Molecular Phylogenetics and Evolution Vol. 17, No. 2, November, pp. 280-293, 2000 doi:10.1006/mpev.2000.0849, available online at http://www.idealibrary.com on IDE~L@.

The Bushlike Radiation of Muroid Is Exemplified by the Molecular Phylogeny of the LCAT Nuclear Gene Johan Michaux and François Catzeflis

Laboratoire de Paléontologie, Institut des Sciences de l'Evolution, UMR 5554 CNRS, Université de Montpellier 2, 34095 Montpellier, France

Received January 11, 2000; revisedJuly 26, 2000

least 1326 spanning more than 281 genera Phylogenetic relationships among 40 extant species (Musser and Carleton, 1993). Thus, this single of rodents, with an emphasis on the taxonomic sam- taxon representsabout 29 and 25% of ail mammalian pling of and Dipodidae, were studied using speciesand genera,respectively. The evolutionary sys- sequences of the nuclear protein-coding gene LCAT tematics of this speciosefamily bas been very difficult (lecithin cholesterol acyl transferase). Analysis of 804 and despitemany attempts (i.e.,Miller and Gidley, 1918; bp from the exonic regions of LCAT confirmed many Simpson,1945; Hooper and Musser, 1964;Chaline et al., traditional groupings in and around Muridae. A 1977; Carleton, 1980), many uncertainties, confusions, strong support was found for the families Muridae and conflicting views have persisted for these animais. (represented by 29 species) and Dipodidae (5 species). For this reason,in their recent review, Musser and Car- Compared with Sciuridae, Gliridae, and Caviomor- pha, the Dipodidae family appeared the closest rela- leton (1993) decidedto keep a prudent state of uncer- tive of Muridae, confirming the suprafamilial Myo- tainty of the hierarchical pattern of muroid suprageneric donta concept. Within the speciose family Muridae, groupsand to divide the family Muridae into 17 subfam- the first branching leads to the fossorial Spalacinae ilies consideredat the sametaxonomic level. These"ma- and semifossorial Rhyzomyinae. The remaining com- jor lineages"within murids are *, Calomysci- ponents of Muridae appear as a polytomy from which nae*, Cricetinae*, Cricetomyinae, *, are issued , Calomyscinae, Arvicolinae, Gerbillinae*, Lophiomyinae,*, Myospalacinae*, Cricetinae, Mystromyinae, , and some Mystromyinae*,Nesomyinae*, Otomyinae*, Petromysci- Dendromurinae ( and ). This nae, Platacanthomyinae, *, Sigmodonti- phylogeny is interpreted as the result of a bushlike nae*, and Spalacinae*(a star denotestaxa investigated radiation at the end of the early Miocene, leading to in this study). Murid speciesand generaare not equally emergence of most living subfamilies. The separation distributed among these subfamilies: most of them are between three additional taxa, Murinae, Gerbillinae, included in the Old World and mice, Murinae (re- and "Acomyinae" (which comprises the genera Aco- spectively,40 and 45%),the New World rats and mice, mys, Deomys, Uranomys, and Lophuromys), has oc- Sigmodontinae(32 and 28%), the and lemmings, curred more recently from a common ancestor issued Arvicolinae (11 and 6.5%),and the gerbils, Gerbillinae (7 from the main basal radiation. As previously shown by and 6%).Some ofthese subfamiliesare very depauperate, other molecular studies, the vlei rats, Otomyinae, are nested within Old World Murinae. ln the same way, such as the manedor crestedrats, Lophiomyinae(1 spe- the , Myospalacinae, appear strongly nested cies),the white-tailed mice, Mystromyinae(1 species),or within the , Cricetinae. Finally, we propose a the -likehamsters, Calomyscinae(1 and 6 sister group relationship between Malagasy Nesomyi- species). nae and south African Mystromyinae. @2000AcademicPress Although BorneMuridae species(i.e., from the genera Key Words: phylogeny; Muridae; Myodonta; radia- Mus, , Mesocricetus, Phodopus, Rattus, tion; LCAT. etc.) have been used often in laboratories for genetic, physiological, or behavioral studies, there remains a strong need to better define the taxonomic boundaries INTRODUCTION of these subfamilies and especially the relationships among them. Many important questions concerning The rodents of the family Muridae (as defined by the evolutionary origins of most of the 17 subfamilies, Musser and Carleton, 1993, and correspondingto the their rates of evolution, or the sister group relation- superfamily of McKenna and Bell, 1997) are ships between Muridae and other rodent families are the most diverse group of , encompassingat not yet adeQuatelyanswered.

280 1055-7903/00 $35.00 Copyright @ 2000 by AcademicPress AIl rights of reproduction in any form reserved. BUSHLIKE RADIATION OF MUROID RODENTS 281

Paleontologicalstudies have provided important in- DNAIDNA hybridization and molecular sequences sights into Sailleofthese questions.For example,fairly (128 rRNA and nuclear ribonucleases, respectively), goodfossil records in Saillelineages were the basis for confirmed the hypothesis of Denys et al. (1992)against estimating the Mus-Rattus dichotomy (= 12 million the monophyly of Murinae and Dendromurinae and years) (Jaeger et al., 1986; Jacobs et al., 1989, 1990; proposedto erect a new "Acomyinae" subfamily. This J acobsand Down, 1994) and the separation between additional muroid lineage clusters genera which were Spalacinae, Rhizomyinae, and the remaining living previously classified as members of Murinae (Acomys, Muridae (= 20 million years) (Flynn, 1990; Hugueney Uranomys,Lophuromys) or Dendromurinae (Deomys). and Mein, 1993). These data are generally used to A recent molecular analysis of the mitochondrial cyto- calibrate the molecular clocks in molecular studies. ln chrome b gene variation in several subfamilies of another way, clear morphological diagnoseswere evi- murids (Jansa et al., 1999),with a special emphasison dencedto define Saillesubfamilies, such as Gerbillinae Malagasy Nesomyinae, suggestedsome suprageneric (gerbils, jirds, and sand rats), Arvicolinae (voles, lem- relationships, but these results are here considered mings, muskrats), Cricetomyinae (pouched rats and doubtful as an inappropriate was selectedfor mice), Spalacinae (blind mole rats), Rhizomyinae rooting highly saturated sequences.Finally, Robinson (bamboo rats and African mole rats), or Cricetinae et al. (1997), in a study of LCAT (lecithin cholesterol (hamsters)(Ellerman, 1940, 1941;Ognev, 1963;Carle- acyl transferase) nuclear gene sequences,confirmed ton, 1980; Carleton and Musser, 1984; Jacobs et al., that 8palacinae and Rhizomyinae were early sepa- 1989; Catzeflis et al., 1992). On the contrary, Saille rated from five other subfamilies that were surveyed. subfamilies are more resistant to a morphological di- The remaining murids appeared as a polytomy from agnosis,such as Sigmodontinae (New World rats and which were issued Gerbillinae, Murinae, 8igmodonti- mice) or Nesomyinae (Malagasy rats and mice). Fi- nae, Cricetinae, and Arvicolinae. naIly, Saille other murid subfamilies might weIl prove However, the former study by Robinsonet al. (1997) para- or polyphyletic, such as Dendromurinae, whose consideredonly 7 of the 17 lineages listed by Musser monophyly was recently challenged by Denys et al. and Carleton (1993). To broaden the picture, we here (1995). enlarge the taxonomiesampling of rodents, with a spe- Dental and cranial characters commonly used in cial emphasis on Dipodidae (sampling 4 of 7 subfami- morphological and paleontological studies are often lies) and Muridae (13 of 17 subfamilies and four rep- subject to parallelism, convergence, and reversaIs resentatives of "Acomyinae").The nuclear gene LCAT events (Carleton and Musser, 1984; Catzeflis et al., was used, as Robinsonet al. (1997) demonstratedthat 1992). These homoplasic characteristics have re- it is a promising marker for this taxonomielevel. This stricted their use for inferring the relationships be- gene codesfor a key enzymein the reverse cholesterol tween aIl subfamilies and have yielded numerous con- pathway and consists of six exonstotaling 1320 nucle- flicting hypotheses, depending on which characters otides in Homo sapiens(McLean et al., 1986;Warden et were emphasized by the various students of muroid al.. 1989). systematics (Simpson, 1945; Honacki and Kinman, 1982; Chaline et al., 1977; Hooper and Musser, 1964). MATERIAL AND METHODS Molecular phylogenetic studies, if based on genes legs subject to homoplasy than traditional morpholog- DNA was extracted and purified from ethanol-pre- ical characters, could produce important complemen- servedtissues taken from the Collection ofMammalian tary information for a better understanding of the evo- Tissueshoused at Montpellier (Catzeflis, 1991).When- lutionary systematics of Muridae. Currently, tao few ever possible,we selectedtwo speciesfor each studied molecular analyseswith the aim of embracingthe sub- subfamily (see Table 1). This biological sampling was familial diversity of murids have been performed. aimed at obtaining an equilibrated representation of Saille studies have provided molecular signatures, eachround lineage and at diminishing a possiblelong- such as the presenceof a repetitive element called Lx branch attraction effect. for defining Murinae (Pascaleet al., 1990; Furano et al., 1994). Catzeflis et al. (1993), through DNA/DNA DNA Sequencingof LCAT Gene hybridization, proposed that cricetines, arvicolines, Two fragments (as in Fig. 1, p. 424, in Robinson et sigmodontines,and possiblyMystromys were clustered al., 1997) of the nuclear gene LCAT were amplified in a clade, separatefrom the murines, gerbillines, and using the PCR primers previously designedby Robin- spalacines. On the basis of 12S rRNA mitochondrial son et al. (1997).AlI PCRs used the following protocol: sequencesand DNA/DNA hybridization, Dubois et al. 5 min at 94°C, 33 cycles(45 s at 94°C, 30 s at 52°C,and (1996) suggestedthe monophyly of Nesomyinaeand a 1 min at 72°C), plus 10 min at 72°C in a Appligen sister group relationship of these with the African Crocodile 3 thermal cycler. Total reaction volume was Cricetomyinae. Chevret et al. (1993a, 1994), Hanni et 100 ILl. PCR products were purified using the Ultra- al. (1995), and Dubois et al. (1999), on the basis of free DNA Amicon kit (Millipore) and directlv se- 282 MICHAUX AND CATZEFLIS

quenced.Sequencing on bath strands was cloneusing a RESULTS dye terminator (Perkin-Elmer) sequencingkit and a ABI 373 (Perkin-Elmer) automatic sequencer. SequencedSpecies SequenceAlignment and Phylogenetic The 21 new rodent sequencesof the LCAT gene are Reconstructions indicated in Table 1, where rodent taxa are listed fol- lowing the taxonomie arrangement of Wilson and Previously known sequenceswere extracted from Reeder(1993). These sequences have beendeposited in GenBank and aligned with the new sequencesusing the EMBL gene bank under AccessionNos. AJ275513 CLUSTAL W (Thompsonet al., 1994) and the ED edi- to AJ275617. tor (MUST package;Philippe, 1993).The phylogenetic The newly determined sequenceshave been com- analysis was conductedon 804 nucleotidescorrespond- pared to 17 rodent sequencesdetermined by Robinson ing to the exonic regions of the two amplified frag- et al. (1997),as weIl as to Mus and Rattus (available in ments. The aligned sequenceswere treated by distance GenBank) (Table 1). We also conducteda few analyses (neighbor-joining, NJ; Saitou and Nei, 1987), maxi- with additional nonrodent taxa as outgroups: two Pri- mum-parsimony (MP), and maximum-likelihood (ML) mates (Homosapiens, GenBank Accession No. X04981; analysesusing, respectively, MUST (NJboot program; Papio anubis L08633) and one Lagomorpha (Oryctola- Philippe, 1993; Tamura and Nei (1993) distance esti- gus cuniculus D13668). mator), PAUP4.0b1 (NJ: ME criterion andTBRbranch swappingoption; MP: heuristic searchand TBR branch Nucleotide Characteristicsof LCAT swapping option (Swofford, 1998)), and PUZZLE ver- sion 4.0 (quartet puzzling procedure; Strimmer and and Analysisof Saturation Von Haeseler, 1996; Tamura and Nei (1993) model of The aligned data matrix includes 43 mammalian evolution and mixed (1 Inv + 8 gamma rates) model of speciesand 804 sites, 456 of which were variable and among-sites rates heterogeneity). The robustness of 317 phylogenetically informative when aIl events inferenceswas assessedthrough bootstrap resampling (transitions (TS) and transversions (TV)) are consid- (BP) (1000 repetitions) with the distance and parsi- ered; 53%of variable sites and 65%of informative sites mony approaches.ln the case of ML, the reliability concern the third position where most synonymous percentage (RP; Strimmer and Von Haeseler, 1996) changestake place. estimated the occurrenceof the nodes in the quartet As already shownby Robinsonet al. (1997),the mean puzzling trees after 1000puzzling steps.Bremer's sup- frequency of nucleotides in the sequencescompared port index (BSI) (Bremer, 1988)was also calculated on shows an overall high GC content in LCAT (22.0%A, the most-parsimonioustree with enforcementof topo- 28.2% C, 24.8% G, 25.0% T, 53% GC). This value is logical constraints. Likelihoods of alternative topolo- more pronounced in the third codon position (61.6% gies were comparedwith MOLPHY 2.3b3 (Adashi and GC3), where it ranges from 54.8 (Myospalax) to 77.9 Hasegawa, 1996) and PUZZLE (Strimmer and Von (Papio sp.). However, as already shown by Robinsonet Haeseler, 1996). According to Kishino and Hasegawa al. (1997),this variation in GC content doesnot seemto (1989), an alternative hypothesis was rejected when affect phylogeneticreconstruction. The averageratio of 81nL> 1.96 SE, where 8lnL is the difference between TSrrv is 1.90, ranging from 0.94 ( ehrenbergil the log-likelihoods of the best and those of the evalu- Spalax leucodoncomparison) to 3.56 (Peromyscus man- ated trees, and SE is the standard error of this differ- iculatuslAcomys cahirinus and Mystromys albicauda- ence. tuslAcomyscahirinus). Relative-RateTest The above differences in base composition and in rates ofTSrrv changesindicate that the data matri.x i8 Relative-rate tests were conductedwith RRTree,ver- heterogeneouswith regard to the different substitution sion 1.0 (Robinsonet al., 1998),which improves the test types at each codon position. To better locate ho- of Wu and Li (1985) by taking into account the taxo- moplasy, we searchedfor evidenceof saturation using nomic representativity and its phylogenetic relation- the method of Hassanin et al. (1998a,b). This analysis ships. Relative-rate tests were performed among ro- aims at determining the relative importance of multi- dents at supra- and intrafamiliallevels; the ML tree ple substitutions by comparing, in scatterplots, the derived from quartet puzzling was chosenas the refer- pairwise numbers of observedvers us inferred changes ence phylogeny. With regard to the different levels of of each of the six substitution types at each codon analyses (see below), various outgroups were chosen: position. Each of the 18 resulting scatterplots can be primates (Homo and Papio) for tests at the suprafamil- characterizedby the consistencyindex (CI) of the most- iallevel and Dipodidae for tests between the Muridae parsimonious tree and by the slope (S) of the linear subfamilies. Relative-rate tests were performed on the regression between observed and inferred changes. proportions of synonymous (Ks) and nonsynonymous Such information provides a rough idea of the level of (Ka) substitutions. saturation for eachkind ofnucleotide substitution (Ta- BUSHLIKE RADIATION OF MUROID RODENTS 283

TABLE 1 References of Rodent Tissues Used for the Experiments

Tissue Suprafamily Subfamily Species sample Geographic origin Collector Accession Nos.

Dipodoidea Sicistinae Sicista kazbegica T-762 Central Caucasus, Cew Valley, M. Baskevitch AJ275513ta AJ525517 RUBsia Allactaginae Allactaga elater T-I045 Turbat Jam, Iran Majad Zadé AJ275518to AJ275522 Zapodinae Napaeozapusinsignis T-240 Nova Scotia,Cumberland County, V. Volobouev AJ275523ta AJ275527 Canada Dipodinae Dipus sagitta T-869 Caucasus Mountains, Russia P. Gambarian AJ275528 ta AJ275532 Jaculus jaculus T-552 Djoudj, Senegal J.-M. Duplantier AJ275533 ta AJ275537 Muroidea Calomyscinae Calomyscus mystax T-I067 Caucasus Mountains, Russia P. Gambarian AJ275538 to AJ275542 Dendromurinae Steatomys sp. T-1167 Sapago, Burkina-Faso J. C. Gautun AJ275543 to AJ275547 Deomysferrugineus T-778 Goumina,Congo L. Granjon AJ275548ta AJ275552 Dendromusmystacalis T-1422 Transvaal, South Africa G. Bronner and D. AJ275553ta AJ275557 Bellars Gerbillinae Tateragambiana T-913 Robinsonet al., 1997 U72297ta U72298 Gerbillus henleyi T-1165 Robinsonet al., 1997 U72295ta U72296 Mystromyinae Mystromysalbicaudatus T-1365 Natal, Nottingham Rd, South G. Bronner AJ275558to AJ275562 Africa Nesomyinae Macrotarsomysingens T-1150 Ampijora, Madagascar D. Rakotondravony AJ275563to AJ275567 Nesomysrufus T-1125 Ranomafana,Madagascar D. Rakotandavony AJ275568ta AJ275572 Sigmodontinae Neotomafuscipes T-385 Monterrey Co, California, USA M. Salvioni AJ275613ta AJ275617 torques T-449 Robinsonet al., 1997 U72303ta U72304 Peromyscusmaniculatus T-142 Robinsonet al., 1997 U72307to U72308 Cricetinae Phodopusroborowski T-714 Laboratory-bred,Gôttingen, 1. Hansmann AJ275573ta AJ275577 Germany Mescocricetusauratus T-1162 Laboratory-bred,montpellier, F. Catzeflis AJ275578to AJ275582 France Cricetulus migratorius T-325 Robinson et al., 1997 U72305 ta U72306 Myospalacinae Myospalax sp. T-394 Unknown locality, Russia P. Gambarian AJ275583 to AJ275587 Arvicolinae Dicrostonyx torquatus T-1337 Taimyr peninsula, Siberia, Russia R. A. Ims AJ275588 to AJ275592 Microtus nivalis T-523 Robinsonet al., 1997 U72301to U72302 Clethrionomysglareolus T-357 Robinsonet al., 1997 U72299to U72300 Murinae Lophuromys sikapusi T-1179 Man, Ivory Coast J.-M. Duplantier AJ275593 ta AJ275597 Rattus norvegicus GenBank X54096 Mus musculus GenBank JO5154 Micromys minutus T-1196 Robinsonet al., 1997 U72293to U72294 Uranomysruddi T-1184 Kédougou,Senegal J.-M. Duplantier AJ275598ta AJ275602 Acomyscahirinus T-1670 Greta island, Greece P. Lymberakis AJ275603to AJ275607 Otomyinae 9tomys angoniensis T-718 Nylsulei, South Africa G. Contrafatta AJ275608to AJ275612 Spalacinae Nanospalaxehrenbergi T-268 Robinsonet al., 1997 U72309ta U72310 Nanospalaxleucodon T-I009 Robinsonet al., 1997 U72311ta U72312 Rhizomyinae Rhizomyspruinosus T-1284 Robinsonet al., 1997 U72313ta U72314 Octodontoidea Octodontidae Octodonlunatus T-I001 Robinsonet al., 1997 U72325to U72326 Dasyproctoidea Myocastoridae Myocastorcoypu T-245 Robinsonet al., 1997 U72323ta U72324 Gliroidea Gliridae Myoxusglis T-1453 Robinsonet al., 1997 U72317to U72318 Eliomys quercinus T-1499 Robinsonet al., 1997 U72315ta U72316 Sciuroidea ~~.~.u~~Q~;..";,Ioo Sciurus vulgaris T-1279 Robinsonet al., 1997 U72321ta U72322 Marmota kamtschatika T-1552 Robinsonet al.. 1997 U72319ta U72320

Note. The taxonomiearrangement follows that of Wilson and Reeder(1993). ble 2). The results show that the C-T and A-G transi- Phylogenetic Reconstructions tions exhibit a lower slope(average, for the three codon positions, of 0.59 for A-G and 0.63 for C-T) and consis- Analyses with 43 Eutherian Mammals tency index (averageof 0.33 for A-G and 0.35 for C-T) A first set of analyses (Table 3, Fig. 1) considered 40 with regard to the transversions, whatever the codon rodents, 2 primates, and 1 lagomorph for aIl coding position. Considering also the fact that the highest sequences (concatenations of exons 2, 3, 4, 5, and 6), nurnbers of informative characters are the result of that is 804 nucleotides, or 268 amino acids. We ex- transitional changes(Table 2), this analysis of satura- cluded three codons corresponding to autapomorphic tion justifies Bornedown-weighting for A-G and C-T insertions (one codon for Sciurus vulgaris and two for changes. Microtus nivalis). Consequently,in parallel to a classical unweighted The maximum-parsimony reconstruction based on parsirnonyanalysis (MP), we performed a secondanal- equal weighting of each nucleotide substitution yielded ysis (MPw) in which we weighted each substitution 24 most-parsimonious trees. Each tree is 1541 steps event according to its slope (taking 1000 tirnes this long, with a consistency index (excluding uninforma- value for each substitution cell to use the "steprnatrix" tive characters) of 0.38 and a retention index of 0.56. option of PAUP). Bootstrapping and Bremer's support index values are 284 MICHAUX AND CATZEFLIS

TABLE 2 Number of Informative Sites, Consistency Index, and Slope of Saturation for Each of the 18 Substitution Caviomorpha Types Sciuridae Number of 81opeof informative Consistency saturation Gliridae sites index (CI) (8) Sicista lst Codon position Jaculus A-C 22 0.53 0.83 Dipus DiDodidae A-G 47 0.32 0.57 Napaeozapus A-T 9 0.56 0.9 Allactaga C-G 24 0.61 0.82 C-T 41 0.38 0.66 G-T 21 0.51 0.83 2nd Codon position A-C 20 0.58 0.92 A-G 42 0.36 0.69 A-T 17 0.81 0.98 Muridae C-G 27 0.76 0.86 C-T 49 0.39 0.71 G-T 19 0.73 0.98 3rd Codon position A-C 26 0.55 0.92 A-G 41 0.32 0.52 A-T 15 0.65 0.83 C-G 19 0.61 0.86 C-T 64 0.29 0.54 G-T 17 0.53 0.84 FIG. 1. Synthetic tree summarlzing the results derived from three approacheson 43 mammalian DNA sequencesof the LCAT gene.The robustnessof each node OabeledA ta H) is describedin Table 3 for maximum-parsimony(bootstrap percentageand Brem- indicated in Table 3 for the ancestral segmentslabeled er's support index), distance analysis (bootstrap percentage),and A to H in Fig. 1. There exists a strong support for the maximum-likelihood (reliability percentages).The tree was rooted families Sciuridae (node D; BP = 99%; BSI = + 12), by the two primate sequences. Gliridae (node E: 100%, + 14), Muridae (node H: 99%; +8), and Dipodidae (node G: 98%; +6), and for the suprafamilial taxon Caviomorpha (node C: 100%; firms the Myodonta concept(Schaub, in Grassé, 1955) +23). This maximum-parsimony analysis also con- including the Dipodidae and Muridae families (nodeF with BP of 96% and BSI of +8). The weighted-parsi- mony analysis with the stepmatrix (taking into ac- TABLE 3 count a different saturation level for each kind of sub- stitution) did not show any interesting difference with Indices of Robustness for the Nodes of the Phyloge- netic Tree Represented in Fig. 1 Using Maximum-Par- regard to the classicalparsimony analysis (compareBP simony (MP), Distance, and Maximum-Likelihood values for MP and MPw in Table 3). To the contrary of (ML) Analyses other studies with mitochondrial genes (Hassanin et al., 1998a,b), this a priori weighting did not improve Nodes MP MPw BSI NJ ML the robustness of the maximum-parsimony tree. Reliability percentages (ML analysis with a esti- A 100 100 +23 100 90 B 96 97 +19 94 82 mated at 0.38 and estimated proportion of invariant C 100 100 +23 100 92 sites of 0.0) and bootstrap percentages(NJ with a = D 99 99 +12 100 85 0.38 according to the puzzle analysis) yield results E 100 100 +14 100 92 similar to those obtained with the parsimony analysis F 96 98 +8 95 89 (Table 3). AlI the ancestral segments that were G 98 99 +6 96 92 H 99 100 +8 97 81 strongly supported through parsimony are also re- trieved with a high robustness by the two other opti- Note. Bootstrap percentagescomputed after standard MP (with mality criteria. equal weighting), MP weighted by slopes of saturation profiles We checkedthat the Myodonta support was not due (MPw), and neighbor-joimng (NJ) on Tamura and Nei (1993) dis- tances with gamma rates (alpha = 0.38) are reported. Reliability to the particular taxonomie sampling of our data set, percentagesdeduced from the quartet puzzling ML methodsare also especially with regard to differences between the Di- given. Finally, Bremer support indices (BSls) are indicated. podidae (5 species)and the Muridae (29 species)rep- BUSHLIKE RADIATION OF MUROID RODENTS 285

Maximum-parsimony. The parsimony reconstruc- tion based on equal weighting of each nucleotide sub- stitution yielded eight most-parsimonioustrees. Each tree is 1096 steps long, with a consistencyindex (ex- cluding uninformative characters) of 0.40 and a reten- tion index of 0.53. Figure 2 indicates that, within Muri- dae, the first dichotomy isolates a clade (node 4 in Table 4) comprising Spalax and . The re- maining Muridae comprise a strongly supported clade (BP of 92% and BSI of +8: node 6). The monophyly of several subfamilies representedby at least two genera is robust: Nesomyinae(node 9), Arvicolinae (node 15), and Gerbillinae (node 20) are supported by BPs be- tween 84 and 100%and by BSIs between +7 and +12. On the contrary, as was already observedby Robinson et al. (1997),the New World rats and mice, Sigmodon- tinae (node 10), representedin this study by Peromys- eus, Neotoma, and Akodon, are poorly defined (BP of 45% and BSI of + 1). Maximum-parsimony analysis also clearly shows (BP of 98% and BSI of +9) that Acomys, Uranomys, and Lophuromys do not belong to the true Murinae (represented here by Mus, Rattus, Micromys) but are clustered in a suprageneric clade

TABLE 4 Indices of Robustness for the Nodes of the Phyloge- netic Tree with Myodonta Only Represented in Fig. 2 Using Maximum-Parsimony (MP), Distance, and Max- FIG. 2. Synthetic tree summarizing the results derived from imum-Likelihood (ML) Analyses three approaches on 34 Myodonta DNA sequences of the LCAT gene. The robustness of each node (labeled 1 to 22) is documented in Table Nodes MP MPw BSI NJ ML 4 for maximum-parsimony (bootstrap percentage and Bremer's sup- port index), distance analysis (BP), and maximum-likelihood (reli- 1 89 96 +4 82 94 ability percentages). The tree was rooted by five Dipodidae se- 2 88 81 +4 85 94 quences. 3 100 100 +10 100 83 4 67 51 +2 96 96 5 100 100 +12 100 98 resentations.Distance analysis (Tamura-Nei, a = 6 92 93 +8 91 70 0.38) was repeated 10 times with a random sample of 7 84 68 +6 93 87 five Muridae sequences,and the robustness of nodes 8 90 92 +6 90 49 9 was addressedby bootstrap. The averageBP was 87.2 66 56 +2 71 34 10 45 51 +1 65 82 for Myodonta (SD 9.9, range 74-98), 92.7 for Dipodidae Il 51 51 +1 57 65 (SD 5.2, range 82-98),and 93.9 for Muridae (SD 8.8, 12 97 93 +7 97 75 range 76-100). Thus, we feel confident that the Myo- 13 79 71 +2 76 72 donta clade is robust and reliable and that this node 14 88 76 +4 61 75 15 97 doesnot rely upon a particular choiceof its represen- 99 +12 99 88 16 96 98 +7 99 93 tative taxa. 17 65 60 +4 59 25 18 Analyses with Dipodidae only as Outgroup 98 98 +9 99 68 19 94 96 +6 89 72 As shawn by the previous analysis, the Myodonta 20 100 100 +10 100 95 21 99 100 +9 100 77 monophyly, as weIl as the naturalness of Dipodidae 22 99 and Muridae, appears weIl established. Thus, for the 99 +8 99 92 purpose of avoiding the use of tao-distant outgroups Note. As in Fig. 1, bootstrappercentages computed after standard with regard to within-Muridae relationships, we per- MP (with equal weighting), MP weighted by slopes of saturation formed a second set of analyses using only the five profiles (MPw), and neighborjoining (NJ) on Tamura and Nei (1993) distanceswith gammarates (alpha = 0.41) are reported. Reliability Dipodidae as outgroup for a monophyletic Muridae percentagesdeduced from the quartet puzzling ML methodsare also (Fig. 2 and Table 4). given. Finally, Bremer support indices (BSIs) are indicated. 286 MICHAUX AND CATZEFLIS

(node 18), which also includes Deomys (traditionally previously obtained by the distance and parsimony classifiedwith Dendromurinae).Following Hânni et al. criteria. (1995,p. 132),we name "acomyines"or [provisionally] Likelihood alternatives to the besttree. The highest- "Acomyinae" as the clade containing the genera likelihood tree was used as the reference topology to Acomys,Deomys, Lophuromys, and Uranomys. apply the test of Kishino and Hasegawa (1989) for A new finding for nonmorphologicalstudies ofmurid assessingthe following clades: (1) the monophyly of systematics is the nesting of Myospalacinae within eachof the "Acomyinae"and Sigmodontinaegroups; (2) Cricetinae (BP characterizing Cricetinae + Myospal- the sister group relationships between (a) Mystromyi- acinae of 97%; BSI of +7) (node 12). As previously nae and Nesomyinaeand (b) Gerbillinae, Murinae, and shownby Chevret et al. (1993b),Otomyinae (Otomys)is "Acomyinae;"and (3) the nested position of (a) Otomyi- included within Murinae (as the sister genus of Mus: nae within Murinae and (b) Myospalacinae within node 22) with a strong support: BP of 99% and BSI of Cricetinae. +8. The ancestral segment(numbered 8 in Table 4 and For doing gO,we tested different alternative topolo- Fig. 2) uniting the Malagasy Nesomyinae(represented gies derived from traditional morphological and pale- by Nesomysand Macrotarsomys)with the South Afri- ontological studies or from various molecular hypoth- can Mystromyinae (represented by Mystromys) is eses:Acomys, Uranomys, and Lophuromys in Murinae strongly supported(BP of 90% and BSI of +6). Finally, (Carleton and Musser, 1984; Musser and Carleton, parsimony suggestsa clade comprising Murinae, Ger- 1993); Deomys in Dendromurinae (Carleton and billinae, and" Acomyinae" (node 17), although with a Musser, 1984; Musser and Carleton, 1993); Peromys- poor support (BP of 60-65% and BSI of +4). cus (Sigmodontinae)with Clethrionomys (Arvicolinae) The weighted parsimony analysis (with stepmatri- (Dickerman, 1992); Neotoma (Sigmodontinae) with ces) gave approximately the same values as the un- Eurasian Cricetinae (Dickerman, 1992); Myospalax weighted analysis and did not improve the robustness with Spalax and Rhizomyinae (Miller and Gidley, of the inferences (compare columns MP and Mpw in 1918); Mystromys with Cricetinae (Carleton and Table 4). An unexpected observation was that Borne Musser, 1984); Otomys as the sister taxon of Murinae nodeswere legs supported (i.e., for node 7: BP of 84% (Thomas, 1896; Misonne, 1971; Chaline et al., 1977); for equal weighted analysis and 68%for weighted anal- Murinae, Gerbillinae, and Acomyinae paraphyletic ysis) when saturation was taken into account. (Ameur, 1984; Flynn et al., 1985). Neighbor-joiningand maximum-likelihood. The clus- AlI these alternative topologies exhibited a signifi- tering of Spalacinaeand Rhizomyinae as the earliest cantly worse log-likelihood than the one measuredfor offshoot from murid ancestorswas strongly supportedby the highest-likelihood tree. Based on these tests, we NJ and ML analyses(respectively, BP and RF of 96%), maintain the previously mentioned relationships, in comparedto the poor robustnessobserved through max- particular the monophyly of each of the two subfami- imum-parsimonymethods (node 4 in Table4). Similarily, lies "Acomyinae"and Sigmodontinaeand the inclusion the monophylyof Sigmodontinaeis alsomore robust with of Otomyinae and Myospalacinaewithin Murinae and maximum-likelihood(RP = 82%). For the NJ analysis, Cricetinae, respectively. branching pattems and bootstrap values were similar Relative-RateTests using either MUST or PAUP. The use of the Tamura and To identify whether differences in rates of LCAT Nei (1993) model of substitution along with a mixed change existed in the major taxa of rodents (Gliridae, model of among-sitesrate heterogeneitycould explain Sciuridae, Dipodidae,Muridae, and Caviomorpha),rel- the better performancesof distanceand likelihood opti- ative-rate tests were conducted with each of them mality criterions, becausesuch approachestake into ac- against the remaining lineages.K. comparisonsdid not countthe heterogeneitiesexisting in the data matrix. The evidencesignificant differencesin relative rate in the higher (Rhizomys)or slower (Sigmodontinae)rates of different groups.To the contrary, Ka comparisons(non- evolution characterizingBorne of these taxa (seebelow) synonymous changes) showed marked differences in could also explain the relative differencein robustness evolutionary rates: Dipodidae and Sciuridae appear to betweenmaximum-parsimony and other approaches.AlI be slowly evolving taxa (respectively, P < 0.001 and other nodesthat were strongly supportedthrough maxi- P < 0.03). Among Sciuridae, more detailed analyses mum-parsimony analysis are also retrieved as strong showedthat only Marmota had a slow rate of evolution ancestralsegments in the maximum-likelihoodand dis- (P < 0.003).On the other hand, within Dipodidae,four tance results (Table 4). of the five studied speciesshowed a significantly lower The highest-likelihood tree (lnL = -6687.93) on 34 rate of nonsynonymouschange (Sicista, Jaculus, Di- Myodonta specieswas identified with PUZZLE (Strim- pus, and Napaeozapus)(P < 0.01).Muridae as a whole mer and Von Haeseler, 1996) among 945 alternative did not have a particularly fast rate of evolution. trees constructed using MOLPHY 2.3b3 (Adashi and Relative-rates tests were then performed among Hasegawa, 1996). This tree has the same topology as Muridae using the slowly evolving Dipodidae as out- BUSHLIKE RADIATION OF MUROID RODENTS 287

TABLE 5 Estimations of the Separation Times of Different Events within the Muroids, on the Basis of the Molecular Data

Calibration point Calibration based on the point based on separation: the separation: Spalax/modern Mus 1Rattus Separation events muroids (in Myr) SE (in Myr) SE Paleontologicalestimations

Spalax/modern muroids 20 21.7 1 20 Myr (Hugueneyand Mein, 1993) Mus/Rattus 11.5 1 12 12 Myr (Jaegeret al., 1986; Jacobsand Down, 1994) Gerbillus / Tatera 7.8 1 8.8 1.1 8-10 Myr (Tong, 1989) Clethrionomys/ Microtus 7.4 0.7 8 0.8 3, 5-6 Myr (Chaline and Graf, 1988) Myospalax/ Phodopus 4.5 0.7 5 0.7 2 Myr (Chaline et al., 1977; Carleton and Musser, 1984) Radiation of modern muroids 16.8 0.5 19 0.5 18 Myr (Tong and Jaeger, 1993) Steatomys 1Dendromus 9.5 1.3 10.7 1.4 8-11 Myr (McKenna and Bell, 1997) Gerbillinae/M urinae/ Acomyinae 15.4 0.7 16.6 0.7 16 Myr (Tong and Jaeger, 1993)

Note. The numbersin boldfacecorrespond to the two calibration points usedfor this analysis:20 Myr for the separationbetween Spalacinae and modem Muroids (Hugueneyand Mein, 1993);12 Myr for the separationbetween Mus and Rattus (Jacobset al., 1986;Jacobs and Down, 1994).SE, standard error. group. As previously, synonymous (K. values) changes ration time between Spalacinaeand aIl the remaining did not show significant differences between the differ- living Muridae estimated at approximately 20 Mybp ent Muridae subfamilies. However, Ka comparisons (Hugueneyand Mein, 1993).The ML distance between showed that Nesomyinae and Sigmodontinae were Mus and Rattus is 0.047,whereas that betweenSpalax slowly evolving (respectively, P < 0.05 and P < and aIl remaining murids is 0.085.These values give a 0.002) and that Rhizomyinae was rapidly evolving rate of 0.0039 (Mus/Rattus) or 0.0042 (Spalax/other (P < 0.02). Within Malagasy rodents, further analyses Muridae) ML distance per million years. When these showed that only had a lower rate of evolu- rather similar rates are applied to the different dichot- tion. Within New World Sigmodontinae, aIl three gen- omieswithin Muridae, the following molecular datings era (Neotoma, Peromyscus, and Akodon) were signifi- are obtained: 16.8 to 19 Mybp for the initial bushlike cantly slowly evolving (respectively, P < 0.001, P < radiation; 15.4to 16.6Mybp for the separationbetween 0.05, and P < 0.02). Murinae, Gerbillinae, and Acomyinae;7.8 to 8.8 Mybp The different results obtained for K. and Ka can be between the gerbil genera Gerbillus and Tatera; 4.5 to explained by the fact that synonymous substitutions 5.0 Mybp between the Phodopus(Cricetinae) saturate at the suprafamilial level, as was already and the Myospalax; 7.4 to 8.0 Mybp betweenthe suggested by the saturation analysis (see above and voles Microtus and Clethrionomys; and 9.5 to 10.7 Table 2). Mybp between the two dendromurines Steatomysand Consequently, to apply a molecular clock and esti- Dendromus. Table 5 provides these values and their mate dates of separation between the murid genera SE in comparisonto paleontologicalestimates. and the subfamilies, we performed another maximum- likelihood analysis with Dipodidae as outgroup and aIl DISCUSSION the Muridae except the slowest and fastest evolving species (Nesomys, aIl Sigmodontinae, and Rhizomys). Molecular EvolutionaryRates The inferred maximum-likelihood distances were the basis for estimating separation times. Two calibration As for another nuclear protein-coding gene which points derived from paleontological data were chosen: was sequencedin representatives of several rodent (1) the Mus/Rattus dichotomy set at 12 millions years families (exon 28 of von Willebrandt Factor gene: Hu- before present (Mybp) (Jaeger et al., 1986; Jacobs et al., chon et al., 1999),LCAT sequencesshow that Muridae 1989. 1990: Jacobs and Down. 1994) and (2) the sepa- are not especiallv rapidlv evolvinŒmammals. This re- 288 MICHAUX AND CATZEFLIS

suIt contrasts with results of other nuclear (DNAIDNA omy at 12 Mybp) suggests that this separation oc- hybridization; Catzeflis et al., 1987;sequences: Li et al., curred approximately 20 Mybp, a dating in good agree- 1987)and mitochondrial (Philippe, 1997)studies, most ment with paleontological inferences (Flynn et al., of them comparing a few murids with nonrodent euth- 1985; Hugueney and Mein, 1993; Mein and Ginsburg, erian mammals. According to Huchon et al. (1999) and 1997). Thus, this study clearly identifies mole rats, Robinsonet al. (1998), part of "this result is probably Spalacinae, and bamboo rats, Rhizomyinae, as the sis- the consequenceof the use of a "topology-weighted" ter group of all remaining murid families, and this procedurein the computation of the relative-rate test." result conHicts with Jansa et al. (1999), who made the Another reason could be that the previous studies an- a priori choice of Calomyscus for rooting several sub- alyzed only a limited number of Muridae subfamilies families of Mrican and Asian murids. (at most Murinae, Cricetinae, and Arvicolinae; Catze- The remaining 12 subfamilies are clustered together flis et al., 1987; O'hUigin and Li, 1992) or, most com- in a strongly supported clade (node 6 on Fig. 2). The monly, just the two laboratory-bred murines Mus and first branching event of this clade is a large polytomy, Rattus. leading to seven lineages, of which three encompass Thus, our study indicates that, if a sufficient sam- more than 1 subfamily: Mystromyinae and Nesomyi- pling of Muridae representatives is considered, this nae (node 8: Fig. 2 and Table 4), Cricetinae and Myo- speciosefamily will exhibit both slowly evolving (Sig- spalacinae (node 12), and Murinae, Otomyinae, Ger- modontinae) and rapidly evolving (Rhizomys) taxa. billinae, and Acomyinae (node 17). Thus, most of the Consequently,in comparison with other rodent fami- "advanced" murid subfamilies appear to be of a polyto- lies (Sciuridae, Gliridae, etc.), murids on average do mous origin, indicating the phenomenon of a spectac- not appear to have a particular pattern of evolution. ular bushlike radiation having led to the majority of them. Because ancestral segments with high bootstrap Relationships between Dipodidae and Muridae: support are retrieved deeper into the trees of Fig. 1 the Myodonta Concept (node H: Muroidea; node F: Myodonta) and Fig. 2 (node The concept of a sister group relationship between 3: Muridae), our inference for the existence of this Dipodidaeand Muridae was proposedfor the first time radiation is probably real and not due to artifacts re- by Schaub(in Grassé,1955) on the basis of morpholog- lated to saturation and homoplasy. On the basis of the ical characters, uniting these taxa into the infraorder molecular data, this event occurred 17-19 Mybp (Table Myodonta. Later, in a comparative myological study, 5). This approximation is in accordance with the Klingener (1964) evidencedtwo exclusive synapomor- records (Hartenberger, 1985; Baskin, 1986; Jacobs et phies for the taxon Myodonta: the lack of differencia- al., 1989; Tong and Jaeger, 1993); the oldest tion of Musculus adductor magnus into M. adductor considered direct ancestors of living murids, such as minimus and M. adductor magnus proprius and the Potwarmus thailandicus, are dated ca. 18 Mybp (Mein separation of M. femorococcygeusfrom M. biceps fem- and Ginsburg, 1997). Other molecular studies (DNA/ oris by the posterior femoral cutaneous nerve. Other DNA hybridization: CatzeHis et al., 1993; nuclear embryological (see review in Luckett and Harten- LCAT gene: Robinson et al., 1997) have also estimated this radiation at "'" 18 Mybp. According to Aguilar et al. berger, 1985) and molecular (Serdobova and Kramerov, 1998) studies confirmed the taxonomie (1996, 1999), this period (end of early Miocene) was value of this infraorder. characterized by changes of climate, which favored the However, until now, nuclear (VWF gene; Huchon et spread in Europe, northern Africa, and the Middle al., 1999)and mitochondrial (12S rDNA; Nedbal et al., East of allochtonous groups such as the extinct cricetid 1996) sequencesgave a weak support for the mono- rodents (Democricetodon), probably coming from Asia. phyly ofthis infraoder. Thus, the phylogeneticsignal of It is also during this period of time that other cricetid the LCAT gene,which strongly clusters Dipodidae and rodents (Mrocricetodontinae) invaded, coming from Muridae, gives additional support tothe morphological Asia, the other Mrican regions, and Madagascar, lead- hypothesis for the Myodonta. ing later to the appearance of modern Mrican subfam- ilies (Lophiomyinae, Cricetomyinae, Dendromurinae, An Early Isolation of Spalacinae and Rhizomyinae and Nesomyinae, Mystromyinae) (Lavocat, 1973, 1978; an Explosive Radiation Leading to dIe Remaining Bernor et al., 1987). Concerning the New World, the extinct genus Copemys, a taxon related to Democricet- Modem Muridae Subfamilies odon (see discussion in Carleton and Musser, 1984), As already shown in Robinson et al. (1997), our re- migrated at the end of early Miocene (Flynn et al., sults confirm the hypothesis derived from the fossil 1985) from the Palearctic to North America, where it record of an early separation of Spalacinae and Rhizo- Hourished and gave rise to the ancestors of Sigmodon- myinae from other Muridae (Flynn et al., 1985; Flynn, tinae (Martin, 1980; Carleton and Musser, 1984); 1990, Hugueney and Mein, 1993). The LCAT molecular Baskin (1986) describes Abelmoschomys simpsoni, a clock (calibrated on the basis of the Mus/Rattus dichot- fossil dated at ca. 9 Mybp, as the oldest direct ancestor BUSHLIKE RADIATION OF MUROID RODENTS 289 for the living New World rats and mice. Thus, combin- aration betweenthem with regard to the other subfam. ing our molecular evidencewith paleontological data ilies of Muridae. and interpretations, we suggestthat the early Miocene bushlike radiation of Muridae was associatedwith and The Existenceof an AcomyineGroup and the Paraphyly immediately followed by a large worldwide spread of of Dendromurinae several ancestral Asiatic cricetid rodents. The LCAT gene study confirms with much robust- ness that Acomys, Lophuromys, Uranomys, and Deo- A Sister Group Relationshipamong Murinae, mys are clustered in a suprageneric clade, which we Gerbillinae,and Acomyinae calI the "Acomyinae"group. This inference is in agree- ment with previous molecular data (Chevret et al., Traditional paleontologicalhypotheses never classi- 1993a; Furano et al., 1994; Hanni et al., 1995; Ver- fied or associatedGerbillinae with Murinae (Simpson, heyen et al., 1996; Dubois et al., 1999), which consid- 1945; Chaline et al., 1977; de Graaf, 1981; Ameur, ered a reduced taxonomie sampling pertaining to this 1984; Flynn et al., 1985), most probably becausetheir question. However, a monophyletic acomyine subfam- dental patterns are so different. A comparative chro- ily is seriously at odds with traditional systematics mosomal study (Viegas-Pequignotet al., 1986) sup- based on comparative morphoanatomy(no morpholog'- ported this view, arguing that a greater similarity in ical signature is known for this group, especially con- karyotypes was observedamong Murinae, Cricetinae, sidering the inclusion of Deomys). and Sigmodontinae than between any of these and The separation between Deomys and the two other , Nesomyinae, Arvicolinae, and Gerbilli- studied Dendromurinae (Dendromus and Steatomys) nae. Molecular studies based on 12S rRNA sequences provides additional evidencefor the paraphyly of this (Hanni et al., 1995; Dubois et al., 1996)also suggested subfamily, as already suggestedby Denys et al. (1995) that Gerbillinae were external to a clade uniting Crice- and Verheyen et al. (1996) through comparative mor- tinae + Murinae. Nevertheless, this topology might phology and molecular data. have been the result of homoplasyrelated to the com- bination of a rapidly evolving mitochondrial gene with OtomyinaeAre Nestedwithin Murinae a too-distant outgroup (Gliridae only, in these refer- As already shown by other molecular (Chevret et al., ences). 1993b,unpublished; Usdin et al., 1995) and paleonto- Although the resampling support is weak (BP values logical (Senegasand Avery, 1998)studies, we confirm a between 59 and 65: node 17 in Table 4), our results closerelationship between Otomyinae (representedby tentatively suggestthat Gerbillinae, "Acomyinae,"and Otomys)and Murinae. This result and results of mor- Murinae had a more recent common ancestor with phological studies (Chevret et al., 1993b;Senegas and regard to the other Muridae subfamilies. The robust- Avery, 1998) suggest that this subfamily should be ness of this cluster improves when LCAT sequences invalidated and that the Otomyinae should be consid- are combinedwith 12SrRNA data (BP of 86% and BSI ered a tribe of Murinae, despite the tremendousdiffer- of Il in MP; P. Chevret et al., unpublished); similarily, encesin the dental patterns ofvlei and karoo rats with the Kishino and Hasegawatest (1989) also confirms a regard to the remaining Old World rats and mice. closer relationship between these three subfamilies. This result is congruent with DNNDNA hybridiza- An African Origin for the MalagasyNesomyinae tion studies (Brownell, 1983; Chevret et al., 1993a, Our results strongly suggesta sister group relation- unpublished; Catzeflis et al., 1992, 1993). From a pa- ship between the two Malagasy Nesomyinae genera leontologicalpoint ofview, recent studies (Jaegeret al., and the South Mrican Mystromyinae. This result is 1985; De Bruyn and Hussain, 1985; Tong, 1989; Tong congruent with the hypothesis of Lavocat (1973, 1978) and Jaeger, 1993) proposedthe separation of Gerbilli- who proposed to ally Nesomyinae with other archaic nae and Murinae from a commonancestor at =16-18 African groups such as Mystromyinae, Cricetomyinae, Mybp. According to these authors, this event appeared and Lophyomyinae; Chaline et al. (1977) united these 2 millions years after an initial separation from other taxa into the family . lndeed, aIl these cricetid muroids. This paleontological scenario is in murids would represent derivatives of an old African accordancewith the divergencetimes estimated in the cricetodontine stock. On the basis of morphological molecular analyses (Table 5), namely 15.4 to 16.6 characters, Carleton and Musser (1984) also proposed Mybp for the Gerbillinae/Murinae split, subsequent the association of the white-tailed hamster, Mystro- from the initial radiation of modern muroids set at 16.8 mys, with N esomyinae. However, other molecular to 19.0 Mybp. studies suggested that the more recent relatives of Thus, our molecular results give some evidencefor Nesomyinaewere Cricetomyinae(12S rRNA: Dubois et confirming a sister group relationship among Gerbilli- al., 1996)or Murinae (cytochromeb gene:Jansa et al., nae. Murinae. and "Acomvinae." imnlving: a later sen- 1999). Neverthele~~- hoth m]T Rnd n]]hoi~'~ pt n[ (l.Q.QR) 290 MICHAUX AND CATZEFLIS studies surfer from a poor taxonomie sampling with A CloseRelationship between Palearctic Myospalacinae regard to Nesomyinae representation. Sequencesof and Cricetinae Cricetomyinae could not be obtained for the LCAT gene. On the other hand, the results and interpreta- Owing to their particular morphology,the Myospal- tions of Jansa et al. (1999) remain doubtful for two acinae have been associatedwith other groups of fos- reasons: their outgroup taxon (Calomyscus)was not sorial rodents such as the mole rats, (Miller adequateto root the different Asian and Mrican sub- and Gidley, 1918), or the bamboo rats, Rhyzomyinae families and most of the cytochromeb variation at such (Tullberg, 1899).Later, Chaline et al. (1977)preferred supragenericlevels is random homoplasy due to satu- to consider this group as a distinct subfamily and pro- ration (data not shown). Clearly, more analyses are posedthat they evolvedfrom Eurasian cricetodontines needed for clarifying these relationships, especially during the Pleistocene.ln the sameway, Carleton and Musser (1984) concludedthat Myospalax "is a primi- sampling, through moderately evolving genes(such as nuclear LCAT or nonsynonymoussubstitutions in mi- tive cricetid that probably becamefossorially adapted tochondrial genes), additional taxa of Cricetomyinae before the Gobi region becamearid." Our results indi- cate that Myospalacinaeare nested within Cricetinae (Cricetomys,, and )and of Ne- (particularly with the Asiatic speciesPhodopus) (see somyinae. Table 4 and Fig. 2). For this reason, we propose to Are the New World Rats and Mice Paraphyletic? invalidate this subfamily and to consider the genus New World rats and mice have been occasionaIly Myospalax (sole living member of "Myospalacinae")as defining a tribe amongthe subfamily Cricetinae. More- grouped together with Palearctic hamsters in the fam- ily (EIlermann, 1941; Simpson, 1945). over, the divergencetime estimation basedon our mo- lecular data suggests that the separation between Later, other paleontologists and morphologists pro- Myospalax and the other Cricetinae appeared during posed to separate the New World species as a distinct Early Pliocene (4-5 Mybp). Until recently, the oldest subfamily (Hesperomyinae of Chaline et al., 1977), fossils attributed with confidence to zokors were at whose name is Sigmodontinae (Reig, 1980; Carleton most of Upper Pliocene (ca. 2 Mybp: Lawrence, 1991). and Musser, 1984). Moreover, Hooper and Musser The synthesis by Zheng (1994) documentedthe fossil (1964) divided this speciose subfamily (about 80 gen- era) into two groups according to the morphology of the Episiphneussinensis, dated at ca. 4 Mybp, as the direct ancestor for living Myospalax spp., thus a temporal glans penis: a simple type, characteristic of North estimate in good agreementwith the molecular dating American species and a complex type distributed ofthis study. Inferencesfrom the LCAT gene disagree among South American species. These two morpholog- with Lawrence's (1991, p. 282) opinion by which "the ical archetypes led to distinct two tribes: the North American Peromyscini and the South American Sigmo- equal division of [morphological] characters between dontini. ln 1980, Reig proposed to designate the tribes plesiomorphic murid features and derived characters associatedwith fossorial adaptation lends support to as subfamilies: Neotominae and Sigmodontinae, re- the proposaI that myospalacinesare derived from a spectively. However, Carleton (1980) cautioned that "formaI recognition of the two assemblages as subfam- primitive murid stock. . .." ilies had not been convincingly demonstrated." On the basis of DNNDNA hybridization, Dickerman (1992) Differencesbetween Datings Estimated by Fossils and Catzeflis et al. (1993) confirmed a phylogenetic and Molecules dichotomy between North American and South Amer- Although most of the separation times estimated on ican cricetids and proposed to consider them two dis- the basis of the molecular data are in good agreement tinct subfamilies. with those obtained through the fossil records (Table Our data involve only three genera, but they confirm, 5), one seemsunclear. The value of7.4 to 8.0 (SE = 0.8) although with a weak support, the monophyly of the Mybp for the split betweenthe genera Clethriono- New World Sigmodontinae. Based on a large taxo- mys and Microtus is much older than the dating at 3.5 nomic sampling of 38 South American and 4 North to 6.0 Mybp suggestedby Chaline and Graf(1988). Our American genera of Sigmodontinae, Smith and Patton estimate is similar to that calculated by Robinsonet al. (1999) evidenced a weak support for the monophyly of (1997),also obtained with the nuclear LCAT gene,but the group based on the mitochondrial cytochrome b is at oddswith other molecular studies basedon DNA/ gene. For the LCAT gene, the alternative topologies DNA hybridization (Catzeflis et al., 1987) or on the contradicting this monophyly always exhibit signifi- nuclear ribonuclease gene sequences(Dubois et al., cantly worse log-likelihood, suggesting confirmation of 1999). The relative-rate tests for LCAT showed that the Sigmodontinae concept. Moreover, the fact that aIl the two vole taxa do not evolve at a particular rate of three Sigmodontinae examined here are characterized evolution with regard to the other Muridae. Thus, the by a slower rate of DNA change corroborates this hy- molecular clock of the LCAT gene seemsalso valid for pothesis. these arvicoline taxa. ln conclusion. we SUf!f!estthat BUSHLIKE RADIATION OF MUROID RODENTS 291 additional taxa related to the Microtus/Clethrionomys remotecountries: Marina Baskevitch,Gary Bronner, GiancarloCon- divergenceshould be examined for their molecular di- trafatto, D. Bellars, Jean-Marc Duplantier, Piotr Gambarian,Jean- vergenceto confirm our 6.5 to 8 Mybp dating. If that Claude Gautun, Laurent Granjon, 1. Hansmann, Rolf Anker Ims, Petros Lymberakis, Daniel Rakotondravony,Marco Salvioni, Majad dating is confirmed, the interpretation of the fossil Zadé,and Vitaly Volobouev.Thanks are also expressedto Emmanuel record leading to the paleontological estimate (3.5 Douzery, DorothéeHuchon, PascaleChevret, Jean-JacquesJaeger, Mybp) should be reconsidered. Jacques Michaux, and Jean-Louis Hartenberger for their helpful discussionsabout analysesof saturation, relative-rate tests, and the phylogenyofMuridae. A specialthanks to Marie-Claire Beckersand CONCLUSIONS Corinne Lautier for their advise in automatic sequencingand to ProfessorF. Gricorescu(Institut Universitaire de RechercheClin- This molecular study was performed on a nuclear ique, Montpellier) for the use of bis automatic sequencer.This is genesequenced for representativesof 13 of 17 Muridae contribution ISEM 2000-098of Institut des Sciencesde l'Evolution subfamilies and 4 of 7 Dipodidae subfamilies. This de Montpellier (UMR 5554 CNRS-UniversitéMontpellier 11). taxonomie sampling led us to confirm the Myodonta infraorder including Muridae and Dipodidae. Within REFERENCES the largest mammalian family, Muridae, the use of the Adachi, J., and Hasegawa,M. (1996). MOLPHY: Programs for mo- LCAT gene evidenced the following results: (1) that lecular phylogenetics Version 2.3 edit. Institute of Statistical murids do not have a faster-evolving rate of change Mathematics, Tokyo. with regard to other rodents such as glirids, sciurids, or Aguilar, J. P., Clauzon, G., de Goer de Herve, A., Maluski, H., caviomorphs;(2) the monophyly of several subfamilies Michaux,J., and Welcomme,J. L. (1996).The MN3 fossil - including the Spalacinae,Nesomyinae, Cricetinae, Ar- bearing locality of Beaulieu (France):Biochronology, radiometric vicolinae, Gerbillinae, and Sigmodontinae;this result dating, and lower age limit of the Early Neogenerenewal of the is neverthelesspreliminary, as most of these taxa were mammalian fauna in Europe.Newsl. Stratigr. 34: 171-191. Aguilar, J. P., Escarguel,G., and Michaux, J. (1999).A successionof represented by too few genera in this study; (3) that Miocene rodent assemblagesfrom fissure fillings in southern Myospalacinae(fossorial zokors) should be invalidated France:Palaeoenvironmental interpretation and comparisonwith and considereda tribe of Cricetinae; in the same way, Spain. Palaeogeogr.-Palaeoclimat. Palaeoecol. 145: 215-230. we propose to abandon the subfamilial rank for Oto- Ameur, R. (1984).Découverte de nouveauxrongeurs dans la forma- myinae and to include them as a tribe within Murinae; tion miocène de Bou Hanifia (Algérie Occidentale).Géobios 17: (4) the confirmation of the Acomyinaesubfamily, which 165-175. comprises genera previously classified in Murinae Baskin, J. A. (1986).The late Mioceneradiation of Neotropical sig- modontinerodents in North America. Contr. Geol.Univ. Wyoming (Acomys,Lophuromys, and Uranomys) or in Dendro- Special Paper 3: 287-303. murinae (Deomys);as a consequence,the traditional Bernor, R. L., Brunet, M., Ginsburg, L., Mein, P., Pickford, M., Rtigl, Murinae and Dendromurinae (as of Carleton and F., Sen,S., Steininger, F., and Thomas,H. (1987).A consideration Musser, 1984; Musser and Carleton, 1993) are of Bornemajor topics concerningOld World Miocenemammalian paraphyletic and/or polyphyletic; (5) the following sis- chronology,migrations and paleogeography.Geobios 20: 431-439. ter group relationships: (a) Spalacinae and Rhizomyi- Bremer, K. (1988).The limits ofamino Reidsequence data in angio- nae, which were early separated from aIl other Muri- sperm phylogeneticreconstruction. Evolution 42: 795-803. dae; (b) Mrican Mystromyinae and Malagasy Brownell, E. (1983).DNA/DNA hybridization studies of muroid ro- dents: Symmetry and rates of molecular evolution. Evolution 37: Nesomyinae;this relationship should be further tested 1034-1051. by including other African archaic murids such as Carleton, M. D. (1980). Phylogenetic relationships in neotomine- Petromyscusand Cricetomyinae;and (c) Murinae, Ger- peromyscinerodents (Muroidea) and a reappraisal of the dichot- billinae, and Acomyinae;and (6) that, when calibrated omy within New World Cricetinae. Misc. Publ. Univ. Michigan by two paleontological data sets within muroids, the 157: 1-146. datings of different separation events estimated on the Carleton, M. D., and Musser, G. G. (1984).Muroid rodents.ln "Or- ders and Families of RecentMammals of the World" (S. Anderson basis of molecular data confirm those documentedby and J. K Jones,Eds.), pp. 289-379. Wiley, New York. the fossil record. Catzeflis, F. M., Sheldon, F. H., Ahlquist, J. E., and Sibley, C. G. FinaIly, our results strongly suggestthat there bas (1987). DNA-DNA hybridization evidence of the rapid rate of been a bushlike radiation, = 18 Mybp, leading to the muroid rodent DNA evolution. Mol. Biol. Evol. 4: 242-253. majority of the Muridae subfamilies. This interpreta- Catzeflis, F. (1991). tissue collectionsfor molecular genetics tion should now be tested by the use of another nuclear and systematics.Trends Ecol. Evol. 6: 168. gene, encompassinga similar biodiversity for an ade- Catzeflis, F. M., Aguilar, J.-P., and Jaeger, J.-J. (1992). Muroid quate representation of this speciosefamily. rodents: Phylogenyand evolution. Trends Ecol. Evol. 7: 122-126. Catzeflis,F. M., Dickerman,A. W., Michaux, J., and Kirsch, J. A. W. (1993). DNA hybridization and rodent phylogeny. ln "Mammal ACKNOWLEDGMENTS Phylogeny" (F. S. Szalay, M. J. Novacek, and M. C., McKenna, Eds.), pp. 159-172. Springer-Verlag,New York. This researchwas funded by a grant from the European Commu- Chaline, J., and Graf, J.-D. (1988). Phylogeny of the Arvicolidae nit y Project ERB-FMRX-CT98-0221.We thank aIl peoplewho pro- (Rodentia):Biochemical and paleontologicalevidence. J. Mamm. vided tissue samples of rodent taxa that they collected. often in 69: 22-33. 292 MICHAUX AND CATZEFLIS

Chaline,J., Mein, P., and Petter, F. (1977).Les grandeslignes d'une Luckett and J.-L. Hartenberger, Eds.), pp. 1-33. Plenum, New classificationévolutive des Muroidea. Mammalia 41: 245-252. York. Chevret,P., Denys,C., Jaeger,J.-J., Michaux, J., and Catzeflis,F. M. Hassanin, A., Pasquet,E., and Vigne, J.-D. (1998a).Molecular sys- (1993a).Molecular evidencethat the (Acomys) is more tematics of the subfamily Caprinae (Artyodactyla, Bovidae) as closelyrelated to gerbils (Gerbillinae)than to true mite (Murinae). determinedfrom cytochromeb sequences.J. Mamm. Evol. 5: 217- Froc. Natl. Acad. Sci. USA 90: 3433-3436. 236. Chevret, P., Denys, C., Jaeger, J.-J., Michaux, J., and Catzeflis, F. Hassanin,A., Lecointre, G., and Tillier, S. (1998b).The "volutionary (1993b).Molecular and fossil aspectsof the tempo and mode of signal" of homoplasy in protein carling gene sequencesand its evolution in Otomys(Otomyinae: Muridae: Mammalia). Biochem. consequencesfor a priori weighting in phylogeny.C. R. Acad. Sei. Syst. Ecol. 21: 123-131. Paris 321: 611-620. Chevret,P., Granjon,L., Duplantier, J.-M., Denys,C., and Catzeflis, Honacki, J. H., and Kinman, K E. (1982). "Mammal Speciesof the F. M. (1994). Molecular phylogeny of the Praomys complex (Ro- World," Allen Press,New York. dentia, Murinae): A study based on DNA/DNA hybridization ex- Hooper,E. T., and Musser, M. D. (1964).The glans penis in Neotro- periments.Zool. J. Linn. Soc. 112: 425-442. pical cricetines(Family Muridae) with commentson classification de Bruyn, H., and Hussain, S. T. (1985).The successionof rodent ofmuroid rodents.Misc. Publ. Univ. Michigan 123: 1-57. faunas from the Lower Mancher Formation, southern Pakistan, Huchon, D., Catzefiis,F. M., and Douzery,E. J. P. (1999).Molecular and its relevance for the biostratigraphy of the Mediterranean Miocène.Paléobiol. Continentale (Montpellier) 14: 191-204. evolution of the nuclear von Willebrand Factor genein mammals and the phylogenyofrodents. Mol. Biol. Evol. 16: 577-589. de Graaff, G. (1981). "The Rodents of Southern Africa," Butter- worths, Durban. Hugueney, M., and Mein, P. (1993). A comment on the earliest Spalacinae(Rodentia, Muroidea). J. Mamm. Evol. 1: 215-223. Denys,C., and Michaux, J. (1992).La troisième molaire supérieure chezles Muridae d'Afrique tropicale et le cas des genresAcomys, Jacobs,L. L., and Downs, W. R. (1994). The evolution of murine Uranomyset Lophuromys.Bonn. Zool. Beitr. 43: 367-382. rodents in Asia. ln "Rodent and Lagomorph Families of Asian Origin and Diversification" (Y. Tomida, C. Li, and T. Setoguchi, Denys, C., Michaux, J., Catzeflis, F., Ducrocq, S., and Chevret, P. Eds.),Vol. 8, pp. 149-156, National ScienceMuseum Monographs, (1995).Morphological and molecular data against the monophyly ofDendromurinae(Muridae: Rodentia).Bonn. Zool. Beitr. 45: 173- Tokyo. 190. Jacobs,L. L., Flynn, L. J., and Downs,W. R. (1989).Neogene rodents of southern Asia. ln "Paperson Fossil Rodentsin Honor of Albert Dickerman, A. W. (1992). "Molecular Systematics of Some New Elmer Wood"(C. C. Black, and M. R. Dawson,Eds.), Science Series World Muroid Rodents,"PhD thesis, University of Wisconsin at No. 33, pp. 157-177. Nat. Hist. Mus. Los Angeles County, Los Madison. Angeles. Dubois, J.-Y., Rakotondravony,D., Hanni, C., Sourrouille, P., and Jacobs,L. L., Flynn, L. J., Downs,W. R., and Barry, J. C. (1990).Quo Catzeflis, F. M. (1996). Molecular evolutionary relationships of vadis, Antemus? The Siwalik muroid record. ln "European Neo- three generaof Nesomyinae,endemic rodent taxa from Madagas- gene Mammal Chronology"(E. H. Linday, V. Fahlbusch, and P. car. J. Mamm. Evol. 3: 239-260. Mein, Eds.), pp. 573-586. Plenum, New York. Dubois,J.-Y., Catzeflis, F. M., and Beintema, J. (1999).The phylo- genetic position of "Acomyinae"(Rodentia, Mammalia) as sister- Jaeger, J.-J., Tong, H., Buffetaut, E., and Ingavat, R. (1985). The group of a Murinae + Gerbillinae clade:evidence from the nuclear first fossil rodents from the Miocene of northern Thailand and their bearing on the problem of the origin of the Muridae. Rev. ribonucleasegene. Mol. Phylogenet.Evol. 13: 181-192. Paléobiol.4: 1-7. Ellerman, J. R. (1940)."The Families and Generaof Living Rodents," Vol. l, Br. Mus. (Nat. Hist.), London. Jaeger, J.-J., Tong, H., and Buffetaut, E. (1986). The age of Mus- Rattus divergence:Paleontological data comparedwith the molec- Ellerman, J. R. (1941)."The Families and Generaof Living Rodents," ular clock. C. R. Acad. Sei. Paris Ser. 2 302: 917-922. Vol. 2, Br. Mus. (Nat. Hist.), London. Jansa, S. A., Goodman,S. M., and Tucker, P. K. (1999). Molecular Flynn, L. J. (1990). The natural history of rhizomyid rodents. ln phylogenyand biogeographyof the native rodents of Madagascar "Evolution of SubterraneanMammals at the Organismal and Mo- (Muridae: Nesomyinae):A test of the single-origin hypothesis. lecular Levels" (E. Nevo, and O. A. Reig, Eds.),pp. 155-183.A. R. Cladistics 15: 253-270. Liss, New York. Kishino, H., and Hasegawa,M. (1989).Examination of the maximum Flynn, L. J., Jacobs,L. L., and Lindsay, E. H. (1985).Problems in likelihood estimate of the evolutionary tree topologiesfrom DNA muroid phylogeny: Relationships to other rodents and origin of sequencedata, and the branching order in Hominoidea.J. Mol. major groups. ln "Evolutionary Relationships among Rodents" Evol. 29: 170-179. (W. P. Luckett, and J.-L. Hartenberger, Eds.), pp. 589-616. Ple- num, New York. Klingener, D. (1964). The comparative myology of four dipodoid rodents (generaZapus, Napeozapus,Sicista and Jaculus). Misc. Furano, A. V., Hayward, B. E., Chevret, P., Catzeflis,F., and Usdin, Publ. Mus. Zool. Univ. Michigan 124: 1-100. K. (1994).Amplification of the ancient murine Lx family of long interspersedrepeated DNA occurredduring the murine radiation. Lavocat, R. (1973). Les rongeurs du Miocène d'Afrique Orientale. J. Mol. Evol. 38: 18-27. Mém. Trav. Ecole Pratique Hautes Etudes 1: 1-284. Grasse,P. P. (1955). Traité de Zoologie(Anatomie, Systématique, Lavocat, R. (1978).Rodentia and Lagomorpha.ln "Evolution of Af- Biologie),Tome XVII, Masson,Paris. rican Mammals" (V. J. Maglio, and H. B. S. Cooke, Eds.), pp. Hiinni, C., Laudet, V., Barriel, V., and Catzeflis, F. M. (1995).Evo- 69-89. Harvard Univ. Press,Cambridge, MA. lutionary relationships of Acomys and other murids (Rodentia, Lawrence,M. A. (1991).A fossil Myospalaxcranium (Rodentia:Muri- Mammalia) basedon complete 12S rRNA mitochondrial gene se- dae) from Shanxi, China, with observations on Zokor relation- quences.Israel J. Zool. 41: 131-146. ships. Bull. Am. Mus. Nat. Hist. 206: 261-286. Hartenberger,J.-L. (1985).The order Rodentia:Major questionson Li, W. K., Tanimura, M., et al. (1987).An evaluation of the molecular their evolutionary origin, relationships and suprafamilial system- clock hypothesisusing mammalian DNA sequences.J. Mol. Evol. atics. ln "Evolutionarv Relationships among Rodents" (W. P. 25: 330-342. BUSHLIKE RADIATION OF MUROID RODENTS 293

McKenna,M. C., and Bell, S. K (1997)."Classification of Mammals Simpson,G. G. (1945).The principles of classificationand a classi- abovethe SpeciesLevel," Columbia Univ. Press,New York. fication ofmammals. Bull. Am. Mus. Nat. Hist. 85: 1-350. McLean, J. C., Fielding, D., Drayna, H., Dieplinger, B., Baer, W., Smith, M. F., and Patton, J. L. (1999). Phylogeneticrelationships Kohr, W., Henzel,W., and Lawn, R. (1986).Cloning and expression and the radiation of sigmodontinerodents in South America: Ev- of human lecithin-cholesterol acyltranferase cDNA. Proc. Natl. idencefrom cytochromeb. J. Mamm. Evol. 6: 89-129. Acad. Sei. USA 83: 2335-2339. Strimmer, K, and von Haeseler, A. (1996). Quartet puzzling: A Martin, L. D. (1980).The early evolution of the Cricetidae in North quartet maximum likelihood methodfor reconstructingtree topol- America.Paleontol. Contrib. 102: 1-42. ogies.Mol. Biol. Evol. 13: 964-969. Mein, P., and Ginsburg, L. (1997). Les mammifères du gisement Swofford,D. L. (1998).PAUP*. PhylogeneticAnalysis Using Parsi- miocèneinférieur de Li Mae Long, Thailande: Systématique,bio- mony (*and Other Methods).Version 4.0b. Sinauer, Sunderland, stratigraphie et paléoenvironnement.Geodiversitas 19: 783-844. MA. Miller,G. S., and Gidley, J. W. (1918).Synopsis of the super-generic Tamura, K., and Nei, M. (1993).Estimation of the number ofnucle- groups of rodents. J. Wash.Acad. Sei. 8: 431-448. otide substitutions in the contraI region of mitochondrial DNA in Misonne,X. (1971).Order Rodentia.ln "The Mammals of Africa: An humans and chimpanzees.Mol. Biol. Evol. 10: 512-529. Identification Manual," (J. Meester,and H. W. Setzer,Eds.), Part Thomas,O. (1896).On the generaofrodents: An attempt to bring up 6, pp. 1-39. Smithsonian Institution Press,Washington, DC. to date the current arrangement of the order. Proc. Zool. Soc. Musser, G. G., and Carleton, M. D. (1993). Family Muridae. ln London, pp. 1012-1028. "Mammal Speciesof the World: A Taxonomic and Geographic Thompson,J. D., Higgins, D. G., and Gibson,T. J. (1994).CLUSTAL Reference,"(D. E. Wilson, and D. M. Reeder,Eds.), pp. 501-755. W: Improving the sensitivity of progressive multiple sequence Smithsonian Institution Press,Washington, DC. alignment through sequenceweighting, positions-specificgap pen- Nedbal, M. A., Honeycutt, R. L., and Schlitter, D. A. (1996).Higher alties and weight matrix choice.Nucleic Acids Res.22: 4673-4680. level systematicsofrodents (Mammalia, Rodentia):Evidence from Tong, H. (1989). Origine et évolution des Gerbillidae (Mammalia, the mitochondrial12S gene.J. Mammal. Evol. 3: 201-237. Rodentia)en Afrique du Nord. Mém. Soc.Géol. France 155: 1-120. Ognev, S. 1. (1963).Mammals of the USSR and adjacent countries: Tong, H., and Jaeger,J.-J. (1993).Muroid rodents from the Middle Rodents.Israel Program Sei. Translat. 5: 1-662. MioceneFort Ternan locality (Kenya)and their contribution to the phylogenyof Muridae. PaleontographicaAbt. A 229: 51-73. O'hUigin, C., and Li, W. H. (1992).The molecular clock ticks regu- larly in muroid rodents and hamsters.J. Mol. Evol. 35: 377-384. Tullberg, T. (1899).Ueber das Systemder Nagetiere:Eine phyloge- netischeStudie. Nova Acta Reg.Soc. Sci. Upsala Ser. 3 18: 1-514. Pascale,E., Valle, E., and Furano, A. V. (1990).Amplification of an ancestral mammalian LI family of long interspersed repeated Usdin, K., Chevret,P., Catzeflis,F. M., Verona,R., and Furano,A. V. DNA occurredjust beforethe murine radiation. Proc. Natl. Acad. (1995).LI (LINE-l) retrotransposableelements provide a "fossil" Sei. USA 87: 9481-9485. record of the phylogenetic history of murid rodents. Mol. Biol. Evol. 12: 73-82. Philippe, H. (1993). MUST, a computer package for management utilities for sequencesand trees. Nucleic Acids Res. 21: 5264- Verheyen,E., Colyn, M., and Verheyen,W. (1996).A mitochondrial 5272. cytochromeb phylogeny confirms the paraphyly of the Dendro- murinae Alston, 1896 (Muridae, Rodentia).Mammalia 60: 780- Philippe, H. (1997).Rodent monophyly: Pitfalls ofmolecular phylog- 785. enies.J. Mol. Evol. 45: 712-715. Viegas-Pequignot,E., Petit, D., Benazzou,T., Prod'homme,M., Lom- Reig, O. A. (1980). A new fossil genus of South American cricetid bard, M., Hoffschir, F., Descailleaux,J., and Dutrillaux, B. (1986). rodents allied to ,with an assessmentof the Sigmodon- Phylogénie chromosomiquechez les Sciuridae, Gerbillidae et tinae. J. Zool. 192: 257-281. Muridae, et étude d'espècesappartenant à d'autres familles de Robinson,M., Catzeflis, F., Briolay, J., and Mouchiroud, D. (1997). Rongeurs.Mammalia 50: 164-202. Molecular phylogenyof rodents,with specialemphasis on murids: Warden, C. H., Langner, C. A., Gordon,J. 1., Taylor, B. A., McLean, Evidencefrom nuclear geneLCAT. Mol. Phylogenet.Evol. 8: 423- J. W., and Lusis, A. J. (1989).Tissue-specific expression, develop- 434. mental regulation, and chromosomalmapping of the lecithin: Cho- Robinson,M., Gouy, M., Gautier, C., and Mouchiroud, D. (1998). lesteraI acyltransferasegene. Evidence for expressionin brain and Sensitivity of the relative-rate test to taxonomie sampling. Mol. testes as weIl as liver. J. Biol. Chem.264: 21573-21581. Biol. Evol. 15: 1091-1098. Wilson, D. E., and Reeder,D. M., Eds. (1993)."Mammal Speciesof Saitou, N., and Nei, M. (1987).The neighbor-joiningmethod: A new the World: A Taxonomic and GeographicReference," 2nd ed., method for reconstructing phylogenetictrees. Mol. Biol. Evol. 4: Smithsonian Institution Press,Washington, DC. 406-425. Wu, C.-I., and Li, W. H. (1985).Evidence for higher rates ofnucleo- Senegas,F., and Avery, D. M. (1998).New evidencefor the murine tide substitutions in rodents than in man. Proc. Natl. Acad. Sci. origins of the Otomyinae (Mammalia, Rodentia) and the age of USA 82: 1741-1745. Bolt's Farm (South Africa). S. Afric. J. Sei. 94: 503-507. Zheng, S. (1994).Classification and evolution of the Siphneidae.ln Serdobova,1. M., and Kramerov,D. A. (1998).Short retroposonof the "Rodentand LagomorphFamilies of Asian Origin and Diversifica- B2 superfamily: Evolution and application for the study of rodent tion" (Y. Tomida, C. Li, and T. Setoguchi,Eds.), Vol. 8, pp. 57-76, phylogeny.J. Mol. Evol. 46: 202-214. National ScienceMuseum Monographs,Tokyo.