Crossing the pond: genetic assignment detects lobster hybridisation

American lobsters (Homarus americanus) imported live into Europe as a seafood commodity have occasionally been released or escaped into the wild, within the range of an allopatric congener, the European lobster (H. gammarus). In addition to disease and competition, introduced lobsters threaten native populations through hybridisation, but morphological discriminants used for species identification are unable to discern hybrids, so molecular methods are required. We tested an array of 79 single nucleotide polymorphisms (SNPs) for their utility to distinguish 1,308 H. gammarus from 38 H. americanus and 30 hybrid offspring from an American female captured in Sweden. These loci provide powerful species assignment in Homarus, enabling the robust identification of hybrid and American individuals among a survey of European stock. Moreover, a subset panel of the 12 most powerful SNPs is sufficient to separate the two pure species, even when tissues have been cooked, and can detect the introduced component of hybrids. We conclude that these SNP loci can unambiguously identify hybrid lobsters that may be undetectable via basic morphology, and offer a valuable tool to investigate the prevalence of cryptic hybridisation in the wild. Such investigations are required to properly evaluate the potential for introgression of alien genes into European lobster populations.

Genetic introgression due to hybridisation with non-native species is a major consequence of human-facilitated introductions that threatens endemic species with reduced fitness and local replacement 1,2 . Even where hybridisation between native and introduced species is rare, rapid and extensive genetic introgression can arise 3 . Population-level introgression and genetic admixture are not necessarily undesirable-managed interbreeding has been proposed to enhance resilience in key species threatened with extinction 4 -but invasive hybridisation is often associated with profoundly harmful effects, including loss of genetic variation and adaptation 1,5 . Nevertheless, assessing the conservation threat posed by hybridisation is not straightforward, and is especially challenging when hybrids themselves are fundamentally difficult to identify 5,6 . Investigation of hybridisation and management of its impacts require tools to distinguish hybrids from pure species strains, and the increasing availability of molecular markers presents a powerful resource to assess the extent of crossbreeding and introgression in wild populations subject to introductions 7,8 .
The European lobster (Homarus gammarus) is renowned for its high value as a seafood commodity, but stock collapses have severely diminished the productivity of fisheries throughout extensive portions of the species' range 9 . In Europe, recent annual landings of H. gammarus of ~5,000 tonnes are dwarfed by those of its transatlantic congener, the American lobster (H. americanus), which supports vast harvests of >150,000 tonnes per year 10 . To satisfy European demand for lobster that native H. gammarus landings cannot fulfil, ~17,000 tonnes of whole American lobsters are imported annually, around three quarters of which are imported live 11 , with the remainder comprised mainly of whole animals which are shipped frozen having already been cooked 12,13 .
Perhaps inevitably, the import of live H. americanus has led to their introduction into European waters via their escape or release 14,15 . In one high-profile case from 2015, some 361 American lobsters were released during a religious ceremony in Brighton, UK, with some introduced females later recovered bearing fertilised eggs and hundreds more never recovered at all 16 . Lobsters captured in European habitats and reported as H. americanus by fishers, managers or scientists have led to concerns that its presence poses a threat to native species, especially H. gammarus 14,17 . Citing their potential for invasiveness and to threaten native lobsters via competition and genetic introgression, Norway banned the import of live H. americanus in 2016, while the same concerns led Swedish officials to launch a bid, ultimately unsuccessful, to add the species to the European Union's list of invasive alien species 11 , which would have prohibited the live trade of American lobsters throughout the EU bloc.
Although concerns that pathogenic syndromes of H. americanus may spread to devastate H. gammarus stocks have been tempered by the latter's apparent disease resilience 18,19 , fears that American lobster introductions may impact European populations through competition and genetic introgression have not been so readily mitigated, particularly following recent observations that Americans both predate on and interbreed with European counterparts 20 . Reports of wild hybridisation are concerning as well as surprising; while long established that Homarus hybrids could be bred in captivity through induced fertilisation 21 , it was also proposed that behavioural characteristics of sexual selection should constrain interspecific mating in the wild 22 . However, this appears not to be the case 20 . Molecular identification of suspected Homarus hybrids has to date been reliant on the microsatellite method of Jørstad et al. ( 23 , cited in 20 ), though neither the sequences nor assignment power of these three loci have ever been published. Reliance on so few loci can also limit power for the detection of hybridisation and introgression following backcrossing 8 .
Most lobsters reported as alien H. americanus in Europe are taxonomically identified on the basis of heterospecific morphological indicators, especially differences in exoskeleton pigmentation and the presence of a spine on the ventral surface of the rostrum that is absent in H. gammarus 23,24 . Hybrids, however, cannot be visually identified since they may display characteristics of either parent or a mixture of both, and these methods also have limitations in distinguishing the pure species; the ventral rostral spine is unreliable, with both species capable of displaying the supposedly diagnostic characteristic of the other, and exoskeletal colouration can vary broadly 14,15,23 . Moreover, colouration is effectively invalidated as a determinant by cooking, which turns the shell of both species red. This presents another means by which H. americanus may impact H. gammarus as a seafood commodity, albeit through economic, rather than ecological, competition. In the UK, a 400 g cooked native European lobster typically costs in excess of £20 25 , whereas an equivalent H. americanus imported from Canada retails at about half this price (e.g. £11 26 ). Given this difference in value, the limitations of morphologically diagnostic characteristics present clear opportunity for exploitative mislabelling 27 , a widespread traceability issue impacting seafood supply chains.
The development of modern molecular resources is required to enable comprehensive assessments of the potential for H. americanus to threaten its congener in Europe through hybridisation in the wild, and to undermine market traceability via mislabelling. In this study, we tested single nucleotide polymorphism (SNP) markers, recently developed to assess H. gammarus population structure 28,29 , for their application in discerning the two Homarus species and their hybrids, including their utility with pre-cooked material, with the aim of developing universal and accurate molecular tools for species and hybrid assignment.

Methods
Clawed lobster tissues were all stored in 95-100% ethanol at −20 °C between sampling and DNA extraction. Tissue samples obtained were as follows: • pleopods of 20 cooked H. americanus, described as originating from Maine, USA, sourced frozen from a UK-based seafood importer and sampled in 2019. Extraction of genomic DNA from these tissues was conducted using the salting-out technique of Jenkins et al. 28 . DNA yields were assessed using a Nanodrop One spectrophotometer and standardised to concentrations of 50-100 ng/μl, before electrophoresis on 1% agarose gels to check fragment quality. As expected, DNA from cooked samples frequently evidenced degradation via electrophoresis. Genotyping and scoring of the 96 SNP loci of Jenkins et al. 28 was carried out on the Fluidigm EP1 system, following the protocols outlined in Jenkins et al. 29 . Two separate genotyping runs were conducted on all samples from cooked lobsters and expected hybrids to test repeatability of results. To provide reference samples and test the potential occurrence of hybrids among wild European lobster populations, these new data were added to a data set comprised of 1,278 H. gammarus genotyped by Jenkins et al. 29 . Quality control and filtering of genotype data followed the protocols outlined by Jenkins et al. 29 , except that the permissible missing data threshold was relaxed from 20% to 40% to ensure retention of three cooked samples that failed the stricter criteria.
Genetic divergence was initially explored using discriminant analysis of principal components (DAPC), executed in the adegenet v2.1.1 package 30 in R 31 . Cross-validation to select the optimal number of principal components to retain was conducted using the xvalDapc function in adegenet. Two analyses were used to test the assignment of species and the detection of hybrids. Firstly, adegenet was used to run snapclust 32 which uses the Expectation-Maximization algorithm to calculate maximum-likelihood estimations of genetic www.nature.com/scientificreports www.nature.com/scientificreports/ clustering and admixture. The number of expected genetic clusters (k) was set to 2, with explicit modelling for the presence of hybrids between these two clusters (hybrids = TRUE) and all other parameters run at default. Secondly, the program StrAuto v1.0 33 was used for parallel processing of the Bayesian clustering algorithm STRUCTURE v2.3.4 34 , implemented across 10,000 MCMC repetitions following a burn-in of 10,000, assuming an admixture model but without prior sampling information (locprior off) and all other parameters run at default. STRUCTURE run replicates were merged using CLUMPP v1.1.2 35 , and the optimal number of genetic groups (K) was defined using ΔK 36 and mean L(K) 34 with the pophelper v2.2.5.1 package 37 in R. In addition, to test the performance of a reduced marker panel that could facilitate cheaper genotyping, the DAPC and snapclust analyses were repeated with a subset of the 12 most informative SNPs, as selected by DAPC loading contributions and inspection of allele frequencies (Fig. 1A). For this subset of loci, snapclust was run as above, both with and without expectations of hybrids being present.  Table S1), across which the mean level of missing data for all new samples was 3.0%. Among sample types, the mean level of missing data was 3.0% for hybrid larvae, 3.8% for cooked lobster samples, and 1.0% for live-sampled H. americanus. Aside from disparities caused by missing data, only 25 allele scores (0.4%) were mismatched between genotyping runs, all but three of which were attributed to inconsistent scoring at one locus.
Of the 79 SNP loci known to be highly polymorphic in H. gammarus 29 , only 12 were polymorphic across all H. americanus samples (n = 38), with 45 polymorphic among the clutch of hybrid siblings (n = 30). All except one of the 12-SNP subset were monomorphic in H. americanus, and the single locus to show heterozygosity did so in only one individual (Fig. 1A). DAPC showed clear and extensive separation of the two Homarus species across all 79 SNPs, with hybrids clustering together in a discrete group between these clades (Fig. 2). Cooked individuals clustered with the correct species clade, while there was also segregation between H. gammarus of Atlantic and Mediterranean origins, as demonstrated by Jenkins et al. 29 .
Using all 79 SNPs, snapclust successfully detected all individuals of H. americanus and hybrid lineages ( Fig. 2; Supplementary Fig. S1), with membership probabilities >0.999 (Supplementary Table S2). All cooked lobsters also assigned to the correct species at probabilities >0.999. Only one of the 1,308 assumed European lobstersan individual from Kåvra in Sweden-was assigned majority hybrid membership by snapclust (to a component of 68%), and was also allocated a pronounced H. americanus component (~25%) using STRUCTURE (Fig. 3). Analysis of STRUCTURE outputs confirmed that K = 3 was the best-supported number of clades, although STRUCTURE placed greater emphasis on the intraspecific differentiation between H. gammarus of Atlantic and Mediterranean origins, so did not allocate a specific cluster to hybrids. Nevertheless, all hybrid individuals were  www.nature.com/scientificreports www.nature.com/scientificreports/ still differentiated, with clade composition defined as approximately half to the H. americanus cluster, and half among the two H. gammarus clusters (Fig. 3).
Using only genotypes from the 12-SNP subset, snapclust assigned all H. gammarus and H. americanus (including cooked samples) to the correct species clade when run without assumptions of hybridisation ( Fig. 1B;  Supplementary Fig. S1). Only four H. gammarus individuals were assigned to probabilities <0.995, and only one of these was <0.95 (Supplementary Table S2). Notably, majority H. americanus composition was also assigned to all known hybrids via this analysis (Fig. 1B). When snapclust assumed the presence of hybridisation, known hybrids were detected and effective assignment of H. americanus was maintained, but 37 H. gammarus individuals (2.8% of the total) were incorrectly assigned with a majority component to the hybrid clade. These results were supported by DAPC, which showed little separation between some outlying H. gammarus individuals and the hybrid cluster using just 12 SNPs (Supplementary Fig. S2).

Discussion
Using 79 SNPs originally developed to investigate intraspecific genetic differentiation among H. gammarus populations, we were instead able to highlight interspecific differences between H. gammarus, its congener H. americanus, and hybrids of the two species. Effective determination of both species and their hybrids were confirmed with the full 79-SNP panel using the snapclust software. Robust assignment was maintained using a subset of the 12 most powerful SNPs when assumptions of hybridisation were omitted; all individuals of both pure species were assigned correctly, and the introduced genetic component of hybrids was still detected, such that hybrids were assigned as H. americanus rather than H. gammarus. These biallelic loci can define species in Homarus because, while highly polymorphic in H. gammarus, the majority (85%) were monomorphic in the 38 H. americanus individuals we tested.
Our results show that these SNP loci present useful genetic tools for accurate taxonomic determination of Homarus lobsters. A subset of just 12 SNPs discriminates H. americanus from H. gammarus, facilitating the use of a genotyping chip such as the Fluidigm Flex Six Genotyping IFC, which can screen 72 individuals across 12 SNPs, but which can be loaded in six discrete sections of only 12 individuals at a time. This technology is therefore useful for processing small sample runs quickly and relatively cheaply (~£4 per individual), so is well suited to applications of routine screening or opportunistic checks of specimens caught and/or sold in Europe which are suspected to be introduced and/or mis-sold H. americanus. We have also demonstrated that this method can successfully genotype and determine species of pre-cooked lobsters, which few consumers or industry personnel could identify visually, even when sold whole. In practical terms, this tool could be used to screen for the genetic signature of H. americanus among putatively European lobsters captured or marketed in Europe. Although this 12-SNP tool lacks the resolution to reliably identify hybrids without generating false positives among pure H. gammarus specimens, it could still be useful in identifying samples warranting further analysis with the full SNP panel to test for hybridisation, since all the hybrids we tested assigned as H. americanus when analytical expectations were for the two pure species only. If necessary, Homarus hybrids could then be distinguished from specimens of the two pure species using all 79 SNPs.
All 30 H. americanus x gammarus offspring from Sweden were assigned hybrid status by snapclust using our 79-SNP tool. Only one assumed H. gammarus (out of 1,306) assigned as a majority hybrid by snpaclust. This individual ('Kav26') was a 103 mm CL male sampled in 2007 from the Kåvra lobster reserve in western Sweden 38 , approximately 20 km from the capture location of the introduced H. americanus female which carried the hybrid larval clutch tested in this study, and in an area which has experienced frequent introductions of American lobsters in recent years 20 . As such, although it seems plausible that this individual's assignment as a hybrid represents an anomalous false positive, we cannot rule out the possibility that this specimen had some H. americanus ancestry via interspecific breeding; that this individual originated from a designated conservation reserve may warrant further investigation. Our study includes only one hybrid sibling cohort, so it would be beneficial to confirm the power of this 79-SNP tool on other suspected hybrid clutches, although the extent of interspecific differentiation we revealed suggests it should be universally applicable for identifying first-generation Homarus crosses.
Currently, it is difficult to ascertain the threat posed to European lobster stocks from hybridisation with H. americanus, since conflicting information exists as to both the likelihood of interspecific mating and the fertility of resultant hybrids. Talbot et al. 21 found that the spermatophores of captive-reared hybrid males lacked sperm, suggesting that they were infertile, but Kitaka (pers comm., as cited by 22 ) achieved interspecific mating in captivity and asserted that both sexes of the resultant hybrid offspring were fertile. Fertile hybrids enable backcrossing and pervasive population-level introgression, although hybridisation is still a conservation threat even where hybrids are infertile, since it represents wasted reproductive effort 5,6 . Mate-choice of H. gammarus females is to favour conspecific males regardless of their size and dominance status 22 , so the discovery of female H. americanus bearing hybrid clutches in the wild in Europe suggests that either American females more readily mate interspecifically than their European counterparts, or that behavioural barriers are not sufficient to prevent interspecific mating where no conspecific mate can be found. Accordingly, we recommend investigation of the propensity for interspecific mating, between both sexes of both species, and of the fertility status of hybrids of both sexes.
The recent motion from some European scientists and regulators to take a conservative and proactive approach to the threats posed by introduced H. americanus by outlawing its live-trade was met with robust resistance from advocates of commerce initially valued at US$260 million a year to North American exporters 12,13 . Indeed, instead of restricting live-trade, a newly-adopted EU trade agreement eliminates 6-20% tariffs on Canadian exporters, providing them with duty-free access to European markets 39 . Although existing regulations prohibit the release of non-native lobsters across most of Europe, this has not been sufficient to prevent increasing numbers of both inadvertent and deliberate introductions of H. americanus, heightening the likelihood of it becoming permanently established and/or damage from hybridisation [14][15][16] . In the wake of our confirmation that interspecific crossbreeding between native and introduced lobsters has occurred in the wild, greater effort should (2020) 10:7781 | https://doi.org/10.1038/s41598-020-64692-z www.nature.com/scientificreports www.nature.com/scientificreports/ be made to educate regulators, commercial traders and the general public as to the potential conservation threats posed by hybridisation between endemic and introduced lobsters.
Although clear opportunity exists to profit from the mislabelling of cheaper H. americanus as the more expensive H. gammarus, the extent to which erroneous species designation-whether inadvertent or fraudulent-currently impacts lobster supply chains is unknown, as is the extent to which hybridisation with introduced H. americanus may be impacting wild European lobster populations. Such investigations can now be implemented using these SNP resources. The cost of genetic techniques is often cited to justify why they are not more widely used to enforce regulatory standards in the seafood sector; in practice, however, the expense of molecular testing is typically far exceeded by financial penalties levied against uncovered incidences of malpractice 40 . Overall, the approaches we have tested offer more powerful assignment of species and hybridisation in clawed lobsters than existing methods based on allozymes 41 , random amplified polymorphic DNA 42 or microsatellites 23 , and enable accurate and rapid sample processing via high-throughput SNP genotyping technology. These assignment techniques can be applied in forensic assessments to address the extent that introduced H. americanus threaten populations of H. gammarus via hybridisation and invasive introgression, and whether fraudulent mislabelling affects clawed lobster seafood products in Europe.