Single nucleotide polymorphism (SNP) discovery in duplicated genomes: intron-primed exon-crossing (IPEC) as a strategy for avoiding amplification of duplicated loci in Atlantic salmon (Salmo salar) and other salmonid fishes

The diversification of the variety of molecular markers available has been an important development in the field of genetics over the past two decades [1], with one of the more recent additions to the 'molecular toolbox' being single nucleotide polymorphisms (SNPs): a variant of traditional DNA sequencing which potentially enables high-throughput analysis of numerous independent (mostly) bi-allelic DNA sequence polymorphisms. The increase in the range of molecular markers partly stems from the realisation that no particular marker type is ideal for all situations, and SNPs are no exception to this. Their beneficial features include having a relatively simple mutation model [2, 3] and a high abundance in the genome (see e.g. [4]). Furthermore, the fact that SNPs occur in coding regions enables assessment of polymorphisms potentially directly affecting the phenotype [5]. On the other hand, as SNPs are normally bi-allelic markers, more loci are needed to obtain sufficient statistical power in certain analyses (see e.g. [6]) and allele frequencies of SNPs are usually skewed in population level analyses [7]. In theory, the limited amount of information in a single SNP locus can be compensated by increasing the number of loci screened, and several high-throughput procedures have been developed to facilitate this need (see [5, 8]). Overall however, it is clear that SNPs are an important class of molecular markers for genomics research and can potentially be applied in a wide range of studies.

While the recognized benefits of SNPs have accelerated their use in studies of model organisms [9–13], the application of SNPs in genetic studies of wild species has been relatively rare. Their potential use in animal genetics has been reviewed by Vignal et al. [14], presenting the usefulness of SNPs in, e.g., parentage assignment, animal tagging (see also [15]) and especially in QTL mapping, but similar studies with natural organisms have only been reported recently, most likely due to a lack of suitable markers. Recently however, new SNP discovery strategies (e.g. [16]) have resulted in characterizations of SNPs in many natural populations of vertebrates to address several evolutionary, ecological and conservation issues. For example, SNPs have been applied for the identification of cryptic vole species [17], to investigate the level of genome introgression in a passerine bird hybrid zone [18, 19], and to study the population genetics of wolves [20].

SNP discovery in model organisms has primarily been performed by comparing genomic information of multiple individuals in the public databases in order to identify putative polymorphic sites (e.g. [21]). This has been a useful approach for species with a wealth of nuclear sequence data available, but is not a very feasible method for the majority of non-model organisms. In species with little published sequence data available, SNP identification has been carried out by sequencing random DNA fragments (e.g. [16, 22, 23]), or by using a targeted gene approach where primers have been designed in conserved regions of orthologous gene sequences from closely related species to amplify less conserved regions like introns, generally termed 'comparative anchor tagged sequences' ('CATS') or the 'exon-primed intron-crossing' ('EPIC') method (e.g. [16, 24–28]). Again, this latter type of SNP discovery may be challenging if the entire taxa of interest lacks published sequence data. In recent years, however, large-scale sequencing and EST projects have provided usable data for a great variety of new species and a particular increase can be seen in fish species, due in part to their potential use as models in environmental genomics [29] as well as the broad variety of species of aquaculture importance. The total number of SNPs needed to trace different strains within a species has been estimated with salmonid fish [30] and, even individual identification would be possible if the population and/or species-specific SNPs were characterized as has been conducted in wolves [20]. Furthermore, Glaubitz et al. [6] estimated in a simulation study that about five times more SNPs than microsatellites are needed to determine pair-wise genetic relationships.

Atlantic salmon (Salmo salar) has been one of the most widely studied fish species in recent decades due to its importance for aquaculture and conservation, but extensive SNP characterization studies with this salmonid species have been scarce. Recently, large-scale sequencing, EST and BAC library projects have made a large amount of data available in the public databases for S. salar (see [31–33]; and also [34–36]). These genomic resources have given rise to the first exploitation of EST databases for SNP discovery (unpublished data, Hayes et al). However, Hayes et al. (unpublished data) speculated that a proportion of the potential SNPs observed in S. salar EST sequences could in fact be a consequence of ancient duplication events in the salmonid genome, and some of the 2,507 putative EST-based SNPs found could actually be sequence differences between ancestral duplicates (i.e. paralogues) rather than true SNPs. Similarly, genome duplication has also been suggested to affect SNP identification in a recent study of Pacific salmon (Oncorhynchus tshawytscha, O. nerka, and O. keta) [37], where one-third of the analyzed loci were suggested to be paralogue sequence variants rather than true SNPs. Potentially sequence differences between duplicons rather than SNPs (see [38]) may emerge especially when the more highly conserved regions (i.e. exons) of the genes are used for primer design as this increases the risk of amplifying both paralogs of the same locus. Thus far, such exon-focused methods (e.g. EPIC) have been exploited in most of the SNP discovery surveys, and no study exists where the more variable, non-coding segments of the genes (e.g. introns) have been utilized to design specific primers aimed at binding to only one of the duplicated loci. This is probably due to fact that more than one sequence copy of the particular gene seldom exists in the databases and not much is known about the extent of potential duplicated genes in different species. However, intron sequences of one known duplicated gene in salmonids – growth hormone – have been used as a source of variation for phylogenetic and population studies [39–41], indicating that the divergence in introns could be sufficient for a discriminative priming strategy between putative paralogs in salmonid species.

The aim of this study was to characterize potential SNPs in the Atlantic salmon genome using gene sequence data for salmonids and other teleost species obtained from GenBank [42]. Initially, PCR primers were designed by utilizing the exonic regions of salmonid or other teleost species (EPIC). However, on observing that numerous duplicated genes had likely been amplified, a new method – termed intron-primed exon-crossing (IPEC) – was developed to circumvent this problem, whereby primers were designed in more variable intronic regions of salmonid genes. The feasibility of this new priming method to avoid amplifications of potential duplicated loci was evaluated, and the proportion of conserved duplicated loci assessed. Polymorphism was assessed by sequencing the fragments of individuals originating from 15 salmon populations spanning the species range. Furthermore, a subset of primers was tested with brown trout (Salmo trutta), arctic char (Salvelinus alpinus) and grayling (Thymallus thymallus) to investigate the usefulness of these loci to produce cross-species sequence data from other salmonids.

The results of this study have important implications for SNP discovery in non-model species with ancestrally duplicated genomes. Exon-targeted primers using sequence data from the same or closely related taxa which have previously been used in SNP characterization studies with non-model species [16, 22, 26, 37] were relatively unsuccessful in Atlantic salmon compared with the IPEC approach proposed here, where less conservative gene regions – i.e., introns – were the target sequences for primer design. The reduced success of the EPIC approach for SNP discovery is most likely due to the duplicated nature of the salmonid genome. This genomic duplication is suggested to have taken place in the ray-finned fish lineage after its divergence from tetrapods (reviewed in [43]) and additional, more recent polyploidization events have also been detected in the salmonid sublineage (see [44]). The subsequent re-diploidization event in the salmonid genome has generated duplicated paralogs, which may diverge from each other due to the relaxation of purifying selection in one of the copies (reviewed in [45]). Thus, assuming that diverged introns evolve even more rapidly due to lower selective pressure, the amplification of potential duplicates could be minimized by focusing on those regions for PCR primer design.

Indeed in S. salar, this new intron-targeted IPEC method clearly outperformed the widely used EPIC (or CATS) approach, which utilizes conserved gene regions in cross-species applications. The proportion of screened loci that revealed polymorphism in S. salar was around six times higher with the IPEC (36.4% polymorphic) than the EPIC (6.4%) method (Table 1a), suggesting that less effort is needed to yield the same number of SNPs than with the EPIC (or CATS) method [28, 37].

Recently, special interest has focused on identifying multi-site variation after duplication from ordinary SNPs in humans [46]. Studies with several salmonid species have also speculated that some of the observed polymorphic sites could actually be variation between retained paralogs of duplicated segments rather than true SNPs ([37, 38]; unpublished data, Hayes et al). The duplication presumably lowered the success rate of the EPIC primers, especially those designed in salmonid genes (Table 1a), but it may have a minor effect on the novel SNPs identified in this study as the IPEC method produced most of the polymorphic loci (71% in total, Table 1a–b). Therefore, this intron-focused approach should be a feasible method to avoid obtaining potential 'duplicated SNPs' when identifying novel polymorphic loci from the species bearing putative duplicated genomic fragments or even an entire duplicated genome.

The observed nucleotide diversity estimates over all loci in S. salar (3.99 × 10^-4) was highly similar to that in European humans [9] and about twofold lower than that observed in a larger scale survey with human genome [4]. On the other hand, the estimations are about ten times less than reported in birds [16] or plants [47, 48] and about three times less than reported in a recent study of the GH1 gene of S. salar [41]. The greater number of base pairs and the number of independent loci sequenced here most likely better represents the overall nucleotide diversity estimate of S. salar genome than that observed in a single locus [41]. A lower nucleotide diversity in S. salar is further supported by the fact that about 60% of all analyzed loci showed no variation (Figure 3), and the overall SNP frequency was lower than in the majority of other organisms (Figure 2 and references therein). A lower level of sequence variation could be a consequence of relatively recent colonization of S. salar in its present habitats in the northern hemisphere after the last glaciation about 10, 000 years ago [49] as such patterns of reduced genetic variability in areas previously glaciated areas has been observed for other northern species (e.g. [50, 51]).

The SNP frequency in non-coding regions was at least threefold higher than coding regions in S. salar, which is to be expected due to the greater selective pressure on exons as observed in a recent human genome study [52]. Studies on disease-associated genes in humans have revealed an even higher proportion of coding SNPs, implying the effects of natural selection [53, 54]. This may also explain the higher frequency of SNPs in the coding region of the transferrin gene, which plays an important role in resistance to bacterial infection in a variety of organisms and was earlier reported to be under positive selection in S. salar [55]. On the contrary, no polymorphisms were detected in the transferrin gene of S. trutta, proposing that the effects of selection may vary considerably within lineages. However this could be due to the selection of the transferrin gene region which was sequenced in this study as considerable molecular variation has been reported in the transferrin gene within European S. trutta populations based on electrophoretic screening [56].

The overall SNP frequencies also varied among the salmonid species examined here (between 1/144 bp in T. thymallus to 1/695 bp in S. alpinus) but were, however, within the range of the SNP frequencies for a range of multi-locus studies with different species (Figure 2 and references therein). The estimates for S. salar and S. alpinus were in congruence with a previous study on S. salar (unpublished data, Hayes et al.), whereas the frequencies for T. thymallus and S. trutta were closer to a recent study with Pacific salmon [37]. Furthermore, in S. trutta and T. thymallus the nucleotide diversities were about six times higher than in S. salar or S. alpinus (Table 2). The high level of diversity in T. thymallus is consistent with the deep divergence between the evolutionary lineages assessed [57]. However, the high level of diversity in S. trutta is more difficult to explain as all individuals analysed originate from the same evolutionary lineage (the Atlantic lineage) proposed by Bernatchez [58]. However it is important to note that no Finnish S. trutta samples were assessed in the study of Bernatchez [58] and thus additional diversity may be harboured in this region.

Exploitation of the exponentially increasing amount of gene sequence data in public databases such as GenBank and recent EST projects is a very useful basis for identifying new polymorphic loci from the genomes of non-model organisms. Applications of SNPs have already been reported in ecological and conservation studies of natural populations [17, 20, 22], and these new types of markers have also been used to identify different Atlantic salmon strains [30]. However, as observed in this study, polymorphisms can be biased toward a relatively small portion of loci (Figure 3) thus increasing the effort required to identify a sufficient number of SNPs for ecological and population genetic applications. Based on a simulation study, the need for independent SNPs is fivefold that of microsatellites [6]. Furthermore, in salmonid fish the genome duplication event has been suggested to reduce SNP validation success ([37]; unpublished data, Hayes et al.), a result supported by this study, which may further hinder the development of a large number of independent loci. Therefore, the new IPEC approach introduced here will be a useful way to identify true SNPs for various applications in species with presumably duplicated genomes.