Hybrid linkage map

We found evidence for a fusion between two RBT chromosomes in the hybrids. Rainbow trout chromosomes Omy20 (a metacentric chromosome in RBT) and Omy28 (an acrocentric chromosome in RBT), were fused into a single linkage group, RYHyb20, in both sexes (Figure 1). Several loci mapping to Omy20 and Omy28 in RBT[21, 33] did not recombine in the F₁ hybrids, indicating that a major portion of RYHyb20 was inherited as a single, non-recombining block of markers. Additional mapping in a male YCT (48 progeny) indicated that the Omy20 and Omy28 homologues were fused in YCT.

Linkage groups RYHyb25 and RYHyb29 (homologues to Omy25 and Omy29, respectively, in RBT) differed in arrangement among parents. The distal mapping loci OMM1301 (on RYHyb25) and Ogo2UW/ii (on RYHyb29) were not linked in Female 1 and Male 2, indicating two independent linkage groups (Figure 2A). However, there was no recombination between these loci in Female 2 and Male 1, suggesting RYHyb25 and RYHyb29 were fused in these parents (noted as RYHyb25_29). The recombination estimate between OMM1301 and OMM1797 in the progeny of a male YCT (Θ = 0.46, LOD = 0.03) confirmed that the RYHyb25 and RYHyb29 homologues are not fused in YCT. We found a similar recombination estimate between the same loci in Male2 (Θ = 0.45, LOD = 0.08), but not in Male 1 (Θ = 0.0, LOD = 12.94).

Marker order was largely conserved among parent-specific and between female- and male-merged maps (Additional file1, Worksheets 2 – 8; Additional files2 and3). However, we observed inconsistent marker order among parent-specific maps for RYHyb25 and RYHyb29 and, therefore, these two linkage groups could not be merged within sex. Further, marker order differed between sexes for distally mapping loci at RYHyb13p, RYHyb19, and RYHyb24. Loci with different order between sexes were removed prior to generating the sex-merged map, and removal did not appear to alter marker order within linkage groups. Marker order differences could be due to marker informativeness within families, insufficient chromosome coverage, or reduced recombination in males.

Assignment of species diagnostic markers to linkage groups

We assigned 114 diagnostic species markers (97 SNPs, 13 indels, and 4 restriction fragment length polymorphisms (RFLPs)) to specific linkage groups (Additional file1, Worksheets 9). Specific mapping locations within linkage groups could not be determined because the F₁ parents were heterozygous at these loci. One indel, OCC-37, was polymorphic within YCT and mapped to a specific location within RYHyb20. All hybrid linkage groups contained at least one species-specific diagnostic marker. Linkage groups RYHyb21 and RYHyb11, both metacentric chromosomes, were each assigned a single diagnostic species marker indicating that only one of the two chromosome arms in both linkage groups had a species-specific marker assigned.

Homeologies

Using duplicated loci, we observed six homeologous hybrid linkage groups (RYHyb12q/RYHyb13q, RYHyb13p/RYHyb17p, RYHyb14p/RYHyb29, RYHyb15q/RYHyb21q, RYHyb10q/RYHyb19p, and RYHyb06p/RYHyb26) that have been previously identified within RBT[34]. We also observed one homeologous pairing that has not been observed in rainbow trout (RYHyb03 centromere region/RYHyb22q). Additionally, for 10 other duplicated loci, only one homeologue could be scored confidently. Linkage groups RYHyb13q and RYHyb19p each contained two of these loci, and RYHyb02p, RYHyb02 centromere region, RYHyb10q, RYHyb10 centromere region, RYHyb18p, and RYHyb21q each contained one of these loci.

Pseudolinkage

We found eight pseudolinkage groups exclusive to both male maps (RYHyb01/RYHyb23, RYHyb02/RYHyb03, RYHyb06/RYHyb26, RYHyb07/RYHyb18, RYHyb10/RYHyb19, RYHyb15/RYHyb21, RYHyb12/RYHyb13, and RYHyb13/RYHyb17). Pseudolinkage was represented in males by statistical linkage between markers that mapped to independent linkage groups in the two female maps. All cases of pseudolinkage were between chromosomes identified as homeologous within RBT[34].

Recombination rates

We identified 13 linkage groups where recombination rates differed significantly between the female consensus F₁ hybrid and RBT maps (Additional file4). We treated RYHyb20 as two separate linkage groups, noted as RYHyb20(Omy20) and RYHyb20(Omy28), because Omy20 and Omy28 are not fused in RBT. Further, RYHyb25 and RYHyb29 represent a fusion/fission polymorphism in the female maps, and so we treated these groups as independent in Female 1 and used the RBT linkage group in Guyomard et al.[33] for comparison. We also treated RYHyb25_29 as a metacentric linkage group in Female 2 and used the map of Rexroad et al.[21] for comparison.

We found five instances of suppressed and one instance of elevated recombination distance across large numbers of loci within several linkage groups in the female hybrid map relative to the female RBT map (Additional file4) (Figure 3). All instances of recombination suppression mapped across centromeres. Linkage group RYHyb11, a metacentric chromosome in RBT, was the shortest linkage group in the female hybrid map. The recombination distance across the hybrid map relative to the RBT map (relative hybrid:RBT recombination distance) was 0.21, indicating the recombination rate was suppressed across all loci mapped in RYHyb11. Linkage group RYHyb20(Omy20) had a relative hybrid:RBT recombination distance of 0.46. The reduction in map distance appeared to be due to six non-recombining loci in the hybrid map, whereas the same six loci mapped over 40 cM to the p- and q-arms in the RBT map. In contrast, the relative hybrid:RBT recombination distance on the other arm of the hybrid map, RYHyb20(Omy28), appeared elevated across the four loci closest to the telomere in the hybrid map (2.44). However, the recombination distance appeared suppressed in the hybrid map between the two markers closest to the centromere. Recombination suppression near the centromere was likely associated with the fusion to RYHyb20(Omy20). Linkage group RYHyb25_29 had a relative hybrid:RBT recombination distance of 0.59; however, we were unable to include the majority of the q-arm in the comparison. We observed five non-recombining loci in the hybrid map, whereas the same five loci mapped over 40 cM in the RBT. In contrast, the single-armed linkage groups RYHyb25 and RYHyb29 had relative hybrid:RBT recombination distance of 0.82 and 1.12, respectively. The reduced relative map distance in RYHyb25_29 compared to RYHyb25 and RYHyb29 suggests that the fusion between the two chromosome arms caused a reduction in the recombination rate across the centromere. Linkage group RYHYb15 had a relative hybrid:RBT recombination distance of 0.66. Although the hybrid map had a reduced recombination distance across the centromere compared to the RBT map, the recombination distance was greater in the hybrid map than in the RBT map near telomeres on both chromosome arms. Linkage group RYHyb14 had a relative hybrid:RBT recombination distance of 0.59; however, we were unable to include the majority of the q-arm in the comparison.

Segregation distortion

Analyses based on a sliding window revealed that 27 of the 286 total 25 cM intervals among all parent-specific maps contained YCT allele frequencies that deviated significantly from Mendelian expectations (Additional file2) (Table 2). The proportion of 25 cM intervals deviating significantly within each parent-specific map was as follows: Female 1, 9/104; Female 2, 8/107; Male 1, 4/35; and Male 2, 6/40. We found no consistent distortion in allele frequencies among all parents. However, both females showed a significant reduction in YCT allele frequencies in the same region on RHYb01p and both males showed a significant increase in YCT allele frequencies across both RYHyb09 arms (Figure 4). Interestingly, homeologues within pseudolinkage groups RYHyb07/RYHyb18 and RYHyb06/RYHyb27 had inverse proportions of YCT allele frequencies in Male 2. Linkage group RYHyb07 had significantly less, and homeologue RYHyb18 had significantly greater, YCT allele frequencies than expected. Similarly, RYHyb27 had significantly less, and the homeologue RYHyb06 had a trend for greater (although not significantly so), YCT allele frequencies than expected.

We found 15 linkage groups where genotypic segregation distortion occurred at P < 0.05 (Additional file5). Approximately 6% of the total loci genotyped were distorted at this significance level. After applying B-Y FDR corrections for multiple tests, the observed genotypic frequencies differed significantly from expected frequencies within only four linkage groups (Additional file5), and represented approximately 1.6% of the total loci genotyped. Linkage groups RYHyb09 and RYHyb18 contained loci with an excess of YCT/YCT genotypes, and RYHyb06 and RYHyb07 contained loci with an excess of RBT/RBT genotypes. Interestingly, none of the 309 loci mapped in common between the two families showed significant genotypic distortion within both families.

Chromosome rearrangements between RBT and YCT

The hybrid map is a product of differences in chromosome rearrangements between RBT and YCT, as well as mixed ancestry of the RBT source stock. Using data from karyotypes[27, 28] and the hybrid map, we suggest that RBT and YCT differ by at least five species-specific chromosome rearrangements. We first consider the fusion between Omy28 and Omy20q in YCT. The most parsimonious explanation for the fusion between the metacentric Omy20 and the acrocentric Omy28 involves two rearrangements. The first rearrangement would have occurred prior to the Omy20-Omy28 fusion as a pericentric inversion of the entire q-arm of Omy20. This inversion would have resulted in Omy20 becoming an acrocentric chromosome and a reduction in the expected 52 haploid chromosome arms[27, 28] to the observed 51 haploid arms in the hybrid map. The possibility that Omy20q could be inverted in YCT relative to RBT is supported in Chinook salmon (O. tshawytscha), as the Omy20 homologue in Chinook is an acrocentric chromosome and the segment homologous to Omy20q appears inverted compared to RBT (unpublished observations, K. Naish). The second rearrangement would have been a Robertsonian type involving centric fusion between the acrocentric Omy20 homologue and Omy28, conserving chromosome arm number. The three remaining differences may be explained by comparing acrocentric chromosome numbers between RBT (seven in the 60 chromosome race) and YCT (twelve). The seven RBT acrocentric chromosomes have been identified as such in the hybrid map, suggesting that they have an acrocentric YCT homologue. The remaining five YCT acrocentric chromosomes likely have homologs with RBT metacentric chromosomes. The third and fourth rearrangement differences probably represent centric fusion/fission events between two RBT metacentric and four of the five remaining YCT acrocentric chromosomes, which would not change the chromosome arm number. The fifth rearrangement difference may have involved a fission event within a RBT metacentric chromosome arm, generating the fifth acrocentric and a submetacentric chromosome in YCT. Fission within a RBT metacentric arm would gain one YCT chromosome arm and restore the 52 haploid chromosome arm number, countering the loss of a chromosome arm by the pericentric inversion.

The variable number of linkage groups identified in the hybrid maps (28 and 29) is likely due to mixed ancestry of the source RBT stock. Linkage analysis in YCT indicated that the Omy25 and Omy29 homologues each represent acrocentric linkage groups in YCT, but these chromosomes are known to be a Robertsonian polymorphism in RBT[28, 34, 35]. The Kamloops stock at Hayspur Hatchery appears to include ancestry from inland RBT (2 N = 58; Omy25 and Omy29 are fused as a metacentric chromosomes) as well as the common hatchery RBT derived from coastal California (2 N = 60; Omy25 and Omy29 are acrocentric chromosomes) (R. F. Leary, Montana Fish, Wildlife and Parks, personal communication). Admixture between the two RBT stocks would generate Robertsonian metacentric (2 N = 58) and acrocentric (2 N = 60) polymorphs, as well as Robertsonian heterozygotes (2 N = 59). Subsequent hybridization between the polymorphic Kamloops stock and YCT would produce F₁ hybrids that were Robertsonian heterozygotes (for example, Female 2 and Male 1) comprised of 28 linkage groups, containing the RBT metacentric fusion Omy25_Omy29 and the YCT acrocentric homologues to Omy25 and Omy29, as well as Robertsonian acrocentric homozygotes (for example, Female 1 and Male 2) comprised of 29 linkage groups, containing RBT acrocentric chromosomes Omy25 and Omy28 and their YCT homologues (Figure 2).

Recombination suppression

Conclusions based on recombination differences between the hybrid and RBT maps have limitations, as rates can vary between related species[36], as well as among individuals within species[18, 19, 24]. Indeed, we found significant differences in pairwise recombination rates between the two RBT maps[21, 33] used to construct the female consensus map across Omy5, Omy8, Omy14, Omy19, Omy20, and Omy22. Furthermore, differences in marker density among hybrid and RBT maps could also account for recombination differences by affecting map distance estimates. However, our objective was to perform a comparative analysis to identify recombination suppression broadly across the genome. Because recombination frequency appears correlated with chromosome arm number[37] and broad-scale recombination rates tend to be conserved between closely related species[36], we might expect similar recombination rates between RBT and YCT. Therefore, although absolute differences between map distances can be affected by number of markers and number of individuals mapped, the trend revealed by the comparative analyses yields interesting insight into recombination suppression in the hybrids, which could indicate the presence of chromosome rearrangements or genic incompatibilities.

Chromosome rearrangements can generate recombination suppression in heterokaryotypes (chromosomal hybrids) through unbalanced gametes which results in non-recombinants being the only viable gametes[38], or by restricting recombination between rearrangements[39]. Although unbalanced gametes cannot be ruled out, we consider them an unlikely cause of suppression in the F₁ hybrids. First, conservation of synteny and marker order between the hybrid and RBT maps[21, 33] suggests that YCT and RBT chromosomes are highly collinear, which would facilitate normal pairing between homologues and alternate disjunction in F₁ hybrids. Second, because RBT and YCT contain the same number of chromosome arms, Robertsonian type rearrangements are considered to have played a significant role in generating the chromosome number differences between the species[34]. Meiotic pairing between homologues that differ by a Robertsonian rearrangement (i.e., pairing between a metacentric chromosome and the two acrocentric homologues) would produce a trivalent in the F₁ hybrids, which would not necessarily cause malsegregation and unbalanced gametes. For example, the rate of nondisjunction was not different in individuals that were heterozygous for Robertsonian rearrangements compared to homozygotes in pink salmon (O. gorbuscha)[40], house mouse (Mus musculus domesticus)[41], and Eurasian common shrew (Sorex araneus)[42], suggesting this type of rearrangement produces balanced gametes.

Recombination was restricted in the F₁ hybrids across several loci spanning chromosome fusion and fission differences between RBT and YCT; the Robertsonian fusion/fission within RYHyb25_29 and RYHyb20. This suggests that chromosome rearrangements did indeed suppress recombination. Rearrangements generate extensive linkage disequilibrium in heterokarotypic hybrids[3] and suppress recombination across genomic regions that extend beyond rearrangements[11, 30, 43]. Therefore, broad-scale recombination suppression across other linkage groups in the F₁ hybrids could indicate the presence of rearrangements. We suggest that three other metacentric hybrid linkage groups, RYHyb11, RYHyb14, and RYHyb15, contained chromosomes that differed by arrangement between RBT and YCT. These linkage groups might represent Robertsonian rearrangements because recombination was suppressed across putative centromeres which could indicate that F₁ hybrids were heterozygous for centric fusions/fissions. However, we found marker order differences within the suppressed regions in each of these three linkage groups compared to RBT linkage maps[21, 33], which could possibly indicate inversions or translocations, or, alternatively, be due to reduced mapping power in regions with low recombination or the number of progeny used to construct the various maps.

Significant differences in recombination rates between several hybrid and RBT linkage groups could indicate the presence of inter-specific, genic incompatibilities. However, our broad-scale analysis hinders inference across smaller genomic scales where incompatibilities have been reported[44–46]. Nevertheless, we found numerous instances where two adjacent markers did not recombine in the female-merged hybrid map, but these same markers recombined in the consensus female RBT map. This might suggest that recombination between particular YCT and RBT genomic regions is maladaptive, or that the difference in adjacent recombination is an artifact of the number of offspring genotyped. Finer scale mapping across these regions that differ in recombination rate could be fruitful for identifying the presence of incompatibilities. Inter-specific incompatibilities may also be inferred from segregation distortion, as genotypes that occur less often than expected may be incompatible[3, 14]. In addition, we expected that inter-specific incompatibilities would show consistent distortion among maps. For example, consistent allelic distortion within both RYHyb01 female hybrid maps and both RYHyb09 male hybrid maps could indicate the presence of incompatibilities. Although segregation distortion was present in several other linkage groups, a lack of consistency across hybrid maps suggests these distorted loci do not reflect genic incompatibilities between RBT and YCT.

Recombination suppression within Robertsonian rearrangements

The recombination suppression pattern across the Robertsonian rearrangement RYHyb25_29 and the presumed Robertsonian rearrangement RYHyb15 differed from patterns reported between chromosome races in the house mouse[47, 48] and common shrew[39]. Recombination in mice and shrews that were homozygous for Robertsonian fusions (i.e., homozygous for the metacentric polymorphism) appeared suppressed near centromeres and elevated toward telomeres in comparison to Robertsonian heterozygotes[39, 47]. However, we observed the opposite; recombination was suppressed in Robertsonian heterozygotes (i.e., hybrids) across the putative centromere and elevated toward telomeres in comparison to fusion homozygotes (i.e., the female RBT consensus map). The mechanisms causing the difference between our results and the house mouse and common shrew are unclear. Given the approximate 3 million year divergence time between RBT and YCT[26] and the absence of historical secondary contact, genomic differences could have evolved near the centromeres of rearranged chromosomes that would restrict crossover events in Robertsonian heterozygotes, such as para- or pericentric inversions or genic incompatibilities. Alternatively, mechanistic processes governing meiotic crossover could differ among taxa. Determining how recombination suppression in the Omy25-Omy29 Robertsonian rearrangement differs between intra-specific (RBT chromosome races) and inter-specific hybrids could indicate the efficiency of this type of rearrangement as a barrier to gene flow in salmonids.

Rearrangements protect genomic regions from recombination

Our results suggest that chromosome rearrangement is the main genomic obstacle for gene exchange between RBT and YCT. Rearrangements have been observed to reduce gene flow between several species pairs, including Drosophila pseudoobscura and D. persimilis[43], Helianthus petiolaris and H. annuus[10], and S. araneus and S. antinorii[49, 50]. Rearrangements protect genomic regions from recombination, enabling genes within or closely linked to the rearranged genome to differentiate between heterokaryotypes while unrestricted gene flow occurs between regions with similar composition[45, 51]. As a result, fitness related genes could accumulate within regions that are protected from recombination and diverge in the face of hybridization[45, 51]. Indeed, genes involved with reproductive isolation have mapped to chromosome rearrangements[31, 32]. The effectiveness of rearrangements as recombination suppressors may be dependent on how chromosomes are reorganized; rearrangements that change gene order (e.g., inversions or translocations) may be more effective in protecting the genome from being disrupted than rearrangements that do not change gene order (e.g., fusions or fissions). However, several studies indicate that Robertsonian type rearrangements restrict gene flow[11, 49, 50, 52], but with extreme interbreeding this type of rearrangement may be an ineffective barrier[11].

Given the hybrid linkage map results, we predict that, within introgressed populations, inter-specific recombination will be restricted in particular genomic regions where chromosome arrangement differs between YCT and RBT. This prediction is supported by studies that have reported reduced gene flow across chromosome rearrangements within stable hybrid zones[10, 11, 50, 52]. Because the majority of hybrid linkage groups appeared to have similar marker order and arrangement with RBT, few chromosomes would be expected to contain regions with restricted recombination, although suppression could potentially extend broadly across these chromosomes[11, 30, 43]. Within a long standing hybrid zone between house mouse chromosome races, linkage disequilibrium persisted among loci mapping near Robertsonian rearrangements[11], indicating that these types of rearrangements can generate and maintain linkage disequilibrium. Given that RBT-YCT introgression is recent and rearrangements produce extensive linkage disequilibrium[3], large genomic regions that flank centric fusion/fission differences could be expected to remain intact within some YCT chromosomes. As a consequence, recombination suppression between heterokaryotypes could protect genomic regions from being disrupted and enable co-adapted gene complexes and/or local adaptations linked to the rearrangements to persist within YCT admixed with RBT.

Segregation distortion mechanisms

The low level of segregation distortion observed in the RBT-YCT hybrid maps was unexpected. Several studies have reported high levels of segregation distortion, greater than 15% of loci at P < 0.05, within hybrid maps across a variety of taxa, including interspecific crosses between Mimulus guttatus and M. nastutus[8], Lepomis cyanellus and L. megalotis[53], and Nasonia giraulti and N. vitripennis[54] and between intraspecific crosses of Salvelinus alpinus[55], M. guttatus[56], Ceratodon purpureus[57], Coregonus clupeaformis[58], and Tigriopus californicus[59]. Divergence time estimates between parental lineages used for several of these maps ranged from 0.1-0.2 MYA to 13–16 MYA[60–63]. We would expect comparable distortion levels to these studies following the estimated 3 million year divergence time between RBT and YCT. Our findings are probably not due to a lack of power for detecting distortion, because we mapped a similar number of markers per linkage group as the studies above, although the number of individuals per mapping panel was less than these studies, except Woram et al.[55]. Nevertheless, it is intriguing that segregation distortion was limited and few consistent distortion patterns emerged between sexes and families. The low level of distortion suggests that few incompatibilities exist between RBT and YCT, and may partially explain why these species readily form hybrid swarms.

Although segregation distortion was limited, we found a few noteworthy cases. A variety of mechanisms may cause distortion[14, 64, 65]. Understanding the causes of distortion typically requires in-depth study, but we discuss two mechanisms. First, pseudolinkage likely caused allelic distortion within two homeologous linkage groups in Male 2 (RYHyb07/RYHyb18 and RYHyb06/RYHyb27). Pseudolinkage may be implicated when homeologous linkage groups both show allelic segregation distortion[14], and we observed this distortion in YCT allele frequencies. Second, meiotic drive could have caused distortion. Meiosis is asymmetric in females and results in one functional haploid gamete per germ cell, compared to males where symmetric meiosis results in four functional gametes per germ cell[64]. Therefore, any process that results in non-random segregation of chromosomes during oogenesis may lead to distortion in females. Female meiotic drive is characterized by competition among centromeres for spindle fiber attachment during meiosis and oogenesis[64]. Because the centromere on one chromosome may outcompete the centromere on the homologue for deposition into the oocyte, allele frequencies in female maps may be distorted at loci mapping near centromeres[66]. Male meiotic drive is characterized by competition between alleles during sperm development and results in differential gamete success. Genomic divergence between species could escalate competition between meiotic drive elements, causing segregation distortion in their hybrid progeny[67]. Consistent distortion patterns within RYHyb01 between female hybrid maps and within RYHyb09 between male hybrid maps suggest that meiotic drive could be acting within the hybrids.

Sex chromosomes

Conservation of synteny with the RBT sex chromosome, Omy1, suggests that RYHyb28 is the sex-linkage group. Indeed, sex has been established as mapping to homologs in YCT and RBT[68]. Conservation of sex-linkage groups is uncommon across several salmonid species[20, 69–71]. Two alternative mechanisms have been proposed to account for the lack of conservation among sex-linkage groups in salmonids: either the sex-determining gene is the same among species but has moved to different chromosomes in different lineages, or sex determination has evolved independently in different lineages[69]. Regardless of the mechanism, the lack of homology among sex chromosomes could be an important factor for restricting inter-specific hybridization, and could explain why hybridization is not commonly observed between many sympatric salmonid species and, conversely, why two species pairs identified as having homologous sex-linkage regions, YCT and RBT[68] and Arctic charr (S. alpinus) and brook charr (S. fontinalis)[72], introgress in the wild[73, 74]. The relationship between conservation of sex chromosomes and introgression is confounded in brown trout (Salmo trutta) and Atlantic salmon (S. salar) because their sex chromosomes lack homology[70], yet the species hybridize in the wild[75]. However, brown trout and Atlantic salmon hybrids suffer reduced viability[76] and introgression is rare[75]. Nevertheless, homology between sex-linkage groups could possibly facilitate introgression between RBT and YCT.

Mapping families

Hybrid F₁ YCT-RBT parents were generated by crossing female YCT collected from Henry’s Lake Fish Hatchery and Fish Management Station, Idaho Department of Fish and Game (IDFG), with male RBT (Kamloops stock) from Hayspur Hatchery (IDFG). Mature F₁ hybrids were collected at Henry’s Lake in March of 2004 and used to generate two F₂ hybrid full-sib crosses (Family 1, N = 54 mapping progeny; Family 2, N = 53 mapping progeny). Because a YCT genetic map does not exist, we also constructed a YCT cross (N = 48 mapping progeny) so that we could clarify linkage anomalies observed between the F₁ hybrid and published RBT maps[21, 33]. Fin tissues were sampled from the parents and we confirmed the hybrid/species status of F₁ hybrid and YCT parents by screening 12 species-specific markers that differentiate RBT and YCT (Additional file1, Worksheet 1). Crosses were reared 10 months post-fertilization at which point the fish were euthanized and fin tissues sampled. DNA was extracted using DNeasy kits (Qiagen Inc., Valencia, CA, USA).

Mapping markers

Rainbow trout genetic linkage maps developed by Guyomard et al.[33] and Rexroad et al.[21] served as templates for F₁ hybrid map construction. We applied 294 microsatellite primers (Additional file1, Worksheet 1), spanning a large portion of each RBT linkage group and ensuring coverage across the centromere for metacentric chromosomes. Microsatellite amplification and PCR product visualization followed the methods of McClelland and Naish[20].

We also applied 169 SNPs (Additional file1, Worksheet 1). SNPs were interrogated using TaqMan 5’ nuclease assays (Applied Biosystems Inc., Carlsbad, CA, USA) or SNPtype assays (Fluidigm Corporation, San Francisco, CA, USA). All genotyping was carried out in 96.96 Dynamic Genotyping Arrays on an EP1 Genotyping System (Fluidigm Corporation), with a pre-amplification step, following manufacturer’s protocols. Genotypes were determined using the Fluidigm SNP Genotyping Analysis software (v 3.0.2), with confidence threshold set to 80%.

In addition, 14 insertion/deletion and four RFLP species-diagnostic primers (Additional file1, Worksheet 1) were included in the map. Amplifications were performed in 20 μl reaction volumes consisting of 15 ng genomic DNA, 1X NH₄ Reaction Buffer (Bioline, Taunton, MA, USA), 1.5-2.5 mM MgCl₂, 200 μM each dNTP, 1.5 pmol of each primer, and 0.5 units Taq polymerase (Bioline, Taunton, MA, USA). PCR products were visualized on 2-4% agarose gels stained in ethidium bromide.

Linkage analysis

Linkage maps were established using LINKMFEX v2.3 software package[14] with an LOD threshold at 3.0. Because male salmonids show less recombination across the genome than females[24], we initially constructed parent-specific linkage maps. Hereafter, parent-specific linkage maps are referred to as Female 1 (Family 1 female parent), Male 1 (Family 1 male parent), Female 2 (Family 2 female parent), and Male 2 (Family 2 male parent). Salmonids have high crossover interference and typically have one or no crossovers per chromosome arm[77]; we therefore used the pairwise recombination fraction, theta (Θ), between adjacent markers to estimate map distances. Male and female maps were compared by generating sex-specific and sex-merged maps using LINKMFEX. The total number of linkage groups in the F₁ hybrid map was determined from sex-merged maps. Linkage maps were graphically represented using the program MAPCHART[78]. Markers that were heterozygous for the same alleles in both parents were excluded from the map because these parental genotypes reduce the number of informative progeny, generate missing data, and reduce mapping power for these markers. All species-diagnostic SNPs, indels, and RFLPs were heterozygous in both of the F₁ hybrid parents, and thus these markers were not ordered within the map. However, we did assign each of these markers to a specific linkage group.

Recombination rate analyses

We estimated the average recombination ratio across the genome and across each linkage group between parents within each family and between parents of the same sex using LINKMFEX. Significant differences in genome-wide and linkage group-wide recombination rates were identified by summing G-test values and degrees of freedom across each comparison.

To determine if introgression suppressed recombination rates, we generated a consensus female RBT map from Guyomard et al.[33] and Rexroad et al.[21] using recombination distances between markers and compared recombination rates to the female-merged F₁ map. The consensus female RBT map was generated using only markers in common with the female-merged F₁ hybrid map (Additional file1, Worksheets 10). We ensured that maps had the same marker order among RBT and female-merged F₁ hybrids by removing markers that differed in their rank in the order. Markers used to compare recombination rates generally covered a substantial proportion of each RBT linkage group and were not restricted to areas of low recombination (centromeric regions) or areas of high recombination (telomeric regions). Kosambi map distances from Rexroad et al.[21] were converted to Θ map distances by applying the formula θ = 0.5(e^4k − 1)/(e^4k + 1)[79], where k = Kosambi distance. Prior to generating the consensus female RBT map, we tested for significant differences in recombination distances between adjacent loci between the two female RBT maps for each linkage group. The two female RBT maps were merged into the consensus map using LINKMFEX. We then tested for significant differences in recombination rate between the female-merged F₁ map and the consensus female RBT map across linkage groups against the null hypothesis of no difference in recombination rate. Significant differences in the recombination rate were identified as described above.

Segregation distortion analyses

We tested for allelic distortion using a sliding window analysis, because loci exhibiting segregation distortion often cluster together within the genome[8]. We established a 25 centiMorgan (cM) interval for the sliding window, because linkage was not supported for loci greater than 25 cM in distance at the LOD = 3.0 threshold. Species-specific markers were used to specify the most likely F₁ parent chromosome phases as either YCT or RBT using LINKMFEX. Locus-specific allele frequencies were tested for deviation from 1:1 Mendelian expectations by summing G-test values and degrees of freedom within each 25 cM interval.

For testing genotypic segregation distortion, we used the most likely allele phases obtained from LINKMFEX in the sliding window analysis to assign locus-specific genotypes in the F₂ hybrid progeny as YCT homozygote (YCT/YCT), RBT homozygote (RBT/RBT), or heterozygote (YCT/RBT). Significant deviation from the expected 1:2:1 Mendelian genotypic proportions was determined for each locus by applying chi square tests followed by multiple comparison corrections using a false discovery rate (B-Y FDR)[80] across all loci.