Multiple bursts of speciation in Madagascar's endangered lemurs

Abstract

Lemurs are often cited as an example of adaptive radiation, as more than 100 extant species have evolved and filled ecological niches on Madagascar. However, recent work suggests that lemurs lack a hallmark of other adaptive radiations: explosive speciation rates that decline over time. Thus, characterizing the tempo and mode of evolution in lemurs can reveal alternative ways that hyperdiverse clades arise over time, which might differ from traditional models. We explore lemur evolution using a phylogenomic dataset with broad taxonomic sampling that includes the lorisiforms of Asia and continental Africa. Our analyses reveal multiple bursts of diversification (without subsequent declines) that explain much of today's lemur diversity. We also find higher rates of speciation in Madagascar's lemurs compared to lorisiforms, and we demonstrate that the lemur clades with high diversification rates also have high rates of genomic introgression. This suggests that hybridization in these primates is not an evolutionary dead-end, but potential fuel for diversification. Considering the conservation crisis affecting strepsirrhine primates, with approximately 95% of species threatened with extinction, this study offers a perspective for explaining Madagascar's primate diversity and reveals patterns of speciation, extinction, and gene flow that will help inform future conservation decisions.

Fig. 1. A species tree of strepsirrhine primates estimated using SVDQuartets.

Node support values were estimated using 1000 bootstrap replicates. Nodes with > 95% bootstrap support are not labeled, while nodes with 75-95% support and <75% support are indicated by gray and white circles, respectively. The actual bootstrap support values for all nodes are provided in Supplementary Fig. 2. Branch lengths are not scaled. Infraordinal names are shown in black bars, and family names are shown in different colors matching the silhouette image of a representative species to the right of the tree. Silhouettes were obtained from PhyloPic.org and are public domain. The numbers next to each silhouette indicate the number of species sampled from each family as a fraction of the total number of described species in the family. Inset map shows the combined distributions of all lorisiform and lemur species in red and blue, respectively. Distribution maps were obtained from the IUCN Red List spatial database.

Fig. 2. Variation in diversification rate (DR) over time and across strepsirrhine phylogeny.

A A time-calibrated phylogeny of Strepsirrhini, with branches colored according to DR (species per million years) estimated using a three-rate model in MiSSE. Full results of this analysis can be found in Supplementary Fig. 6. Dashed vertical lines distinguish geological epochs (Pal. Paleocene, Pl. Pliocene, Q. Quaternary). Fully detailed time-calibrated phylogenies showing tip labels, node confidence intervals, and outgroups are shown in Supplementary Fig. 4. Three genera with particularly high speciation rates (Eulemur, Lepilemur, and Microcebus) are indicated on the tree and are discussed in the text. B Lineages-through-time plots of lemurs (blue) and lorisiforms (red). Each line represents a distinct tree generated using six different fossil calibration sets and accounting for incomplete taxonomic sampling, which was done by stochastically adding missing taxa to the proper genus 1000 times. The complete set of 6000 trees is available on Figshare (10.6084/m9.figshare.28699742). Note that the y-axis representing the number of lineages is log-transformed. C Variation in DR between lorisiforms and lemurs is visualized as the distribution of tip DR (λ_DR) values for all species within each taxon. Darker colored distributions represent the empirical λ_DR values across the set of 6000 stochastically resolved trees, while light gray distributions represent simulated λ_DR values, in which trees of the same species richness as each Family were simulated using a rate-constant birth-death model. The empirical and simulated values shown in these plots were calculated using a custom script (TipDr_Calculation.R; see Code Availability statement). D MiSSE results, identical to panel A, but estimated using a two-rate model. See Supplementary Fig. 7 for the full visualization of this analysis, and Supplementary Data 6 for MiSSE model testing results. E Results of our CRABS analysis which estimated ten different diversification models in the congruence class (the set of speciation and extinction functions with equal likelihoods) under three different extinction treatments (left, models where extinction increased over time, center, models where extinction decreased over time, and right, models where extinction fluctuated randomly over time). Each line represents a different model in the congruence class. Across all analyses, an increase in net diversification rate was detected ~5-6 mya (indicated by a yellow vertical bar). Full details of this analysis are provided in Supplementary Fig. 8.

Fig. 3. Discordance between nuclear (left) and mitochondrial (right) phylogenies of strepsirrhine genera.

Both phylogenies were estimated using IQTree with all available individuals, and then species-level branches were manually collapsed to visualize one branch per genus. Fully detailed phylogenies from these analyses are available in Supplementary Figs. 9 and 13. The nuclear and mitochondrial trees are discordant in three locations labeled with gray numbered circles. Gray arrows on the nuclear phylogeny indicate three locations, where gene flow was inferred using QuIBL, with the proportions of introgressed loci labeled. Full details from QuIBL are found in Supplementary Fig. 10. Families are colored to match Fig. 1.

Fig. 4. Historical introgression events estimated in five lemur clades.

Phylogenies with reticulate relationships estimated by PhyloNet for five strepsirrhine genera (A–E) that had poorly resolved nodes (bootstrap support values < 75%) in our species tree analyses (Fig. 1). Arrows indicate the reticulation events (H) and are labeled with the estimated inheritance probabilities. Note that several models received similar support (Supplementary Fig. 11; source data in Supplementary Data 13), and here we show the models with the lowest H among the well-supported models. Colors and silhouette images for each genus match the Family-level formatting from Fig. 1. *Note that a recent paper proposed synonymizing M. mittermeieri and M. lehilahytsara as a single species; however, we treated these as distinct species as our data did not support a sister relationship.