The geographical context of speciation is related to levels of gene flow between populations during divergence (Mayr 1963; Gavrilets 2004; Bolnick & Fitzpatrick 2007). At one extreme, gene flow is totally absent during ‘allopatric speciation’. At the other extreme, there are no physical barriers to gene flow at all. This represents ‘sympatric speciation’, which is an extreme case of speciation with gene flow. Intermediate scenarios with partial, but incomplete, physical restrictions to gene flow are also likely (e.g. parapatric speciation). Speciation in the face of gene flow is generally thought to be difficult, because gene flow constrains population differentiation, thereby preventing the evolution of strong reproductive isolation (Mayr 1963; Coyne & Orr 2004). Although theoretical models have now demonstrated that speciation with gene flow is feasible under particular conditions (Gavrilets 2004; Bolnick & Fitzpatrick 2007), empirically demonstrating that speciation occurred in the face of gene flow is difficult. Nonetheless, a number of convincing examples of speciation with gene flow have now emerged, owing to methodological advances and creative approaches to studying the phenonemon (Hey 2006 for review).
An increasingly common test for divergence with gene flow involves using coalescent-based analyses to isolate the roles of time vs. gene flow on levels of genetic divergence, in an ‘isolation with migration’ model (the IM approach; Hey 2006). A study of speciation in salamanders, published in this issue of Molecular Ecology (Niemiller et al. 2008), used the IM approach to test whether divergence between spring and cave salamanders occurred with or without gene flow (Fig. 1). In such salamanders, two main speciation scenarios have been advanced: a ‘climate-relict’ hypothesis which invokes allopatric speciation after cold-adapted species become stranded in caves during climate change, and an ‘adaptive-shift’ model in which parapatric speciation occurs in the face of gene flow, with divergent selection between cave and surface habitats driving differentiation. Niemiller et al. (2008) applied coalescent-based analyses to DNA sequence variation from two mitochondrial and one nuclear gene, and strongly rejected a model of divergence with zero gene flow, thereby supporting an ‘adaptive-shift’ scenario involving gene flow.
Other studies employing slightly different approaches suggest that the salamander study, and others employing the IM approach, are not extreme anomalies. For example, studies reporting highly discordant patterns in gene geneologies from different loci support divergence with gene flow (Wang et al. 1997; Dopman et al. 2005; Bull et al. 2006; Putnam et al. 2007), because greater geneological discordance is expected during divergence with gene flow (Hey 2006). Niemiller et al. (2008) also report such geneological discordance. Other studies have focused on the criteria for sympatric speciation proposed by Coyne & Orr (2004) in their book on speciation (e.g. Barluenga et al. 2006; Savolainen et al. 2006). Finally, other work has inferred divergence with gene flow using the heterogeneity of genetic divergence among loci within the genome (Emelianov et al. 2004), geographical patterns of divergence in allopatry vs. sympatry (Grant et al. 2005), site-specific genetic differentiation (Panova et al. 2006), or patterns of linkage disequilibrium within populations (Nosil et al. 2006). In short, methods for detecting gene flow are improving, and their creative use has increased the available examples of speciation with gene flow.
Given convincing cases where gene flow occurred at some point in the speciation process, three major questions emerge: the timing of gene flow relative to the initiation of speciation, the factors which constrain gene flow to faciliate divergence in the face of it, and the generality of the process.
First, a rarely addressed issue is the timing of gene flow. In particular, at what point in the speciation process did gene flow occur? At one extreme, gene flow may have been ongoing during the entire process. At the other extreme, gene flow ensued upon secondary contact, after a long period of allopatric divergence in which partial reproductive isolation accumulated unimpeded by gene flow. In the latter scenario, although there may be some gene flow upon secondary contact, speciation was essentially allopatric. Of course, intermediate scenerios are possible (Rundle & Nosil 2005). The study by Niemiller et al. (2008) advances our general understanding of speciation with gene flow by testing when, in this continuum of possible points in time, gene flow occurred. They extend an approach used to address the divergence of chimpanzee subspecies (Won & Hey 2005), and show that the timing of reconstructed migration (i.e. gene flow) events between salamander species was broadly distributed through time. This result is more consisent with continuous or recurrent gene flow during divergence than with secondary contact following a long period of allopatric divergence (Fig. 2).
Second, elicudating the factors that facilitate divergence with gene flow will increase our understanding of how the process occurs, and the conditions that favour it. For example, ecological divergence between populations is likely to be of key importance in faciliating speciation with gene flow. If so, a postive correlation between the degree of adaptive phenotypic population divergence and levels of neutral genetic differentiation, independent from geographical distance, is expected (Nosil et al. 2008), because such a pattern indicates that adaptive divergence restricts gene flow. Likewise, tests for the role of genetic factors, such as chromosomal inversions, in facilitating divergence with gene flow are warranted (Noor et al. 2001; Rieseberg 2001).
Third, as more tests for divergence with gene flow emerge, it will be interesting to assess the generality of the process. The existing evidence, bolstered by the Niemiller et al. (2008) study, indicates that divergence with gene flow certainly can occur. But how common and taxonomically widespread is the process? The use of standardized methods (e.g. the IM approach) indicates that meta-analytic techniques can one day be used to address questions concerning generality. Collectively, these avenues of future research will likely increase our understanding of the process of speciation.