Sampling and species identification

Multimammate mice were collected in the framework of the Joint Ethio-Russian Biological Expedition (JERBE) (1986-2018) and the Czech-Ethiopian expeditions (2012-2018). For specimens collected in the early years of the JERBE activity, species identification was carried out using cytogenetic methods. Individual Mastomys species of eastern Africa are well distinguished by their karyotypes: M. awashensis – 2n = 32, FNa = 54; M. erythroleucus – 2n = 38, FNa = 50-56; M. natalensis – 2n = 32, FNa = 52-54 (Green et al. 1980, Granjon et al. 1997, Lavrenchenko et al. 1998). Two species with 32 chromosomes, M. natalensis and M. awashensis, differ by the number of metacentric and submetacentric elements, C-banding pattern and the form of the Y-chromosome (see Lavrenchenko et al. 1998). Another diagnostic method was based on differences in the relative mobility of haemoglobin electromorphs (see details in Dobrokhotov 1982, Robbins et al. 1983, Lavrenchenko et al. 1992). More recent samples were barcoded at the mitochondrial cytochrome b gene (cytb), which is a suitable marker for identification of Mastomys species (Lecompte et al. 2002, Colangelo et al. 2010). DNA was extracted using a standard phenol-chloroform method or commercial kits. Cytb sequences (1140 bp) were amplified using primers and a protocol described by Lecompte et al. (2002). To determine the phylogenetic position of Ethiopian samples within the genus, 30 cytb sequences of all Mastomys species possibly occurring in Ethiopia were retrieved from GenBank (Table S1). Nucleotide sequences were edited and aligned using Bioedit 7.0.5.3 (Hall 1999). A phylogenetic tree was constructed using Bayesian inference in MrBayes version 3.2.6 (Ronquist & Huelsenbeck 2003). For the detection of partitions and a choice of appropriate models of sequence evolution we used PartitionFinder version 2.1.1 (Lanfear et al. 2012). Three heated and one cold chain were employed (nchains = 4) with the runs initiating from random trees; two independent analyses (nruns = 2) were conducted with 5 million generations each; the trees and parameters were sampled every 2000 generations. Results were checked for convergence using Tracer version 1.6 (Drummond & Rambaut 2007). For each run, the first 20% of sampled trees were discarded as burn-in. For rapid genotyping and species identification of larger numbers of Ethiopian Mastomys without sequencing, we also developed a simple molecular PCR assay, allowing the discrimination of the three species based on cytb fragment length (Martynov & Lavrenchenko 2018). A cytb haplotype network was constructed on the basis of a subset of nucleotide sequences from 83 specimens and trimmed to 688 bp to avoid missing data. Haplotypes were generated using DNAsp version 5.10.01 (Librado & Rozas 2009), the median-joining network was produced in PopART (downloaded on 15.5.2016 from  http://popart.otago.ac.nzhttp://popart.otago.ac.nz). The level of genetic divergence between mitochondrial groups was estimated using the number of base substitutions per site (uncorrected p-distances) in MEGA version 7.0 (Kumar et al. 2016). In total, using the methods described above, we genetically identified 59 specimens of M. awashensis, 230 M. erythroleucus, 88 M. natalensis and one M. kollmannspergeri from Ethiopia (Table 1).

Table 1.

List of localities with genetically identified Mastomys species. Localities closer than 3 km were merged. Species identification was performed by sequencing the mitochondrial cytb gene (S), PCR assay (P), cytogenetic analysis of karyotypes (K), and electrophoresis of haemoglobin (H). (No. of localities correspond to Fig. 1).

Continued

Species distribution modelling

Ecological niche modelling can predict species distributions based on various factors (e.g. bioclimatic variables) and is a rapidly developing approach used for understanding geographical biodiversity patterns (Lyet et al. 2013, Feuda et al. 2015, Vences et al. 2017). We modelled the potential distribution of three Ethiopian Mastomys species by using a maximum entropy approach in MaxEnt version 3.4.1 (Phillips et al. 2006, Phillips & Dudík 2008). This algorithm combines the geographic coordinates of particular species occurrences with environmental parameters and characterizes the ecological niche. The resulting model can be used to predict species' distributions (Hernández et al. 2006). For ecological niche modelling we used the following data layers: 19 bioclimatic variables from the WorldClim database ( http:www.worldclim.org; version 2.0), Global Aridity Index and Potential Evapotranspiration Climate Database v2 ( http://www.cgiar-csi.org/), Percent Tree Cover and Land Cover (Table S3) ( https://globalmaps.github.io/). We used layers with 30 arc seconds spatial resolution and the WGS 84 datum. For all analyses, a region of sub-Saharan Africa was cut from the layers between 21° and –35° latitude, –18° and 52° longitude. To create the models in MaxEnt software for each species 75 % of the occurrence localities were used as training data, and the remaining 25 % were reserved for testing the resulting model. In order to exclude possible errors in the ecological niche model associated with training and testing data, a bootstrap with 500 replications with random seeding was used. Other settings were default. Model quality was measured using the area under the curve (AUC) derived from receiver operating characteristic (ROC) plots. AUC values of 0.5 indicate no greater fit than expected by chance, 0.75 are considered useful and 1.0 show a perfect model fit (Hanley & McNeil 1982, Swets 1988, Elith 2000, Jiménez-Valverde 2012).

We estimated bioclimatic variables that are most important for species distribution with the Jackknife test (Elith et al. 2011). Each variable was excluded in turn, and a model created with the remaining variables. In the next step a model was created using each variable separately. Finally we constructed a model using all variables. ENMtools version 1.4 (Warren et al. 2010) was used to measure predicted range overlaps between three Ethiopian Mastomys species (suitability threshold for presence: 0.2) (Fitzpatrick & Turelli 2006). Overlap values are in the range from 0 (no overlap) to 1 (absolute overlap). The transformation of GIS-layers into the work formats, excluding the areas necessary for research and visualisation of distribution models for each species, was carried out using QGIS 3.4 (Palomo et al. 2017).

Diversity and distribution of multimammate mice in Ethiopia

The genus Mastomys is widespread in Ethiopia, especially across the western and southern part of the country. It is absent only from the highest mountains (e.g. the Bale and Arsi Mountains) and dry areas like the Afar triangle and the Somali region of the eastern part of the country (Fig. 1). All cytb sequences of Mastomys from Ethiopia unequivocally (posterior probability PP > 0.95) clustered with one of the four mitochondrial clades of the phylogenetic tree, which corresponded to four species (Fig. 2). Detailed exploration of the results of phylogenetic analysis shows that only phylogroup A-III of M. natalensis (sensu Colangelo et al. 2013) and phylogroup D of M. erythroleucus (sensu Brouat et al. 2009) are represented in Ethiopia (Fig. 2). Haplotype network analysis (Fig. 3) indicates two subclades of phylogroup D of M. erythroleucus in Ethiopia (differing by up to 2.5% at cytb), geographically separated from each other by the western side of the Ethiopian Rift Valley (ERV). The “West” subclade is distributed in the western part of the country, while the “East” is in the bottom of the ERV and in the southeastern part of the country (Fig. 1). When combined with additional species identification methods, we documented the presence of M. awashensis at 18 localities, M. erythroleucus at 32 localities, and M. natalensis at 13 localities (Fig. 1; see details in Table 1). The fourth species, M. kollmannspergeri, was identified based on a single cytb sequence that significantly clustered with sequences of this species from northern Cameroon and Chad (Fig. 2), and represents a new species for Ethiopia.

Fig. 1.

Distribution of four Mastomys species in Ethiopia based on genetic identification of species. The numbers of localities correspond to Table 1 (ERV – Ethiopian Rift Valley).

Fig. 2.

Bayesian phylogenetic tree of four Mastomys species occurring in Ethiopia. The numbers above branches represent posterior probabilities of nodes. The abbreviations A-I to B-VI and A-D indicate intraspecific lineages of M. natalensis (sensu Colangelo et al. 2013) and M. erythroleucus (sensu Brouat et al. 2009), respectively. Ethiopian Mastomys are in bold and in colour.

Fig. 3.

Median-joining network for Ethiopian Mastomys. The abbreviations H_1 to H_59 correspond to haplotypes in Table S1 in the Supplementary information.

The four Ethiopian Mastomys species have partly overlapping distributions (Fig. 1). The Ethiopian endemic species M. awashensis is widely distributed on the north-western plateau of the Ethiopian highlands, the ERV near Lake Koka and in the Awash NP. In the south-eastern plateau it was documented only from the Babille Elephant Sanctuary (see also Lavrenchenko et al. 2010). It occurs at the highest elevations of any Mastomys species, ranging from 967 at Mai-Temen to 3145 m a.s.l. in the Semien Mountains NP (Fig. 4). Mastomys erythroleucus is common in the southern part of the country (Fig. 1). Additionally, the species was recorded from two localities (vicinities of Lakes Koka and Ziway (Adamy-Tulu)) in the central part of the ERV. The distribution of M. erythroleucus at the northwestern plateau is sporadic, but it was relatively abundant in the Alatish NP (Fig. 1). This species was found within an elevational range from 380 m a.s.l. in the River Omo Valley to 1945 m a.s.l. in the Bulcha Forest on the slopes of the ERV (Fig. 4). Mastomys natalensis is widespread at the northwestern plateau. It was found in a single locality (Lake Koka) at the bottom of the ERV and was never documented in the south-eastern plateau of the Ethiopian highlands. The elevational range of the species varies from 417 m a.s.l. in the Gambela NP to 2191 m a.s.l. in Ambo (Fig. 4). The area near Lake Koka is exceptional because three Ethiopian Mastomys species co-occur there. The fourth species, M. kollmannspergeri, was found at a single locality (Mai-Temen 967 m a.s.l.) in the northeast of the country (Fig. 1), reaching here its eastern distributional limit.

Fig. 4.

The distribution of elevation profiles of Ethiopian Mastomys. Altitudinal range groups by steps of 300 m. The vertical axis shows the number of localities in each group. Note that the fourth species, M. kollmannspergeri, was found only at a single locality at 967 m a.s.l.

Species distribution modelling

All models obtained using MaxEnt had high mean AUC values (0.996 ± 0.001 for M. awashensis, 0.994 ± 0.002 for M. erythroleucus, and 0.987 ± 0.007 for M. natalensis), indicating good performance of the model. Based on the results of the Jackknifetest (Table S2) there are no specific bioclimatic variables that strictly limit the distribution of particular species, but their combination appears to be important. Nevertheless, it is possible to define variables that best describe the model species distribution. The environmental variables with highest gain when used in isolation are Isothermality (BIO2/BIO7) (*100) (AUK = 0.9082), Mean Temperature of Wettest Quarter (AUK = 0.8666) and Temperature Seasonality (standard deviation *100) (AUK = 0.8467) for M. awashensis, Temperature Seasonality (standard deviation *100) (AUK = 0.8788), Precipitation of Driest Quarter (AUK = 0.8664) and Aridity Index (AUK = 0.8599) for M. erythroleucus, Precipitation of Coldest Quarter (AUK = 0.8755), Mean Temperature of Wettest Quarter (AUK = 0.8440) and Aridity Index (AUK = 0.8409) for M. natalensis. The environmental variables that most significantly decrease the gain when omitted and, therefore, have the most information that is not present in other the variables are Min Temperature of Coldest Month (AUK = 0.9826) for M. awashensis, Precipitation of Coldest Quarter (AUK = 0.9783) for M. erythroleucus and Precipitation of Coldest Quarter (AUK = 0.9595) for M. natalensis. The predicted distributions under the current climate conditions for M. awashensis, M. erythroleucus and M. natalensis are shown in Fig. 5, and they clearly suggest that the arid Afar and Somali regions are not suitable for any Mastomys species. The most suitable conditions for M. awashensis are predicted at medium elevation in the most north-eastern part of the north-western plateau, but also in the ERV around Lake Koka and Awash NP, along the Chercher Mountain ridge, as well as in a small part of Eritrea. The distribution of M. erythroleucus is predicted mainly in the south of the country. For this species, suitable habitats may also occur in the eastern part of the Tigray region, in a narrow area of central Eritrea near the border with Ethiopia, where this species is absent (or possibly replaced by M. kollmannspergeri, though further research is required). On the other hand, the model indicates low/medium suitability for the broad belt including the Alatish NP and the Dinder NP on the border with Sudan, though we demonstrated the occurrence of M. erythroleucus in this area. There are almost no suitable conditions in the highlands and the northern part of the ERV. The predicted distribution for M. natalensis largely overlaps with that of M. awashensis. The highest probability of its occurrence was found at medium elevation around Lake Tana, but also in other parts of the northwestern plateau, especially in its westernmost part, and in the central part of the ERV around Lakes Koka, Ziway, Abijatta and Shalla. The model also predicts suitable conditions at the south-eastern plateau around the Arsi and Bale Mountains, but currently there is no confirmed record of the species from this area.

Fig. 5.

Ecological niche models for present-day conditions for (A) M. awashensis, (B) M. erythroleucus, and (C) M. natalensis. Colours (from blue to red) represent suitability of environmental conditions and black dots depict locations of specimens used to generate models.

Predicted distributions for particular Mastomys species partly overlap. The most similar are models for M. awashensis and M. natalensis (0.743). This finding is in accordance with available records of these species (Fig. 1). We documented the coexistence of both species at four localities: Gibe NP, Lake Koka, Dhati-Welel NP and the River Gumara (Table 1). There is a medium degree of overlap between M. natalensis and M. erythroleucus models (0.595). Both species co-exist at two localities: River Alvero Valley and Lake Koka. The smallest overlap in environmental requirements is observed between M. awashensis and M. erythroleucus (0.501). Nevertheless, these two species were found in sympatry at two localities with an average level of suitability: the eastern shore of Lake Koka and the Didessa River Valley.

The highest diversity of multimammate mice is in Ethiopia

In this study we used a combination of genetic approaches to describe the diversity and distribution of multimammate mice in Ethiopia. We provide evidence of four Mastomys species in Ethiopia, more than any other African country. Such diversity is apparently associated with the heterogeneous natural conditions. The two major blocks of the Ethiopian Highlands, separated by the ERV, provide a significant elevational gradient in ecological conditions (the elevation in Ethiopia ranges from –125 to 4533 m a.s.l). Further, savannahs surrounding the highlands are divided into at least two major biogeographical types: Sudanian in the northwest and Somali-Masai in the ERV and southeast (Linder et al. 2012). This makes the country an outstanding hotspot of biodiversity, not only for numerous highland endemics, but also for taxa living in open savannah-like ecosystems. Until the end of the last century, all Ethiopian Mastomys were considered a single species (called e.g. “Praomys natalensis” by Yalden et al. 1976). With increasing use of (cyto)genetic methods, the number of species has increased. Lavrenchenko et al. (1989) and Orlov et al. (1989) found Mastomys with 2n = 32 chromosomes in Ethiopia, and Yalden et al. (1996) in their revised Ethiopian check-list of mammals report two species of multimammate mice (called Praomys erythroleucus with 2n = 38, and P. hildebrandtii with 2n = 32). Later, Lavrenchenko et al. (1998) described a new Ethiopian species, M. awashensis, distinguished on the basis of its genital morphology, spermatozoal structure, allozyme electrophoresis and unique features of its 2n = 32 karyotype. Finally, we provide here the first evidence for the occurrence of a fourth species, M. kollmannspergeri, which is known from Sudanian savannahs in central Africa, with one (presumably isolated) locality in Sudan (Dobigny et al. 2008). This record means that the distribution range of M. kollmannspergeri is expanded to the northeast by 380 km.

Two species of Ethiopian Mastomys, M. natalensis and M. erythroleucus, are among the most widespread rodent species in sub-Saharan Africa. Both have pronounced intraspecific phylogeographical structure with six described lineages in M. natalensis (AI-AIII in Sudanian and BIV-BVI in Zambezian savannahs; Colangelo et al. 2013) and four lineages in M. erythroleucus (all of them in Sudanian savannah; Brouat et al. 2009). The parapatric distribution of particular lineages and the presence of narrow zones of hybridization at their zone of contact (e.g. Gryseels et al. 2017) suggest that their evolutionary history was influenced by Pleistocene climatic changes. During unsuitable (humid) periods their populations were fragmented (e.g. by riverine or montane forests; see Colangelo et al. 2013) and differentiated in allopatry in discrete savannah refugia. In phases when the savannah ecosystem expanded, their populations also increased and currently we can observe secondary contacts of differentiated lineages, sometimes located along large rivers (Brouat et al. 2009), or mountains (e.g. Eastern Arc Mountains; Colangelo et al. 2013). In Ethiopia, both species reach the north-eastern margin of their distribution. All genotyped individuals of M. natalensis belong to the clade A-III, which has hitherto been known only from south-western Kenya (Colangelo et al. 2013), and we have, therefore, significantly extended the known distribution of this species northwards (compare with the map in Colangelo et al. 2013 and Denys et al. 2017). Similarly, all Ethiopian individuals of M. erythroleucus had mtDNA of the easternmost lineage D, previously known only from one locality in southern Ethiopia and one in Sudan (Brouat et al. 2009). This lineage is further divided into “West” and “East” subclades, geographically separated by unfavorable mountain ridges at the west margin of the ERV and at the bottom of the ERV, which likely acted as a barrier to gene flow during humid periods of the Pleistocene when was filled by large palaeolakes (a similar phylogeographic pattern is also found in other savannah rodents, e.g. gerbils of the genus Gerbilliscus; Aghová et al. 2017). Visual inspection of distributional patterns (Fig. 1) and ecological niche models (Fig. 5) suggest that the two species overlap in Ethiopia to only a limited extent. While M. erythroleucus is common in the relatively arid southernmost part of the country (plus the Alatish NP in north-western Ethiopia), M. natalensis is widespread in the lower to middle elevations of the north-western highlands. More intensive sampling in western Ethiopia (e.g. Benishangul-Gumuz region) is still needed to assess the real level of their distributional and ecological overlap.

The third species, M. awashensis, was originally described as a narrow endemic in the River Awash Valley (Lavrenchenko et al. 1998). Genotyping of material collected in more recent surveys shows it is also found in eastern Ethiopia (Lavrenchenko et al. 2010) and Tigray (Colangelo et al. 2010), suggesting that it is more widespread than previously thought. Here we report numerous additional localities, especially from the west-central highlands, where it was found in sympatry with both species discussed above (Fig. 1). On the other hand, it seems to be the only multimammate mouse species occupying large parts of the northern highlands and eastern Ethiopia. We can speculate that the species differentiated in a long-term savannah refugium separated from other similar open habitats (inhabited by other Mastomys species) by mountain ranges. Such a refugium may have been located at the margin of the Afar lowlands, where other endemic taxa of savannah rodents were recently found, e.g. the genera Gerbilliscus (Aghová et al. 2017), Acomys (Aghová et al. 2019) and Arvicanthis (Bryja et al. 2019). Spreading deeper and higher into the highlands, especially along the river valleys, may have been facilitated by recent intensive deforestation of the landscape by humans. More detailed phylogeographic studies are needed to shed light on the evolutionary history of this taxon. Currently we consider it as endemic to Ethiopia, though it may also be present in neighbouring parts of Eritrea. The ecological niche model also predicts suitable environmental conditions for this species in northern Uganda and western Kenya, but it has never been confirmed in these countries. Its presence there seems unlikely because the belt of semi-desert savanna, inhospitable for all Mastomys, effectively separates southern Ethiopia and the predicted regions of Kenya and Uganda. While many species in the highlands are endemic to Ethiopia, this is a much rarer phenomenon among savannah taxa. In addition to M. awashensis there are only few such species, e.g. Mus proconodon (Bryja et al. 2014, Lavrenchenko & Bekele 2017), or savannah taxa from the margin of the Afar Triangle mentioned above.

We documented M. kollmannspergeri from a single locality (Mai-Temen) in the northeast of the country, where it was found in sympatry with M. awashensis (but not captured in the same trap line). Mastomys kollmannspergeri was originally described as a subspecies of M. natalensis from Ighazer (Niger) (Petter 1957). On the basis of chromosomal analysis of samples from the type locality, Chad and Southern Sudan, it was proposed that these specimens, referred to as M. cf. kollmannspergeri, should receive species designation (Viegas-Péquignot et al. 1987, Dobigny et al. 2002, Granjon et al. 2004). Mastomys verheyeni was described as a species with a narrow habitat surrounding Lake Chad in northeastern Nigeria and northern Cameroon (Robbins & Van der Straeten 1989, Lecompte et al. 2002, 2005). Further studies unequivocally showed that M. verheyeni is a junior synonym of M. kollmannspergeri (Denys et al. 2002, Dobigny et al. 2008). This species was reported from relatively few localities in the central part of the Sahelo-Sudanian savannah (Niger, northern Cameroon and Chad) and a single locality in eastern Africa in south-eastern Sudan (Dobigny et al. 2008). This is a new species for Ethiopia and the present record suggests that it may be more widespread in under-sampled regions of the eastern half of the Sahelo-Sudanian savannah belt.

Co-existence of multiple species of Mastomys

A notable finding of the present study is evidence for the co-existence of multimammate mice species as well as divergence in ecological niche occupancy. Dobrokhotov (1982) showed that M. erythroleucus and M. natalensis are sympatric in most of Nigeria. However, M. erythroleucus lives exclusively in the houses and outbuildings of humans, while M. natalensis is known only from natural biotopes (secondary forests, agricultural land and untouched savannah). In the north of Nigeria, in a semi-desert savannah where M. erythroleucus is absent, M. natalensis also lives in human buildings. Thus, populations of both species, occupying different ecological niches, and apparently do not co-occur with each other. In Sierra Leone the habitat preferences of these species are the opposite to those observed in Nigeria: M. erythroleucus dominates the natural biotopes, and M. natalensis prefers a commensal lifestyle (Bellier 1975, Hubert 1977). A similar situation is observed in Senegal, where M. natalensis lives strictly in human habitations, while M. erythroleucus is opportunistic but prefers savannah-like habitats, and a third species, M. huberti, inhabits natural and cultivated wetlands (Duplantier et al. 1990, 1996, 1997, Brouat et al. 2007). In Coastal Guinea, in Mankountan, M. erythroleucus is found both in wet rice fields and in houses, M. huberti prefers wet rice fields, where it co-exists with M. erythroleucus, but it never enters human buildings (Denys et al. 2012). In northern Cameroon, both M. erythroleucus and M. kollmanspergeri inhabit savannah and fields but are rarely found inside houses (in Kossa and Doué, respectively). In contrast, most M. natalensis were trapped inside houses, and only few were found outdoors (Dobigny et al. 2011). In Tanzania, M. natalensis shows both commensal and free-living populations and it was shown that commensalism could quickly lead to genetic differentiation (Gryseels et al. 2016).

Several Mastomys species can live in sympatry in Ethiopia. However, their co-existence shows a different character to that seen in West Africa, where one species is usually commensal, while another lives in natural savannah-like habitats. In Ethiopia, all three widespread species prefer natural habitats, but this observation can be biased by the relatively low intensity of sampling in human habitations. The only case of clear habitat separation was observed in the Alvero River Valley, where M. natalensis lives strictly in human settlements while M. erythroleucus occurs both in buildings and in savannah habitat (Lavrenchenko 1995). In nearby localities near Lake Koka, both M. natalensis and M. erythroleucus co-exist in natural biotopes and the results of distributional modelling predict a high level of suitability for both species in this area. A detailed study of commensal Mastomys in Ethiopia has not yet been conducted. Our pilot data show that M. erythroleucus can inhabit both the savannah and human settlements in southern Ethiopia (Arero Forest and Borena NP). M. awashensis demonstrates a lesser tendency towards commensalism. This species is rare in crop fields but is documented in household compounds in three hamlets in the Tigray region (Meheretu et al. 2013). This pattern is similar to the situation in West Africa, where M. natalensis and M. erythroleucus can be commensal, and M. huberti expresses a preference for natural habitats.

Ethiopia has the highest number of Mastomys species of any African country, one of which is endemic (M. awashensis). Further research based on a geographically extended data set will help to reduce the gaps in our understanding of the distribution of multimammate mice in Ethiopia and demonstrate their co-existence and level of commensalism. This information is of particular importance, because Mastomys rodents are important agricultural pests, and each of the sibling species can be a specific reservoir for pathogens potentially harmful to humans. Furthermore, detailed information on the distribution of Mastomys species will be necessary for the adoption of the most promising species-specific pest-control methods (e.g. Massawe et al. 2018).