The Opinion article of Wang (2010) focuses on the definition and application of phylogeography and landscape genetics. Because different genetic metrics and molecular markers integrate over different lengths of time, it is important that researchers interpret their results in terms of the correct temporal scale. Wang advocates restriction of the term ‘landscape genetics’ to specific types of genetic data, analyses and temporal scales and states that the ‘gap between phylogeography and landscape genetics cannot be thoroughly investigated with current methods.’ He twice cites one of our studies (Vandergast et al. 2007) as being confused on these points. In this response, we clarify some technical issues from his overview of the field, and elaborate on the approaches taken in our study.
Wang (2010) defines landscape genetics as being focused on contemporary processes, clearly distinct from the study of historic processes in the field of phylogeography. Within this framework, he has several critiques of intraspecific genetic studies. He first concerns the types of markers that are used to study contemporary processes. After acknowledging that population genetic structure in neutral markers is the product of mutation, drift and gene flow, Wang cautions that chloroplast DNA and mitochondrial DNA (cp/mtDNA) evolve ‘too slowly to be useful for inferring most recent and ongoing microevolutionary processes. Although some noncoding cp/mtDNA regions evolve at relatively faster rates than coding regions, these still experience considerably lower substitution rates than typical microsatellite loci.’ The substitution rate is the rate at which new mutations become fixed, which is simply equal to the mutation rate. Thus, Wang implicitly focuses on analytical methods that require populations to be fixed for unique mutations. His concerns about substitution (or mutation) rate for organellular sequence data would also apply to nuclear sequence data.
Both cp/mtDNA and nDNA sequence data are analysed with a variety of methods. The fixation of new mutations in isolated populations is almost never a criterion of interest in the studies of contemporary processes, regardless of the marker used. More commonly, analyses focus on allele frequency differences and the genealogical relationships (evolutionary distances) between alleles. Mutation, drift and population-specific patterns of gene flow collectively influence allele frequencies over a range of temporal scales that always begin with the most recent generation. In fact, cp/mtDNA markers may be better suited for some studies of contemporary evolutionary processes than nuclear markers. If sufficiently variable sequence data are available, drift will be faster in organellular genomes because of their lower effective population size. It is possible that Wang’s critique is very narrowly aimed at the interpretation of traditional phylogeographic methods (such as labelling gene trees with geographic locations), and if so, his point is valid.
Wang’s second concern is that at least some methods for analysing intraspecific genetic data are being misinterpreted in terms of the wrong temporal scale. This concern has traditionally been expressed in terms of the amount of time a particular metric or marker requires to reach equilibrium. The recognition and study of nonequilibrium conditions has been a prominent part of the literature over the past three decades (e.g., Slatkin 1985; Wade & McCauley 1988; Boileau et al. 1992; Bossart & Pashley Prowell 1998; Bohonak & Roderick 2001; Excoffier et al. 2009; Allen et al. 2010). It can be summarized briefly as follows:
- 1 Genealogy-based analyses that require new, unique mutations to reach detectable levels or even fixation in local populations must necessarily focus on the oldest evolutionary processes in the gene genealogy. Thus, phylogeographic analyses (Avise et al. 1987) often aim to infer historic isolating events from the accumulation of fixed differences among lineages over time. Similarly, cladistic gene flow estimates (Slatkin & Maddison 1989) and coalescent-based models (e.g., Beerli & Felsenstein 1999; Hey & Nielsen 2004) implicitly assume that parameters such as gene flow and drift have been constant across the entire gene genealogy. Thus, contemporary changes in gene flow rates or patterns are not readily inferred from these methods.
- 2 Analyses that use frequency-based similarity or distance measures (amova, Excoffier et al. 1992; e.g., IBD, Slatkin 1993) may reach drift–gene flow equilibrium on shorter timescales than coalescent-based analyses, typically requiring at least tens and possibly thousands of generation. The specific amount of time required depends on whether gene flow (or effective population size) has increased or decreased and the magnitude of the change. Frequency-based genetic distance measures can provide information even prior to a drift–gene flow equilibrium when comparative approaches are employed (e.g., Bohonak & Roderick 2001; Keyghobadi 2007 and references therein). For example, the levels of genetic differentiation can be compared in recently fragmented and contiguous habitats.
- 3 Clustering algorithms commonly applied to microsatellite data (e.g., structure, Pritchard et al. 2000; baps, Corander et al. 2003) are used to define contemporary gene pool boundaries. These analyses utilize linkage disequilibrium across loci, which is statistically detectable for only a few generations after a unique genotype immigrates.
Clearly, these three categories of analysis overlap in terms of the time periods for which they provide useful information. Nonetheless, Wang (2010) states that the temporal gradient between phylogeographic studies and landscape genetics ‘is an oversimplification’ and that there is ‘a gap between phylogeography and landscape genetics which cannot be thoroughly investigated with current methods.’ In contrast, we believe that the sharp dichotomy is an oversimplification, because many analyses bridge that gap by separating historical from current population processes. Perhaps Wang is most concerned about the use of terminology: ‘studies should adhere to the definitions of phylogeography and landscape genetics… particularly in regard to their explicit temporal distinctions, and employ the correct terminology so that readers clearly understand the validity of the inferences made in those studies.’ Although we agree that population geneticists must understand the temporal bounds of their analyses, we disagree as to how often published studies make this mistake or are in turn misinterpreted by other researchers in the field.
Our interpretation of Wang’s opinion piece is that he wishes to restrict the term ‘phylogeography’ to describe DNA sequence data analysed with genealogical approaches (category 1 above) and the term ‘landscape genetics’ to microsatellite data analysed with allele frequency-based statistics or clustering algorithms (categories 2–3). Coalescent models are mentioned in his review, but their placement into either phylogenetics or landscape genetics is ambiguous. Wang also marginalizes the dominant role that frequency-based analyses of DNA sequences have played in developing our understanding of both recent and historic evolutionary processes since the 1980s. Although Wang cites Manel et al. (2003) as providing a model definition for landscape genetics that should not be ‘diluted’, he overlooks the fact that Manel et al.’s review discussed multiple studies that only used mtDNA. In contrast to Wang’s assertion, we believe that mtDNA data can be used to make accurate inferences about contemporary processes such as spatial patterns of gene flow. We illustrate this point using our recent study that he cites as an example of misapplication.
Vandergast et al. (2007) studied the populations of the flightless Jerusalem cricket Stenopelmatus ‘mahogani’ in southern California using mtDNA sequence data. The sampling sites were located throughout a landscape that included both large and small habitat fragments surrounding the Los Angeles basin. We analysed these data using four approaches:
- 1 As is common, the regional sampling locations were labelled on a Bayesian gene genealogy and on a parsimony network to highlight deep historical isolation. The genealogy was dated using a molecular clock. Regions were not reciprocally monophyletic, and individual sampling locations often shared alleles within a region.
- 2To separate recent population isolation from that which occurred prior to human influence, we generated a spatial (GIS-based) model of habitat fragmentation by marine inundation during the Holocene/Pleistocene. The effects of distance, ancient and current habitat fragmentation were analysed using partial Mantel tests and a novel regression approach that is described in the study. The results clearly demonstrate that both contemporary and historic landscape features influence current patterns of mtDNA differentiation. This approach has been advocated in several recent review articles in Molecular Ecology (e.g., Anderson et al. 2010).
- 3 Correlations between genetic diversity and contiguous habitat area were considered in terms of both prehistoric and current fragment size.
- 4 One large historic fragment in our study has been recently bisected by a highway. Genetic divergence across the highway is higher than for comparable distances on the same side of the highway. Computer simulations showed that increased mtDNA divergence across the highway matches theoretical expectations for complete cessation of gene flow on a timescale corresponding to the highway’s construction.
In summary, Vandergast et al. (2007) interpreted mtDNA diversity using multiple approaches that include both phylogeography and landscape genetics. Beyond our explanations of the timescale relevant for each analysis, we actually validated our interpretations with a dated gene genealogy, historic and contemporary GIS models and computer simulations. Although the study could be criticized for using only one genetic marker (a point with which we do not disagree), Wang’s primary concern is instead that the article’s title includes the term ‘landscape genetics’. His implication that we are confused and incorrectly interpreted our analyses seems unwarranted.
Wang (2010) cites several other studies in support of his general argument, but we could not find fault with their data interpretation either. Swart et al. (2009) and Measey & Tolley (in press) both used mtDNA data to study ancient evolutionary events in terms of landscape-level processes. Each listed the keyword phrase ‘landscape genetics’, and both interpreted their data in terms of the appropriate temporal scale. Wang also cited Koscinski et al. (2009) because their study relied on mtDNA but included the keyword phrase ‘landscape genetics’. Like Vandergast et al. (2007), Koscinski et al. (2009) analysed their data in terms of both historic processes and current landscape variables. We believe that ‘landscape genetics’ is not a misleading study descriptor in this case.
In summary, we welcome the additional attention that Wang’s review has brought to the appropriate timescales (time to equilibrium) for different types of analyses and molecular markers. However, we disagree strongly with the implication that contemporary processes cannot be analysed using DNA sequence data. We also suggest that it is not fruitful to restrict the term ‘landscape genetics’ to studies that only focus on contemporary processes. A major goal of landscape genetics is to test for the effects of recent habitat change, and this can be carried out by comparing current and past landscape features using approaches like those described in Vandergast et al. (2007), regardless of marker type.