Phylogeny and biogeography of African Murinae based on mitochondrial and nuclear gene sequences, with a new tribal classification of the subfamily

Background Within the subfamily Murinae, African murines represent 25% of species biodiversity, making this group ideal for detailed studies of the patterns and timing of diversification of the African endemic fauna and its relationships with Asia. Here we report the results of phylogenetic analyses of the endemic African murines through a broad sampling of murine diversity from all their distribution area, based on the mitochondrial cytochrome b gene and the two nuclear gene fragments (IRBP exon 1 and GHR). Results A combined analysis of one mitochondrial and two nuclear gene sequences consistently identified and robustly supported ten primary lineages within Murinae. We propose to formalize a new tribal arrangement within the Murinae that reflects this phylogeny. The diverse African murine assemblage includes members of five of the ten tribes and clearly derives from multiple faunal exchanges between Africa and Eurasia. Molecular dating analyses using a relaxed Bayesian molecular clock put the first colonization of Africa around 11 Mya, which is consistent with the fossil record. The main period of African murine diversification occurred later following disruption of the migration route between Africa and Asia about 7–9 Mya. A second period of interchange, dating to around 5–6.5 Mya, saw the arrival in Africa of Mus (leading to the speciose endemic Nannomys), and explains the appearance of several distinctive African lineages in the late Miocene and Pliocene fossil record of Eurasia. Conclusion Our molecular survey of Murinae, which includes the most complete sampling so far of African taxa, indicates that there were at least four separate radiations within the African region, as well as several phases of dispersal between Asia and Africa during the last 12 My. We also reconstruct the phylogenetic structure of the Murinae, and propose a new classification at tribal level for this traditionally problematic group.


Background
Rodents are the most speciose mammalian order and comprise almost half of all mammalian species diversity [1]. Within Rodentia, the most diverse assemblage is the superfamily Muroidea, with a global membership of 1300 living species and a natural distribution that includes all continents except Antarctica and all but the most remote islands. This remarkable group also includes the commensal rats and mice, long despised as human pests and agents of disease [2], but now highly valued as model organisms for research related to human health [3,4].
Not surprisingly, morphology-based classifications of muroid rodents were beset by problems of parallel evolution, with many common adaptations evolving independently on different landmasses. Molecular phylogenetic analyses are much less constrained by this problem and recent studies using slowly evolving nuclear genes have done much to clarify the membership and structure of Muroidea [5][6][7]. Recent classifications of this group recognize five or six family level lineages [7,8]. The speciose family Muridae Illiger, 1811 (150 genera and 730 species) is divided by Musser and Carleton [8] into five subfamilies, of which the Murinae Illiger, 1811 is the most diversified (126 genera, 561 species). Within the family Muridae, there is strong molecular support for three subfamilies (Deomyinae, Gerbillinae, Murinae) [subfamily Leimacomyinae of Musser and Carleton [8] has not yet been surveyed], and for a link between Deomyinae and Gerbillinae, with these as a sister clade to Murinae (this latter subfamily encompassing otomyines) [5][6][7].
The subfamily Murinae has a natural distribution that spans the Old World, including all of Africa and Eurasia, and extending to Australia, New Guinea and many islands of the western Pacific (we do not consider here the human-mediated distribution of a few commensal rodents of the genera Mus and Rattus in the Americas and throughout oceanic islands). More than 500 species are currently recognised [8], with centers of diversity and endemism in each of Tropical Africa, Southeast Asia, and the Australo-Papuan region [9,10]. Despite the obvious significance of this group for biogeographic studies, previous molecular studies have either had specific regional foci (e.g. Africa [11][12][13]; Philippines: [14]; Australia: [15,16] ; Eurasia: [17,18]) or employed immunological methods of uncertain reliability [10]. These studies have encouraged regionally-based classifications at tribal or subfamilial level, especially within the Australasian and Philippine regions where various higher level groupings are sometimes recognized (e.g. Anisomyini, Conilurini, Hydromyini, Phloeomyinae, Pseudomyinae, Rhynchomyinae). In Africa, Ducroz et al. [12] designated a tribe Arvicanthini for one well-supported monophyletic group. The most recent, global classification of Murinae [8] abandons the tribal level of classification in favour of a less formal arrangement of genera into divisions, following and improving a system already employed by Misonne [9]. Specifically, Musser and Carleton [8] (2005: pages 902 -905) organize the 126 genera of the subfamily Murinae into 29 divisions, and consider the living taxa Myotomys, Otomys, and Parotomys as members of the subfamily Otomyinae.
Africa supports more than 25% of all living murine species including representatives of 32 endemic genera [8]. All African murines are endemic at species level and only two genera are shared between Africa and Eurasia. One of these is the genus Mus, which is widespread across Eurasia and is represented in Africa by an endemic subgenus, Nannomys, the African pigmy mice [19][20][21]. The second is the primarily African genus Myomyscus which has one species (M. yemeni) native to the Arabian Peninsula. A single origin for all African Murinae, except possibly Dasymys, was proposed by Watts and Baverstock [22] based on their analyses of albumin microcomplement fixation. In contrast, Chevret's [23] studies using the DNA/DNA hybridization method found a minimum of three ancient African lineages within Murinae, each associated with Eurasian taxa. Later studies using direct sequencing methods supported the notion of polyphyly for African Murinae, e.g. [12][13][14]16,24]. Jansa et al. [14] identified three distinct groups: the 'Arvicanthines' (sensu Ducroz et al. [12]), a 'Praomys group' (sensu Lecompte et al. [25]) and the genus Malacomys. The 'otomyines', a dentally distinctive African lineage with three genera (Myotomys, Otomys,Parotomys), are variously associated in molecular studies with either the Praomys group [10] or the arvicanthines [6,11,12,16,24]. Ducroz et al. [12] suggested recognition of this group at tribal rank, as Otomyini. However, Musser and Carleton [8] follow more traditional practice by recognizing a distinct subfamily Otomyinae within Muridae.
Numerous questions thus remain unresolved concerning the pattern and timing of African Murinae diversification. In particular, the relationships of the various African lineages with Asian genera are enigmatic, and the timing of most cladogenic events remains poorly resolved or understood. The latter issue is critical to understanding the history of faunal interchange via the Arabian plate following the collision of Africa with Asia around 16 and 20 Million years ago (Mya) [26,27]. Notably, the murine palaeontological record attests to the presence of some shared genera in Africa and Asia during the late Miocene and the Pliocene [28][29][30], but whether this is due to multiple faunal exchanges between Asia and Africa, to the presence of ancient shared lineages followed by vicariance, or else to convergent evolution, remains a matter of conjecture.
To more adequately assess the pattern and timing of faunal exchanges between Africa and Asia, it is necessary to first establish a more complete phylogenetic framework including all of the key African and Eurasian lineages, and then to derive reliable estimates of divergence times. The main objectives of our study are: (1) to provide a robust and comprehensive phylogeny of the extant African murines and to infer their relationships with the Asian Murinae using mitochondrial and nuclear gene sequences, (2) to provide a new systematic framework that accurately reflects the phylogeny of Murinae; (3) to estimate times of origin and diversification for the African murines lineages; and (4) to place this phylogeny in an historical and geographical context to gain insight into the origin and maintenance of African murine diversity.

Phylogenetics
The final alignments included 1140 sites and 81 taxa for cyt b, 931 sites and 62 taxa for GHR, 1233 sites and 79 taxa for IRBP, and 3304 sites for 83 taxa for the concatenated dataset. The best-fitting substitution models were TVM+G+I for the GHR and IRBP data sets, and GTR+G+I for the cyt b and combined data set ( Table 1). Analysis of the combined dataset produced a single ML tree (Figure 1, lnL = -50270.78386), the supports obtained for each node and each gene are presented in the additional files 1 (ML analysis) and 2 (Bayesian analysis). Monophyly of Murinae is strongly supported but only with inclusion of the two 'otomyine' taxa (100% BP; 1.0 PP). Ten primary lineages can be recognized within Murinae, all with strong nodal support (Figure 1, BP ≥ 97%; PP = 1.0). African murines are polyphyletic and divided among five lineages. We here describe the different lineages to highlight the relationships among the African murines.
The most basal lineage (Lineage 1) consists of the genera Phloeomys and Batomys, both Philippine endemics. There is very strong support (100% BP; 1.0 PP) for reciprocal monophyly of Lineage 1 and all other Murinae.
The fourth lineage consists of the genus Mus (Lineage 4, 100% BP; 1.0 PP), represented by all four subgenera including the African Nannomys. The relationships among the four Mus subgenera remain unresolved as the position of Mus (Nannomys) minutoides and Mus (Coelomys) crociduroides is unstable between ML and BI analyses [see additional files 1 and 2]. These values were estimated from maximum-likelihood analysis of each gene separately (cytochrome b, IRBP, and GHR, respectively) and of the combined data set.
Maximum likelihood tree for the combined dataset    The fifth murine lineage is a diverse and robustly supported African assemblage (Lineage 5, 100% BP; 1.0 PP) that corresponds to the 'Praomys group' of Lecompte et al. [13]. The monophyly of Lineage 5 is further supported by a shared insertion of 6 bp (TTGCCT) at position 893 of the GHR gene alignment. Although the basal nodes within Lineage 5 are poorly supported, it appears likely that Mastomys and Myomyscus are both paraphyletic. The order of branching between sublineages is unresolved and incongruent between ML and BI analyses [see additional files 1 and 2]. However, several terminal groups have strong support: 1) Myomyscus verreauxii + Colomys The ninth murine lineage (Lineage 9, 100% BP; 1.0 PP) consists of the African 'otomyines' Parotomys and Otomys. As noted earlier, Musser and Carleton [8] included these taxa in a separate subfamily -Otomyinae.
Relationships among the ten lineages are partially resolved under each of ML and BI but nodal support values are only moderate to strong. The best support is observed for a diverse Afro-Asian large group comprising Lineages 4 to 7, which we here call Clade A (93% BP; 1.0 PP). Monophyly of Clade A is further supported by an insertion of 6 bp (YGGAYG) at position 86 of the GHR alignment. Within this group, Lineages 6 and 7 are identi-fied as sister lineages but with only moderate support (77% BP; 0.68 PP); and Lineages 4 and 5 form a second sister pair, also with only moderate support (77% BP; 0.69 PP). Lineages 8, 9 and 10, also representing a mix of both African and Asian taxa, are united on the ML tree with moderate to strong support (87% BP, 1.00 PP) in what is named Clade B. Lineages 9 and 10 are sister taxa, with a very strong nodal support (100% BP; 1.0 PP).
Clades A and B are identified as sister lineages on the ML tree, and build up what we refer to Clade C, albeit with very low support (51% BP). This clade C includes all the African murines. A different topology was obtained under BI [see additional file 2] in which Lineage 3 (Philippine and Australo-Papuan groups) forms the sister group of Clade B, once again with low support (0.60 PP). This was the only discrepancy in branching order among the primary lineages of Murinae observed between the two methods.

Molecular divergence estimates
Estimated divergence times are indicated on the ML topology in Figure 2. A detailed chronogram is provided in the additional file 3. The standard deviations of all estimates fall between 0.5 to 0.7 Million years (My); this error value is implied in all divergence estimates indicated below. Divergence time estimations, standard deviations and credibility intervals calculated by multidivtime for the main nodes are indicated in the additional file 4, both for the combined dataset and for each gene separately. There is good congruence between the various estimations but with larger standard deviations for the ones based on one gene than for the values obtained with the combined dataset.
The earliest cladogenic event (to Lineage 1) is dated to 12.3 Mya. Emergence of the Clade C containing all African taxa as well as many Eurasian lineages is dated 11.1 Mya. Cladogenesis of the Afro-Asian Clades A and B is dated to 11 Mya. Divergences between each of Lineages 4 + 5, 6 + 7 and Clade B all fall within the interval 10.1-10.3 Mya. However, while these lineages originated more or less simultaneously, their subsequent diversification was unbalanced and asynchronous. Five of the seven lineages comprise only one or two genera (Lineages 4, 6, 7, 8 and 9), while the two most diverse and well-sampled lineages, corresponding to the main part of the African diversity, radiated somewhat at different times, at about 8.4 Mya (Lineage 10: 'arvicanthines') and 7.6 Mya (Lineage 5: 'Praomys group'), respectively. As we have a good sampling within these African groups (14 of 18 genera in the 'arvicanthines' and 8 of 9 genera in the Praomys group), we are confident that our results accurately reflect the diversification histories of these lineages.
Simplified chronogram with the main murine groups Figure 2 Simplified chronogram with the main murine groups. For each group the oldest fossil is indicated by an arrow according to [51,52,65,66,71,73,76,104,105,108,[134][135][136]. Black area represents African taxa, light grey the Australasian taxa, and dark grey the Eurasian ones.  The phylogeny shows strong geographic structure (shown Figure 2) with most primary lineages restricted to a single biogeographic area. Notable exceptions are the genus Mus (Lineage 4), which includes both Eurasian and African sub-lineages, Lineages 9+10 which are predominantly African ('otomyines' and 'arvicanthines') but also includes the Asian genus Golunda, and the African Praomys group (Lineage 5) which also includes the Arabian species Myomyscus yemeni.
Three near-basal cladogenic events within Murinae correspond to separations between 'mostly Asian' and 'mostly African' lineages.

Phylogenetic relationships of African Murinae and a new suprageneric taxonomy
Many of our ten primary lineages of Murinae were also identified by other scholars in previous molecular phylogenetic studies of Murinae [13,14,16,23,24,32]. However, our enlarged taxon sampling has improved the support for some relationships, which were tentatively identified in previous studies and also identified new primary lineages and associations. Based on these robust results and on the geographical structure of the phylogeny, we propose to formalize a tribal level of classification within Murinae (see Table 2), for convenient use above the informal rank of division employed by Musser and Carleton [8].
Tribe Phloemyini (Lineage 1): A basal division within Murinae between certain Philippine 'Old Endemics' and all other murines was first suggested by Watts and Baverstock [10] based on microcomplement fixation of albumin, and strongly supported since then by numerous nuclear and/or mitochondrial gene phylogenies ( [6,7,14,16,24], this study). Broader membership of this group includes two other endemic Philippine murine genera, Carpomys and Crateromys [14]. All members of this group are morphologically specialised in different ways but they do share at least one clearly derived dental traitan unusually complex anteroconid morphology on the first lower molar [33].  [24]. Jansa et al. [14] recovered a clade that includes the Philippine members of this group but their study did not include any Australo-Papuan murines. Ford [15], using a combination of mitochondrial and nuclear intron sequences, demonstrated the close affinity of all Australian murine genera (Rattus excluded) but did not include any Philippine taxon in his study. Watts and Baverstock [10] included the majority of Australian and New Guin- The "divisions" of Musser and Carleton [8] are indicated as well as taxa not included in our analyses. Taxa for which there is independent molecular or morphological evidence of phylogenetic position are underlined (see text for details). All the other ones should be treated as Murinae incertae sedis. † : fossil genus. ean murine genera in their microcomplement fixation study of albumin but they had poor coverage of Philippine murines. They failed to recover a single lineage that includes all Australo-Papuan murines. Studies of sperm ultrastructure also point to monophyly of the majority of Australo-Papuan murines, albeit with some notable exceptions [39,40]. Rowe et al. [16] included a wide array of Australo-Papuan and Philippine murines in their multilocus analysis, including representatives of the four suprageneric taxa recognised in previous studies of these regional faunas (i.e. uromyines, conilurines, hydromyines and anisomyines  Table 2). Use of one tribal name -Hydromyini -for this expended Australo-Papuan and Philippine murine radiation serves to draw attention to the phylogenetic connection between these geographically isolated assemblages. Musser and Carleton [8] divided members of our tribe Hydromyini among seven divisions ( Our Clade C contains a highly heterogeneous and geographically disparate assemblage, including all the African murines. Although this lineage has a poor basal support, a comparable assemblage was recently recovered with strength by Rowe et al [16], whose study clearly indicates that Vandeleuria also belongs to that clade. Within this group, we identify a total of seven primary lineages (Lineages 4-10), each well supported and geographically unified; and we note that the same seven lineages were recovered by Rowe et al. [16]. Our division of Clade C into two major sections [Clades A (Lineages 4-7) and B (Lineages 8-10)] is also supported by the results of previous multi-gene analyses [16,24], and by the presence of diagnostic indel events in the GHR alignment for several nodes (basal for Clade A; basal for Lineage 5), and we are confident as to the essential correctness of the topology.
In terms of taxonomy, we might assign all members of Clade C to a single tribe, for which the earliest available name would be Murina Illiger, 1811. However, we prefer a more expansive tribal classification that recognises the huge taxic and ecomorphological diversity contained within Clade C. Accordingly, we propose to represent a total of seven tribes for each of Lineages 4-10. The result is an overall tribal classification of Murinae that is concordant in large measure with geographic partitioning and also has strong morphological expression.  [13,25,48], that each of Myomyscus and Mastomys are paraphyletic within the Praomyini. As in previous molecular and morphological analyses [13], the genus Praomys appears to be monophyletic with inclusion of P. verschureni and P. daltoni, although support is still quite low. Our enlarged dataset also resolves some relationships within the Praomyini, especially at the base of the clade, where resolution was poor in previous analyses [13]. Tribe Malacomyini (Lineage 6): Malacomys has long been regarded as an isolated and enigmatic genus, whether assessed on dental morphology ([9]: 106) or on chromosomes [49]. Its isolated position is confirmed by our results and other molecular multilocus analyses [16,32]. The taxon Malacomyini tribe nov. is based on type genus Malacomys Milne-Edwards, 1877.
Tribe Apodemini (Lineage 7): Apodemus is among the most thoroughly studied of all murine genera, both from a molecular perspective, e.g. [17,38,50], and based upon the rich fossil record of western Eurasia, e.g. [51,52]. A close relationship between Apodemus and Tokudaia was suggested on dental morphology, e.g. [9], but molecular supporting data were only recently obtained [16,17,38]. Our analysis confirms a sister relation between Apodemus and Tokudaia but also highlight the considerable antiquity of their generic divergence. The taxon Apodemini tribe nov. is based on type genus Apodemus Kaup, 1829.
Our analysis identifies Malacomys as a possible sister lineage to Apodemus + Tokudaia. Although nodal support is rather poor (77% BP; 0.69 PP) on our tree, we note that a comparable grouping of these lineages was observed in various other multi-locus analyses [13,16,32]. An exception is the multi-gene topology of Steppan et al. [24] in which Malacomys occupies a more basal position within a group corresponding to our Clade A. Musser and Carleton [8] recognised separate Apodemus and Malacomys divisions and we follow their lead in treating each of Lineages 6 and 7 as separate murine tribes. Moreover, since no included genus has previously formed the basis of a family level name, we propose two new names at tribal rank for these lineages. Although both lineages have limited generic diversity, we note that the genus Apodemus, despite being morphologically conservative, contains far greater molecular diversity than many other murine genera. Musser and Carleton [8] included the recently extinct genus Rhagamys from Corsica and Sardinia in the Apodemus division, based on paleontological interpretations of its dental morphology, e.g. [52], and we follow this lead.
All remaining murines examined in this study fall into our Clade B. Key members of this group are the Indian Millardia + Cremnomys and the African 'arvicanthines' and 'otomyines'. Phyletic association of Otomys + Parotomys with the 'arvicanthines' is robustly supported by numerous other molecular analyses and must now be considered as proven [6,11,12,14,16,24]. Association of Millardia + Cremnomys with this group is a more controversial finding, although we note a comparable topology in the DNA/ DNA hybridization results of Chevret [23] and partial support from several recent molecular analysis [16,32].

Ducroz et al. ([12]: p 200) found no evidence from analyses of mitochondrial DNA of close relationship between
Millardia and African arvicanthines, while Watts and Baverstock ([10]: p111) concluded from their albumin immunology that "Millardia appears to be a monogeneric lineage arising early in the history of the murines". Rowe et al. [16] identified conflict among the three genes available for the position of Millardia. Our analysis differs mainly in the inclusion of two Millardia species and a representative of the genus Cremnomys and this wider taxon sampling may account for the improved support for the sister group relationship of this lineage with the 'arvicanthines' and 'otomyines'. However, conflict with previous analysis highlights the need for further testing of this relationship using sequences from other slowly evolving nuclear genes.
Consistent with our treatment of Clade A, we propose to recognize three separate tribes within Clade B, an arrangement that in our view best reflects the taxic and morphological diversity, and the geographic partitioning of this assemblage. Tribe Otomyini (Lineage 9): Traditional recognition of a subfamily Otomyinae for the African genera Otomys and Parotomys reflects the extreme specialization of the cheek-teeth of these taxa, especially among members of the genus Otomys. Despite compelling molecular [6,11,12], and paleontological [53][54][55] evidence that otomyines not only belong within Murinae but are specifically associated with arvicanthines ( [14,16,24], this study), the notion of taxonomic isolation maintains an inertia that is difficult to break, e.g. Musser and Carleton [8]. Like some previous authors [55,56], we advocate recognition of this lineage at tribal level, as Otomyini Thomas, 1896 with type genus Otomys Cuvier, 1824.
Tribe Arvicanthini (Lineage 10): Ducroz et al. [12] proposed a tribe Arvicanthini but failed to explicitly designate a type genus. As indicated by Musser and Carleton [8], their name is a nomen nudum and nomenclaturally unavailable. We here formalise the Arvicanthini tribe nov. with type genus Arvicanthis Lesson, 1842. The tribe corresponds in large part to Misonne [9] 's 'Arvicanthis division' but with notable additions (Oenomys, [11,12,24], this study) and exceptions (Bandicota and Nesokia, both close relatives of Rattus, [32,57], this study). The arvicanthine affinity of the Indian genus Golunda was promoted on dental criteria by each of Misonne [9] and Musser [58], and was weakly supported by the 12S and 16S mitochondrial gene phylogenies of Ducroz et al. [12] and by the IRBP and cytochrome b phylogeny of Michaux et al. [32].
Our results confirm this association, with moderately strong nodal support, and provide, for the first time, a basal position for Golunda within the tribe.  Table 2). Our phylogeny for Arvicanthini is the first one based on nuclear genes and it also features enlarged taxon sampling. We confirm earlier mtDNA evidence [12] of a clade containing Arvicanthis, Desmomys, Lemniscomys, Mylomys, Pelomys, and Rhabdomys, and for sister-group relationships between Mylomys and Pelomys, and between Desmomys and Rhabdomys. Our results depart from previous interpretations in the wellsupported grouping of Arvicanthis and Lemniscomys as sister taxa (Lemniscomys occupied a basal position within the clade in previous analyses [12]). The inclusion of previously unsampled taxa in our phylogeny also produced new insights into Arvicanthini phylogeny, most notably the basal position of Golunda, followed by the divergence of Oenomys then by the highly supported clade containing Stochomys and Hybomys. The basal position of Oenomys among the arvicanthini was also proposed in a recent molecular study [16] despite sparse sampling within the tribe. The other associations identified here are not supported by previous analyses and they require further testing with sequences from other slowly evolving nuclear genes.
Some genera, not yet available for molecular phylogenetic studies, can be associated with the Arvicanthini on morphological criteria. For example, the rare African genus Dephomys shares dental and cranial morphometric traits with Hybomys [9,59], and was included in the Hybomys division by Musser and Carleton [8]. Similarly, the monotypic genus Lamottemys, described after the work of Misonne, is thought be closely related to Oenomys [60,61], and was included in the Oenomys division by Musser and Carleton [8]. Malpaisomys, an extinct genus from the Canary Islands, was also included in the Oenomys division by Musser and Carleton [8], based on morphological studies by Lopez-Martinez et al. [62] and their own assessment. These authors also suggest that Canariomys, the other murine endemic from the Canary Island, might be a member of this divison but that morphological reexamination of the specimens is needed. Finally, the Manipur bush rat, genus Hadromys, was included within the Arvicanthis division by Misonne [9] but regarded as potentially distinct from this lineage by Musser [58]. Musser and Carleton [8] placed this Indian genus in its own monotypic division and we follow suite by listing it as incertae sedis within Murinae ( Table 2).

Timing of cladogenesis among African lineages
Several authors have estimated divergence times among muroids from molecular data [7,11,12,14,[16][17][18]38,63]. Our divergence time estimates are consistently older than those calculated by Chevret et al. [11,63] [16] and our results (for example, Mus/Rattus at 9.7 ± 0.5 versus 11.3 ± 0.5 Mya). As rightly pointed by Steppan et al. [7] and Rowe et al. [16], these differences most obviously reflect the nodal assignment on the topology of the crucial transition from fossil Antemus to fossil Progonomys at 12.1 Mya. In addition, the differences may also reflect selection of other calibration points, and the differences in taxon sampling.
Several molecular studies on Apodemus suggest an early divergence between Tokudaia and Apodemus as well as between the main lineages within Apodemus [17,37,38,50]. We derived estimates of 10.2 Mya (± 0.5) for the separation of Apodemus and Malacomys, 9.6 (± 0.5) for Apodemus/Tokudaia and 8.6 (± 0.5) for the earliest divergence within Apodemus. Similar estimates were found by Michaux et al. [17] but Sato and Suzuki [38] obtained highly variable times for the Apodemus/Tokudaia divergence with each of their five data sets, ranging from 6.5-7.6 My for IRBP to 11.3-13.2 Mya for mitochondrial cyt b.
The genus Mus has been subjected to extensive phylogenetic study, e.g. [18,20,45], though in most studies the African Nannomys was underrepresented. We estimated the initial divergence of extant Mus [including Nannomys] lineages to 6.6 Mya (± 0.7), with Nannomys as the earliest offshoot. Catzeflis and Denys [19] [21]) timing for the initial divergence of subgenera within the genus Mus (inclusive of Nannomys).
Jansa et al. [14] presented divergence time estimates for murines that are considerably older than our own. For example, based on IRBP sequences they estimated the divergence date between our Hydromyini and our Murini+Praomyini+Arvicanthini at 15.8-20.5 Mya, depending on calculation method used. These values are much older than our estimate of 11.1 ± 0.5 Mya for this divergence. We suspect that Jansa et al. [14] systematically overestimated divergence times within Murinae through their use of fossil calibration points placed on more basal nodes in the Rodentia as well as in the general mammalian tree, leading to an increased likelihood of partial saturation at mutational hotspots. Jansa et al. [14] defended their divergence estimates by referring to the incompleteness of the fossil record, especially the fact that large parts of the Old World have almost no relevant small mammal fossil record.
To further explore this conflict in interpretation, we tested our molecular divergence framework within the Murinae against the relatively good fossil record of this group in Europe, Africa, and South Asia. As shown on Figure 2, the earliest first fossil occurrences of various lineages all fall within the time ranges suggested by our divergence date estimates. Moreover, we note that the oldest fossil Murinae from South Asia and Africa, estimated to be about 12-14 Mya and 10-11 Mya, respectively (Asia: [65,66]; Africa: [67][68][69][70][71]) are not attributable to extant genera (e.g. Progonomys: [72]; Karnimata: [70,73]); or only tentatively so (c.f. Stenocephalemys, c.f. Parapelomys: [71]; c.f. Lemniscomys: [74,73]). Conversely, representatives of modern genera are not definitely recorded prior to 5-7 Mya [73,[75][76][77] which is consistent with our dating of murine evolution but difficult to reconcile with a much longer evolutionary time frame. Even more convincingly, our divergence estimates are consistent with first appearance of murines in the fossil records of Africa around 12 Mya [30,71,72] and in Europe around 11 Mya [78,79].

Biogeographic implications for African murines
Our molecular phylogeny contributes in several ways to an improved understanding of the pattern and timing of initial murine colonization of Africa. The earliest, generally accepted murine fossils occur in the sedimentary record of the Siwalik Hills of Pakistan, and date to around 14 Mya [65,66,80,81]. In contrast, the earliest murine fossils from anywhere in Africa date to less than 12 Mya [68], despite the fact that other groups of muroid rodents (including the genus Potwarmus, a taxon of uncertain subfamilial affinity) are represented in older fossil deposits, e.g. [69,71,82]. Similarly, the abundant fossil record of Europe contains no evidence of murines prior to 11 Mya, at which time they appear fully differentiated and undergo rapid diversification [78,79]. This disparity between the various regional fossil records suggests that Murinae originated in Asia and colonized both Africa and Europe during a common period of dispersal [30,72]. Our molecular phylogeny of Murinae is consistent with this scenario to the extent that each of the three basal branches on our phylogeny (Phloeomyini, Rattini and Hydromyini) is almost entirely restricted to Asia and/or the major islands of the western Pacific (i.e. Philippines and Australasia). The major exceptions are Micromys, an extant genus with a wide Palearctic distribution [8] but with no known African fossil record [83], and the fossil genus Karnimata, which is best known from the Siwalik sequence but is also reported from late Miocene localities in southern and eastern Africa [77]. Karnimata is a possible stem genus for our Rattini [65,72], and its presence in Africa, if confirmed by further study of the fossils, would imply that some early immigrant lineages died out without leaving modern descendants.
Jacobs et al. [80] postulated that dispersal of murines from Asia to Africa started around 11.8 Mya, following establishment of a vegetation corridor between Africa and Asia across the recently established Arabian peninsula [30,76,[84][85][86][87]. The best evidence of intercontinental dispersal by mammals during this period is the sudden appearance of equids ('Hipparion') in the African fossil record [86,88,89]. Significantly, the earliest African hipparionines and murines occur together in sites dated to around 11 Mya in Algeria [68] and 10 Mya in Ethiopia [86,90]. Just how many murine lineages crossed from Eurasia into Africa during this early period of dispersal is less certain, with somewhat contradictory indications coming from each of the fossil record and the molecular phylogeny.
The earliest fossil murines from African localities are referred to the genus Progonomys [68,86,90,91]. Slightly younger localities in Namibia and East-Africa, dated to around 9-10 Mya contain more diverse murine faunas with Karnimata sp., Aethomys, c.f. Parapelomys sp. and c.f. Stenocephalemys sp. [69][70][71]92]. As noted above, Karnimata is a typical Asian Miocene genus but the other taxa suggest an early period of in situ diversification leading to each of the endemic African praomyine and arvicanthine lineages. In apparent contradiction to this scenario, our molecular phylogeny suggests that each of three early branches of the African murine radiation (Praomyini, Arvicanthini+Otomyini and Malacomyini) has a sister lineage among Eurasian Murinae (Murini, Millardini and Apodemini, respectively). The obvious interpretation is that each of these lineages was differentiated prior to their dispersal into Africa, and arrived around the same time as part of a broader episode of faunal interchange. Our divergence time estimates would place this period of faunal interchange followed by regional differentiation in the interval 11-10 Mya -a very good fit with the fossil record of Africa and Asia. However, an alternative scenario, only marginally more complex, could posit an early dispersal to Africa, followed by differentiation and back dispersal of three lineages from Africa to Eurasia (ancestral Murini, Apodemini and Millardini). A detailed reassessment of the earliest African murine fossils, looking for evidence of phyletic continuity versus disjunction, might resolve this issue. Until this is done, we must be content with the notion of a shared biogeographic province spanning the 'Arabic Corridor' across which various early murines referrable to Progonomys, Karnimata and possibly other genera made their way between southwest Asia and northern Africa, starting around 11 Mya. These populations presumably included basal members of the Apodemini + Malacomyini, the Murini + Praomyini, and our Clade B (stem group of Millardini + Otomyini + Arvicanthini).
The earliest African fossil faunas of fully modern aspect (i.e. with species confidently assigned to extant genera) date to the interval 7-5 Mya [73,[75][76][77][92][93][94][95]. However, due to sizable gaps in the African fossil record, it is currently unclear whether these later murines were derived from the earliest colonists or from a later wave of colonization from Asia, or perhaps from a combination of both. Certainly, the appearance around 7-9 Mya in the African record of distinctively Asian lineages of Bovidae [96], Elephantoidea [97] and non-murine rodents [30,76,98] is strong evidence for habitat continuity and dispersal between Asia and Africa during the terminal Miocene. However, the rise to dominance of the Gerbillinae in the fossil record of the Middle East during the interval 7-8 Mya also suggests increasingly arid conditions on the Arabian Peninsula [84,99]. This may have presented a barrier to dispersal by murine rodents, and and hence, caused the onset of independent diversification of the African and Asian murine faunas. Direct evidence for murine dispersal into Africa during this period is limited by the paucity of the fossil record.
We estimate the timing of diversification of modern Arvicanthini + Otomyini at 8.6 ± 0.6 Mya, and of modern Praomyini at 7.6 ± 0.6 Mya. Diversification of the modern African murine genera thus seems to narrowly postdate the disruption of the Arabic Corridor.
After 6 Mya, there is renewed evidence of faunal interchange between Africa and each of Southwest Asia and Western Europe [28,76,91,[100][101][102][103][104][105]. This coincides with a period of global sea level depression [106], and with the combination of eustatic and tectonic events in the Mediterranean region that precipitated Messinian salinity crisis [84,107]. Fossil evidence from the circum-Mediterranean region through this period documents significant dispersal and associated mammalian turnover [28,84,100,102,108,109]. Among murine rodents, a species of Mus probably entered Africa from Asia around this time, somewhere between 6.6 ± 0.7 Mya (the divergence estimate for the subgenus Nannomys within Mus) and 4.0 ± 0.8 Mya (the earliest cladogenic event within subgenus Nannomys [20,21]). The earliest fossil occurrence of Mus in Africa comes from Kenya, dated to 4.5 Mya [76]. Around the same time, a species of Myomyscus (Praomyini) evidently spread to the Arabic region, giving rise to the modern species M. yemeni. We estimate the time of divergence of this species from its East African sister species (M. brockmani) at 5.1 ± 0.6 Mya, which also coincides locally with the opening of the Red Sea. In North Africa, the western European fossil genus Occitanomys is recorded for the first time in a section younger than 5.32 Mya [91]. Finally, the fossil record also provides some examples of late Tertiary murine dispersal between Asia and Africa. Most notably, African sites of latest Miocene-Pliocene age reportedly contain several 'Indian' genera (Millardia and Golunda) [91,98,110], while Asian localities of latest Miocene and early Pliocene age have produced several genera of possible arvicanthines. One such lineage is the extinct arvicanthine genus Saidomys, with a stratigraphic range that extends back to the late Miocene in Africa [76,104], to the early Pliocene in Pakistan and Afghanistan [28,100], and to the latest Pliocene in Thailand [111]. The extinct genus Parapelomys, known from several South Asian localities of latest Miocene and early Pliocene age, is also touted as possible arvicanthine [28,112].
Environmental changes after 3 Myr probably caused the regional extinction of some lineages and generally shaped the modern continental faunas [113][114][115]. The genera Millardia and Golunda may have disappeared from Africa, while Saidomys and Occitanomys went to global extinction. Over the same period, numerous groups of African murines radiated to fill newly emerging habitats. However, few were quite so successful as the African pigmy mice (18 living species are recognized for the subgenus Nannomys [8]), which appear to have found a largely underexploited set of niches below the body size range of other African murines.

Conclusion
Our molecular dataset for Murinae, which includes the most complete sampling so far of the African murines, gives compelling evidence for five phyletically separate radiations within the African region, as well as several phases of dispersal between Asia and Africa during the late Miocene to early Pliocene. Through our expanded taxon sampling, which also includes a good coverage of Eurasian taxa we also reveal many new details concerning the overall phylogenetic structure of the Murinae, and this forms a basis for rational classification at tribal level of this traditionally problematic group. Further studies of Murinae should target the few remaining African genera that were not available in our dataset (including Thallomys, Lamottemys and Muriculus), as well as various unsampled Asian taxa (e.g. Hapalomys, Lenothrix) including those that have been associated with the African Arvicanthini on morphological grounds (e.g. Hadromys). Dense taxon sampling of the Australo-Papuan Hydromyini was recently provided by Rowe et al. [16], although a few important gaps remain for this region. On a broader level, a comparison of the phylogenetic structure of Murinae with that of other co-distributed groups of small mammals, such as Gerbillinae and Soricidae, might shed even greater light on the history of the faunal interchange and extinction across Africa and Asia during the last 15 My.

Taxon and gene sampling
We obtained sequences from 83 species including representatives of 49 murine genera from most previously identified major murine lineages, as well as eight genera of Deomyinae and Gerbillinae (Table 3) for use as outgroups [5][6][7]. Our sampling for African Murinae and otomyines covers 25 out of 32 living African genera and includes representatives of all the four previously identified lineages. Most genera are represented by a single species but multiple representatives are included for highly diversified or potentially paraphyletic genera.
Sequences were obtained for two single-copy nuclear genes (growth hormone receptor exon 10: GHR; and interphotoreceptor retinoid binding protein exon 1: IRBP) and one mitochondrial-coding gene (cytochrome b apoenzyme: cyt b). Specimen identification and sequence data are listed in Table 3.
The nuclear genes were chosen because of their proven utility for understanding muroid relationships and the presence of an existing sequence dataset for this group [6,7,14,24,116,117]. The GHR and IRBP genes are not genetically linked and their location is variable, on chromosomes 15 and 14 in Mus, and chromosomes 2 and 16 in Rattus [118]. The mitochondrial cytochrome b gene was chosen because it provides a third independent marker that evolves at a faster rate than either of the two nuclear genes, and also is well represented in previous datasets.
Most taxa are represented by sequences from two or three genes, the one exception being Parotomys for which we have only GHR sequence (Table 3). All ingroup genera are represented by sequences from the same species and where possible, by sequences from the same DNA sample. Chimeric data (i.e. different sequences deriving from more than one species of a genus) were used only for two outgroup taxa: Acomys (A. cahirinus and A. ignitus) and Meriones (M. unguiculatus and M. shawi).

DNA extraction and sequencing
Total genomic DNA was extracted from tissues preserved in ethanol using a CTAB protocol [119] or a QiaAmp extraction kit (Qiagen). The cytochrome b (1140 bp) gene was amplified as described in Lecompte et al. [25] or Montgelard et al. [120]. PCRs used the following thermal cycling parameters: one step at 94°C for 4 min, followed by 35 cycles (40 s at 94°C, 45 s at 50°C, 1 min at 72°C). The final extension at the end of the profile was at 72°C for 10 min.
Double-stranded PCR products were purified directly from the PCR product or from agarose gel using the Min-Elute purification kit (Qiagen) or Amicon Ultrafree-DNA columns (Millipore) and sequenced directly on both strands using an automatic sequencer CEQ2000 (Beckman) or an ABI 310 (PE Applied Biosystems).
The new sequences were deposited in the EMBL data bank. Accession numbers for all sequences used in this analysis are listed in Table 3.

Phylogenetic reconstruction
Sequences were manually aligned with the ED editor of the MUST package version 2000 [123]. Nonsequenced positions and gaps were coded as missing data. Phylogenetic reconstructions were performed on the complete DNA data set by maximum likelihood (ML) with PAUP* (version 4 beta 10) [124], and by Bayesian inference (BI) with MrBayes (version 3.1.2) [125].
Modeltest 3.7 [126] was used to determine the sequence evolution model that best fits our data using the Akaike Information Criterion (AIC). This program examined the fit of 56 models, with either a proportion of invariable sites (I), a gamma distribution of substitution rate variation among-sites (G), or a combination of both (I + G).
To avoid excessive calculation times, our PAUP* ML analyses were conducted in two steps. A ML heuristic search was first conducted by Tree Bisection Reconnection (TBR) branch swapping to identify the optimal tree under parameters estimated by Modeltest. This tree was re-used  for a new round of parameter estimation/branch swapping. This procedure was repeated until there was a stabilization of both topologies and parameters. The robustness of nodes was estimated in PHYML [127] with ML bootstrap percentages (BP ML ) estimated from 1000 pseudoreplicates using as a starting tree the best ML tree obtained from PAUP. PHYML was preferred over PAUP* for bootstrap analyses because of its rapidity. We also performed Bayesian Inference, as calculated by MrBayes, and report Posterior Probabilities (PP) for recovered nodes.
For the Bayesian analysis we used 9 partitions, one for each codon position of each gene.

Estimating dates of divergences
Divergence times were estimated for the optimum ML topology. The hypothesis of a constant molecular clock was tested by a Likelihood Ratio Test as proposed by Felsenstein [128] and calculated in PAUP*4.0b10. We used a relaxed Bayesian molecular clock approach as implemented in MultiDivTime [129], using parameter estimates derived with PAML [130] as described by Yoder and Young [131]. Divergence times were estimated with two fossil-based calibration intervals: 1) the Mus/Rattus divergence set to between 10-12 Mya [65,66,132,133]; and 2) the divergence between Apodemus mystacinus and all the species of subgenus Sylvaemus (A. flavicollis and A. sylvaticus) set to a minimum of 7 Mya [51,78].