A decade of amphibian population genetic studies: synthesis and recommendations
- First Online:
- Cite this article as:
- Emel, S.L. & Storfer, A. Conserv Genet (2012) 13: 1685. doi:10.1007/s10592-012-0407-1
- 857 Views
Amphibians are declining globally, and a comprehensive understanding of the spatial distribution of genetic diversity will inform conservation efforts. However, studies that estimate amphibian population genetic structure and gene flow have not yet been synthesized. Our search of literature from 2001 to 2010 yielded 552 amphibian population and landscape genetic studies, of which 139 explicitly estimated gene flow or genetic structure. We examined these works for general trends and conducted a meta-analysis of reported FST values. The majority of studies took place in temperate forests in North America and Europe, with no studies of caecilians and few studies of direct-developing species. Among landscape genetic studies, rivers, roads, and mountain ridges were the predominant barriers identified. Conservation status was the only factor that showed a significant relationship with FST, with the least concern IUCN status differing significantly from the near threatened (NT) status as well as from any combination of IUCN statuses that included NT. Recent technological advances will help researchers fill taxonomic and geographic research gaps, thereby facilitating management plans that address a greater diversity of amphibian species.
KeywordsAmphibianGene flowPopulation geneticsLandscape geneticsConservation
In the face of global amphibian declines, it has become increasingly critical to identify trends in amphibian population genetic structure to inform conservation planning. An estimated 32 % of amphibian species are currently threatened with extinction, with over 42 % showing population declines (IUCN, Conservation International, NatureServe 2011). The primary factors implicated in these declines include habitat alteration, overharvest, invasive species, climate change and disease (Collins and Storfer 2003; Stuart et al. 2004). Gene flow maintains population connectivity and evolutionary potential, allowing future response to environmental change. Fortunately, the tools and analyses for assessing gene flow and population genetic structure, such as highly variable molecular markers, are increasingly available (Ouberg et al. 1999; Vignal et al. 2002; Morin et al. 2004; Guichoux et al. 2011). Larger datasets and more statistically powerful analyses are now commonplace, resulting in identification of highly diverged or even “cryptic” populations, estimation of effective population sizes, and testing for bottlenecks (Allendorf and Luikart 2007). Meanwhile, landscape genetic analyses can identify barriers to gene flow, landscape factors that facilitate gene flow, and determine the necessary size of corridors between habitat patches to maintain genetic connectivity (Segelbacher et al. 2010). Although the feasibility of amphibian population genetics studies has dramatically increased, syntheses of such studies are rare.
In this review, we surveyed the peer-reviewed literature for studies of amphibians that examined gene flow from 2001 to 2010. The goals of this study were to: 1) determine overall trends in amphibian genetic studies, 2) determine whether there are research gaps in groups of species, geographic areas covered or particular study questions, 3) perform a meta-analysis of available FST estimates to determine whether there are particular factors that correlated with the level of genetic structure, 4) summarize results of landscape genetics studies, and 5) suggest avenues for future study.
We conducted a literature search for studies published between 2001 and 2010 using the ‘Web of Science’ (http://www.isiknowledge.com) with the keywords (found in title, keywords, or abstract): [amphibian* OR (salamander* OR caudata*) OR (frog* OR anura*)] AND [population genetic* OR conservation genetic* OR landscape genetic* OR gene flow* OR genetic variation OR genetic structure]. A separate search for (caecilian* OR gymnophiona) yielded zero applicable results.
The search yielded 552 studies. These articles were then screened manually; those that did not examine amphibians or estimate gene flow were excluded. The final dataset consisted of 139 studies, for which we collected title, year, authors, journal, species studied, geographic range, type of larval/adult habitat, and type of molecular marker(s) used (dataset available upon request). We also noted whether explicit hypotheses were tested and summarized the main question and conclusions of each study. Additionally, studies were categorized based on their broad focus: landscape genetic studies, which quantify the relationship between spatial landscape variables and genetic structure (Storfer et al. 2007); population/conservation genetic studies, which quantify genetic structure but not in a spatial context; or both, if both a spatial and conservation focus exists. For landscape genetic studies, we recorded the analyses used, landscape features tested, and the results of those analyses.
Meta-analysis of FST
A subset of 68 studies reported mean FST with standard error or provided data from which the mean and standard error were calculated. Because weighting mean FST per study by the standard error or the global mean plus the standard error did not affect the results, the unweighted mean of FST/(1 − FST) was used to standardize values among studies (Rousset 1997).
Species were categorized into either Order Anura (frog) or Caudata (salamander). Type of larval/adult habitat was determined from the study or through the AmphibiaWeb (2011) species account. International Union for Conservation of Nature and Natural Resources (IUCN) status (least concern (LC), vulnerable (VU), near threatened (NT), endangered (EN), critically endangered (CR)) was also determined (IUCN, 2010). Because some studies reported FST for two species, multiple molecular marker types, or study locations, these 68 studies yielded 75 total data points (63 LC, 5 NT, 6 VU, 1 EN, 1 CR). We created univariate linear mixed-effects models using the program SAS 9.3 (SAS Institute 2011) with taxon, habitat type, and IUCN status as the fixed effects, species as a random effect, and FST/(1 − FST) as the response variable. Due to the small sample sizes in all categories other than LC, we also performed the analyses by combining the category VU with NT and EN with CR, as well as by combining all categories of some concern (i.e., all except LC). Because of the potential for the populations studied to be of greater concern than is reflected by the IUCN status of the species as a whole, the above analyses were run both for all studies in the meta-analysis and for the subset of studies that did not indicate that the study populations were endangered due to habitat fragmentation, existing at the edge of the species’ distribution, or through recent genetic bottlenecks. In the case of a significant fixed effect in the main model, a Tukey–Kramer test was performed to assess significance of treatment levels.
The molecular markers used included microsatellites, mtDNA, allozymes, RAPDs, AFLPs and nDNA (Fig. 1). Twenty-three studies used more than one marker types and are included twice. Microsatellites were the most popular marker, used in 80 studies. Forty studies included analyses of mtDNA, while the remaining markers were used less frequently.
Most studies examined either North American species (73) or European (44) species. South America, Africa, Asia, and Australia were represented by a small number of studies (17).
Fewer than half of the studies surveyed tested an explicit hypothesis. Typical hypothesis-free foci included: characterization of population genetic structure, definition of conservation units, or identification of landscape genetic associations.
While the number of studies per year increased over the decade, the proportion of landscape genetic studies also increased. Overall, there were 46 studies with a landscape genetic focus; 14 sought to identify barriers to gene flow, primarily testing roads, rivers, and mountain ridges, and 10 of which found support for their hypothesized barrier. Additionally, 11 of 14 studies found a significant effect of fragmentation on genetic connectivity due to generally unsuitable habitat, roads, and/or railways. The remaining 93 largely aimed to identify levels of gene flow, characterize genetic structure, and inform management practices. The hypotheses most frequently tested in this group included whether bottlenecks, hybridization, and/or range contraction or expansion has occurred.
Meta-analysis of FST
Based on the mixed-effects models, only IUCN status had a significant effect on FST for all combinations of IUCN levels when all studies were examined and for two out of three combinations of IUCN statuses when the subset of study populations not indicated to be of greater concern than the species were examined (online resource 1). For these significant main effects, subsequent Tukey–Kramer tests revealed that the LC species differed significantly from NT as well as from any combination of IUCN statuses that included NT, with LC having lower parameter estimate for FST in each case. There was no significant effect of taxonomic group or type of larval/adult habitat on FST.
Four main points arose from our review of amphibian population genetics studies published over the past decade: 1) tropical species, especially caecilians, are understudied, 2) there is a general lack of population genetic information on direct-developing species, 3) the majority of studies do not test explicit hypotheses, and 4) roads, rivers, and mountain ridges are the most frequent barriers to gene flow. We identified a significant effect of conservation status on FST but no significant effect of either order or larval/adult habitat.
The great majority of studies surveyed took place in temperate forests of North America and Europe, despite the consensus that amphibian biodiversity and the number of threatened species is highest in the tropics (IUCN, Conservation International, NatureServe 2011). Thus, the high amphibian diversity of tropical forests is greatly understudied, and management plans devised for temperate species may not be appropriate for conservation of tropical species. Caecilians were not represented at all, likely due to their aquatic or subterranean habits that make caecilians difficult to study.
Population genetics of direct-developing species is also not well understood, with only 18 studies focusing on direct-developing species. Biphasic amphibians that commonly breed aggregately in ponds or streams presumably differ in rates of gene flow relative to direct-developing species, which are strictly terrestrial and do not breed aggregately. However, we found no quantitative or qualitative difference between the two groups, potentially due to the small sample size of studies of direct-developers. Nonetheless, the substantial difference in life history between direct-developing and metamorphosing amphibians warrants further study.
Meta-analysis of FST
The meta-analysis of FST yielded significant results for the relationship between FST/(1 − FST) and IUCN status, with LC consistently differing from NT and the combinations of IUCN statuses that included NT. We expected a positive correlation between level of conservation concern and FST values, due to the effects of increased habitat fragmentation on population connectivity. While these analyses failed to find such an effect, our statistical power was likely limited by low sample sizes among treatments (63 FST values from species of LC, while all other IUCN statuses had five or fewer FST values, likely due to the difficulty studying vulnerable or endangered species). Thus, the significant differences in FST between species of little to no conservation concern (LC) and those of potential future concern (NT) may be indicative of a trend that initial declines in species are often characterized by loss of populations in fragmented areas, and may signal a major effect from transitioning from LC to NT. Nonetheless, given the low sample sizes of each of the categories of concern, other trends may emerge as sample sizes increase in other IUCN categories of concern.
One important caveat is that various criticisms have been raised about the validity of FST for estimating gene flow (e.g., Whitlock and McCauley 1999; Marko and Hart 2011), and some have suggested a preference for coalescent methods that do not share some of the limiting assumptions inherent in the infinite island model used to estimate FST. However, few coalescent-based amphibian gene flow studies exist; our hope is to re-analyze amphibian gene flow studies in the future and include new gene flow estimates as they are increasingly used.
The focus on characterizing levels of gene flow and genetic diversity in amphibians has clearly increased over the past decade due to increased availability of molecular markers and development of new analytical methods. Despite these advances fewer than half of the studies surveyed tested explicit hypotheses. Landscape genetics has emerged as a field that tends to be more hypothesis-driven, with the goal to test the influence of landscape features on gene flow in an explicitly spatial context (Segelbacher et al. 2010; Storfer et al. 2007, 2010). In the 46 studies surveyed with a landscape genetic focus, roads, rivers, and mountain ridges were the landscape variables most often tested and found to be barriers to gene flow. Hopefully, future studies will test the influence of additional landscape features on amphibian population genetic structure, and researchers will begin to fill the taxonomic and geographic gaps that currently exist to better inform amphibian conservation and management.
The Washington State University Libraries provided access to Web of Science. We would like to thank David Crowder and Jesse Brunner for statistical assistance with the meta-analysis. We also thank Daryl Trumbo and Steven Micheletti for comments on the manuscript.