Rhinolophus darlingi, Andersen, 1905

David S. Jacobs, Hassan Babiker, Anna Bastian, Teresa Kearney, Rowen van Eeden & Jacqueline M. Bishop, 2013, Phenotypic Convergence in Genetically Distinct Lineages of a Rhinolophus Species Complex (Mammalia, Chiroptera), PLoS ONE 8 (12), pp. 1-16 : 12-13

Systematics of R. darlingi

Eastern R. darlingi , which includes the holotype of R. d. barbetonensis (TM 2476), are genetically distinct from western R. darlingi , the two groups occurring in different clades. Eastern R. darlingi is embedded in the fumigatus clade that includes R. fumigatus , R. eloquens and R. hildebrandtii . Western R. darlingi form a monophyletic clade which diverged from the fumigatus clade ~9.7 Mya ( Figure 2 ). Furthermore, the percentage sequence divergence for cyt b within eastern R. darlingi (1.3%) and within western R. darlingi (2.1%) was much lower than between the two clades (8.1%). This percentage sequence divergence falls within the range used to distinguish other species of bats (3–27%; [ 20, 45]). This divergence was supported by differences in the size and shape of the bacula, characteristics which may act as crucial pre-zygotic reproductive barriers [ 60].

Ecological niche modelling showed substantial differences in the ecology of the two lineages. The distribution of the western lineage ( damarensis ; Figure 5 ) is restricted to the drier, hotter western half of the subcontinent ( Figures 5b and 6b ) characterisedbyanannualprecipitationof <500 mmanda mean temperature in the wettest quarter of about 28°C ( Figure 6b ). The distribution of eastern lineage ( R. darlingi ; Figure 5a ) is restricted to the more mesic eastern half of the subcontinent ( Figures 5a and 6a ) suggesting a limited tolerance to arid conditions and a general absence from regions with an annual precipitationof <100 mmandatemperatureseasonality <10% and> 55% ( Figure 6a ).

Thus, there are species level genetic differences supported by differences in bacula morphology as well as substantial ecological differences between R. darlingi in the east and R. darlingi in the west, suggesting that a taxonomic revision is required despite their substantial convergence in echolocation frequency and skull and post-cranial morphology. Since the type specimen of R. darlingi was collected in Zimbabwe (eastern part of the subcontinent) we retain the name R. darlingi for the eastern lineage ( Figure 1 ). Henceforth we refer to individuals in the western lineage as R. damarensis because the holotype of R. d. damarensis (TM9474, locality Namibia; Table S2 and Appendix S 1 in File S1 Supporting Information) is associated with this group ( Figure 3 ). The distribution of R. darlingi is restricted to the mesic eastern parts of the subcontinent ( Figure 5a ) and is described in detail in Monadjem et al. [ 42]. It appears to occupy mainly woodland and grassland biomes ( Figure 5a ; [ 41, 42]). The distribution of R. damarensis is restricted to the xeric regions from south-western Angola, through northern Namibia, southwards as far as Carnarvon in south-western South Africa and occupies mainly arid savanna, Succulent- and Nama-Karoo, shrubland and desert ( Figure 5b ; [41,42]). The eastern limits of the distribution of R. damarensis appears to be demarcated by the area around Taung, the eastern most locality for a specimen in R. damarensis (TM48040, Table S2 and Appendix S 1 in File S1 Supporting Information; Figure 1 ) and close to the boundary between the Savanna biome in the west and the Grassland biome in the east [ 41].

The R. damarensis clade comprises two lineages, a northern lineage restricted to the more mesic regions of northern Namibia and a southern lineage with a distribution across several biomes in central and north-western South Africa, extending as far north as central Namibia ( Figure 1 ). The genetic differentiation between the two lineages in R. damarensis is of the same magnitude as that used to infer cryptic species in other rhinolophids e.g. R. arcuatus [ 61]. In combination with ecological divergence the genetic divergence reported here suggests that R. damarensis may itself consist of cryptic species and further taxonomic revision of both this clade and R. darlingi (sensu lato) is required.

Morphological and acoustic convergence

Despite the relatively ancient split (~9.7 Mya), marked genetic differentiation and the occupation of different biomes there was convergence in the cranial and post-cranial measurements of R. damarensis and R. darlingi ( Figure 3 , Table S1 and Appendix S 1 in File S1 Supporting Information) as well as in the noseleaf width and resting echolocation frequency ( Table S1 and Appendix S 1 in File S1 Supporting Information). None of these parameters are thus taxonomically informative with respect to differentiating R. damarensis from R. darlingi . The phenotypic similarity between the two darlingi lineages is greater than that between any other pair of species in our analyses ( Table 2). Furthermore, R. darlingi is more similar to R. damarensis in body size and echolocation frequency than it is to any of the other species in the fumigatus clade ( Figure 2 ) which all have bigger body sizes and lower resting echolocation frequencies (than both R. darlingi and R. damarensis ) ranging from 48–67 mm (forearm length) and 32– 54 kHz, respectively [ 31, 42]. They are also more similar to each other than either is to the ancestral character state at node B ( Figure 2 ; forearm length = 58.7 mm; resting frequency = 60.1 kHz [ 39]).

Such convergence involving morphology and echolocation in non-sibling species deviates from the pattern normally found in cryptic species of bats in general and rhinolophids in particular. All cryptic bat species uncovered so far have similar morphology but divergent echolocation frequencies which differed by up to 13 kHz (e.g. [ 19, 20, 62], however see 61). These cryptic species all co-occurred and such differences in sensory traits may be important isolating mechanisms between species [ 63] leading to resource partitioning and subsequent genetic divergence. At lower echolocation frequencies, where differences in frequency translate into large differences in wavelength [ 64], habitat and insect prey may be partitioned [27,36,65,66]. At higher frequencies, where differences are unlikely to equate to marked differences in wavelength, resource partitioning may be mediated by the selection for discrete frequency bands to facilitate intraspecific communication [ 29, 33, 34, 64]. The call frequency of one or more of the co-existing species may shift so that individuals are more sensitive, and will respond preferentially, to the calls of their own species [ 27, 29], facilitating intraspecific communication. Divergence in echolocation calls in sympatry may therefore be a consequence of competition leading to character displacement in at least one phenotypic trait that permits resource partitioning and coexistence. There is at least one example where sympatric bat lineages converge in both morphology and echolocation [ 61], attributed to either novel niche partitioning or recent contact. Here lineages within the rhinolophid R. arcuatus may have partitioned their niches in novel ways along dimensions not previously considered. Alternatively, convergence may have evolved in allopatry with the two lineages recently making contact [ 61].

Strong convergence in allopatry may be a consequence of lineages evolving in the absence of competition from ecologically similar species; their phenotypes being the result of neutral evolution or shaped by selection pressures resulting from occupying similar niches albeit in different biomes. The disjunct distribution of R. damarensis and R. darlingi may allow their morphology and echolocation to converge because they do not compete for foraging space, prey or discrete frequency bands. Such convergence may result from one or more of several processes including inheritance from a common ancestor, adaptation to similar local environments, random genetic drift and shared constraints [ 5, 67]. Inheritance from a recent common ancestor is unlikely to explain the phenotypic convergence between R. darlingi and R. damarensis . They are placed in different, albeit sister clades: R. damarensis in its own clade and R. darlingi in the fumigatus clade ( Figure 2 ). The two lineages last shared a common ancestor ~9.7Mya, giving rise to numerous lineages comprising individuals that are bigger in size and echolocate at lower frequencies than either R. darlingi and R. damarensis viz. R. fumigatus , R. eloquens and R. hildebrandtii [ 42]. Similarly, the fact that there are species that share a common ancestor with R. damarensis but that are nevertheless divergent in both morphology and echolocation appears to exclude constraints as an explanation for the convergence. Local adaptation also appears to be an unlikely explanation for the convergence because the two species occur in different biomes and it would be expected that local adaptation would lead to divergence not convergence. It is therefore likely that convergence may be the result of random genetic drift especially since rates of convergence can be high when lineages are diverging only under the influence of genetic drift [ 6]. Testing this hypothesis would require thorough and integrated analyses of genetic and phenotypic variation in both the damarensis and the fumigatus clades ( Figure 2 ). Nevertheless, there is some evidence that founder effect and random genetic drift may be implicated in the evolution of different body sizes during the diversification of the R. hildebrandtii species-complex [ 31], one of the lineages in the fumigatus clade – this clade also includes R. darlingi ( Figure 2 ). If so, smaller body size in R. darlingi may have evolved through genetic drift resulting in the convergence of body size between it and R. damarensis , assuming that the ancestral body size of R. damarensis is similar to its current body size. Similar body sizes, coupled with the unique flutter-detection system of rhinolophids [ 68], would require similar detection distances and levels of flight manoeuvrability that could lead to convergence in wing morphology and echolocation frequency and possibly also insect prey types. This may be especially so given the well-established correlations between body size on the one hand and wing loading, echolocation frequency and bite force, on the other, in bats [63,69–71]. Bite force is in turn correlated with diet [ 70].

The split between the two damarensis lineages provides further insight into the role of random genetic drift in the evolution of rhinolophids in southern Africa. The split occurred ~5 Mya ( Figure 2 ) which is similar to divergence times reported in many co-distributed taxa including the African four-striped mouse ( Rhabdomys pumilio , [ 72]), the southern rock Agama ( Agama atra , [ 73]) and the gecko, Pachydactylus rugosus [ 74]. Similarity in the timing of evolutionary diversification amongst co-distributed but diverse taxa is likely a consequence of vicariant evolution [ 75] and in southern Africa this has been attributed to climate change and subsequent vegetation shifts during the Plio-Pleistocene and Miocene [72–74,76] together with the Plio-Pleistocene uplift of southern Africa’s great escarpment and interior plateau [ 77]. Diversification across these lineages coincided with a period of increased aridity in southern Africa as a result of the interaction in the Miocene between global cooling [ 78] and tectonic uplift that resulted in a topography which sloped from east to west causing a rainshadow effect across the region [ 79, 80]. This in turn resulted in an east-west gradient of rainfall and subsequent changes in vegetation which included the contraction of forests, and the expansion of savanna woodlands, grasslands and shrublands [ 81 – 83] towards the end of the Miocene (7–5Mya). Such climatic oscillations and habitat fluctuation/fragmentation promote diversification of lineages. The diversification of R. damarensis into two distinct mitochondrial lineages may have been caused by disruption to gene flow associated with these changes in biomes especially since the lineages currently occupy separate geographic regions. Given that rates of convergence can be high when lineages diverge under the influence of genetic drift [ 6] convergence in phenotype in the two damarensis lineages would not be surprising if drift was the dominant process acting during their initial divergence. Testing this hypothesis and the relative influence of the different processes that could bring about convergence can only be elucidated through thorough and integrated analyses of both genetic and phenotypic variation using multiple rapidly and slowly evolving genetic markers, within the context of historical biogeography.

In conclusion, cryptic lineages in R. darlingi (sensu lato) appear to have arisen independently and in isolation of each other allowing convergence in both morphology and echolocation. Similarly, cryptic lineage diversification within R. damarensis also appears to have arisen more recently in response to changes in biome boundaries during the Miocene. Although this might be due to vicariant evolution the role of other processes such as adaptation as a result of occupying similar niches cannot be excluded at this stage.

Figures

Figure 1. The distribution of R. darlingi (shaded areas) and sample localities (symbols). Squares = eastern R. darlingi; Circles (southern lineage) and triangles (northern lineage) = western R. darlingi (R. damarensis; Fig. 2). TM 9474 = holotype R. d. damarensis Namibia), TM 2476 = holotype of R. d. barbetonensis South Africa. doi: 10.1371/journal.pone.0082614.g001

Figure 2. Bayesian consensus topology of Rhinolophus spp. based on cytochrome b. Values above nodes indicate Bayesian posterior probabilities (> 0.80) and below nodes report parsimony bootstrap support. Branch lengths are proportional to average number of substitutions per site. Divergence times (millions years ago, Mya) were estimated for three nodes of interest: node A = 4.8 Mya (95% HPD; 2.69–7.52), B = 9.68 Mya (95% HPD; 5.87–10.04), C = 10.51 Mya (95% HPD; 6.65–15.43). The split between the Hipposideridae (outgroup) and the Rhinolophidae together with fossil dates for R. ferrumequinum were used as calibration points; calibration points are indicated on the tree with an asterisk *. Sequences from the type localities for R. darlingi darlingi (TM2476) and R. darlingi damarensis (TM9474) are indicated on the tree in bold. The eastern R. darlingi darlingi subspecies forms a single clade together with R. fumigatus, R. eloquens and R. hildebrandti (blue) while R. damarensis damarensis, distributed in the western region of southern Africa, is composed of two very well supported clades that subdivide into northern and southern lineages (red) that diverged in the late Miocene. doi: 10.1371/journal. pone.0082614.g002

Figure 3. Plot of canonical scores extracted by Discriminant Function Analysis from 24 skull and external parameters. Species abbreviations are the same as those in Table S1 (Supporting Information). Museum accession numbers relate to holotypes (See legend of Figure 1). doi: 10.1371/journal.pone.0082614.g003

Figure 5. Probability of occurrence for A) R. darlingi and B) R. damarensis in southern Africa. The predictions (Low-High) show a suitability gradient from 0–1 based on environmental suitability where 0 (low) indicates a zero probability of occurrence and 1 (high) indicates maximum probability of occurrence. White circles indicate locality records used for modelling the species. Dotted white triangles indicate locality records from where DNA sequences were obtained. Variables used in the final model include annual mean temperature, isothermality, temperature seasonality, temperature annual range, mean temperature of the wettest quarter, mean temperature of the driest quarter, mean temperature of the warmest quarter, mean temperature of the coldest quarter, annual precipitation, precipitation seasonality, precipitation of the wettest quarter, precipitation of the driest quarter, and precipitation of the coldest quarter.

Figure 6. The log response curves for R. darlingi. A) R. darlingi; (i) annual precipitation, (ii) precipitation of the wettest quarter, (iii) temperature seasonality when all locality records were used to build the model (temperature seasonality (Coefficient of Variation) is the standard deviation of the monthly temperature estimates expressed as a percentage of the mean of those estimates (i.e., the annual mean); and B) R. damarensis; (i) annual precipitation, (ii), precipitation of the wettest quarter, and (iii) mean temperature of the wettest quarter when all locality records were used to build the model. doi: 10.1371/journal. pone.0082614.g006