Introduction

A key feature of the speciation process is the evolving structure of genetic differentiation between populations that culminates in genetic divergence across much of the genomes of daughter taxa. Indeed, genomic patterns of evolutionary divergence across the speciation continuum promise to reveal not only the functional nature of adaptive evolution and reproductive isolation (Storz, 2005) but also the processes underlying speciation (Pinho and Hey, 2010). Consequently, a major research theme has emerged that aims to characterize genome-wide patterns of differentiation between diverging populations, races and species. Pockets of divergence or ‘genomic islands of speciation,’ as well as larger ‘continents,’ have been uncovered and could reflect the effect of genes contributing to reproductive isolation (‘barrier loci’) (Dopman et al., 2005; Turner et al., 2005; Feder, Egan et al., 2012). However, connecting these fundamental genetic signatures with their biological cause remains elusive (Noor and Bennett, 2009; Dopman, 2011; Nachman and Payseur, 2012; Cruickshank and Hahn, 2014).

Chromosomal rearrangements such as inversions are thought to facilitate speciation and promote the origin of evolutionary divergence by suppressing recombination. When speciation occurs in the face of gene flow, suppressed recombination in a heterokaryotypic hybrid offspring may cause the entire inverted segment to behave as a single linked unit (Noor et al., 2001; Rieseberg, 2001). Consequently, only a small number of barrier loci might be required to protect a large chromosomal band, possibly several megabases in length, from gene flow. Because of restricted gene flow, barrier and locally adaptive alleles are thought to accumulate within inversions leading to increasing divergence until speciation is complete (Navarro and Barton, 2003). Finally, adaptive alleles that are captured by inversions may rapidly drive their fixation, creating long-range hitch-hiking across the nonrecombining rearranged interval (Kirkpatrick and Barton, 2006; Kirkpatrick, 2010; Feder et al., 2011). In the absence of structural mutations, reduced recombination in collinear regions could also lead to evolutionary divergence; however, the footprint of genetic differentiation and opportunity for the accumulation of barrier loci might be limited (Butlin, 2005; West and Via, 2008; Feder et al., 2012). Thus, changes in recombination rate because of structural mutations might be the key to speciation with gene flow, and they may therefore help to explain many instances of genomic islands of speciation.

In this study, we evaluate the role of suppressed recombination for the evolution of a modestly sized genomic island of speciation between a pair of moth lineages. The ‘Z’ and ‘E’ strains of European corn borer (ECB, Ostrinia nubilalis) are incipient species that colonized North America ~100 ya without an appreciable bottleneck from allopatric locations in Italy (mixture of E and Z strains) and Hungary (exclusively Z-strain; Smith, 1920; Caffrey, 1927; Klun and Cooperators, 1975; Dopman, 2011). Strains are recently diverged (~100 000 ya; Malausa et al., 2007) and are textbook examples of speciation (Coyne and Orr, 2004), in which one species is splitting into two through the evolution of many forms of reproductive isolation (Dopman et al., 2010). The sex (Z) chromosome harbors several known genetic factors for adaptation and/or barrier traits, including those underlying behavioral isolation (Roelofs et al., 1987; Dopman et al., 2004) and temporal isolation (Glover et al., 1992; Dopman et al., 2005). Current and historical gene exchange between ECB strains is likely, as indicated by hybridization in Europe and North America (Dopman et al., 2010; Coates et al., 2013), geographic variation in strength of reproductive isolation (ranging from 0.91 to 0.99; Dopman et al., 2010) and molecular evidence for an isolation-with-migration model of divergence (2 Nm~10; Malausa et al., 2007). Whereas most loci show extensive shared polymorphism (for example, FST<0.05), a possible island of speciation along an ~1-cM swath of the Z chromosome consists of four genes (Tpi and three olfactory receptors, ORs) and reveals ECB strains as nearly reciprocally monophyletic (for example, FST>0.7; Dopman et al., 2005; Dopman, 2011; Lassance et al., 2011). Evolutionary divergence can be explained by several tightly linked barrier loci on the sex chromosome or by a regional selective sweep at these or other loci. However, such interpretations must be viewed as incomplete without proper tests of suppressed recombination and its repercussions for the genomic zone of influence for adaptation and/or barrier loci. Indeed, a possible signature of an inversion between Z and E strains was detected on the Z chromosome (Dopman et al., 2004), in which the lowest estimated recombination rate across the entire 1697-cM genetic linkage map was found in the vicinity of the Tpi/ORs loci (Figure 1).

Figure 1
figure 1

Local recombination rates (cM/Mb, black line) and marker densities (count) on the ECB sex (Z) chromosome. The dashed red line indicates the genome-wide average. Genetically differentiated loci Tpi and ORs (FST>0.7) are located in the low-recombination region (~17 cM). Adapted from Dopman et al. (2004).

If inversions are important during speciation and the evolution of genetic differentiation, at least three testable predictions have been made (Faria and Navarro, 2010). First, patterns of gene flow should be higher within collinear versus rearranged regions (Rieseberg, 2001; Feder et al., 2005; Kulathinal et al., 2009; McGaugh and Noor, 2012). Second, traits involved in reproductive isolation should map to regions located within chromosomal rearrangements (Noor et al., 2001; Feder et al., 2005; Lowry and Willis, 2010). Finally, genotyping hybrid offspring in the laboratory ought to reveal evidence for suppressed recombination (Feder et al., 2003; Kulathinal et al., 2009). By using a combination of genetic crosses, phenotyping of a trait contributing to reproductive isolation, and population genetic analysis, we set out to test these predictions for the ECB sex chromosome.

Materials and methods

Recombination suppression

To estimate variation in recombination frequencies surrounding the genetically differentiated region of the ECB sex chromosome, we generated three-point test crosses within and between Z and E strains. Crosses used insects collected as caterpillars, pupae and adults from the New York State, USA (n500 males and n500 females each from East Aurora, Geneva and Bouckville). Laboratory populations were maintained by mass rearing ~100 males and 100 females each generation. The female sex-pheromone blend is strain-specific. Z-strain females produce a 3:97 ratio of E/Z11-14:OAc, whereas E-strain females use a 99:1 ratio. Strain identities of all parental stocks were confirmed by genotyping pgFAR, the autosomal locus that determines the female sex-pheromone blend (Lassance et al., 2010; Supplementary Table S1).

Recombination is restricted to males in many Lepidoptera (Dopman et al., 2004). Therefore, in backcross pedigrees only F1 males are informative. Here, between-strain backcrosses (referred to as ‘Z × E’) involved a Z grandmother, an E grandfather, an F1 male and a recurrent Z-strain mother. Within-strain crosses (referred to as ‘Z × Z’) had a similar design but were exclusively from the Z-strain.

The ECB diverged from Bombyx mori (the silk moth) ~100 Mya (Pringle et al., 2007); however, the species pair shows a high degree of conservation in gene order like many other Lepidoptera (Pringle et al., 2007; d'Alencon et al., 2010; Kroemer et al., 2011). Therefore, we relied on the B. mori genome to identify evenly spaced sex-linked mapping loci for the ECB, whose genome is incomplete. Markers consisted of Tpi at 9 Mb, Per at 13 Mb and Ldh at 17.4 Mb (International Silkworm Genome Consortium, 2008). We adopted methods developed elsewhere for Tpi and Ldh (Dopman et al., 2005), whereas initial Per sequence for the ECB was obtained using degenerate PCR.

DNA from mapping family individuals was extracted from adult legs following using a DNeasy Blood and Tissue kit (QIAGEN, Valencia, CA, USA). Genotyping assays for offspring were developed from segregating polymorphisms and consisted of diagnostic PCR products that amplified polymorphic indels and/or restriction enzyme cut sites (EcoR1, Alu1 and Bsr1 for Tpi, Per and Ldh, respectively; Supplementary Table S1). Only female offsprings were genotyped and, because female Lepidoptera are the heterogametic sex and possess only one Z chromosome, one allele per locus was detected. Genotyping results were confirmed by Sanger sequencing (n~100 sequenced alleles per marker). One Z × E cross had already been genotyped for Tpi and Ldh in a prior study (Dopman et al., 2005).

Colocalization of reproductive isolation

As prior results (Figure 1) and preliminary data suggested evidence for reduced recombination near Tpi in F1 hybrids, we tested for colocalization of traits involved in reproductive isolation. A subset of Z × E backcross offspring was mapped for a major sex-linked factor (named Pdd) controlling diapause emergence time (Glover et al., 1992; Dopman et al., 2005), defined as the time to pupation for over-wintering caterpillars under environmental conditions conducive to breaking diapause. Differences at Pdd contribute to temporal isolation between Z and E strains in North America by conferring an ~30-day shift in emergence and thus adult-mating flights (Wadsworth et al., 2013). Shifts in adult flights between univoltine Z and bivoltine E populations eliminate as much as 85% gene flow in nature (Dopman et al., 2010).

We measured diapause timing variation by inducing and then breaking diapause in recombinant ECB caterpillars. Following earlier studies (Glover et al., 1992; Dopman et al., 2005), diapause was induced by a 12:12 light:dark photoperiod and then broken 35 days later with a 16:8 light:dark cycle. Diapause emergence time was noted every 2 days. As females are hemizygous and Pdd is inherited as a Mendelian locus, backcross females show patterns of emergence consistent with either their grandfather (here, E-strain) or grandmother (here, Z-strain; Glover et al., 1992; Dopman et al., 2005). Hence, recombinant backcross females show E-like earlier emergence times and pupate in less than 22 days (range=8–22 days, mean=15.5 days, s.e.±0.55), or they show Z-like later emergence times and pupate in more than 18 days (range=18–80, mean=43.94 days, s.e.±1.64; Glover et al., 1992).

Higher introgression in collinear regions

We evaluated patterns of genetic variation among field-caught female ECB: 18 insects were E-strain, 18 were Z-strain and at least one insect was from an outgroup species, the Asian corn borer (ACB, O. furnacalis). ECB caterpillars, pupae or adults were collected over a large geographic range, and thus genetic similarities should be viewed as indicating a combination of ongoing or recent gene exchange. Samples included insects from USA (n=7 E-strain and n=5 Z-strain from New York; n=7 E-strain from North Carolina and n=4 Z-strain from North Carolina; n=4 Z-strain from Iowa), Italy (n=4 E-strain and n=3 Z-strain) and Hungary (n=2 Z-strain). Moths were classified into strain by diagnostic gas chromatographic profiles of female sex-pheromone blend (that is, 99:1 versus 3:97 ratio of E/Z11-14:OAc; for example, Dopman et al., 2004) and/or by genotyping the pgFAR locus.

Molecular markers were developed from genes that were evenly distributed across the B. mori sex chromosome. Using ECB transcripts that were developed from a separate study, reciprocal BLASTs were performed to obtain ECB–B. mori gene pairs with predicted locations on the ECB sex chromosome (B. mori nucleotide CDS, BLASTn, e-value<1e−40). ECB transcripts were then searched with tBLASTx against the matching B. mori genomic sequence to identify predicted introns. Primer pairs for 23 loci (Supplementary Table S2) that had interlocus intervals ranging from 0.1 to 2.5 Mb (mean=0.96 Mb, s.e.±0.16) were designed to amplify ECB introns using PrimerBlast (NCBI). Sequences from field-caught insects were aligned and edited by eye using Genious Pro 5.5 (Biomatters Inc., San Francisco, CA, USA). Markers were named based on Bombyx gene identifiers.

Measures of genetic variation (π, θw), genetic differentiation (da, dxy, FST (Hudson et al., 1992), Snn (Hudson, 2000)) and the allele-frequency spectrum (Tajima's D (Tajima, 1989b), Fay and Wu's H (Fay and Wu, 2000)) were calculated using DnaSP (Rozas et al., 2003). The genealogical sorting index (gsi; Cummings et al., 2008) was calculated to quantify the degree of exclusive ancestry of Z- and E-strain moths. gsi ranges from 0 (no exclusivity) to 1 (monophyletic), and was calculated in R (Team RDC, 2013) on the Tufts high-performance computing research cluster using the genealogicalSorting library (www.genealogicalsorting.org). Genealogies were constructed using neighbor-joining with a Tamura-Nei 93 model of evolution to calculate distances. Statistical significance of gsi was assessed through permutation tests (n=10 000) on each of 1000 bootstrap replicates for a total of 10 000 000 replicated genealogies. Indel polymorphisms were considered a fifth base when calculating genetic differentiation and genetic exclusivity. All figures were constructed using the ggplot2 library (Wickham, 2009) in R.

Results

Recombination suppression

We used three-point crosses within- and between-strain to test for evidence of suppressed recombination associated with a high-frequency chromosomal rearrangement between Z and E strains of ECB. No significant differences in recombination frequency were observed among different within-strain backcross families (Z × Z, n=2 families) or among between-strain backcross families (Z × E, n=7 families) and therefore families were combined for each backcross type. Within strain, we observed six recombinants between the Tpi and Per pair (7%, n=86 offsprings), six between the Per and Ldh pair (7%, n=86 offsprings) and twelve between Tpi and Ldh (14%, n=86 offsprings; Figure 2a). Hence, in agreement with that found in B. mori, the predicted gene order within the Z-strain of ECB is Tpi—Per—Ldh (Figure 2b).

Figure 2
figure 2

Suppressed recombination on the sex chromosome in ECB hybrids. (a) Percent recombinant offspring between Tpi, Per and Ldh in within-strain (Z × Z) and between-strain (ZxE) crosses. (b) Inferred linear order of markers in Bombyx mori and the two ECB crosses. ***P-value<0.001.

Between-strain crosses revealed heterogeneity in recombination frequency compared with that seen within-strain (Figure 2a). ZxE crosses yielded slightly elevated recombination frequencies for the PerLdh gene pair (seven recombinants, 9.1%, n=77 offsprings; Fisher's exact test, P-value=0.7739) and slightly reduced recombination between Tpi and Ldh (seven recombinants, 9.1%, n=77 offsprings; Fisher's exact test, P-value=0.4643). However, a statistically significant suppression in recombination was observed between Tpi and Per. Zero recombinants were found in over 600 offsprings (n=666 offsprings; Fisher's exact test, P-value=1.91e−06).

Colocalization of reproductive isolation

We phenotyped a subset of backcross offspring to determine the linkage relationship between the nonrecombining TpiPer gene pair and Pdd. Diapause emergence timing of 333 hemizygous females from backcross families ranged from 2 to 100 days; yet, a clear bimodal distribution emerged (Figure 3). Eight offspring (2.4%) had emergence times that fell within the period of overlap of parental Z and E strains (day 18–22) and their Pdd genotype was considered indeterminate. The remaining 325 offspring had emergence times consistent with pure-strain insects. Of these, one female with E-strain markers at the Tpi–Per gene pair displayed Z-strain like diapause and emerged later, and one female with Z-strain markers displayed E-strain like diapause and emerged earlier. Thus, two females (0.62%) were apparent recombinants between the Tpi–Per gene pair and the Pdd locus. In the other 323 offspring (99.38%), Pdd was perfectly associated with their genotype at the Tpi–Per gene pair.

Figure 3
figure 3

Diapause emergence time maps to a putative sex-linked inversion. Bimodal distribution of diapause emergence time for female backcross offspring genotyped for Tpi and Per. Note that TpiE/PerE are E-strain alleles and TpiZ/PerZ are Z-strain alleles. Dashed lines indicate parental distribution values.

Higher introgression in collinear regions

We characterized genetic variation for 23 loci to test for differences in patterns of genetic variation between loci with predicted locations inside (referred to as ‘rearranged’) and outside (referred to as ‘collinear’) of the nonrecombining Tpi–Per chromosomal region. A total of 779 sequences were obtained, resulting in ~9 kb of aligned data and 224 segregating sites. Seven loci fell within the Tpi–Per interval and 16 were outside (8 upstream and 8 downstream).

We tested for variation in patterns of genetic differentiation using four different indices. Rearranged loci exhibited statistically significantly elevated evolutionary divergence for both FST (median rearranged=0.14990, median collinear=0.06418, W=88, P-value=0.01764) and Snn (median rearranged=0.6569, median collinear=0.5742, W=90, P-value=0.01258; Figure 4). Similarly, the number of net substitutions per site between strains (da, median rearranged=0.09%, median collinear=0.029%, W=74, P-value=0.121) and the average number of substitutions per site between strains (dxy, median rearranged=0.76%, median collinear=0.66%, W=54, P-value=0.5647) were both higher for rearranged loci (Figure 4), although these differences were not statistically significant.

Figure 4
figure 4

Elevated evolutionary divergence between ECB strains. (a) FST and Snn. (b) da and dxy. *P-value<0.05.

We also used gsi to explore gene tree topology measures of genetic differentiation (Table 1). Loci with significant values of gsi ranged from 0.01 to 0.68 with an average of 0.22. E-strain ECB had seven loci with significant genealogical exclusivity, whereas the Z-strain had 15. There were no significant differences in the magnitude of gsi for rearranged and collinear loci (t=−1.3504, degrees of freedom=19.019, P-value=0.1927); however, all rearranged loci possessed significant genealogical exclusivity in either one or both strains, in contrast to ~50% of collinear loci (Fisher's exact test, P-value=0.0574).

Table 1 Genetic exclusivity between ECB strains for colinear and rearranged loci

To evaluate patterns of nucleotide polymorphism, we compared π and θw between groups. Average genetic variation for rearranged loci was approximately one-half of that for collinear loci (Table 2), although only statistically significant (or trending) for E-strain moths (mean rearranged π=0.386%, mean collinear π=0.812%, t=−1.7158, degrees of freedom=16.839, P-value=0.05227; mean θw=0.347%, mean collinear θw=0.916%, t=−2.3966, degrees of freedom=15.098, P-value=0.01496). For both strains Tajima's D showed a slight skew toward an excess of intermediate-frequency alleles compared with low-frequency alleles in rearranged loci (that is, more positive values; Table 2), whereas Fay and Wu's H revealed a relative excess of high-frequency derived alleles compared with intermediate-frequency alleles (that is, more negative values; Table 2). However, neither estimate was statistically significant.

Table 2 Patterns of genetic variation among rearranged and colinear loci

Discussion

Although our post-genomic age has produced billions of base pairs of DNA sequence data for many organisms across the speciation continuum, we often lack adequate biological explanations for the evolution of fundamental genomic features such as genetic divergence. Inversions are thought to have far-reaching impacts on evolutionary divergence during speciation, especially when divergence occurs in the face of gene flow. However, demonstrating the importance of inversions to the speciation process requires evidence for (i) recombination suppression in hybrid offspring, (ii) localization of barrier or adaptation loci in rearranged regions and (iii) enhanced sequence divergence (Faria and Navarro, 2010). In this study, we combined genetic crosses (Figure 2), phenotyping of traits underlying reproductive isolation (Figure 3) and observational studies of DNA polymorphism in wild populations (Figure 4, Table 1) to show that Z and E strains of ECB meet all three criteria. Although we cannot physically confirm the presence of an inversion yet through cytology, these results are consistent with the presence of an inversion that has contributed to accumulation of ecologically adaptive alleles and genetic differentiation across roughly 20% of the ECB sex chromosome (~4 Mb).

Inversions and reproductive isolation

A major appeal of chromosomal rearrangement models of speciation is their ability to help explain apparent examples of speciation with gene flow (Pinho and Hey, 2010). Indeed, a serious theoretical weakness of nonallopatric speciation is that recombination randomizes associations between genes conferring local adaptation (hybrid unfitness) and assortative mating (Felsenstein, 1981; Gavrilets, 2003; Butlin, 2005). Although the specifics of chromosomal rearrangement models of speciation differ (Noor et al., 2001; Rieseberg, 2001; Navarro and Barton, 2003; Kirkpatrick and Barton, 2006; Faria and Navarro, 2010), a common feature is the removal of this antagonism and a buildup of adaptation and barrier loci within rearrangements during a nonallopatric phase(s) that then allows for reinforcement and the completion of speciation. Thus, a critical prediction for speciation with gene flow is that rearrangements such as inversions should be present and contain multiple prezygotic and/or postzygotic barrier loci with major effects on reproductive isolation. Determining the degree to which these expectations are borne out in nature is challenging. First, we have limited data on quantitative measures of reproductive isolation, defined as RI=1−(between species fitness)÷(within species fitness) and ranging from 0 (no effect) to 1 (complete RI; Ramsey et al., 2003; Coyne and Orr, 2004). Second, the genomic positions of the underlying barrier loci is often unknown. Monkeyflowers are a rare exception, in which loci of large reproductive isolation effect that contribute to floral traits, flowering time and hybrid male sterility map to chromosomal rearrangements in the Mimulus genome (Ramsey et al., 2003; Lowry and Willis, 2010; Fishman et al., 2013).

The putative sex-linked inversion we document in Ostrinia appears consistent with the large-effect, multicomponent prediction of chromosomal rearrangement models of speciation. First, multiple ‘speciation phenotypes’ (Shaw and Mullen, 2011) in ECB have known sex-linked factors, including male pheromone response and diapause timing (McLeod, 1978; Dopman et al., 2004; 2005, Wanner et al., 2010; Ikten et al., 2011). Second, a recent study indicated that large isolation effect sizes are common between ECB strains (RI0.5 for 4 of 7 forms of isolation; Dopman et al., 2010). Given this prior work and the large size of the putative inversion (7 cM or ~4 Mb; Figure 2b), there would appear to be good reason to expect diverse and potent barrier loci in an inversion on the ECB sex chromosome. Indeed, we found that only two of 325 offspring were recombinants (0.62%) between the nonrecombining region (Tpi–Per loci, Figure 2b) and the factor(s) determining diapause emergence time (Pdd; Figure 3). Thus, a major cause of temporal isolation (mean RI=0.65) is likely inside or near the chromosomal rearrangement. Similarly, the sex-linked ORs are known to underlie strain-specific male antennal response to Z versus E-strain female pheromones (that is, 3:97 versus 99:1 ratio of E/Z11-14:OAc) and they are tightly linked to Tpi and thus the sex-linked rearrangement (Wanner et al., 2010; Lassance et al., 2011). Hence, a rearrangement may have promoted coupling between multiple genes conferring ecological (Pdd) and behavioral (ORs) divergence, thereby facilitating speciation. Clearly, there is a need for formal experimentation to determine whether chromosomal rearrangements in Ostrinia and in other systems harbor multiple loci of large isolation effect that contribute to varied components of reproductive isolation.

Evolution of genetic divergence

Results from several systems including Ostrinia and Drosophila (McGaugh and Noor, 2012) imply high genetic divergence within inversions and minimal divergence outside of them, and a reciprocal relationship for polymorphism. Such patterns may be common in the early stages of speciation because of the coupling of selection and recombination. Globally beneficial mutations and adaptive barrier alleles that arise within fixed inversions are expected to get trapped and cannot easily migrate between lineages (Navarro and Barton, 2003). Further, according to the Kirkpatrick–Barton model (Kirkpatrick and Barton, 2006), selective establishment and fixation of inversions occurs because locally adaptive alleles at loci within inversions (for example, underlying ecological and behavioral divergence) cannot recombine with chromosomes from other populations and thus they are favored because they do not suffer the ‘Achilles' heel’ of being located on the same chromosome as immigrant disadvantageous allele(s). Hence, genetic hitch-hiking associated with either selective fixation of inversions or their internal mutations might lead to inversion-specific reductions in polymorphism within one or both lineages, while also creating enhanced differentiation for relative divergence measures that are sensitive to within lineage variation (for example, FST; Charlesworth, 1998).

An additional role of suppressed recombination for evolutionary divergence involves repeated bouts of gene exchange and disruptive selection. Specifically, barrier loci within inversions under the Navarro–Barton model (Navarro and Barton, 2003) might produce a molecular signal of differentiation that is spread across the inversion or at least near its breakpoints (Rieseberg, 2001; Noor et al., 2001; Strasburg et al., 2009; McGaugh and Noor, 2012) because of selective purging of maladapted (and nonrecombining) regions encompassed by inversions upon their introduction into sister taxa. Such a process might be analogous to ‘differential introgression,’ in which gene flow is restricted at individual barrier loci but is high elsewhere (Barton and Bengtsson, 1986; Wu, 2001; West and Via, 2008; Harrison, 2012). Over the long term, differential introgression is expected to produce specific signatures. For example, increases in absolute genetic divergence (for example, dxy) as new mutations arise and fix independently in daughter taxa (Charlesworth et al., 1997; Charlesworth, 1998; Nachman and Payseur, 2012; Cruickshank and Hahn, 2014).

Despite limitations of our current sampling scheme (18 samples per strain from both allopatric and sympatric localities), results from Ostrinia suggest that selection and suppressed recombination were important for evolutionary divergence. Compared with ‘collinear’ loci, patterns of genetic variation at ‘rearranged’ loci are consistent with a selective sweep(s), including 50% lower genetic variation (primarily in E-strain moths; Table 2) and 2–3 × greater differentiation for measures that are sensitive to selection at linked sites (FST and da; Figure 4). Support for differential introgression is less obvious. The putative inversion shows modest increases in absolute divergence (1.15 × greater values of dxy) and also genealogical exclusivity (1.14 × greater values of Snn). Moreover, although patterns in the allele-frequency spectrum (D and H) are consistent with rare migration within the inversion and higher migration elsewhere (Table 2; Tajima, 1989a; Simonsen et al., 1995; Przeworski, 2002; Fay and Wu, 2005), these changes are relatively minor. Lack of genetic evidence for differential introgression runs counter to observations in nature. An appreciable number of hybrid offsprings occur at sampled sympatric localities in New York (5–15%; Dopman et al., 2010; Coates et al., 2013) and yet differentiation between Z and E strains at Tpi has been maintained at these sites for more than 40 generations (Glover et al., 1991; Dopman et al., 2005; Dopman, 2011). Disruptive selection following introgression of the nonrecombining region that includes Tpi might be important in maintaining divergence at New York localities, where sweeps and incomplete lineage sorting alone cannot easily account for patterns of nucleotide variation across loci. Equivocal genetic support for differential introgression should not be surprising because statistical power is limited during early stages of speciation (for example, for dxy; Cruickshank and Hahn, 2014). Hence, lineages that are in the early stages of divergence like ECB strains may be too closely related to show strong statistical signs of differential gene flow even at genome regions where it occurs.

More extensive investigation of Ostrinia moths and other systems might validate the scenario we propose in which evolutionary divergence during speciation occurs by two complementary but distinct mechanisms—one driving the transient origin of differentiation through local adaptation and selection on linked sites, and one determining its stable persistence through reproductive isolation and/or differential introgression (see also Dopman, 2011; Nachman and Payseur, 2012; Cruickshank and Hahn, 2014). If an inversion is present as a low-frequency polymorphism in allopatry and greatly differentiates lineages only in geographic regions where they co-occur (for example, as in the ‘mixed-geographic model’ (Feder et al., 2011)), this might suggest a direct role of recombination suppression in promoting evolutionary divergence by local adaptation and selective sweeps (for example, to eliminate the creation of less fit hybrids or to keep high fitness loci together). In contrast, the presence of an older, high-frequency inversion that differentiates lineages regardless of geography could indicate an indirect role of recombination suppression in facilitating accumulation of additional genetic changes once inversions establish or fix. That is, the rearrangement could protect chromosomal regions from gene flow and allow for additional genetic changes to accrue (for example, through local sweeps of internal beneficial mutations or differential introgression of the region encompassed by the rearrangement). Broader comparisons between recently sympatric Z and E strains of Ostrinia living in North America and their allopatric, European ancestors may help distinguish between direct and indirect roles of recombination suppression during divergence.

Conclusions

Our studies of recombination, reproductive isolation and genetic variation in the wild imply that structural mutations have facilitated speciation with gene flow between Z and E strains of ECB. Our results add to the limited number of studies that provide empirical support for this notion, including apple maggot flies (Feder et al., 2003; 2005, Michel et al., 2010; Powell et al., 2013), fruit flies (Noor et al., 2001; Kulathinal et al., 2009), stickleback fishes (Kitano et al., 2009), sunflowers (Rieseberg, 2001; Strasburg et al., 2009) and monkeyflowers (Lowry and Willis, 2010; Fishman et al., 2013). Our finding that structural mutations contribute to speciation may be the first of many for Lepidoptera. The rate of rearrangement in lepidopteran genomes appears to be one of the fastest among eukaryotes at ~2 breakages/Mb/My, or ~3 × faster than nematodes and more than an order of magnitude faster than flies, mammals and plants (Ranz et al., 2001; Coghlan and Wolfe, 2002; d'Alencon et al., 2010). As lepidopterans (and nematodes) have diffusely organized centromeres (holocentric), these high evolutionary rates may stem from an increased likelihood of reintegration of double-strand break fragments (d'Alencon et al., 2010). Hence, tens or even hundreds of fixed rearrangements could characterize moth or butterfly species pairs that diverged just several hundred thousand years ago, implying that structural mutations may have had a disproportionate role over the evolutionary history of the second largest order of insects, and may commonly promote adaptive divergence and speciation in moths and butterflies.

Data archiving

Data available from the Dryad Digital Repository: http://doi.org/10.5061/dryad.2vg05