Differential introgression and the maintenance of species boundaries in an advanced generation avian hybrid zone

Background Evolutionary processes, including selection and differential fitness, shape the introgression of genetic material across a hybrid zone, resulting in the exchange of some genes but not others. Differential introgression of molecular or phenotypic markers can thus provide insight into factors contributing to reproductive isolation. We characterized patterns of genetic variation across a hybrid zone between two tidal marsh birds, Saltmarsh (Ammodramus caudacutus) and Nelson’s (A. nelsoni) sparrows (n = 286), and compared patterns of introgression among multiple genetic markers and phenotypic traits. Results Geographic and genomic cline analyses revealed variable patterns of introgression among marker types. Most markers exhibited gradual clines and indicated that introgression exceeds the spatial extent of the previously documented hybrid zone. We found steeper clines, indicating strong selection for loci associated with traits related to tidal marsh adaptations, including for a marker linked to a gene region associated with metabolic functions, including an osmotic regulatory pathway, as well as for a marker related to melanin-based pigmentation, supporting an adaptive role of darker plumage (salt marsh melanism) in tidal marshes. Narrow clines at mitochondrial and sex-linked markers also offer support for Haldane’s rule. We detected patterns of asymmetrical introgression toward A. caudacutus, which may be driven by differences in mating strategy or differences in population density between the two species. Conclusions Our findings offer insight into the dynamics of a hybrid zone traversing a unique environmental gradient and provide evidence for a role of ecological divergence in the maintenance of pure species boundaries despite ongoing gene flow. Electronic supplementary material The online version of this article (doi:10.1186/s12862-016-0635-y) contains supplementary material, which is available to authorized users.


Background
Hybrid zones are excellent model systems for evolutionary studies as they provide a diversity of recombinant genotypes through generations of mutation, recombination, and gene flow [1,2]. Growing empirical evidence indicates that natural hybrid zones occur across a range of taxonomic groups at rates greater than previously estimated [3] and that hybridization and introgression are important forces that can shape the evolutionary trajectory of a species [4][5][6]. Studies of hybridizing taxa that maintain genetic distinction with ongoing gene flow provide insight into the speciation process [6,7] and offer a direct measure of reproductive isolation. Because hybrid zone studies allow for the quantification of differential patterns of introgression of foreign alleles, hybrid zones provide the opportunity to identify the genetic and phenotypic traits influencing species divergence [8].
Hybrid zones are thought to be semi-permeable boundaries between genomes as differential fitness of hybrids can result in reduced introgression of those regions important in maintaining reproductive isolation, while introgression of regions free of selection is permitted [1,9,10]. Loci with no influence on hybrid fitness should display uninhibited movement across a hybrid zone, whereas regions underlying local adaptation or that are responsible for genetic incompatibilities remain differentiated, often in the presence of gene flow [1,8,11]. Rates of introgression have been found to vary among genetic and phenotypic markers across a number of natural hybrid zones [12][13][14]. These observations have been linked to numerous demographic and selective processes, including genetic incompatibilities [15], ecological divergence [16], differential fitness [17], and variations in mate preference and behavior [11].
Sampling a diversity of genetic and phenotypic markers provides an unbiased view of introgression and genetic structure across a hybrid zone [18,19]. Understanding these patterns can offer valuable insight into the mechanisms responsible for restricting gene flow across species' boundaries [11,[20][21][22], as differential introgression may be indicative of ecological or evolutionary dynamics in the focal gene regions [13,23]. For example, neutral microsatellite markers should diffuse freely across the hybrid zone, resulting in widespread movement of alleles. Conversely, diagnostic markers (i.e. markers that are fixed or highly differentiated between two parental species) are predicted to be under divergent selection, exhibiting reduced introgression [19], as the elevated divergence typically associated with diagnostic markers suggests association with genomic regions under selection [24]. Differential introgression of sexlinked and mitochondrial markers relative to autosomal loci is often attributed to Haldane's rule, which predicts greater fitness reductions in hybrids of the heterogametic sex [25]. This pattern has been observed in a number of avian [26][27][28] and mammalian systems [12].
Morphological traits also provide insight into extrinsic selection and demographic events shaping a hybrid zone [29]. Bimodal distribution of phenotypes, or an abrupt clinal transition, can be indicative of high dispersal, differential selection, hybrid zone movement [13,30], or assortative mating [29,31]. Assessing introgression of secondary sexual characteristics (e.g., plumage) can also aid in identifying patterns of asymmetrical introgression [11]. Divergence in plumage characteristics can be particularly important in driving pre-zygotic isolation in birds [32], as these traits play an important role in mate selection, providing a range of important cues to females including individual and territory quality [33,34] and offspring attentiveness [35].
Here we investigated patterns of introgression in an avian hybrid zone between two recently diverged marsh endemics, the Saltmarsh (Ammodramus caudacutus) and Nelson's (A. nelsoni) sparrow (~600,000 years; [36]). In the USA and Maritime Canada, the two species are restricted to a linear ribbon of tidal-marsh habitat along the Atlantic seaboard with a subspecies of caudacutus (A.c. caudacutus) predominantly inhabiting coastal salt marshes from southern Maine to New Jersey and a subspecies of nelsoni (A.n. subvirgatus) predominantly inhabiting brackish and tidal marshes from the Canadian Maritimes to northern Massachusetts [37,38]. Current knowledge suggests that the two taxa (hereafter caudacutus and nelsoni, respectively) overlap and hybridize in tidal marshes along a 210 km stretch of the New England coast between the Weskeag River estuary in South Thomaston, Maine and Plum Island in Newburyport, Massachusetts [39][40][41].
Recent work in the caudacutus-nelsoni hybrid zone indicates extensive introgression with a high proportion of backcrossed sparrows in sympatric populations [42]. Despite high rates of admixture, very few individuals are recent generation (F1/F2) hybrids (3 %; [42]), indicative of an advanced generation hybrid zone characterized by high rates of recombination [43,44]. Accordingly, there is no intermediate hybrid phenotype, and complex patterns of morphological variation preclude discrimination of pure and admixed sparrows from morphology alone [42]. While backcrossing is extensive between caudacutus and nelsoni, variation in habitat affinity, morphology, and behavior suggest a role for isolating mechanisms in this system. Abrupt environmental gradients across the marine-terrestrial ecotone within tidal marshes present adaptive challenges to terrestrial vertebrates (e.g. tidal inundation and osmoregulatory demands [45,46]). A. caudacutus is a narrow niche specialist, reliant exclusively on salt marshes in both its breeding and wintering habitat; it has been associated with salt marshes over a longer evolutionary time frame [47] compared to A. nelsoni, which, in allopatry, uses a broader range of habitats including brackish and fresh water marshes and hay fields. In the hybrid zone, a mosaic of fine-scale habitat typescoastal, bay, and upriver, tidal marshesoccurs, and the spatial structuring of pure and hybrid individuals follows a patchy distribution consistent with these local habitat differences [48]. Due to these differences in niche specificity, there may be stronger selection for adaptive traits in pure caudacutus individuals, driving ecological divergence. Tidal marsh adaptations may also influence morphology and plumage coloration in pure caudacutus and nelsoni [42,49,50] with potential reinforcement of these traits through sexual selection. Numerous behavioral differences between caudacutus and nelsoni males, including differences in flight displays, song, aggressiveness, and mating strategy [37,38,51] further have the potential to shape asymmetries in mate selection within the hybrid zone.
The aim of this study was to characterize the genetic structure, including patterns of differential introgression and selection, across the caudacutus-nelsoni hybrid zone and to test the hypothesis that adaptive traits are important in maintaining pure species boundaries despite ongoing gene flow. We conducted extensive, systematic sampling across the full extent of the caudacutus-nelsoni hybrid zone, coupled with population genetic analyses and geographic and non-geographic cline analyses to characterize genetic variation, quantify introgression across genetic and morphological markers, and identify the width and center of the hybrid zone. We used plumage features and a diversity of genetic markers, including anonymous (putatively neutral) microsatellites, diagnostic (species-specific and potentially under selection; [52]) microsatellites, mitochondrial, and sex-linked markers, to compare introgression patterns across potentially variable selective processes. The diagnostic markers used in this study were identified from a genome-wide comparison of microsatellite loci between caudacutus and nelsoni [52] and were selected because they showed elevated divergence between allopatric caudacutus and nelsoni individuals (F ST = 0.4667) compared to neutral, anonymous microsatellites (F ST = 0.15). Several of the loci are linked to genes of known function, indicative of divergent selection for functional traits that differ between the species [52]. The potential influence of selection and introgression patterns for these markers across the geographic extent of the hybrid zone are of yet unknown. We predicted that the gene-associated diagnostic markers would show reduced introgression and more abrupt clines compared to the neutral microsatellite markers. We also predicted selection would occur for sex-linked and mitochondrial markers in accordance with Haldane's rule. In birds, females are the heterogametic sex (ZW), and thus Haldane's rule predicts reduced introgression of both sex-linked markers and mitochondrial markers (due to maternal inheritance) compared to autosomal markers. Lastly, we predicted strong selection for features related to plumage darkness, as increased melanin is thought to be an adaptation to tidal marshes (salt marsh melanism; [49,53,54]), which we hypothesized to be under stronger selection in caudacutus.

Results
We characterized genotypic data at 24 microsatellite loci and DNA sequences from two mitochondrial, 2 Z-linked and one autosomal gene from 286 sparrows from 32 marshes across the caudacutus-nelsoni hybrid zone and surrounding allopatric populations (Fig. 1). We obtained morphological data (plumage, bill, and structural measurements) from 254 of these individuals. Microsatellite loci were highly polymorphic with allelic richness ranging from 4 to 33 alleles per locus (mean = 12.7). Allelic richness was greater in pure caudacutus populations (mean = 8.5 alleles per locus, range = 2 -21) than in pure nelsoni populations (mean = 7.3 alleles per locus, range = 1 -14).
Mean observed heterozygosities ranged from 0.531 to 0.704 (Table 1), with heterozygosity generally increasing from North to South. Six of the 24 microsatellite markers (Ammo008, Ammo012, Ammo015, Ammo016, Ammo030, and Ammo036) were candidates for positive selection, likely a result of their association with coding regions (Additional file 1: Figure S1). All other microsatellite markers were within neutral expectations. We detected significant deviations (Bonferroni adjustment; α = 0.05, P = 0.001) from Hardy-Weinberg in 7 out of 32 (22 %) marshes (Table 1). We did not observe significant linkage disequilibrium in any of our populations.
STRUCTURE assigned individuals to one of two genetic clusters (Fig. 3) based on ΔK (Additional file 2: Figure S2), which corresponded to caudacutus and nelsoni populations. Consistent with previous findings, we found few intermediate individuals (F1 hybrids) and pure and backcrossed individuals appeared to be patchily distributed across sympatric populations (Fig. 3). Individuals sampled from allopatric nelsoni populations had a low probability (mean Q value ± SD = 0.007 ± 0.01) of being assigned to the caudacutus cluster, while individuals sampled from allopatric caudacutus populations had a high probability of being assigned to the caudacutus cluster (mean Q ± SD = 0.995 ± 0.006). Sympatric populations had intermediate Q values and hybrid indices (mean Q ± SD = 0.667 ± 0.450, Range = 0 -1 and mean HI ± SD = 0.66 ± 0.38, Range = 0 -1); however pure caudacutus and pure nelsoni individuals inhabiting the same marshes largely drove this pattern (Fig. 3). Twelve individuals (4 %) had Q values ranging from 0.1 to 0.9 (recent generation hybrids); these 12 individuals were dispersed across the sampled marshes (i.e., there were no marshes with a disproportionately high number of recent generation hybrids). There were 94 individuals (42 %) with a hybrid index ranging from 0.1 to 0.9 (indicating they were not pure parental genotypes). Mean site-specific interspecific heterozygosity ranged from 0 to 0.76 (mean ± SD = 0.26 ± 0.12), with the greatest interspecific heterozygosities found on sites near the center of the hybrid zone (Table 1). We observed significant genetic differentiation among sampled marshes (F ST ), with values ranging from 0 to 0.375 (θ = 0.1; Additional file 3: Figure S3). The largest F ST values were generally observed between the allopatric nelsoni populations and all other marshes. We also detected significant F ST values between sympatric marshes that were predominantly composed of nelsoni individuals (Maquoit Bay, Cousins River, and Rye Beach) compared to all other marshes.

Genomic and geographic analyses of introgression
Genomic clines revealed that introgression patterns were variable among markers (Additional file 4: Figure S4). Sixty-six percent (19) of the 29 markers showed deviations from patterns of neutral introgression (meaning they either exhibited more gradual or more abrupt patterns compared to neutral expectation; Fig. 4, Additional file 5: Figure S5). Clines were steeper than neutral expectation for 12 of these markers, including six of the diagnostic microsatellite loci (Ammo markers 001, 003, 006, 008, 027, 036), three of the anonymous microsatellites (Escμ1, Asμ15, Aca08), two mitochondrial markers (ND2/ND3), and SLC30A5. Six neutral microsatellite markers and RAG-1 displayed more gradual clines than neutral expectation. Comparison of individual loci to multilocus expectation using the logit-logistic model revealed variations in cline slope and position among the Fig. 1 The location of 32 marshes along the northeastern coast of the United States where A. caudacutus and A. nelsoni individuals were sampled. Black circles represent allopatric populations from which putatively pure individuals were used for calculating a hybrid index. White circles represent marshes that are outside of the currently hypothesized overlap zone, yet were treated as sympatric populations due to their close proximity to the hybrid zone and evidence of introgressed individuals [50]. Gray circles represent marshes within the hybrid zone and the red star represents the approximate center of the zone, based on geographic cline estimates. The boxplot represents the distribution of hybrid index values for each of the sympatric marshes sampled; the center of the hybrid zone is colored in red. To demonstrate phenotypic differences between the parental species, representative photographs are shown for allopatric nelsoni (top) and caudacutus (bottom). Hybrids do not exhibit a clear intermediate phenotype, but rather display complex combinations of phenotypic traits that do not vary predictably by genotypic class ( [42]; see text) 29 genetic markers (Fig. 5, Table 2). We detected overall patterns of asymmetrical introgression with 66 % (19) of the markers shifted toward caudacutus and 34 % (10 loci) shifted toward nelsoni. Five markers (Ammo006, Ammo036, ND2, ND3, and SLC45A2) displayed stronger selection (more abrupt slopes; Table 2); all five of these markers exhibited asymmetrical introgression toward caudacutus. Twenty-four markers displayed more gradual slopes (weaker selection) than multilocus expectation.
Geographic cline analyses revealed variation in estimates for cline width (mean = 392 km, range = 248 -969) but more consistent estimates for cline center (mean = 330 km, range = 229 -421; Table 3) across marker types. Based on these estimates, cline center was consistently predicted to be around sampling location 10 (Cousins River -Yarmouth, Maine). Estimates for cline width were the smallest for mitochondrial (264 km) and z-linked markers (299 and 358 km for SLC45A2 and  Table 4). Estimates for cline width were narrower for traits related to the darkness and definition of plumage (349 km) compared to traits related to the amount of plumage streaking (380 km; Table 4). Estimates for cline center were similar (only 8 km difference in mean) between morphological traits and genetic markers. Phenotypic variance fluctuated across sympatric marshes. For the five morphological traits, we observed peaks in phenotypic variance that fell approximately between 350 and 450 km along the sampling transect, consistent with cline estimates for the center of the hybrid zone (Fig. 6). We observed peaks in phenotypic variance near the estimated center of the hybrid zone for weight and for traits related to definition/darkness of streaking; further, variance in weight exceeded V max near the approximate  4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 (Fig. 6). Variance in bill length in sympatric populations was greater than the variance observed in allopatric populations for all sampled marshes. The degree of introgression (ratio of V obs and V max ) was higher for wing chord (V obs /V max = 0.5) and for plumage traits related to coloration and amount of streaking (V obs / V max = 0.46) than it was for weight (V obs /V max = 1.1) and plumage traits related to the definition and darkness of streaking (V obs /V max = 0.82).

Discussion
Species boundaries can remain distinct in the face of ongoing introgression, even if only a few regions of the genome remain differentiated while other regions become homogenized. Within the caudacutus-nelsoni hybrid zone, we found patterns indicative of strong selection (more abrupt slopes compared to a multi-locus average) for 5 out of 29 genetic markers despite extensive introgression in sympatric populations. We identified 42 % of the sampled individuals as admixed (hybrid index ranging from 0.1 to 0.9). The majority of these admixed individuals were backcrossed, with the very low proportion of recent generation hybrids in this system indicative of an advanced generation hybrid zone [42,44]. The distribution of pure and admixed individuals appeared patchy across sympatric populations, with neighboring marshes exhibiting noticeable differences in genotypic compositions. Increased heterozygosity and F IS at select marshes across the zone, including Weskeag and Chapman's Landing (which are 112 and 128 km north and south from the center, respectively) support the idea that certain marshes facilitate mixing more than others. The evolutionary history of caudacutus and nelsoni is complex; however, the leading hypothesis suggests that the current overlap zone is an area of secondary intergradation following a split during a Pleistocene glaciation event [51]. Consistent with this hypothesis, the results of this study provide evidence for secondary contact and contemporary introgression as opposed to incomplete lineage sorting. We found strong divergence across all markers in allopatric populations and high levels of admixture and a noticeable peak in phenotypic variation in sympatric populations. Greater genetic differentiation in allopatry (average F ST between allopatric caudacutus and nelsoni = 0.313; locus-specific F ST as high as 0.71) than in sympatry (average F ST between sympatric caudacutus and nelsoni = 0.24; locus-specific F ST as high as 0.61) suggests geographic structuring of alleles. Incomplete lineage sorting, alternatively, would manifest in random geographic distribution of ancestral alleles [55,56]. Furthermore, the occurrence of recent generation hybrids in sympatric marshes, although in low frequency, points toward contemporary hybridization events between these species.
Estimates for cline width were highly variable among markers, ranging from 248 to 970 km, and were, on average, most narrow for mitochondrial and z-linked genes. Estimates for cline center, however, were consistent among marker types (genetic and morphological) falling around Yarmouth, Maine (328 km from locality 1). Previous field surveys identified caudacutus and nelsoni individuals co-occurring from Weskeag, Maine to Newburyport, Massachusetts (~208 km overlap zone; Hodgman et al., [39]). Consistent with the field estimates of the overlap zone, three of the markers analyzed in this study (ND2, SLC45A2, and Ammo006) exhibited cline widths in the 250 -300 km range. The remaining markers had substantially larger cline widths, indicating extensive introgression and recombination within and well outside of the overlap zone. Clines were compared using the logit-logistic model of Fitzpatrick [99]. Two loci are presented in each plot; black and gray points represent the raw data points for each marker. Line color indicates whether markers are shifted toward nelsoni (blue) or caudacutus (red) based on estimates of u. Line type indicates whether markers show gradual (dotted) or abrupt slopes (solid) based on estimates of v Clines varying in width but constrained to the same center are indicative of differential introgression across the hybrid zone. This is consistent with predictions that hybrid zones act as a semi-permeable barrier for the exchange of genetic material between taxa [1]. We found differential introgression consistent with our a priori predictions for each marker type, including comparatively narrow cline estimates for mitochondrial, sexlinked, and select gene-associated, diagnostic markers relative to wide clines for neutral loci. This variable introgression across markers suggests that while most traits exhibit uninhibited movement, there are certain traits that do not freely cross the species' boundaries and therefore may be important in reproductive isolation. The observed patterns can be explained by both selection against hybrids and adaptive divergence along a tidal marsh gradient as active mechanisms in shaping species boundaries between caudacutus and nelsoni.
Consistent with Haldane's rule, we found that on average, mitochondrial and z-linked markers show reduced introgression compared to autosomal markers (including neutral and selected loci). Haldane's rule predicts that fitness reductions should occur more often in hybrids of the heterogametic sex [25]; these differential fitness reductions appear to play an important role in speciation [57]. Reduced introgression of mitochondrial or sexlinked markers in organisms with ZW sex determination is an expectation of the dominance theory of the Dobzhansky-Muller incompatibility model [58][59][60]. This theory predicts that fitness reductions arise through the interaction of incompatible alleles, which evolved in allopatry. If these incompatible alleles are recessive, fitness reductions will be greater for the heterogametic sex if these genes are located on the sex chromosomes. In systems where females are the heterogametic sex, Haldane's rule also predicts reduced introgression of mitochondrial markers because they are maternally inherited. There is extensive empirical support for Haldane's rule [61], increasingly so in avian systems, including sterility (Ficedula hypoleuca and F. albicollis; [62]) and lower survival rates (Larus argentatus and L. cachinnans; [63] of hybrid females, and reductions in female-mediated gene flow (Larus occidentalis and O. glaucescens; [64] and Aquila clanga and A. pomarina; [65]). Accordingly, adaptive behavioral differences in pure caudacutus and nelsoni females associated with nesting synchrony in relation to tidal cycles [66] suggest a potentially important influence of differential fitness among pure and admixed females in shaping zone dynamics. Evidence for reduced survival in F1/F2 females provide further support for Haldane's Rule in this system [67]. Nonetheless, other causes of the observed patterns of restricted introgression of mitochondrial and sex-linked genes cannot be discounted. Differences in marker-specific inheritance patterns, effective population sizes, genetic drift, and sex-biased dispersal can generate disparate rates of gene flow across the genome and lead to differential introgression across markers [1].
Only one marker (diagnostic microsatellite marker Ammo006) exhibited narrower clines than the z-linked and mitochondrial markers. Based on annotation with the zebra finch genome, Ammo006 was found to be associated with a gene that codes for a mitogen-activated protein kinase (MAPK; 52). Specifically, the MAPK Table 2 Summary of v and u estimates from concordance tests. These parameters were estimated based on the comparison of each focal locus to a multilocus expectation using a logit-logistic model. Parameter estimates are presented for 24 microsatellite loci, 2 mitochondrial genes, 2 z-linked genes, and 1 autosomal locus. Perfect concordance between a focal locus and the multilocus expectation is a diagonal line (u = 0 and v =1)  superfamily consists of three distinct signaling pathways with roles linked to numerous cellular functions including immune responses, host-parasite interactions, and adaptive responses to thermal, osmotic, and oxygen stresses [68,69]. Of particular interest is the response of MAPK to osmotic stress, which has been documented in a range of organisms [69], including in mammalian kidney [70] and liver astrocytes [71] and in the osmosensory signaling pathways of fish (Fundulus heteroclitus; [72]). MAPKs therefore may have a critical role in salinity adaptation [69] and may serve an important role in osmoregulatory functions of A. caudacutus. The transition from upland and brackish habitat (nelsoni) to salt marsh (caudacutus) presents major adaptive challenges to terrestrial vertebrates [45], and adaptive divergence across this salinity gradient may thus play an important role in reproductive isolation between the species [73]. Pathways related to osmotic stress (i.e., MAPK) would arguably be under strong selection in this system. The MAPK gene region likely plays an important ecological role for A. caudacutus, which exhibits a pre-Pleistocene association with tidal salt marshes [37,47], and thereby a longer time to evolve adaptations to salt marshes compared to A. nelsoni, which exhibits a broader ecological niche, breeding also in grassland and brackish marshes and a more recent association with tidal marshes [38,51,74]. Restricted introgression of additional molecular and phenotypic traits provided evidence for selection on increased melanin in A. caudacutus, consistent with the hypothesized adaptive role for melanin in vertebrates that inhabit saltmarsh ecosystems [49]. Here we present two lines of support for this hypothesis. First, we found narrow cline estimates for the z-linked marker SLC45A2 (299 km) along with a more abrupt transition in slope compared to a multilocus average (+1.41). SLC45A2 (solute carrier family 45, member 2, protein) is associated with melanin-based pigmentation and has been linked to plumage phenotypes in birds, including silver and cinnamon colored phenotypes (Gallus gallus and Coturnix japonica; [75]) and the gray plumage of hooded crows (Corvus cornix; [76]). Similarly, mutations in SLC45A2 may relate to the differences in plumage coloration between caudacutus and nelsoni. A. caudacutus individuals have dark chestnut streaking patterns on the breast and flanks and dark chestnut backs, while A. nelsoni have gray streaking on the breast and flanks and more gray on the back [40,42,51]. Walsh et al. [42] found that plumage traits related to plumage darkness (particularly in the breast and flanks) are more strongly correlated with genotype. Secondly, we found that the introgression of traits related to plumage darkness was reduced (V obs /V max = 0.82) compared to traits related to streaking amount (V obs /V max = 0.46). Natural selection for the adaptive benefits of salt marsh melanism (including reduced predation risk and resistance to mechanical and bacterial degradation; [49,77,78]) may be reinforced by sexual selection [34]. The darkness of streaking in the breast and flanks may offer strong visual cues for individuals during mate selection.
We detected strong patterns of asymmetrical introgression across the 29 genetic markers, with 19 showing patterns of asymmetrical introgression toward A. caudacutus and 10 markers showing patterns of asymmetrical introgression toward A. nelsoni. A majority of these markers, including all of the markers that showed asymmetries toward nelsoni, displayed gradual slopes indicative of weak selection. Five markers exhibited abrupt clines and all of them showed patterns of asymmetry toward caudacutus. These findings are consistent with previous work suggesting that backcrossing is asymmetrical and biased toward A. caudacutus [40,41], possibly due to differences in mating systems [40] or population density.
Both species exhibit an unusual mating system among emberizines, characterized by non-territoriality, lack of male parental care, and high levels of promiscuity facilitating intense male-male competition for receptive females [51,79]. The two species differ in their mating tactics, however. Nelsoni males spend substantial time mate guarding and have more distinctive song and flight displays for attracting females [51,66,80]. Caudacutus males are highly polygamous and exhibit a scramble competition mating system whereby males search for and attempt to mate with multiple receptive females [37,79]. Size differences between nelsoni and caudacutus males (14.9 -19.2 g versus 19 -24 g, respectively) may thus place nelsoni at a substantial competitive disadvantage when competing with caudacutus males to secure mates in sympatric marshes. Admixed females are thus Table 4 Parameter estimates for the best fitting clines for 5 morphological traits using the package HZAR. For each trait, we present the best model, cline center (c), cline width (w), and AICc more likely to backcross with caudacutus males leading to asymmetries. This is particularly true in sites toward the southern portion of the hybrid zone, where caudacutus males outnumber nelsoni males by approximately 4:1 [67]. Cline estimates for weight coupled with a peak in weight variance near the center of the zone provide supportive evidence that size is an important factor in shaping zone dynamics in this system. The cline for weight was the most abrupt of the five morphological traits analyzed, indicative of strong selection against intermediately sized individuals, which may be ineffective in securing mates using either of the mating tactics (direct male-male competition or flight displays and mate guarding). Furthermore, variance in weight at the center of the hybrid zone exceeded variance in allopatry, which may be indicative of character displacement with smaller nelsoni and larger caudacutus in sympatric versus allopatric populations.
Asymmetrical introgression has been documented in a number of avian contact zones [81][82][83] and may also be indicative of hybrid zone movement or of one species being displaced by the other. Moving hybrid zones leave tails of clines of unlinked neutral markers in their wake, giving the appearance of asymmetrical introgression [30]. Distinguishing hybrid zone expansion from asymmetrical introgression poses a challenge, and is best addressed with temporally replicated sampling. However, multiple alleles introgressing in one direction offers additional support for zone movement [30,84]. Previous research has documented a potential southward expansion of nelsoni into the range of caudacutus, with nelsoni alleles documented as far south as Rhode Island [41]. Extensive field surveys also suggest a more pronounced decline in caudacutus abundance across their range in comparison to nelsoni (Correll et al., unpublished data); however, a direct temporal comparison of genetic data is required to test hypotheses of a hybrid zone expansion.

Conclusions
In conclusion, we found support for hybrid zones acting as semi-permeable boundaries to foreign alleles across a tidal marsh gradient. While a majority of the markers used for this analysis showed patterns of weak selection and uninhibited movement across the hybrid zone, we found evidence for strong selection for a few molecular markers and plumage characteristics, consistent with evolutionary processes contributing to reproductive isolation. Specifically, we detected reduced introgression of mitochondrial and z-linked markers, providing evidence for Haldane's rule, along with divergent selection for traits conferring adaptive benefits to tidal marshes. Despite the overall low genetic differentiation between caudacutus and nelsoni, niche differentiation may be driving ecological speciation between the species, with strong selective pressures for a few critical gene regions playing an important adaptive role. We conclude that adaptive divergence across a tidal marsh ecotone may promote isolating mechanisms and prevent the erosion of pure species boundaries in this system.

Geographic transect and sample collection
For this study, we used a previously published phenotypic and microsatellite genotypic dataset collected from the caudacutus-nelsoni hybrid zone [42], along with new DNA sequence data from two mitochondrial, 2 Z-lined, and one autosomal marker. The sampling design captured the extent of genetic and phenotypic variation across the hybrid zone, via 34 tidal marshes sampled along a linear, coastal transect from Lubec, Maine (44°4 9′22 N, 66°59′20 W) to Madison, Connecticut (41°15′ 46 N, 72°33′00 W; Fig. 1; Table 1) during the 2012 and 2013 breeding seasons (June -August). The marshes sampled for this study represent a gradient of tidal marsh habitat types (including coastal, bay, and upriver marshes) and exhibited a range of variation in salinity, vegetation structure, isolation, and tidal amplitude [48]. Within the hybrid zone, marshes were sampled approximately every 10 km; allopatric marshes were also sampled north and south of the hybrid zone (Fig. 1). We sampled 290 sparrows and collected a suite of morphological measurements from 254 of them [50]. Due to inadequate sample sizes (n = 2) in two of the sampled locations, we used data from 286 individuals from 32 sites for all analyses (Table 1; Fig. 1). Of these individuals, 37 sparrows were sampled from putatively pure nelsoni populations (n = 4 marshes), 52 individuals were sampled from putatively pure caudacutus populations (n = 6 marshes), and 197 individuals were sampled from sympatric populations (n = 22 marshes). Because this was the first extensive sampling and genetic evaluation of the hybrid zone, we initially defined all marshes outside of the currently hypothesized overlap zone of Hodgman et al. [39] to be putatively allopatric. We scored each individual sparrow for 13 plumage traits developed for evaluating levels of phenotypic introgression [40,42]. Briefly, the plumage scores capture basic phenotypic differences between the species and include the color of the bill, the color and definition of the face and back, the width and definition of the whisker line and crown, and the amount and definition of the streaking on the back and flanks. We used digital calipers to measure tarsus length and bill length (nares to tip; mm), a wing-chord ruler to measure unflattened wing chord (mm), and a digital scale to measure weight (to the nearest 0.1 g). We collected blood samples (10 -20 μl) from the brachial vein and transferred drops to Nobuto blood filter strips (Sterlitech, Kent, Washington) for storage at room temperature until later genetic analysis.

Analysis of molecular markers
The genotypic dataset consisted of 24 microsatellite loci, including 12 putatively neutral, anonymous microsatellite loci [85][86][87] and 12 diagnostic microsatellites developed specifically to differentiate caudacutus, nelsoni, and hybrids [52]. Information on these markers and PCR amplification conditions can be found in Walsh et al. [42]. Owing to elevated divergence in the diagnostic marker panel relative to the anonymous microsatellites (F ST = 0.4667 and F ST = 0.15, respectively; [52]), we considered these two marker sets separately for some analyses and interpretation. We also amplified each individual at two mitochondrial genes [1100 bp of NADH dehydrogenase subunit 2 (ND2); 356 bp of NADH dehydrogenase subunit 3 (ND3)], two z-linked genes [183 bp of solute carrier family 45, member 2 (SLC45A2), 724 bp of solute carrier family 30 (SLC30A5)], and one autosomal marker [900 bp of recombination activating gene 1 (RAG1); Primer information for these markers can be found in Additional file 6: Table S1]. PCR reactions included the following: 3 μl of eluted genomic DNA, 0.5 μM of each primer, 2.0 mM MgCl 2 , 1X PCR buffer (Promega, Madison, Wisconsin), 0.12 mM of deoxyribonucleotides, and 1 unit of Taq DNA polymerase (Promega). Cycling conditions were as follows: 35 -40 cycles of 94°C for 30 s, 46-60°C for 45 s, 72°C for 90 s, and a final extension step at 72°C for 5 min. For the two shorter fragments (ND3, SLC45A2), we sequenced all individuals sampled along the geographic transect; sequences were visually inspected in 4Peaks (Nucleobytes, Amsterdam, NL) and aligned in Geneious Pro 4.7.6 (Biomatters Ltd, Auckland, NZ). We assigned sequences to one of two haplotypes (caudacutus or nelsoni) based on visual inspection of species-specific polymorphic sites identified in putatively allopatric populations. For the three longer fragments (ND2, SLC30A5, and RAG1), we sequenced a subset of 14 putatively pure individuals (based on morphology and microsatellite data) and designed a Restriction Fragment Length Polymorphism (RFLP) analysis to identify species-specific haplotypes in the PCR-amplified fragments. We digested amplified products in 25 μl reactions, with 10 μl template DNA, 0.2 μl of enzyme TseI, PstI, and MwoI (for ND2, SLC30A-5, and RAG-1 fragments respectively), and 2.5 μl of NEBuffer (New England BioLabs, Ipswich, MA, USA) and incubated according to manufacturer protocols. We resolved the resulting fragments on a 2 % agarose gel and assigned haplotypes (see Additional file 7: Table S2 for protocols and fragment patterns). For SLC30A-5 and RAG-1, the RFLP method allowed us to assign individuals to one of three haplotypes (caudacutus, nelsoni, or hybrid) based on the combination of observed banding patterns. We checked the validity of the RFLP assay using the 20 sequenced individuals (see above) and found no discrepancy between RFLP and sequence haplotypes.

Population structure
For each site, we characterized genetic diversity using standard population genetic metrics. Specifically, we calculated unbiased estimates of expected and observed heterozygosities and tested for deviations from Hardy-Weinberg equilibrium in GENEPOP V4 [88]. We also calculated genetic diversity metrics, including F IS , number of alleles, and allelic richness in FSTAT [89]. To quantify patterns of admixture for each site, we estimated a hybrid index (proportion of caudacutus alleles in an individual) and interspecific heterozygosity (proportion of an individuals' genome with alleles inherited from both parental populations) for all individuals using the R package introgress [90,91]. To identify markers under selection, we performed selection tests for all loci using an F ST outlier approach [92] in the program LOSI-TAN [93]. To test for genetic differentiation among populations, we calculated pairwise F ST values and performed significance testing using 1000 permutations in FSTAT. To characterize genetic structure of the caudacutus-nelsoni hybrid zone, we used the Bayesian clustering approach of STRUCTURE, version 2.3.2 [94]. STRUCTURE uses membership proportions to assign all individuals into appropriate population clusters (K). We conducted five runs for each value of K = 1-5; each run consisted of a 300,000 burn-in followed by 200,000 iterations. We also ran STRUCTURE with the 12 anonymous microsatellites separately to ensure that the geneassociated diagnostic markers were not biasing the results. Because we detected the same patterns with both marker sets, we ran all subsequent analyses with the full set of 24 microsatellites. We used the admixture model and assumed correlated allele frequencies [95]. We determined the most likely number of population clusters (K) using the ΔK method of Evanno et al. [96]; STRUCTURE output was visualized using the program STRUCTURE HARVESTER [97]. Lastly, we tested for linkage disequilibrium using GENEPOP. Pvalues for multiple comparisons were adjusted with the Bonferroni correction.

Patterns of differential introgression: genomic clines
We used the R-package introgress to estimate genomic clines for each locus using a multinomial regression to estimate individual clines for each locus along an admixture gradient (represented by the hybrid index, calculated as the proportion of caudacutus alleles in an individual in introgress). Calculating hybrid index requires a priori definition of pure individuals of each parental species. In doing so, we took a conservative approach to minimize the potential for including introgressed individuals in our parental samples; we defined pure individuals as those sampled from allopatric populations >115 km north and south of the currently recognized overlap zone ( Fig. 1; [42]). To identify loci that displayed deviations from neutral expectations, we compared the likelihoods of the regression models to a null model of neutral introgression. Null models were generated using parametric simulations described in Gompert & Buerkle [90]. Using this approach, a large simulated admixed population is generated based on expected genotype frequency distributions (estimated using hybrid index and heterozygosity values equal to the observed data). We simulated 2000 admixed individuals and adjusted all significance thresholds using the false discovery rate procedure [98]. Deviations from neutrality were summarized as either gradual clines (homozygote excess or deficit and/or heterozygote excess) or abrupt clines (heterozygote deficit indicative of disruptive selection or assortative mating; 90).
To test for concordance among genomic clines, we compared genomic clines at individual loci against the multilocus expectation using the logit-logistic model of Fitzpatrick [99]. This approach compares the mean hybrid index over all loci to a hybrid index for a focal locus. The logit-logistic model estimates two parameters: u gives the relative difference in cline position (positive values indicate a shift of the cline toward caudacutus and negative values indicate a shift toward nelsoni) and v gives the relative difference in slope (values greater than one indicate a more abrupt slope and stronger selection whereas values less than one indicate a more gradual slope and weaker selection compared to the multilocus average). Perfect concordance between a focal locus and the mean hybrid index would result in u = 0 and v =1. Thus, the expectation for equal introgression over all loci lies on the diagonal. We fit the parameters u and v using the function "mle2" in the R package bblme.

Patterns of differential introgression: geographic clines
To evaluate the distribution of caudacutus and nelsoni alleles across the transect, we used the Metropolis-Hastings Markov chain Monte Carlo algorithm employed in the R package HZAR [100,101] to fit a series of geographic cline models to allele frequencies for each genetic marker and a suite of morphological traits. We reduced the variation observed in the 24-microsatellite loci to a two-allele system using species-specific compound alleles [29,102,103]. Using this approach, each allele was assigned to a species group based on its coordinates on the first axis of a multiple correspondence analysis (MCA), conducted using the CA package in R. We ran fifteen separate models for each genetic marker, all of which estimated cline center (distance from sampling location 1, c) and width (1/maximum slope, w). The tested models included all possible combinations for fitting tails (none fitted, left only, right only, mirror tails, or both tails estimated separately) and for estimating allele frequencies at the cline ends (pMin, pMax; fixed to 0 and 1, observed values, or estimated values).
To compare introgression patterns between genetic and phenotypic data, geographic clines were also fitted to five morphological traits, including: bill length, wing chord, weight, and two separate groups of plumage traits. We used plumage traits predominantly related to 1) the amount of streaking and the width of plumage features observed on an individual (including crown width, malar width, face definition, streaking amount on the breast, flanks, and back, and color of the back) and 2) traits predominantly related to the darkness and definition of plumage traits on an individual (including bill darkness, face color, and definition of plumage on the crown, malar, breast, and flanks). Traits related to plumage streaking amount and those related to plumage darkness and definition were previously found to differ with respect to their correlation with genotype [42]. We ran five separate models for each morphological trait, all of which estimated trait mean and variance (right, left, and center) along with cline center and width; models varied in how the tails were fitted. We compared all models using Akaike information criterion corrected for small sample sizes (AICc) and considered models with the lowest AICc score as the best-fitting model [104].