Study area

Rodents were collected from five mountain massifs in Ethiopia: four located west of the Great Rift Valley (Aboye Gara, Borena Saynt, Mt. Choqa, and Guassa Community Conservation Area, hereafter referred to as Guassa) and one located east of the Rift (Bale Mountains). The five sites are covered by Afroalpine habitats at high elevations (above 3,000 m a.s.l.). Additionally, an Afromontane Ericaceous belt occurs in transitional elevation zones, while various types of forests are found at lower elevations (Fig. 1, see Lillesø et al. 2011, Bryja et al. 2019b for further details).

The five mountain massifs are situated in the Eastern Afromontane Biodiversity Hotspot, which extends from Ethiopia's highlands in the north to Mozambique in the south (Mittermeier et al. 2004, CEPF 2012). This hotspot is primarily composed of high-altitude plateaus and isolated mountain peaks and exhibits notable levels of species endemism. For example, approximately 21% of the approximately 490 mammalian species discovered in this hotspot are endemic (CEPF 2012).

Taxon sampling

The study analysed 138 individuals from the following ten species across four rodent genera. For details about the capturing sites, altitude range, number of captured females and males, mean body mass, and mean standard body measurements, refer to Table 1. Five out of the six species from the genus Stenocephalemys were included: white-tailed Ethiopian rat S. albocaudatus, Zima's Ethiopian rat S. zimai, Sokolov's Ethiopian rat S. sokolovi, gray-tailed Ethiopian rat S. griseicauda, and white-footed Ethiopian rat S. albipes. All species are nocturnal and considered predominantly herbivorous, with 80% plant matter in their diet (information available for S. albipes only, Bantihun & Bekele 2015). Stenocephalemys albipes is the smallest and most widespread species of the genus. It prefers forest types at lower elevations (Bryja et al. 2019b). Stenocephalemys albocaudatus is the largest species of the genus (Mizerovská et al. 2020) and highly specialised for life in open Afroalpine grasslands in Bale and Arsi Mountains and Guassa (3,000-4,500 m a.s.l.). Stenocephalemys griseicauda is only found on the eastern side of the Rift Valley in the Bale and Arsi Mountains at 3,000-3,800 m a.s.l., where it inhabits mainly Erica shrublands. Stenocephalemys zimai is an Afroalpine specialist that inhabits moorland and grassland habitats (3,600-4,000 m a.s.l.) in the northern part of the Highlands (Bryja et al. 2019b, Mizerovská et al. 2020). Stenocephalemys sokolovi inhabits grasslands and mosaic Afroalpine habitats with Erica shrubs in the western part of the Rift Valley (Mizerovská et al. 2020). Arvicanthis abyssinicus is a diurnal (100% of captures by the end of the daylight phase, Lövy et al. 2023), exclusively herbivorous species (Wilson et al. 2017) adapted to high-elevation open habitats (Bryja et al. 2019a, b). The Heller's vlei rat Otomys helleri is a true herbivore with prevailing diurnal activity (80% of captures occurring by the end of the daylight phase, Lövy et al. 2023). The last three species belong to the genus Lophuromys. The buff-spotted brush-furred rat L. flavopunctatus inhabits elevations from 2,600 to 3,300 m a.s.l. in central Ethiopia (Komarova et al. 2021). It is primarily diurnal (71%, Lövy et al. 2023) and likely omnivorous with a prevailing invertebrate diet (38% and 52% plant matter in the diet during the wet and dry season, respectively: Bantihun & Bekele 2015). The black-clawed brush-furred rat Lophuromys melanonyx is a highland diurnal, primarily herbivorous (98 [95-100]% plant matter in diet: Yalden & Largen 1992) that inhabits Afroalpine moorland above 3,200 m a.s.l. in Bale Mts. (Wilson et al. 2017). The Simien brush-furred rat Lophuromys simensis is mostly found at higher elevations above 3,000 m a.s.l. but has also been captured at 1,800 m a.s.l. (see Bryja et al. 2019b) and its biology is largely unknown. As L. simensis skulls were not available for this study, the species was only included in the analyses of long bones (see below). Our sampling does not cover all species that comprise the Afroalpine rodent communities at the study sites. Some species, such as the genera Dendromus, Megadendromus, and some Otomys species, are relatively rare. Trapping success depends on the season or the type of traps used. Finally, no truly fossorial species from the genus Tachyoryctes were included in the study due to their distinct morphology reflecting their fossorial life, in contrast to the species analysed, which exhibit regular surface activity.

Fig. 1.

Map showing sampling localities (depicted as blue circles) in the Ethiopian Highlands where rodents were captured for this study. Each photo shows a typical habitat found in a particular locality.

To summarise, the species included in the study primarily inhabit open Afroalpine grassland habitats, with some also found in neighbouring ericaceous vegetation. For comparison, we included a single forest species, S. albipes. The species of the individuals were determined using a combination of morphometric diagnostic traits (Monadjem et al. 2015), molecular barcoding based on the mitochondrial gene for cytb (Bryja et al. 2019a, Mizerovská et al. 2020, Komarova et al. 2021), and their locality of origin (cf. Bryja et al. 2019b).

Table 1.

Capture sites, the number of captured animals, and their mean body mass for nine species of Ethiopian rodents were used for this morphological study of craniodental and long bone morphology.

Fig. 2.

Cranial, mandibular, and front and hind limb morphological traits examined in this study. For a detailed description of each trait, see Table S1.

Morphological traits studied

The study analysed two sets of functional traits: skull and mandible dimensions, which mainly reflect sensory perception and diet, and limb bone lengths, which reflect locomotion style. A total of 15 cranial and mandibular traits and four limb bone traits were analysed (see Table S1 for the complete list and Fig. 2 for their definition). As the species differ in their body size, we standardized by the condylobasal length of the skull and mandible length, respectively. Zygomatic width, for instance, is a proportional width of the skull relative to its length. The mandible length was standardised by dividing by the skull length. The only variables left in their original units were the condylobasal length of the skull, which measures body size, and the joint-to-angular process angle of the mandible. Two mandible variables (Z_I and A_D) were designed as proxies of bite force (Freeman & Lemen 2008). They were calculated from cross-sectional dimensions of the lower incisor and the mandible at its diastema (the gap between incisors and molars). These dimensions are defined in Table S1.

Three locomotory indices were calculated based on the lengths of limb bones: the brachial index (BI), the crural index (CI), and the intermembral index (IMI). The BI was calculated as the length ratio of the distal and proximal parts of the forelimbs and hindlimbs. The CI was calculated as the maximal length of the tibia (measured from the proximal border to the most distal part of the medial malleolus) divided by the length of the femur (measured from the femoral head to the distal condyles). The IMI was calculated as the ratio of the forelimb (the length of the ulna plus the length of the humerus) to the hindlimb (the length of the tibia plus the length of the femur) lengths. For the reference see Samuels & Van Valkenburgh (2008).

Fig. 3.

The ordination of species in the space of skull and mandible traits reduced to the bgPC1,2 plane. The large, filled circles indicate species means; the small, empty circles are individuals, post hoc projected to the axes. The arrows represent variable loadings (magnified three times for better visibility). The colour legend is provided in the figure; variable abbreviations correspond to those in Table S1.

Statistical analyses

The morphological traits measured were subjected to between-group principal component analysis (bgPCA); PCA performed on group (= species) mean values. With K species, such PCA defines up to K –1 principal components, and individual data are then projected onto them post hoc (Yendle & MacFie 1989). To account for the heterogeneous nature of the variables, all were scaled so that the variance of species means was unity. The number of retained principal components was estimated using the elbow method of Salvador & Chan (2004). The ordination of species means and the size and shape of individual scatters reveal the packing of species in this reduced phenotype space. The variable loadings, which define the principal components, allow for the functional interpretation of directions in this space.

Fig. 4.

Differences in CBL, ZW, IOW, OD, ABL and ABW in the studied Afroalpine rodents. Black, dark grey and light grey circles indicate significant results for the comparison between two sister branches of the phylogenetic tree at P < 0.001, 0.01 and 0.05, respectively; white arrows indicate the relative increase of the respective trait for a given sister pair.

Fig. 5.

Differences in ML, MD, JLM1 and JAPrA in the studied Afroalpine rodents. Black, dark grey and light grey circles indicate significant results for the comparison between two sister branches of the phylogenetic tree at P < 0.001, 0.01 and 0.05, respectively; white arrows indicate the relative increase of the respective trait for a given sister pair.

Finally, we subjected the functional traits to separate univariate analyses. In each of them, we used an analysis of variance (ANOVA) with a particular trait as a response variable and species as the only categorical predictor. If the effect of species was significant, we followed the hierarchy of their phylogenetic relationships, and for each split in the tree (reported in Fig. S1), we tested for the differences between sister clades. We achieved this by setting up a priori contrasts to compare each pair of sister branches while excluding the rest of the species. The first contrast, the deepest one, was between Lophuromys and the murine genera, with no species left aside, and the last contrast was the shallowest one, between S. albipes and S. zimai. Topology and divergence times in the tree (Fig. S1) were obtained from previously published phylogenies (Aghová et al. 2018, Komarova et al. 2021, Mikula et al. 2021, Nicolas et al. 2021). All tests were performed in R (R Core Team 2023).

Morphological variation in skull and mandible shape

The first pattern distinguishes O. helleri from the genera Stenocephalemys and Lophuromys, with A. abyssinicus in between (Fig. 3). Otomys helleri and, to a much lesser extent, A. abyssinicus, differ from the others mainly due to their relatively long molar rows, massive (long and deep) mandible, wide joint-to-angular process angle, and strong bite force (indicated by higher values of the cross-sectional area at the diastema and the section modulus of the lower incisors). The observed morphological variations suggest dietary differentiation, as both O. helleri and A. abyssinicus are specialized herbivores. This finding aligns with prior studies on rodents, which have shown that herbivorous rodents typically have more robust mandibles (e.g. Michaux et al. 2007, Samuels 2009, Grossnickle 2020) and greater bite force compared to faunivorous taxa (Freeman & Lemen 2008, Verde Arregoitia et al. 2017). Differences in the joint-to-angular process angle are reflected in other cranial characteristics, such as gape angle, molar dimensions, and jaw articulation, that influence the insertion of chewing musculature (e.g. Verde Arregoitia et al. 2017). A similar correlation between prey hardness and bat cranial morphology was observed in a vespertilionid bat community inhabiting a mountain ecosystem in Mexico (Ayala-Berdon et al. 2023). This finding suggests that species feeding on harder prey have evolved a skull shape characterised by more developed muscle attachment areas (Nogueira et al. 2009).

Fig. 8.

Differences in BI, CI, and IMI in the studied Afroalpine rodents. Black, dark grey and light grey circles indicate significant results for the comparison between two sister branches of the phylogenetic tree at P < 0.001, 0.01 and 0.05, respectively; white arrows indicate the relative increase of the respective trait for a given sister pair.

The phylogenetic relationships among the examined taxa indicate functional similarity is likely due to evolutionary convergence. Notably, the omnivorous Stenocephalemys, which lacks significant differences in traits reflecting feeding ecology (e.g. UMR and LMR, see Fig. 6), is more closely related to the herbivorous pair Arvicanthis and Otomys than to the omnivorous Lophuromys. Regarding Lophuromys, the two studied species differ in the proportion of animal matter in their diet. While L. flavopunctatus is omnivorous/insectivorous, L. melanonyx is a true herbivore (Yalden & Largen 1992). Consequently, L. melanonyx has a broader skull and higher values of AD (the cross-sectional area at the diastema; a predictor of bite force – Freeman & Lemen (2008)) compared to L. flavopunctatus. These results support the general trend that specialised herbivores, primarily consuming vegetative plant parts, tend to have a broader skull and higher bite force (e.g. Samuels 2009, Verde Arregoitia et al. 2017).

Our data exemplify phylogenetic niche conservatism. The genus Otomys and the tribe Arvicanthini shared a common ancestor around nine million years ago, according to Mikula et al. (2021). The Otomys lineage appears to have been herbivorous since then, supported by fossil evidence demonstrating the gradual evolution of its distinctive laminated molars (Pocock 1976, Denys 1989) adapted to chewing grassy food. This conclusion is evidenced by a higher complexity of the occlusal surface (Evans et al. 2007, Tiphaine et al. 2013) and a specialised digestive tract for herbivory in extant species (Bronner et al. 1988). Given the dietary diversity within the tribe Arvicanthini, herbivory emerges as a less common trait, suggesting both retention of an ancestral state and convergence.

Multivariate analysis revealed a second prominent pattern, highlighting interspecific differentiation within the genera Stenocephalemys and Lophuromys. This pattern indicates specialization for life in different habitats spanning different elevational zones within the Afroalpine ecosystem (see Mizerovská et al. 2020, 2023). The species of the genus Stenocephalemys best exemplified this differentiation. Stenocephalemys albipes is a forest-dwelling species with a relatively small skull and small orbits separated by a wide interorbital constriction. On the opposite end of the spectrum is S. albocaudatus, an Afroalpine specialist with a relatively larger skull and orbits and a narrower interorbital constriction. The other species fall between these two extremes. Those inhabiting ericaceous bushy vegetation consistently display a slightly more ‘forest-like’ morphology than their Afroalpine counterparts, as reflected in a relatively wider interorbital constriction. East of the rift, S. griseicauda aligns more closely with S. albipes than S. albocaudatus, and on the western side, the same relationship is observed between S. sokolovi and S. zimai (or again, in the specific case of the Guassa plateau, with S. albocaudatus). Similarly, the two species of Lophuromys show differentiation along the same line as the five species of Stenocephalemys. The highland specialist L. melanonyx from Bale Mts. shifts towards the high-elevation side of the gradient, primarily owing to its narrower interorbital constriction, and only marginally overlaps with L. flavopunctatus (see Fig. 3). This suggests that the latter species is primarily adapted to life in more forested habitats at relatively lower elevations.

The variations in size and shape associated with this elevational trend corroborate the findings of a geometric morphometric study that included a wider range of species from the genera Stenocephalemys, Lophuromys, and Arvicanthis across different elevational zones (Mizerovská et al. 2023). Likewise, Peterhans et al. (1998) observed a narrow-headed (i.e. stenocephalic) rodent guild inhabiting high-altitude open moorland in the Rwenzori Mountains in Uganda. This guild comprised vole-like herbivorous rodents such as O. typus dartmouthi, O. denti denti, and Dasymys montanus. The stenocephalic phenotype, characterised by large orbits and a relatively narrow interorbital constriction facilitating the shift of eyes to the top of the head, is recognised as an adaptation to life in open Afroalpine habitats. In such environments, rodents benefit from improved and faster detection of aerial predators (Peterhans et al. 1998, Beolchini & Corti 2004, Mizerovská et al. 2023). The influence of altitude on shaping mammalian morphology has also been demonstrated in mammals from Tibet. High-altitude populations of several rodent species showed a larger and broader skull and a larger bulla compared to populations from lower altitudes (Feijo et al. 2019). In this context, it is important to emphasise that the observed morphological changes in size predominantly indicate the influence of altitude. Relatively larger body size is advantageous in mitigating the increased thermoregulatory demands associated with colder temperatures at high-altitude environments. Conversely, morphological changes in shape, such as enlarged orbits and a narrower interorbital constriction, primarily represent adaptations to life in open habitats. These specific morphological changes are considered adaptations that enhance the ability to thrive in open environments.

When comparing high-altitude species within the genus Stenocephalemys, it is noteworthy that S. albocaudatus occupies a slightly different position in the defined morphospace than S. zimai, S. griseicauda, and even more so than S. sokolovi. This distinction results from its larger and wider average skull size (CBL and ZW) (see Figs. 3 and 4). Furthermore, S. sokolovi has a significantly wider interorbital constriction than all other Afroalpine species. These morphological differences may influence microhabitat selection within the Afroalpine rodent community. In areas where S. albocaudatus occurs sympatrically with either S. griseicauda (Arsi and Bale Mts.; in the latter, they occupy different elevational zones) or S. sokolovi (Guassa), the relatively robust skull and narrow interorbital constriction of S. albocaudatus suggest a superior adaptation to life in open Afroalpine habitats compared to its congeners. It is plausible that the larger size of S. albocaudatus confers a competitive advantage over the smaller S. sokolovi and S. griseicauda. This conclusion is supported by observations in areas where S. sokolovi occurs alone, such as in Borena Saynt National Park, where it is commonly found in the highest Afroalpine habitats (Mizerovská et al. 2020), whereas the opposite pattern is observed in Guassa (based on our observations). We hypothesize that the coexistence of S. sokolovi and S. albocaudatus in the Afroalpine ecosystem is primarily driven by niche partitioning resulting from interspecific competition, with S. albocaudatus being a superior competitor due to its larger size and morphological adaptations for life in open habitats. A more in-depth study of the ecological requirements and morphology of these species is warranted to gain a comprehensive understanding of their coexistence dynamics.

Morphological variation in locomotion-related indices

The rodent species studied also showed differentiation in locomotion indexes derived from the lengths of long bones of the front and hind limbs. Despite the relatively low variability in the locomotion indices, indicating a generalised (also called ambulatory) murid phenotype with no apparent specialisation for a particular locomotory style (see Samuels & Van Valkenburgh 2008), finer-scale differences were discerned among the rodents studied. Parallel to the patterns observed in cranial traits, a notable observation is the marked morphological distinctiveness of O. helleri, highlighted by its isolated position within the morphospace examined. This distinctiveness primarily stems from its relatively higher IMI compared to all other species examined (see Fig. 7). A relatively higher IMI suggests that rodents tend to use forelimbs and hindlimbs equally (Samuels & Van Valkenburgh 2008), indicating a limb configuration adapted to quadrupedal terrestrial locomotion (Howel 1965). Indeed, Otomys species are recognised as quadrupedal, employing a diagonal sequence of alternating limb movements during locomotion, as exemplified by O. irroratus (Davis 1972). In addition, they have been observed to run rapidly with the body held flat and close to the ground, probably an adaptation for rapid movement along vegetation runways (Davis 1972). Comparable IMI values to Otomys (IMI = 0.769 ± 0.01) were found in the California vole Microtus californicus (IMI = 0.786, Samuels & Van Valkenburgh 2008), whose locomotor mode resembles that of Otomys. Interestingly, recent research has shown that the quadrupedal running locomotion in O. helleri coincides with an enlarged spleen, leading to the hypothesis that this adaptation may be advantageous during endurance running. The enlarged spleen in O. helleri could release more red blood (Lövy et al. 2023) cells for aerobic physiology, suggesting a potential advantage during rapid movements, especially when disturbed during foraging and relocation within a territory (Lövy et al. 2023). Our results and field observations support this scenario and suggest that O. helleri relies on fast quadrupedal running, which may be advantageous in response to disturbances during foraging and territorial movements.

The remaining rodent species fell into two clusters (see Fig. 7). The first cluster consisted of Stenocephalemys species with relatively higher CI and BI. In contrast, the second cluster included A. abyssinicus and three species of Lophuromys, which had relatively lower BI and CI values. Mode of locomotion is an important factor influencing a rodent's ability to coexist within communities, particularly in microhabitat occupancy and predation pressure (Taraborelli et al. 2003). For instance, saltatory bipedality is common in desert rodents, helping minimise predation risk when foraging in open habitats. This mode of locomotion offers advantages such as high speed, better manoeuvrability, and the ability to cover a more extensive home range in open habitats (e.g. Randall 1993, Kotler et al. 1994). Although bipedal rodents are not found in open Afroalpine habitats, our findings suggest that Stenocephalemys species are better adapted for jumping, as evidenced by their elongated tibiae (relatively high CI) and short humeri (relatively high BI) (see Table S5). These limb proportions are conducive to quadrupedal hopping. This locomotory pattern likely provides rapid acceleration and faster responses to predator attacks, similar to what has been observed in bipedal saltatory rodents (Taraborelli et al. 2003). Furthermore, locomotory pattern diversification, namely the differentiation between quadrupedal and bipedal rodents, may have played an essential role in shaping rodent communities in desert ecosystems (e.g. Moore et al. 2017). In contrast, Arvicanthis and Lophuromys rely more on quadrupedal running, although not to the extent of O. helleri, as described above. Rodents employing quadrupedal running to evade predators tend to live in areas with vegetation closer to their burrows than saltatory species (Schröpfer & Klenner-Fringes 1991). Indeed, A. abyssinicus is most abundant in the Helichrysum-Festuca habitat in the Simien Mountains and Guassa (M. Lövy, pers. observ.). This habitat provides enhanced protection against predation due to the up to 1 m high and wide canopy of Helichrysum splendidum (Wubie 2018).

At the interspecific level, differences in locomotory indices were generally less pronounced and, in many cases, not statistically significant within the genera Stenocephalemys and Lophuromys (see Table S7). This variation is likely due to the relatively small sample sizes for some species examined. Notably, S. albipes and S. sokolovi stand out from the other congeners by having a considerably longer tail relative to body length (see Table 1). Although this parameter was not analysed in the current study, as it does not pertain to the cranial, mandibular, or locomotory indices, a longer tail is indicative of enhanced climbing ability in these two species (see e.g. Verde Arregoitia et al. 2017). This ability was not explicitly revealed in our analyses. Enhanced climbing ability could be advantageous for both S. albipes, a species adapted to forested environments, and S. sokolovi, which, despite inhabiting more open habitats at higher altitudes, shows a preference for areas with Erica shrubs (Mizerovská et al. 2020, M. Lövy, pers. observ. from Guassa). In the case of Lophuromys, L. melanonyx differed from the other two species by its higher IMI. Although we can only speculate at this point, it is conceivable that this difference reflects morphological adaptations associated with the predominant diurnal activity and foraging on vegetation, similar to the herbivorous O. helleri. However, more detailed research is needed to clarify this aspect.