New species can form under conditions of geographical overlap or separation [1
], and through the action of different evolutionary processes. Thus, geographical mode and mechanism of speciation are a focus of speciation research [3
]. Mayr and others [5
] argued that geographical isolation promotes speciation by reducing homogenizing gene flow, and indeed numerous examples of allopatric speciation exist [3
]. Additionally, the geographical arrangement of populations can have consequences beyond affecting gene flow. For example, increased geographical separation might expose populations to greater divergent selection [3
]. This could occur if populations that are farther apart are exposed to stronger divergent selection along environmental gradients or a greater diversity of selection pressures. In such instances, selection might cause ‘ecological speciation’ [8
]. In other instances, geographical contact between taxa might actually promote, rather than constrain, speciation. One possibility is ‘reinforcement’, in which selection against unfit hybrids in geographical regions of contact drives the evolution of premating isolation [12
]. Thus, a variety of speciation mechanisms are known and individual genes implicated in adaptation and speciation have been described [15
In contrast to the classical study of mode and mechanism of speciation, the emerging field of speciation genomics is in its infancy [19
]. The field has usefully revealed that divergence is highly heterogeneous across the genome, but is still in a phase where explicit and non-overlapping predictions associated with many hypotheses have yet to emerge [20
]. Thus, genomic data from populations that are variably isolated on ecological and spatial scales have the potential to advance our conceptual understanding of the interaction and relative importance of different processes in shaping genomic divergence. Our goal here is to take advantage of recent advances in DNA sequencing and an ecologically characterized system that allows replicated comparisons across populations to conduct an analysis of the genomic consequences of mode and mechanism of speciation.
Although non-overlapping predictions concerning the specific causes of patterns of genomic divergence are still being developed, a number of a priori
predictions can be made for the manner in which divergence will be affected by particular processes [19
]. For example, although many processes affect the distribution of genomic differentiation across loci, gene flow could do so in predictable ways. Speciation with gene flow, either primary or following secondary contact, might be characterized by divergence in only a few regions that harbour genes under strong divergent selection and those causing reproductive isolation [23
], while the rest of the genome is homogenized by gene flow. This is predicted to generate an ‘L-shaped’ frequency distribution of genetic differentiation across loci (i.e. most loci have low FST
values). In contrast, allopatric speciation might be characterized by divergence across more of the genome, leading to a different distribution of genomic differentiation than observed with gene flow [19
]. Additionally, processes other than gene flow might leave predictable genomic patterns. For example, gene regions involved in ecological speciation should be strongly differentiated between ecologically divergent, but not ecologically similar, population pairs. Likewise, reinforcement should result in some loci being strongly differentiated only between adjacent, hybridizing population pairs.
Our goal here is to assess these predictions and their applicability in population comparisons of Timema cristinae
stick insects. We collected genomic data from a mosaic of populations where past work clearly indicates that numerous factors, such as divergent host adaptation, geographical separation, gene flow and reinforcement all affect speciation [25
]. The genotyping-by-sequencing approach was to use restriction enzymes to cut up the genome into DNA fragments that are distributed across the genome, sequence tens of millions of these fragments on the Illumina next-generation sequencing (NGS) platform, align the fragments to discover genetic variation (facilitated by specimens being individually barcoded), and then conduct population genetic analyses on the resulting 86 130 single nucleotide polymorphisms (SNPs). This approach is thus aimed at surveying genome-wide patterns of genetic differentiation across the approximately 1.3 gigabase genome of T. cristinae
, rather than focusing in on specific genes that causally affect adaptation and speciation. Our Bayesian analyses incorporate genotype uncertainty, and thus account for the uneven coverage among gene regions and individuals inherent in NGS data.
The results demonstrate interplay between different speciation processes, each of which is associated with different patterns of genomic divergence. Surprisingly, we find that the effects of host use on genomic divergence are subtle, despite the strong effects of host use on ecological speciation, and that factors other than host use have pronounced effects on genomic divergence. The results highlight how genomic data have great potential for advancing understanding of speciation.