Population diversification in the frog Mantidactylus bellyi on an isolated massif in northern Madagascar based on genetic, morphological, bioacoustic and ecological evidence

Abstract

In the processes that give rise to new species, changes first occur at the population level. But with the continuous nature of the divergence process, change in biological properties delimiting the shift from "individuals of divergent populations" towards "individuals of distinct species", as well as abiotic factors driving the change, remain largely ambivalent. Here we study diversification processes at the population level in a semi-aquatic frog, Mantidactylus (Brygoomantis) bellyi, across the diverse vegetation types of Montagne d'Ambre National Park (MANP), Madagascar. Genetic diversity was assessed with seven newly developed microsatellite markers as well as mitochondrial DNA sequences and concordance with patterns of ecological, morphological, and bioacoustic divergence evaluated. We found M. bellyi lacking mitochondrial differentiation within MANP, while microsatellite datasets partitioned them into three highly differentiated, geographically separated subpopulations (with indications for up to five subpopulations). The molecular grouping-primarily clustering individuals by geographic proximity-was coincident with differences in mean depth and width of waters, suggesting a possible role of fluvial characteristics in genetic exchange in this stream-breeding species. Genetic clustering not consistent with differences in call properties, except for dominant call frequencies under the two-subpopulations model. Morphological divergence was mostly consistent with the genetic clustering; subpopulations strongly differed by their snout-vent length, with individuals from high-elevation subpopulations smaller than those from populations below 1000 m above sea level. These results exemplify how mountains and environmental conditions might primarily shape genetic and morphological divergence in frog populations, without strongly affecting their calls.

Fig 1. Habitat survey sites and their respective vegetation cover intervals.

A] Map showing the location of Montagne d’Ambre National Park and the different collection localities of ecological variables within the park are represented by orange stars for plots and light blue triangles for streams. The localities were grouped a priori into six sites based on geographical proximity as emphasized by the differently colored large circles, which also range from dark red at lowest elevations to dark blue at higher elevations, except the site on the western slope, which is colored in navy blue. B] Vegetation cover at sites 2–6 obtained from the two replicates of a 100-m-long linear transect at each site. Satellite imagery of the park were taken from Microsoft® Bing™ Maps.

Fig 2. Sampled populations of Mantidactylus bellyi and their genetic cluster assignment.

A] A male specimen of Mantidactylus bellyi and a map of Montagne d’Ambre, showing the a priori grouping into six sites based on geographical proximity as emphasized by large circles. Each dot represents a specimen of M. bellyi collected at a specific locality. The dots were colored according to the assignment of the individual to a genetic subpopulation cluster as in the K = 3 STRUCTURE models. B–E] Individual assignments of 236 individuals of M. bellyi to genetic clusters as inferred by STRUCTURE from a data set of seven microsatellites. The clustering scenarios of subpopulations K = 2–3 were depicted for the B] Admixture model (K = 3), C] No admixture model (K = 3), D] Admixture model with LOCPRIOR information (K = 3) and E] No admixture model with LOCPRIOR information (K = 2). Satellite imagery of the park were taken from Microsoft® Bing™ Maps.

Fig 3. Non-metric Multidimensional Scaling (NMDS) results depicting sample points and the convex hulls for ecological datasets.

Three datasets were used: variables of breeding streams of Mantidactylus bellyi and forest areas surrounding them analyzed together in the first row, separate analyses of the breeding streams variables in the second row, and separate analyses of the forest variables in the third row. Dots represent ecological sample points. They were colored according to their assignment to one of the six sites in the first column, to a genetic subpopulation cluster as in the K = 3 STRUCTURE models in the second column, to a genetic subpopulation cluster as in the K = 2 STRUCTURE model (No admixture/LOCPRIOR) in the third column.

Fig 4. Analyses of call parameters of Mantidactylus bellyi.

A] Principal Component Analysis (PCA) of the temporal and spectral call parameters, corrected for body temperature (T_b) for temporal call characters and body size (SVL) for dominant call frequency by regression. Convex hulls were drawn for call scores of specimens belonging to different sites on the first two principal components. Each dot symbolizes average values of multiple calls per individual. The colors correspond to those in Fig 2A. B] Variation of the residuals of dominant call frequency regressed against body size of calling Mantidactylus bellyi between different study sites in Montagne d’Ambre National Park.

Fig 5. Multivariate analysis of the morphology of Mantidactylus bellyi.

A] and B] show the variation of the residuals of morphological variables A] in males and B] in females between the six sites, corrected for the effect of size by regression, except SVL. Each small point symbolizes an individual’s sampling, with bigger dots representing the group mean point. Ellipses correspond to the 95% confidence interval. The colors refer to those in Fig 2A. We used the residual values of the variables for the analysis, except for SVL. The figure shows a detachment of individuals from sites 4 and 5 from the rest of the individuals in either case. SVL = snout–vent length, HL = head length, HW = maximum head width, ED = horizontal eye diameter, TD = tympanum diameter, LHU = humerus length, FOL = forearm length, THL = thigh length, TL = tibia length, TRL = tarsus length, TOE_L = length of the third toe.