotype (2ND50/FND66). Karyo-type is mostly stable within species. Major intraspeciWc var-iation is observed only in C. lepidus for which a substantialdiVerence in diploid number (2ND36 and 2ND 44) havebeen found in individuals from places geographically farapart (Espinosa et al., 1997). IntraspeciWc variation in thefundamental number is also minimal, with the exception ofC. musculinus (2ND 38/ FND48or 56; Forcone et al., 1980;Gardenal et al., 1977; Lisanti et al., 1996; Massoia et al.,1968) and C. laucha (2ND 64, FND 68–72; Brum-Zorillaet al., 1990; Gardenal et al., 1977; Mattevi et al., 2005; Pear-son and Patton, 1976). Minor deviation from the standardkaryotype of the species has also been reported in a few

460 F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466

individuals of C. expulsus and C. tener (Mattevi et al.,2005).

The proposed hypotheses for karyotypic evolution inCalomys are always associated to a trend towards thereduction from a high diploid number as believed to be theancestral condition of the genus (Espinosa et al., 1997;Pearson and Patton, 1976; Salazar-Bravo et al., 2001).These hypotheses follow a general model of karyotypicevolution proposed for sigmodontines (Pearson and Pat-ton, 1976). As shown by Salazar-Bravo et al. (2001) andcorroborated by our results, the reduction in the diploidnumber is not a unidirectional trend as had been previouslysuggested (Pearson and Patton, 1976; Vitullo et al., 1990;Espinosa et al., 1997), but has rather occurred several timesindependently throughout the diversiWcation of Calomys.The inclusion of C. expulsus in the phylogeny showed thatthe ancestral condition of high diploid number is found inall basal taxa in each main clade of the phylogeny. Majorreductions in the diploid number seem to have happened atleast four times: once in the ancestral of C. lepidus and C.musculinus, and the other three times in the “large-bodied”clade, in Calomys sp. nov., the clade formed by C. venustus,C. fecundus, and Calomys sp. (from Beni), and the cladeformed by C. callosus and C. tocantinsi. Further reductionshappened within each of the latter two clades. The descrip-tion of C. expulsus, C. tocantinsi, and Calomys sp. nov. kar-yotypes corroborate the suggestion that Robertsonianfusion is the main mechanism of karyotype evolution in thegenus, since most species share very similar fundamentalnumbers. This is especially apparent in the evolution of thelarge-bodied clade where major changes in the diploidnumber are combined with almost total conservation ofFN. The only exception for the general conservation of FNin the genus is C. musculinus.

4.4. Biogeography

In this section, we analyze the biogeographic patternsexhibited by the species of Calomys, drawing attention tothe understudied and endangered Neotropical, nonhumidforest and grassland ecosystems. Habitat data for most ofthe Brazilian specimens were obtained by either directobservation (C.R. Bonvicino) or personal communicationwith the other collectors. Data for the non-Brazilian speci-mens was inferred from literature records and the projec-tion of coordinates on the Ecoregions of the NeotropicalRealm map (WWF, 1999; Olson and Dinerstein, 2005). Thelocalities plotted on the maps (Figs 3 and 4) represent col-lection sites of specimens sequenced by us or by Salazar-Bravo et al. (2001, 2002) as well as of a few additionalkaryotyped specimens (C.R. Bonvicino, unpublishedresults; Hurtado de Catalfo and Wainberg, 1974; Lisantiet al., 1976; Vitullo et al., 1983). The geographic distributionof Calomys presents a unique pattern among sigmodon-tines, being characterized by its exclusive occurrence in dryforests, savannas, shrublands, and grasslands, as well as adisjunct distribution to the south and north of Amazonia,

in the “diagonal of open formations” (Vanzolini, 1963) andin the Venezuelan Llanos, respectively. In fact, among sigm-odontines, Calomys is the only non-forest genus found inthese two areas. A similar disjunct distribution of non-for-est endemic mammals is recorded only for Lutreolinacrassicaudata, a marsupial, and Cavia aperea, a Cavi-omorph rodent (Voss, 1991). Calomys species are alsofound in a large range of altitudes, including high altitudelocations in the Peruvian and Bolivian Andes, where theyare mostly associated with the Puna habitat.

A major dichotomy can be established within the genusCalomys based on vegetation type and body size. The Wrstgroup contains a paraphyletic assemblage of all speciesdiverged before the cladogenesis of C. expulsus and includessmall-bodied species found in strict or mixed grassland andsavanna habitats. Within this group, C. hummelincki is dis-tributed in the northwestern semiarid plain and savannas ofVenezuela as well as in the Islands of Curaçao and Aruba(Martino et al., 2001). C. lepidus and C. sorrellus inhabit themontane grasslands of the Peruvian and Bolivian Altiplano,the former being found at higher elevations than the latter(Pearson and Patton, 1976). C. musculinus and C. laucha areopportunistic species inhabiting the central grassland plainsof Argentina (the Pampas). They are abundant in cropWelds, although they seem to diVer in micro-habitat prefer-ence (Busch and Kravetz, 1992a,b). C. tener occurs indiverse vegetation types of the Cerrado, including the eco-tone region with the Atlantic Forest. The second group con-tains a monophyletic assemblage of large-bodied Calomysfound in shrub lands, savannas, woodland savannas, anddry forests. C. callosus presents a wide distribution, usuallyassociated with the Chaco biome and enclaves of this habi-tat in adjacent areas, but is also found in the Chiquitano dryforests (Salazar-Bravo et al., 2002). Here we extend its distri-bution to the Cerrado of Mato Grosso and Mato Grosso doSul States, Brazil. Calomys sp. from Beni is restricted to theBeni savannas, characterized by seasonally Xooded grass-lands with patches of humid forest (Salazar-Bravo et al.,2002). This ecoregion is isolated from other open-vegetationareas by moist forests. Calomys fecundus occurs in ecotoneareas of the Bolivian montane dry forest and the AndeanYungas with the Chaco. According to Salazar-Bravo et al.(2002), C. venustus is associated if the Espinal biome inArgentina, a thorny deciduous shrub land forest. However,map projections show that C. venustus also occurs in theecotone between the Chaco and the Andean Yungas.

The inclusion of the Brazilian species C. tocantinsi, C.expulsus and Calomys sp. nov. (from Minas Gerais) in thephylogeny of Calomys revealed an interesting pattern. Thetwo latter species are basal species within the monophyletic“large-bodied” clade and occupy the Cerrado ecoregion,which is the most extensive savanna in South America. Cer-rado is a general term and actually encompasses a mosaicof vegetation types ranging from grassland (“campolimpo”) to dry forest (“cerradão”), including also swamps(“veredas”) and gallery forests (Eiten, 1972; Oliveira-Filhoand Ratter, 1995). Calomys expulsus is a rather prevalent

F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466 461

species in this biome, where it is sympatric with C. tener,occurring in most types of vegetation (Bonvicino andAlmeida, 2000; Marinho-Filho et al., 2002). C. expulsus isalso found in the drier Caatinga biome. Adjacent to its dis-tribution, to the east, we Wnd Calomys sp. nov. with a muchnarrower distribution in seasonal semi-deciduous forests(Bahia Interior Forests), and adjacent Cerrado (RosanaGentile, personal communication). Calomys tocantinsi, alsowith a narrow distribution, occurs west of the distributionof C. expulsus. It has been found in diVerent kinds of Cer-rado vegetation but always in the proximity of gallery for-ests (F.S. Lima, personal communication), and/or intransitional areas between this biome and the Amazon For-est (Fagundes et al., 2000; Bonvicino et al., 2003). Thesethree latter species were never found in sympatry. There-fore, when observing the temporal and spatial patterns ofthe monophyletic “large-bodied”clade, we can infer that ithas probably originated in the Cerrado biome and that sub-sequent speciation events occurred in ecoregions peripheralto the Cerrado and Chaco or in ecotones after the clado-genesis of C. expulsus. Furthermore an increasing tendencyof association with dry forests can be observed as a wholein this group. Even C. expulsus, found throughout the Cer-rado, seems to prefer more closed vegetation types than thesmaller sympatric C. tener in some parts of their distribu-tion (Henriques et al., 1997).

4.5. Tempo and mode of diversiWcation of Calomys

As shown by our results, molecular clock estimatesdepend not only on the statistical methods used but also,and mostly, on the calibration points. The large range ofour estimates indicates the need of further paleontologicaldiscoveries to improve divergence time estimates in sigm-odontines. Salazar-Bravo et al. (2001) used as the main cali-bration point the split between Mus and Rattus at 12 Ma.Since this split represents an internal node in the Muridaephylogeny and Sigmodontinae is considered a subfamily ofCricetidae (Steppan et al., 2004), this point should not beideally used as a calibration point for recent sigmodontineclades. Moreover, this date is based on the Wrst fossil withmodern murine dentition, which most likely appeared sometime before the split between Mus and Rattus (Jacobs et al.,1994). While this calibration point is often used due tosequence availability, divergence dates based on it are likelyoverestimated (Steppan et al., 2004). In fact, the datesobtained by Salazar-Bravo et al. (2001) based on the Mus-Rattus split are much older than those based on theBolomys-Thaptomys split in the same study and the datespresented here. For reasons already discussed in theResults, we will take into account only divergence time esti-mates obtained with the PL method.

The divergence between Akodontini and Phyllotini can-not be placed in a clear geological epoch, but our estimatessuggest that the two tribes started diverging before the clo-sure of the Panamanian Isthmus (between 3.5 and 8.8 Ma).Our estimates suggest a late Miocene—early Pliocene diver-

gence of the two tribes, in rough agreement with Marshall’s(1979) proposal of the sigmodontine radiation taking placebetween 5 and 7 Ma. The divergence of Calomys from otherPhyllotini was estimated in 2.8–7.1 Ma, with a mean of4.9 Ma. The interval between the Wrst and third quartile ofour estimates (3.7–5.9 Ma) is similar to the standard devia-tion obtained by Steppan et al. (2004) for the same node(3.97–5.46 Ma), but much more recent than the estimates bySalazar-Bravo et al. (2001), even considering our maximumNPRS estimates. These dates are coincident with the begin-ning of major uplift phase of the Andes in the late Mioceneand early Pliocene (Gregory-Wodzicki, 2000). Fossil evi-dence supports the presence of savannas in the proto-Andes in Argentina and Bolivia during this time (Webb,1978).

The two basal cladogenesis in Calomys, the highland-lowland split and the Amazonian split (C. hummelincki inthe Venezuelan Llanos and all the other lowland species),seem to have happened close to one another (within some50 to 200 thousand Ma), sometime around the beginning ofthe Pliocene or at the very end of the Miocene (< 5.3 Ma).Early in its history, the genus Calomys might have divergedinto a highland and a lowland clade following the Andesuplift between the late Miocene and middle Pliocene. Thedispersal of C. hummelincki has been correlated with Mar-shall’s hypothesis on the appearance of an open-vegetationcorridor in the eastern slopes of the Andes connecting thesavannas in the north and south of the Amazon Forest ataround 3.5 Ma (Marshall, 1978). Despite the lack of fossilevidence supporting Marshall’s hypothesis, our estimatesfor the separation of C. hummelincki roughly coincides withthe proposed timing for the corridor. Marshall (1978) pro-posed that the genus originated in the Venezuelan Llanosand migrated southward to reach its current distribution inthe Andes and the “diagonal of open formations”. Weagree with the Salazar-Bravo et al. (2001) on their proposalof a south of Amazon origin of the genus and a northwardmigration of C. hummelincki ancestor. An alternativehypothesis (although Marshall did not made clear if thecorridor was a highland or a lowland route) is that the dis-persal of C. hummelincki occurred through the Andes(Webb, 1978), in which case the invasion of lowlands wouldhave happened twice. The formation of the Amazon basinin the Pliocene promoted an increase in rainfall and theextension of broadleaf evergreen forests (Hooghiemstraand van der Hammen, 1998), creating the current disjunctdistribution of lowland Calomys species.

The lowland species that evolved south of the AmazonForest have their origins associated with the Cerrado. Thisis in agreement with the proposal that this biome is proba-bly the most stable savanna environment of South Amer-ica, as indicated by its high rates of plant endemism (Eiten,1972). The appearance of C. tener, an endemic, widespreadspecies from the Cerrado, is followed by the appearance ofa clade containing C. laucha and the all the large-bodiedspecies. C. laucha is originally endemic to the Pampasof southern Brazil, Uruguay, and Argentina, suggesting a

462 F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466

connection between this biome and the Cerrado by themiddle-late Pliocene similar to the one that occurred duringthe Last Glacial Maximum (Behling, 2002). Currently,these two biomes are separated by the Araucaria MoistForests and the Paraná Interior Forest. The “large-bodied”clade had its origins in the late Pliocene and its most basalspecies are also found in the Cerrado. The appearance of alarge-bodied Calomys gave rise to a speciation burst duringthe Pleistocene, leading to at least 6 species: C. expulsus,Calomys sp. nov. (from Minas Gerais), C. venustus, C.fecundus, Calomys sp. (from Beni), C. tocantinsi, and C.callosus (not counting C. callidus, 2ND 48, that probablybelongs to this clade as suggested by its body size).

The estimated dates of cladogenesis in the “large-bod-ied” clade fell in a period subjected to cycles of climaticchanges and vegetation shifts between dry vegetation andmoist forests that began in the late Pliocene and extendsthroughout the Pleistocene (Webb, 1978). Although thePleistocene glaciations have been suggested as a causal fac-tor for the tropical biodiversity in South America (HaVer,1969; Diamond and Hamilton, 1980), molecular clock esti-mates indicate that very few taxa seem to have diversiWedduring this time (Costa, 2003; Moritz et al., 2000; Penning-ton et al., 2005). Most of the studies on these vegetationshifts and their consequences for the associated fauna, havefocused on forest taxa or taxa that are present in both sav-annas and moist forests (e.g. Costa, 2003; Hooghiemstraand van der Hammen, 1998). Thus, in contrast with themany studies and hypotheses on forest taxa biogeography,no vicariant scenario has been proposed for the diagonal ofopen formations of South America. Palynological evidencepoint to the expansion of forests during warm, humid peri-ods in the northeast (where now we Wnd the dry Caatinga)and southwest of the diagonal, causing a reduction of thecore dry area, and allowing a connection between theAtlantic coast and the Amazon (Auler et al., 2004; Behlinget al., 2000; Costa, 2003; de Vivo and Carmignotto, 2004;Por, 1992). During those humid periods, a third connectionbetween the two rainforests might have arisen in centralBrazil, where there is a current northward extension of theParaná Forest (Oliveira-Filho and Ratter, 1995; Por, 1992).The existence of this ancient forest corridor is supported bypalynological evidence from several sites in central Brazil(Salgado-Labouriau, 1997) and computer simulations ofPleistocene vegetation (Mayle, 2004). This forest corridorwould have acted as a barrier between populations locatedeast and west of it. This scenario could explain the splitbetween Calomys sp. nov. and its sister group at one timeand the separation between C. callosus and C. tocantinsi atanother, more recent time.

Another factor to take into consideration is the tendencytowards a habitat shift from open savanna to dry forest orforest-savanna ecotone in the evolution of the “large-bod-ied” clade, as evidenced by the distribution of C. callosus,Calomys sp. nov. (from Minas Gerais), C. tocantinsi, C.venustus, and C. fecundus. This pattern suggests that parap-atric speciation in ecotones may be an important mecha-

nism in the diversiWcation of this group. Vanzolini andWilliams (1981) proposed the vanishing refuge model toexplain this mechanism, focusing on savanna species thathave their sister species in a neighboring forest. This modelpostulates that populations trapped in patches of forestmay diVerentiate into species by directional selectiontoward tolerance to ecotones or the surrounding dry habi-tats, as rainforest patches become too small to retain viablepopulations. In the case of Calomys, this model can beapplied but in a reverse way, with the ancestral speciesoccupying open habitats instead of forests. Random karyo-typic changes in isolated populations could also haveplayed a role in the Wnal stage of speciation, as suggested bythe karyotypic diversity in the group. This mechanism canexplain the cladogenesis of Calomys sp. nov. and theappearance of the clade formed by C. venustus, C. fecundus,and Calomys sp. (from Beni). Since the habitats of the latterthree species are not continuous, speciation in this clade canbe Wt into a vicariant scenario with two consecutive north-south splits: the Wrst split between C. venustus in the southand the other two species in the north, and a second splitbetween C. fecundus and Calomys sp.

5. Conclusions

The inclusion of cytochrome b sequences of Brazilianspecimens of Calomys in the reconstruction of the evolu-tionary relationships of the genus has provided a robustphylogenetic hypothesis for the genus. Although the taxonsampling of 13 species is quite inclusive, some forms are stillmissing from molecular systematics studies on Calomys.Some of the missing forms have taxonomic problems, suchas C. bimmaculatus, C. murillus, and C. boliviae, urging fora thorough revision of the genus (Cordeiro-Estrela et al., inprep). The inclusion of these species and other missing inthis study (e.g. C. callidus) would allow for an even moredetailed study of biogeography and karyotypic evolution.

The results of the relaxed clock models suggest that thebasal split of the genus probably occurred in the Pliocene,thus earlier than previous suggestions, and that, among thelowland species, the large-bodied species group diversiWedrelatively recently, in the Pleistocene, despite its highestdiversity in karyotypes and habitats. The inclusion of Bra-zilian specimens in the analysis of the historical biogeogra-phy of the genus showed that diversiWcation patterns aremore complex than previously thought, as the core Cerradoarea seems to be the center stage of speciation dynamicswith adjacent ecosystems. This Wnding is of great impor-tance at a time when the Cerrado biome, a hot spot of bio-diversity (Machado et al., 2004; Klink and Machado, 2005),is being deforested at rates even higher than those aVectingthe Amazon Forest. The Cerrado and its bordering ecotonezones already account with a transformed or cleared areaof 800,00 km2, 3 times the deforested area of Amazonia. Wehighlight that systematics studies and conservation of eco-tone biomes might be a priority for description and conser-vation of biological diversity.

F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466 463

Acknowledgments

This work would not be possible without the kinddonation of samples by S.M. Lindbergh, E.M. Vieira, A.P.Carmignotto, J.A. Oliveira, A. Palma, A.M.R. Bezerra,C.E.V. Grelle, P.S. D’Andrea, and R. Gentille, (samplesfrom Morro dos Cabeludos, Campinas, Pantanal, Ara-guaia National Park, Aquidauana, Capitão Andrade,João Costa, Emas Nacional Park, and Lagoa Santa). Wethank a number of people that helped us in the Weld and

Leslie Skalak for her help in the lab. We are in debit withRob DeSalle for providing the lab facilities where part ofthe sequences was generated and for his helpful commentson an earlier version of the manuscript and M. Wekslerfor his help with the Wgures. Collecting licenses weregranted by IBAMA (Instituto Brasileiro do Meio Ambi-ente). Work supported by CNPq, CNPq/PRONEX. PCEwas supported by CAPES Grant bex0982016. FCA wassupported by CAPES and the Henry McCraken fellow-ship (New York University).

Appendix A

Samples included in this study. In parentheses in the IdentiWcation column is the number of additional individuals withidentical haplotype and collected in the same locality as the specimen identiWed here. The identiWcation numbers of theseadditional specimens are available upon request. The last column corresponds to the numbers assigned to localities inFigs. 3 and 4.

Taxon GenBank IdentiWcation Locality #

C.tocantinsi DQ447277 ARB12 (1) BRA: TO, PARNA Araguaia 1C.tocantinsi DQ447278 ARB55 BRA: TO, PARNA Araguaia 1C.expulsus DQ447284 CRB2374 BRA: GO, Mimoso de Goiás 2C.expulsus DQ447285 CRB2307 BRA: GO, Mimoso de Goiás 2C.expulsus DQ447293 CRB500 BRA: GO, Corumbá de Goiás 3C.expulsus DQ447291 CRB759 BRA: GO, Terezina de Goiás 4C.expulsus DQ447292 APC574 BRA: GO, Mineiros, Emas National Park 5C.expulsus DQ447286 CRB1676 BRA: BA, Jaborandi 6C.expulsus DQ447283 CRB2733 (1) BRA: BA, Correntina 7C.expulsus DQ447287 LBCE1548 BRA: BA, Caetité 8C.expulsus DQ447288 LBCE1547 BRA: BA, Caetité 8C.expulsus DQ447289 LBCE1542 BRA: BA, Caetité 8C.expulsus DQ447290 LBCE1245 (1) BRA: PI, João Costa 9Calomys sp.nov. DQ447273 LBCE5583 BRA: MG, Capitão Andrade 10Calomys sp.nov. DQ447274 LBCE5556 (2) BRA: MG, Capitão Andrade 10Calomys sp.nov. DQ447275 LV-CEG41 BRA: MG, Lagoa Santa 11Calomys sp.nov. DQ447276 LV-CEG40 BRA: MG, Lagoa Santa 11C.callosus DQ447279 JAO356 (1) BRA: MT, Barão de Melgaço 12C.callosus DQ447280 LBCE3054 (5) BRA: MS, Corumbá 13C.callosus DQ447282 LBCE5682 BRA: MS, Corumbá 13C.callosus DQ447281 LBCE4474 (2) BRA: MS, Aquidauana 14C.callosus AY033188 NK72378 PAR: Boqueron, FiladelWa 15C.callosus AY033181 NK72351 PAR: Boqueron, Monte Palma 16C.callosus AY033187 NK72344 PAR: Boqueron, Monte Palma 16C.callosus AY033186 NK22523 PAR: Amabaya, PN Cerro Corá 17C.callosus AY033185 NK22532 PAR: Amabaya, PN Cerro Corá 17C.callosus AY033184 NK23306 BOL: Santa Cruz, Zanja Honda 18C.callosus AY033180 NK23320 BOL: Santa Cruz, Zanja Honda 18C.callosus AY033183 NK11590 BOL: Santa Cruz, San Miguel Rincón 19C.callosus AY033182 NK13009 BOL: Santa Cruz, San Ramón 20C.callosus AY033179 NK21176 BOL: Santa Cruz, Las Cruces 21C.callosus AY033178 NK12310 BOL: Santa Cruz, Santiago de Chiquitos 22C.callosus AY033177 NK12308 BOL: Santa Cruz, Santiago de Chiquitos 22C.venustus AY033176 AK15337 ARG: Santiago del Estero 23C.venustus AY033174 TK49115 ARG: Córdoba, 2km S Espinillo 24Calomys sp. AY033157 NK37739 BOL: Beni, La Republica 25C.fecundus AY033170 NK23354 BOL: Tarija, Camantindy 26C.fecundus AY033159 NK21508 BOL: Chuquisaca, Monteagudo 27C.fecundus AY033173 AK15276 ARG: Tucuman 28C.tener AF385595 NK21054 BOL: Santa Cruz, Santa Rosa de la Roca 29C.tener AF385597 NK42183 BRA: SP, Tupi Paulista 30C.tener AF385596 NK42140 BRA: SP, Tupi Paulista 30C.tener DQ447294 EM1135 BRA: SP, Campinas 31C.tener DQ447296 CRB1219 BRA: SP, Pedreira 32

(continued on next page)

464 F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466

1- Three letter abbreviations refer to countries: ARG, Argentina; BOL, Bolivia; BRA, Brazil; PAR, Paraguay; PER, Peru; VEN, Venezuela. Two letterabbreviation refer to Brazilian States: BA, Bahia; GO, Goiás; MG, Minas Gerais; MT, Mato Grosso; MS, Mato Grosso do Sul; PI, Piaui; SP, São Paulo;TO, Tocantins.

Appendix A (continued)

Taxon GenBank IdentiWcation Locality #

C.tener DQ447297 CRB1220 BRA: SP, Pedreira 32C.tener DQ447302 CRB2382 BRA: GO, Mimoso de Goiás 2C.tener DQ447295 CRB503 BRA: GO, Corumbá de Goiás 3C.tener DQ447298 CRB1549 BRA: BA, Jaborandi 6C.tener DQ447299 CRB1558 (2) BRA: BA, Jaborandi 6C.tener DQ447300 CRB1584 BRA: BA, Jaborandi 6C.tener DQ447301 CRB1590 BRA: BA, Jaborandi 6C.laucha AY033190 NK72376 PAR: Boqueron, FiladelWaC.laucha AY033189 NK25158 BOL: Tarija, Estancia BolivarC.hummelincki AF385598 AM V001 VEN: Falcón, IsiroC.sorellus AF385608 FMNH107709 PER: Arequipa, CayllomaC.musculinus AF385599 RE 126 ARG: Chubut, Puerto MadrynC.lepidus AF385606 NK14656 BOL: Tarija, 1 km E IscayachiC.lepidus AF385607 NK31032 BOL: La Paz, Reserva de Fauna Ulla-Ulla

Appendix B. Supplementary data

Supplementary data associated with this article can befound, in the online version, at 10.1016/j.ympev.2006.07.005.

References

Alho, C.J., Pereira, L.A., 1985. Population ecology of a Cerrado rodentcommunity in Central Brazil. Rev. Bras. Biol. 45, 597–609.

Auler, A.S., Wang, X., Edwards, R.L., Chencg, H., Cristalli, P.S., Smart,P.L., Richards, D.A., 2004. Quaternary ecological and geomorphicchanges associated with rainfall events in presently semi-arid north-eastern, Brazil. J. Quat. Sci. 19, 693–701.

Behling, H., 2002. South and southeast Brazilian grasslands during latequaternary times: a synthesis. Palaeogeogr. Palaeoclimatol. Palaeoecol.177, 19–27.

Behling, H., Arz, H.W., Pätzold, J., Wefer, G., 2000. Late Quaternary vege-tational and climate dynamics in the northeastern Brazil, inferencesfrom marine core GeoB 3104-1. Quat. Sci. Rev. 19, 981–994.

Bonvicino, C.R., Almeida, F.C., 2000. Karyotype, morphology and taxo-nomic status of Calomys expulsus (Rodentia: Sigmodontinae). Mam-malia 64, 339–351.

Bonvicino, C.R., Lima, J.F.S., Almeida, F.C., 2003. A new species of Calo-mys Waterhouse (Rodentia: Sigmodontinae) from the Cerrado of cen-tral Brazil. Rev. Bras. Zool. 20, 301–307.

Braun, J.K., 1993. Systematic relationships of the tribe Phyllotini (Muri-dae: Sigmodontinae) of South America. Special Publication, Okla-homa Museum of Natural History, Norman.

Brum-Zorilla, N., de Catalfo, G.H., Degiovanangelo, C., Wainberg, R.L.,de Fronza, T.G., 1990. Calomys laucha chromosome from Uruguayand Argentina. Caryologia 43, 65–77.

Busch, M., Kravetz, F.O., 1992a. Competitive interactions among rodents(Akodon azarae, Calomys laucha, C. musculinus and OligoryzomysXavescens) in a two-habitat system. I. Spatial and numerical relation-ships. Mammalia 56, 45–56.

Busch, M., Kravetz, F.O., 1992b. Competitive interactions among rodents(Akodon azarae, Calomys laucha, Calomys musculinus, and Oligoryzo-mys Xavescens) in a two-habitat system. II. EVect of species removal.Mammalia 56, 541–554.

Carroll, D.S., Mills, J.N., Montgomery, J.M., Bausch, D.G., Blair, P.J.,Burans, J.P., Felices, V., Gianella, A., Iihoshi, N., Nichol, S.T., Olson,J.G., Rogers, D.S., Salazar, M., Ksiazek, T.G., 2005. Hantavirus pul-monary syndrome in Central Bolivia: relationships between reser-voir hosts, habitats, and viral genotypes. Am. J. Trop. Med. Hyg. 72,42–44.

Cione, A.L., Tonni, E.P., 2001. Correlation of Pliocene to Holocene south-ern American and European vertebrate-bearing units. Boll. Soc. Paleo-tol. Ital. 40, 167–173.

Corti, M., Merani, M.S., de Villafañe, G., 1987. Multivariate morphomet-rics of Vesper mice (Calomys): preliminary assessment of species, popu-lation and strain divergence. Z. Saügetierkunde 52, 236–242.

Costa, L.P., 2003. The historical bridge between the Amazon and theAtlantic Forest of Brazil: a study of molecular phylogeography withsmall mammals. J. Biogeogr. 30, 71–86.

De Vivo, M., Carmignotto, A.P., 2004. Holocene vegetation changeand the mammal faunas of South America and Africa. J. Biogeogr. 31,943–957.

Diamond, A.W., Hamilton, A.C., 1980. The distribution of forest passerinebirds and quaternary climate change in tropical. Africa. J. Zool. 191,379–402.

Eiten, G., 1972. Cerrado vegetation of Brazil. Bot. Rev. 38, 205–341.Engel, S., Hogan, K., Taylor, J., Davis, S., 1998. Molecular systematics and

paleobiogeography of the South American sigmodontine rodents. Mol.Biol. Evol 15, 35–49.

Espinosa, M.B., Lasserre, A., Piantanida, M., Vitullo, A.D., 1997. Cytoge-netics of Vesper mice, Calomys (Sigmodontinae): a new karyotypefrom the Puna region and its implication for chromosomal phylogeny.Cell. Mol. Life Sci 53, 583–586.

Fagundes, V., Sato, Y., Silva, M.J.J., Rodrigues, F., Yonenaga-Yassuda, Y.,2000. A new species of Calomys (Rodentia: Sigmodontinae) from Cen-tral Brazil identiWed by its karyotype. Hereditas 133, 195–200.

Forcone, A., Luna, M., Kravetz, F., Lisanti, J.A., 1980. Bandas C y G deCalomys musculinus (Rodentia, Cricetidae). Mendeliana 4, 57–65.

Gardenal, C.N., de Juarez, N.T., Gutierrez, M., Sabatini, M.S., 1977. Con-tribución al conocimiento de tres especies de genero Calomys (Roden-tia: Cricetidae). PHYSIS.I. Estudios Cytogeneticos (Buenos Aires) 36,169–178.

Geise, L., Hingst, E., Weksler, M., Cerqueira, R., 1996. A new karyotype ofCalomys (Rodentia: Sigmodontinae), with taxonomic considerations.Rev. Bras. Genet. 19 (suppl.), A45.

Gregory-Wodzicki, K.M., 2000. Uplift history of Central and NorthernAndes: A review. Geol. Soc. Am. Bull. 112, 1091–1105.

Guindon, S., Gascuel, O., 2003. A simple, fast, and accurate algorithm toestimate large phylogenies by maximum likelihood. Syst. Biol. 52, 696–704.

HaVer, J., 1969. Speciation in Amazonian forest birds. Science 165,131–137.

Henriques, R.P.B., Bizerril, M.X.A., Kohlsdorf, T., 1997. Abundância eriqueza e seleção de habitat de pequenos mamíferos dos cerrados doBrasil central. In: Leite, L.L., Saito, H. (Eds.), Contribuição ao conheci-mento ecológico do Cerrado. Universidade de Brasília.

F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466 465

Hershkovitz, P., 1962. Evolution of Neotropical Cricetine rodents (Muri-dae) with special reference to the phyllotine group. Fieldana Zool. 46,1–524.

Hooghiemstra, H., van der Hammen, T., 1998. Neogene and Quaternarydevelopment of the neotropical rain forest: the forest refugia hypothe-sis, and a literature overview. Earth Sci. Rev. 44, 147–183.

Huelsenbeck, J.P., Rannala, B., 1997. Phylogenetic methods come of age:testing hypotheses in an evolutionary context. Science 276, 227–232.

Huelsenbeck, J.P., Ronquist, F., 2001. MRBAYES: Bayesian inference ofphylogenetic trees. Bioinformatics 17, 754–755.

Hurtado de Catalfo, G.E., Wainberg, R.L., 1974. Citogenética de Calomyscallosus (Rengger, 1830)(Rodentia, Cricetidae). Análisis metrico delcariotipo somático. Physis. Secc. C3, 215–219.

Jacobs, L.L., Downs, W.R., Tomida, Y., 1994. The evolution of murinerodents in Asia. In: Li, C., Setoguschi, T. (Eds.), Rodent and lagomorphfamilies of Asian origins and their diversiWcation. National ScienceMuseum Monograph, Tokyo, pp. 49–156.

Klink, C.A., Machado, R.B., 2005. Conservation of the Brazilian Cerrado.Conservation Biol. 19, 707–713.

Lacher Jr., T.E., Mares, M.A., Alho, C.J.R., 1989. The structure of a smallmammal community in a central Brazilian savanna. In: Redford, K.,Eisenberg, J. (Eds.), Advances in Neotropical Mammalogy. SandhillCrane Press, Gainesville, FL, pp. 137–162.

Langley, C.H., Fitch, W., 1974. An estimation of the constancy of the rateof molecular evolution. J. Mol. Evol. 3, 161–177.

Levis, S., Garcia, J., Pini, N., Calderon, G., Ramirez, J., Bravo, D., St Jeor,S., Ripoll, C., Bego, M., Lozano, E., Barquez, R., Ksiazek, T.G., 2004.Hantavirus pulmonary syndrome in northwestern Argentina: circula-tion of Laguna Negra virus associated with Calomys callosus. Am. J.Trop. Med. Hyg. 75, 658–663.

Lisanti, J., de Barale, G.D., Pinna Senn, E., Bella, J.L., 1996. Chromosomalcharacterization of Calomys musculinus (Rodentia, Cricetidae). Caryo-logia 49, 327–334.

Lund, P.W., 1841. K. Danske. Vidensk. Selsk. Afhandl. VIII, p. 280.Machado, R.B., Ramos Neto, M., Pereira, P., Caldas, E., Gonçalves, D.,

Santos, N., Tabor, K., Steininger, M., 2004. Estimativas de perda deárea do Cerrado brasileiro. Conservation International do Brasil.

Marinho-Filho, J., Rodrigues, F.H.G., Juarez, K.M., 2002. The Cerradomammals: diversity, ecology and natural history. In: Oliveira, P.S.,Marquis, R.J. (Eds.), The Cerrados of Brazil: Ecology and Natural His-tory of a Neotropical Savanna. Columbia University Press, N.Y., pp.266–284.

Marshall, L.G., 1978. A model for paleogeography of South Americancricetine rodents. Paleobiology 5, 126–132.

Martino, A.M.G., Capana, E., Filippucci, M.G., 2001. Allozyme variationand divergence in the phyllotine rodent Calomys hummelincki (Husson,1960). Genetica 110, 163–175.

Massoia, E., Fornes, A., Wainberg, R.L., De Fronza, T.G., 1968. Nuevosaportes al conocimiento de las species bonaerenses del género Calomys(Rodentia-Cricetidae). Rev. Invest. Agrop. 1, Biol. Produc. An. 5, 63–92.

Mattevi, M., Haag, T., Oliveira, L.F.B., Langguth, A., 2005. Chromosomecharacterization of Brazilian species of Calomys (Sigmodontinae,Rodentia) from Amazon, Cerrado and Pampas domains. Arq. Mus.Nacional 63, 157–181.

Mayle, F.E., 2004. Assessment of the Neotropical dry forest refugiahypothesis in the light of palaeoecological data and vegetation modelsimulations. J. Quat. Sci. 19, 713–720.

Mills, J.N., Ellis, B.A., McKee, K.T., Ksiazek, T.G., Oro, J.G., Maiztegui,J.I., Calderón, G.E., Peters, C.J., Childs, J.E., 1991. Junin virus activityin rodents from endemic and nonendemic loci in central Argentina.Am. J. Trop. Med. Hyg. 44, 589–597.

Mills, J.N., Ellis, B.A., McKee, K.T., Calderón, G.E., Maiztegui, J.I., Nel-son, G.O., Ksiazek, T.G., Peters, C.J., Childs, J.E., 1992. A longitudinalstudy of Junin virus activity in the rodent reservoir of Argentine hem-orrhagic fever. Am. J. Trop. Med. Hyg. 47, 749–763.

Mills, J.N., Ellis, B.A., Childs, J.E., McKee, K.T., Maiztegui, J.I., Peters,C.J., Ksiazek, T.G., Jahrling, P.B., 1994. Prevalence of infection with

Junin virus in rodent populations in the epidemic area of Argentinehemorrhagic fever. Am. J. Trop. Med. Hyg. 51, 554–562.

Moritz, C., Patton, J.L., Schneider, C.J., Smith, T.B., 2000. DiversiWcationof rainforest faunas: An integrated molecular approach. Annu. Rev.Ecol. Syst. 31, 533–563.

Musser, G.G.,Carleton, M.D., 1993, 2005. Family Muridae. In: Wilson,D.E. and D.A.M. Reeder (Eds.), Mammals species of the World, 2ndand 3rd eds. , Smithsonian Institution, Washington. pp. 501–753.

Nylander, J.A.A., 2004. MrAIC.pl. Program distributed by the author.Evolutionary Biology Centre, Uppsala University.

Olds, N., 1988. A Revision of the genus Calomys (Rodentia: Muridae).Unpublished PhD. Thesis. City University of New York.

Oliveira-Filho, T.A., Ratter, J.A., 1995. A study of the origin of centralBrazilian forests by the analysis of plant species distribution patterns.Edinburgh J. Bot. 52, 141–194.

Olson, D.M., Dinerstein, E., 2005. The Global 200: Priority ecoregions forglobal conservation. Annals of the Missouri Botanical Garden 89, 125–126.

Pardiñas, U.F.J., D’Elia, G., Ortiz, P.E., 2002. Sigmodontinos fósiles(Rodentia: Muridae: Sigmodontinae) de América del Sur: estadoactual de su conocimiento y prospectiva. Mastozoologia Neotropical 9,209–252.

Pearson, O.P., Patton, J.L., 1976. Relationships among South Americanphyllotine rodents based on chromosome analysis. J. Mammal. 57,677–686.

Pennington, R.T., Prado, D.E., Pendry, C.A., 2005. Neotropical seasonallydry forests and Quaternary vegetation changes. J. Biogeogr. 27, 261–273.

Pérez-Zapata, A., Vitullo, A.D., Reig, O.A., 1987. Karyotipic and spermdistinction of Calomys hummelincki from Calomys laucha (Rodentia,Cricetidae). Acta Cient. Venez 38, 90–93.

Por, F.D., 1992. Sooretama: The Atlantic Forest of Brazil. SPB AcademicPublishing.

Porzecanski, A.L., Cracraft, J., 2005. Cladistic analysis of distributions andendemism (CADE): using raw distributions of birds to unravel the bio-geography of the South American aridlands. J. Biogr. 32, 261–275.

Posada, D., Crandall, K.A., 1998. Modeltest: testing the model of DNAsubstitution. Bioinformatics 14, 817–818.

Reig, O.A., 1978. Roedores cricetidos del Plioceno superior de la provinciade Buenos Aires (Argentina). Publicaciones del Museo Municipal deCiencias Naturales de Mar del Plata “Lorenzo Scaglia” 2, 164–190.

Reig, O.A., 1986. Diversity patterns and diVerentiation of high Andeanrodents. In: Vuilleumier, F., Monasterio, M. (Eds.), High altitude tropi-cal biogeography. Oxford University Press, p. 649.

Ruedas, L.A., Salazar-Bravo, J., Dragoo, J.W., Yates, T.L., 2000. Theimportance of being earnest: what, if anything, constitutes a “specimenexamined? Mol. Phylogenet. Evol. 17, 129–132.

Salazar-Bravo, J., Dragoo, J.W., Tinnin, D.S., Yates, T.L., 2001. Phylogenyand evolution of the Neotropical rodent genus Calomys: Inferences frommitochondrial DNA sequence data. Mol. Phylogen. Evol. 20, 173–184.

Salazar-Bravo, J., Ruedas, L.A., Yates, T.L., 2002. Mammalian reservoirsof arenaviruses. Curr. Top. Microbiol. Immunol. 262, 25–63.

Salgado-Labouriau, M.L., 1997. Late Quaternary palaeoclimate in the sav-annas of South America. J. Quat. Sci. 12, 371–379.

Sanderson, M.J., 1997. A nonparametric approach to estimating divergencetimes in the absence of rate constancy. Mol. Biol. Evol. 14, 1218–1231.

Sanderson, M.J., 2002. Estimating absolute rates of molecular evolutionand divergence times: a penalized likelihood approach. Mol. Biol. Evol.19, 101–109.

Sanderson, M.J., 2003. r8s: inferring absolute rates of molecular evolutionand divergence times in the absence of a molecular clock. Bioinformat-ics 19, 301–302.

Simpson, G.G., 1969. South American Mammals. In: Fittkau, J.J. (Ed.),Biogeography and ecology in South America. Dr. W. Junk N.V. Publ.,The Hague, pp. 879–909.

Smith, M.F., Patton, J.L., 1999. Phylogenetic relationships and the radia-tion of sigmodontine rodents in South America: evidence from cyto-chrome b. J. Mammal. Evol. 6, 89–128.

466 F.C. Almeida et al. / Molecular Phylogenetics and Evolution 42 (2007) 449–466

Smith, L.J., Braylan, R.C., Nutkis, J.E., Edmundson, K.B., Downing, J.R.,Wakeland, E.K., 1987. Extraction of cellular DNA from human cellsand tissues Wxed with ethanol. Anal. Biochem. 160, 135–138.

Steppan, S., 1993. Phylogenetic relationships among the Phyllotini(Rodentia: Sigmodontinae) using morphological characters. J. Mam-mal. Evol. 1, 187–213.

Steppan, S., Adkins, R., Anderson, J., 2004. Phylogeny and divergence-dateestimates of rapid radiations in muroid rodents based on multiplenuclear genes. Syst. Biol. 53, 533–553.

SwoVord, D.L., 2002. PAUP*. Phylogenetic Analysis Using Parsimony(*and other methods). Version 4. Sinauer Associates. Sunderland, Mas-sachusetts.

Vanzolini, P.E., 1963. Problemas faunísticos do Cerrado. In: Ferri, M.G.(Ed.), Simpósio sobre o Cerrado. São Paulo: Ed. Universidade de SãoPaulo. pp. 307–320.

Vanzolini, P.E., Williams, E.E., 1981. The vanishing refuge: a mechanismfor eco-geographic speciation. Pap. Avulsos Zool. 34, 251–255.

Videla, C., Kajon, A., Carballal, G., Weissenbacher, M., 1989. Calomys cal-lidus as a potential Junin virus reservoir. J. Med. Virol. 27, 238–243.

Vitullo, A.D., Hodara, V.L., Merani, M.S., 1983. Caracterización citogene-tica de tres especies de roedores (Rodentia, Cricetidae) de la RepublicaArgentina. Rev. Zool. Mus. Arg. Cs. Nat. 13, 491–498.

Voss, R.S., 1991. An introduction to Neotropical genus Zygodontomys.Bull. Am. Mus. Nat. Hist. 210, 1–113.

Webb, S.D., 1978. A history of savanna vertebrates in the New World PartII: South America and the Great Interchange. Ann. Rev. Ecol. Syst. 9,393–426.

Williams, R.J., Bryan, R.T., Mills, J.N., Palma, R.E., Vera, I., Velasquez,F.D., Baez, E., Schmidt, W.E., Figueroa, R.E., Peters, C.J., Zaki, S.R.,Khan, A.S., Ksiazek, T.G., 1997. An outbreak of hantavirus pulmonarysyndrome in western Paraguay. Am. J. Trop. Med. Hyg. 57, 274–282.

Yahnke, C.J., Meserve, P.L., Ksiazek, T.G., Mills, J.N., 2001. Patterns ofinfection with Laguna Negra virus in wild populations of Calomyslaucha in the central Paraguayan chaco. Am. J. Trop. Med. Hyg. 65,768–776.

Zuckerkandl, E., Pauling, L., 1962. Molecular disease, evolution andgenetic heterogeneity. In: Kasha, M., Pullman, B. (Eds.), Horizons inbiochemistry. Academic press.