This past summer, two groups of authors published reports on the phylogeography of the only anole native to the continental United States: Anolis carolinensis. Each report sought to characterize genetic diversity across this species’ range by identifying genetically distinct populations, inferring historical demographic events, estimating the absolute timing of diversification events, and testing the hypothesized impact of riverine barriers and Pleistocene glaciation on geographic differentiation.
Because these two reports effectively appeared simultaneously (Tollis et al.’s report appeared on June 7th in PLoS ONE and Campbell-Staton et al.’s report was accepted for publication on June 18th at Ecology and Evolution), and do not cite or discuss one another’s work, I thought it would be worth writing a post that compares and contrasts their results and conclusions. I’m going to focus in particular on three specific results reported by both groups of authors: (1) diagnosis of geographically and genetically distinct populations, (2) inference of historical demographic processes within populations, and (3) estimates for the timing of A. carolinensis diversification. While the two studies largely agree on the first two results, they appear to disagree somewhat on third.
1. Diagnosis of geographically and genetically distinct populations
Tollis et al. and Campbell-Staton et al. identify similar boundaries between geographically and genetically distinct populations. This shouldn’t be too surprising given that both studies rely extensively on range-wide sampling of the same molecular marker: DNA sequences of the mitochondrial encoded ND2 gene. Campbell-Staton et al. report 171 new ND2 sequences from 42 localities while Tollis et al. report 190 new ND2 sequences from 49 localities (average of ~4 individuals per locality in each study). Both studies combine these sequences with some previously published sequences, but only Campbell-Staton et al.’s combined dataset includes all of the other ND2 haplotypes previously sequenced from across the ranges of A. carolinensis‘ close relatives in the northern Caribbean.
Tollis et al. recovered four major mitochondrial phylogroups, including an Everglades clade from extreme southern Florida (purple in figure above), a Suwannee clade from the crook of Florida’s panhandle (blue), a widespread Gulf Coast clade found from the Atlantic coast of Georgia and South Carolina across into Texas and southern Louisiana (green), and a North Carolina clade (yellow). This last clade was rather curious because it included not only haplotypes sampled from North Carolina, but also a single haplotype sampled from the eastern coast of Florida (yellow arrow in figure above). As noted by Tollis, this result initially seemed like it might be an anomaly: “The sole anole we collected near the central Atlantic coast of FL may represent a poorly sampled mitochondrial clade endemic to that region or, more simply, an introgressed individual with a mitochondrial haplotype derived via human-mediated dispersal. More sampling is needed to address this issue.” In other words, with just a single haplotype from a single individual, Tollis et al. couldn’t rule out the possibility that a fourth grader brought a pet anole home with her to Florida from vacation in North Carolina, and that Tollis and colleagues later sampled her escaped pet or one of its offspring for their study.
Fortunately, Campbell-Staton et al.’s dataset confirms Tollis et al.’s main discoveries and somewhat clarifies the status of the eastern Florida populations. First, Campbell-Staton et al. recover four major mtDNA haplotype clades that closely correspond with the Everglades, Suwannee, Gulf Coast and North Carolina clades identified by Tollis et al. Second, Campbell-Staton et al. recover the same geographically discontinuous clade comprised of haplotypes sampled from both North Carolina and eastern Florida. Having sampled multiple haplotypes from multiple individuals in eastern Florida, Campbell-Staton are able to diagnose the haplotypes sampled from this region as a fifth major haplotype clade. Campbell-Staton’s work also extends the range of the Gulf Coast clade into Texas, as far south as Brownsville along the Mexican border.
More importantly, Tollis et al.’s somewhat more limited sampling of 10 nuclear loci confirms that the four main mtDNA haplotype clades they identified are not merely artifacts of mtDNA haplotype evolution by recovering four distinct genotypic clusters that closely correspond with the mtDNA haplotype clades. Tollis et al.’s nuclear results also provide a potential explanation for the somewhat unusual geographically disjunct distribution of closely related mtDNA haplotypes sampled from North Carolina and eastern Florida by suggesting that gene flow may be ongoing across the Atlantic coast. More of this type of data is clearly needed to understand the nature of geographic genetic differentiation in A. carolinensis. Obtaining markers shouldn’t be a problem given the availability of the A. carolinensis genome, but a simple AFLP study might a be quick, cheap, and easy first step toward getting a better handle on geographic genetic variation in the species.
2. Historical demography of Anolis carolinensis
Both studies assess historical demographic processes such as range expansions or bottlenecks using a number of standard methods, including estimates of nucleotide diversity, Tajima’s D, and Bayesian skyline plots. None of these methods are perfect and all suffer from a range of assumptions and caveats, but both studies consistently recover evidence supporting at least one seemingly important demographic event: recent range expansion in the widely distributed Gulf Coast clade. Both studies report low genetic diversity in this clade relative to the Florida clades. Both studies also recovered significantly negative values for Tajima’s D from this widespread clade, consistent with a recent expansion in the size of this population. After recovering a significant relationship between nucleotide diversity and longitude, Tollis et al., suggest a westward expansion in this clade rather than the type of south to north expansion that might have been expected under post-glacial recolonization of more northern climes. Campbell-Staton et al. suggest the possibility that the North Carolina population is a relict population that persisted in a glacial refugium.
3. The age of Anolis carolinensis
Historically, some authors hypothesized that Anolis carolinensis diverged from its closest relatives on Cuba species sometime in the Pleistocene, but this was largely just a guess made in the absence of fossil data. A previous phylogenetic study of mtDNA haplotypes from A. carolinensis and its close relatives rejected the Pleistocene origination hypothesis in favor of pre-Pleistocene divergence. In that study, we used simple estimates of uncorrected genetic pairwise distances between A. carolinensis and related taxa and a molecular clock calibration for the ND2-region sequence evolution of 1.3% pairwise divergence/million years. Anolis carolinensis differs from all of its relatives by uncorrected distances exceeding 9.6%, suggesting that it diverged from its closest relative at least 7 MYA. Because this study examined uncorrected distances, the estimated ages of diversification events were likely underestimates of the true age of diversification events, but this simple approach was considered appropriate in that study given that the primary interest was in rejecting the Pleistocene origination hypothesis (using uncorrected distances is almost certainly a conservative approach to testing this hypothesis).
Neither Tollis et al. or Campbell-Staton et al. have used particularly novel or cutting edge approaches to further our understanding of the timing of diversification in A. carolinensis and related taxa, and neither group does a particularly good job explaining exactly how they implemented their molecular clock analyses. Both groups estimate ages for divergence events in A. carolinensis by applying relaxed molecular clock methods that rely on the widely used external calibration for the ND2 region reported by Macey et al. for Asian Laudakia.
Campbell-Staton et al. say that they used a relaxed molecular clock model after setting the rate of molecular evolution to 1.3% pairwise divergence per millions years. This is a fairly straightforward approach, but Campbell-Staton don’t provide much detail and only explicitly tell us that this analysis was done in BEAST in the legend of Figure 1. In any case, Campbell-Staton et al. estimate a split between A. carolinensis and A. porcatus at ~6.8-17.8 MYA and diversification within A. carolinensis beginning ~6.8-12.6 MYA.
Tollis et al. seem to have used an even more complicated approach that I had a harder time disentangling. They seem to have used the Macey et al. calibration to estimate ages for the four main mtDNA clades in their study, possibly using simple pairwise distance estimates. Tollis et al. subsequently used these age estimates to calibrate their relaxed clock analysis in BEAST. They also mention calibrating the split between A. carolinensis and A. porcatus to 6 MYA based on the results in Glor et al. 2005, but I’m not sure how they came up with this number from that paper. Tollis et al. ultimately infer a much younger age for diversification in A. carolinensis than previous studies when they suggest that diversification in this species began around 2 MYA. I think I’d have to reanalyze their dataset to understand why they are coming up with this relatively low age estimate, but my first impression is that their approach is somewhat less direct than the approach used by Campbell-Staton et al.
Overall, its great to finally have some nice phylogeographic data on one of the most widely known of all anole species. Both Campbell-Staton and Tollis et al. have provided a strong foundation for subsequent work on the species, and I hope we see more such work in the near future.