RADseq data reveal a lack of admixture in a mouse lemur contact zone contrary to previous microsatellite results

Abstract

Microsatellites have been a workhorse of evolutionary genetic studies for decades and are still commonly in use for estimating signatures of genetic diversity at the population and species level across a multitude of taxa. Yet, the very high mutation rate of these loci is a double-edged sword, conferring great sensitivity at shallow levels of analysis (e.g. paternity analysis) but yielding considerable uncertainty for deeper evolutionary comparisons. For the present study, we used reduced representation genome-wide data (restriction site-associated DNA sequencing (RADseq)) to test for patterns of interspecific hybridization previously characterized using microsatellite data in a contact zone between two closely related mouse lemur species in Madagascar (Microcebus murinus and Microcebus griseorufus). We revisit this system by examining populations in, near, and far from the contact zone, including many of the same individuals that had previously been identified as hybrids with microsatellite data. Surprisingly, we find no evidence for admixed nuclear ancestry. Instead, re-analyses of microsatellite data and simulations suggest that previously inferred hybrids were false positives and that the program NewHybrids can be particularly sensitive to erroneously inferring hybrid ancestry. Combined with results from coalescent-based analyses and evidence for local syntopic co-occurrence, we conclude that the two mouse lemur species are in fact completely reproductively isolated, thus providing a new understanding of the evolutionary rate whereby reproductive isolation can be achieved in a primate.

Keywords: Madagascar; RADseq; admixture; cryptic species; microsatellites; mouse lemurs.

Figure 1.

Distributions and sampling sites of murinus and griseorufus in southern Madagascar. The distribution of murinus is shown in purple and that of griseorufus in gold. A population in southeastern Madagascar was recently split from murinus as M. ganzhorni, but is here included within murinus s.l. The range of M. murinus extends to the north of the area shown in the map, whereas the entire distribution of M. griseorufus is shown. Inset: overview of sampling in the contact zone area (corresponding to the study site of [10]), showing two parapatric (Hazofotsy with griseorufus and Ambatoaba with murinus) and two sympatric (Mangatsiaka and Tsimelahy) sites. Microcebus illustrations courtesy of Stephen Nash. (Online version in colour.)

Figure 2.

No evidence for hybridization in the contact zone. Nuclear RADseq data from the contact zone area was used for all analyses, including 12 individuals that had been identified as admixed in a previous microsatellite study (dark grey in panels a and b). (a) Admixture results. Top: a cross-validation error plot identifies K = 2 as the optimal number of clusters. Bottom: ancestry components for each individual for K = 2 reveal a lack of admixture: all individuals were inferred to have 100% ancestry from only a single species. Individuals were previously characterized using mitochondrial DNA (mtDNA) (bottom bars) and microsatellites (labels at top). (b) A PCA analysis reveals two clusters that are well-separated along PC1, corresponding to griseorufus and murinus, with no individuals that are intermediate along this axis. (c) Map showing spatial distribution of murinus and griseorufus individuals at the two contact sites. (Online version in colour.)

Figure 3.

Re-analysis of microsatellite data and analysis of simulated individuals. (a) Re-analysis of microsatellite data with NewHybrids (NH; top row) and Structure (STR; bottom row). Among the 12 individuals previously identified as hybrids (green background bars), NewHybrids now identifies only a single individual as a hybrid (black dot), with several further griseorufus individuals showing non-significant signs of admixed ancestry (yellow ancestry). (b) Analysis of simulated individuals. Dots indicate detected hybrids. Using SNPs (bottom two rows), both NewHybrids and Structure correctly inferred ancestry for all individuals. Using microsatellites (top two rows), NewHybrids was prone to falsely inferring hybrids (4 out of 40 unadmixed individuals), and false negatives occurred both with NewHybrids (2 out of 20) and Structure (6 out of 20). (Online version in colour.)

Figure 4.

Demographic inferences using G-PhoCS and BPP. (a–c) Summary of results for G-PhoCS models without (a) and with (b) gene flow and for BPP (c; with gene flow). Each box represents an extant (bright colours: gold for griseorufus, purple for murinus) or ancestral (faded colours) lineage, with box width indicating N_e and box height indicating time. Gene flow was estimated reciprocally between three pairs of lineages, as depicted by the arrows (using the same units as panels d and e). (d) Point estimates and 95% highest posterior densities (HPDs) of BPP introgression probabilities (phi). (e) Point estimates and 95% HPDs of G-PhoCS population migration rates (2Nm). (Online version in colour.)