Population Genomics and Connectivity of the Blue Mussel Species Complex: Insights From a North‐East Atlantic Hybrid Zone

ABSTRACT

Blue mussels (Mytilus spp.) are ecologically and economically important bivalves widespread in both hemispheres. Their relevance to coastal ecosystems and the aquaculture industry has made them extensively studied. The Mytilus complex consists of distinct genetic lineages, including Mytilus edulis , Mytilus galloprovincialis , Mytilus trossulus , and their fertile hybrids. In overlapping areas, they create complex hybrid zones, which have been investigated along European coasts, employing multi‐marker approaches. However, knowledge gaps still exist in the North‐east Atlantic region, in the middle of their hybrid zone around the island of Ireland, regarding their genomic composition, population structure and connectivity. This study addresses these gaps by genotyping 781 individuals from 26 sites encompassing Ireland's hybrid zone, including both wild and farmed stocks from varying environmental conditions. Using a selected panel of 72 SNP markers we examined relationships among genotypic composition, genetic diversity, isolation by distance (IBD) and environmental variables to identify drivers of Mytilus genetic structure. Results confirmed two distinct genetic lineages and their hybrids, with a clear geographic pattern: the east coast of Ireland is dominated by pure M. edulis genotype populations, while the south, west and north coasts exhibit varying degrees of admixture with M. galloprovincialis genotype. Pure M. galloprovincialis populations were identified at specific sites on the west and north coast. Sea current resistance and wave height were significant predictors for both genotype composition and genetic differentiation. This study corroborates previous findings and provides the first comprehensive investigation of Irish Mytilus spp. population structure and connectivity using a multi‐marker approach. The findings highlight the importance of understanding the Mytilus complex's composition and population dynamics to inform sustainable aquaculture practices and monitor potential climate change‐driven shifts in the North‐east Atlantic region.

1. Introduction

The blue mussel (Mytilus spp.) is an economically and ecologically important bivalve, commonly distributed both in the northern and southern hemispheres (Gardner et al. 2021; Gosling 2021; Larraín et al. 2018; Mathiesen et al. 2017). It is often described as an ecosystem engineer and environmental indicator, providing several important ecosystem services (Barrett et al. 2022; van der Schatte Olivier et al. 2020). Moreover, blue mussels are an important species in the shellfish aquaculture industry worldwide (FAO 2024). In 2022, European aquaculture production yielded around 1.1 million tonnes of aquatic organisms worth €4.8 billion, of which mussels production accounted for more than 35% by weight and almost 10% by value (Eurostat, Aquaculture Statistics retrieved 6th February 2025). In northern Europe, mussel farming mostly relies on two culturing techniques: seabed culture and rope culture (Avdelas et al. 2021). Notwithstanding the effects of dredging and the addition of physical structures to promote settlement, mussel farming has low environmental impacts relative to other food production systems, making this aquaculture sector particularly sustainable and providing high‐quality animal proteins rich in Omega‐3 fatty acids (Avdelas et al. 2021; Barrett et al. 2022; Cooney et al. 2025; Yaghubi et al. 2021).

In the temperate Northern Hemisphere, the blue mussel complex comprises three congeneric species: Mytilus edulis (Linnaeus 1758), Mytilus galloprovincialis (Lamarck 1819) and Mytilus trossulus (Gould 1850). M. edulis , commonly referred to as the Atlantic blue mussel, is a cold‐temperate species with a distribution range that covers the eastern and western coasts of the North Atlantic, and the coasts of northern Europe up to the Arctic region (Beaumont et al. 2008; Bierne et al. 2003; Diz and Skibinski 2024; Gosling et al. 2008; Mathiesen et al. 2017; Nascimento‐Schulze et al. 2023; Simon et al. 2021). M. galloprovincialis , often referred to as the Mediterranean mussel, has a distribution range from the Black Sea, through the Mediterranean Sea to the North‐east Atlantic up to the Arctic region (Bierne et al. 2003; Gosling et al. 2008; Kijewski et al. 2011; Mathiesen et al. 2017; Vendrami et al. 2020). It is divided into two main lineages: the Atlantic and the Mediterranean (Bierne et al. 2003; del Rio‐Lavín et al. 2022; Kijewski et al. 2011; Lynch et al. 2020; Mathiesen et al. 2017; Simon et al. 2021; Vendrami et al. 2020). Moreover, M. galloprovincialis is also present in both the southern and northern hemispheres, on the Pacific and Atlantic coasts, highlighting its great capacity to settle and establish in regions outside its original distribution range (del Rio‐Lavín et al. 2022; Gardner and Westfall 2012; Larraín et al. 2018; Nascimento‐Schulze et al. 2023; Zbawicka et al. 2022). Finally, M. trossulus is originally from the Pacific Ocean, and it can be found on North Atlantic coasts, in the Baltic Sea and in the Arctic (Braby and Somero 2006; Dias, Piertney, et al. 2011; Mathiesen et al. 2017; Nascimento‐Schulze et al. 2023; Väinölä and Strelkov 2011; Vendrami et al. 2020). These three species are morphologically difficult to distinguish, making their population structure complex to investigate, especially when their distribution ranges overlap (Gosling and Wilkina 1981; Seed 1974).

In these overlapping areas, different lineages of blue mussels hybridise extensively, producing fertile offspring and a composite genomic make‐up, with varying levels of introgression and backcrossing with local populations (Bierne et al. 2003; Gosling et al. 2008; Vendrami et al. 2020). Their hybrid zones have been extensively studied in recent decades, with particular interest in the evolutionary mechanisms of speciation, adaptation and the impact of hybridisation on the aquaculture industry (Bierne et al. 2003; Dias, Malgrange, et al. 2011; Fraïsse et al. 2014; Michalek et al. 2016; Simon et al. 2021). In Europe, a well‐established hybrid zone between M. edulis and M. galloprovincialis extends from the north coast of France and south‐west England to the southwest and west of Ireland (Bierne et al. 2003; Coghlan and Gosling 2007; Diz and Skibinski 2024; Hilbish et al. 2002; Simon et al. 2019). Hybrid zones between M. edulis , M. galloprovincialis and M. trossulus are mostly present in northern European coasts in Scotland, the Baltic Sea, North Sea, Norwegian Sea, parts of the Barents and White Seas, and up to the Arctic Ocean in Greenland (Beaumont et al. 2008; Dias, Piertney, et al. 2011; Mathiesen et al. 2017; Väinölä and Strelkov 2011). Local environmental conditions such as temperature, salinity and wave exposure have been proposed as key drivers of Mytilus spp. local population structure (Bierne, David, Boudry, and Bonhomme 2002; Coghlan and Gosling 2007; Diz and Skibinski 2024; Gosling and Wilkina 1981; Hilbish et al. 2002; Lynch et al. 2020).

Genetic structure in the Northern Europe hybrid zones is temporally dynamic, with frequencies of pure types and their hybrids varying throughout the decades (Fly et al. 2015; Lynch et al. 2020). Several studies suggested that local environmental conditions and ongoing climate change could play a key role in these dynamics, via both pre‐zygotic mechanisms (e.g., effects on reproductive) and post‐zygotic mechanisms (e.g., differential effects on hybrid versus pure type survival, especially at the spat stage) (Beaumont et al. 2004; Doherty et al. 2009; Kenchington et al. 2020; Lynch et al. 2020; Shields et al. 2008). Climate change effects in the North‐east Atlantic region have been observed in the past few decades, with rising sea surface temperatures impacting the geographic distribution of cold‐water species (ICES 2024). However, this warming trend is interrupted by periodic cool spells and increased storminess, which contribute to greater freshening of coastal waters (ICES 2024; Nolan et al. 2023). Moreover, coastal currents play a central role in the connectivity and settlement of mussels and changes in these coastal currents can have a great impact on the shellfish industry (Avdelas et al. 2021; Demmer, Neill, et al. 2022; Demmer, Robins, et al. 2022).

A general decline in mussel production and the unreliability of natural spat supply (e.g., scarcity of natural mussel beds) have been observed around all European coasts since the 1990s, especially in Northern countries (Avdelas et al. 2021). According to Baden et al. (2021), in the past few decades, mussel beds in the North‐east Atlantic sheltered littoral and sublittoral zones have declined by more than 50%. In most areas, the decline is generally attributed to poor recruitment success, often linked to the increasing frequency of extreme weather events driven by climate change (Avdelas et al. 2021; Baden et al. 2021; Little et al. 2024). In France, recurrent mass mortalities in hybrid zones between M. edulis and M. galloprovincialis have been observed by Benabdelmouna and Ledu (2016), who hypothesised cytogenic anomalies linked to hybridisation as a potential cause. Moreover, in the review by Lupo et al. (2021), a higher mortality risk between hybrids of M. edulis and M. galloprovincialis compared to pure strains was reported, as well as a link with lower heterozygosity of M. edulis . Given the decline in wild and farmed mussel stocks, the environmental pressures of climate change, and the complexity of mussel population structure and hybridisation dynamics, an in‐depth investigation of mussel population genomics and connectivity in the North‐east Atlantic and northern Europe is essential to establish a baseline for future monitoring.

In the past 10 years, new tools and methodologies have been developed to investigate the population structure of the blue mussel complex (Fraïsse et al. 2016; Mathiesen et al. 2017; Nascimento‐Schulze et al. 2023; Simon et al. 2019; Wilson et al. 2018). The advent of new and increasingly affordable multi‐marker approaches, such as single nucleotide polymorphisms (SNPs), has allowed more powerful and comprehensive analyses of species and population genetics, especially for investigating the level of hybridisation and introgression in hybrid zones (Mathiesen et al. 2017; Simon et al. 2019), compared to the traditional nuclear single marker approach (e.g., Me15/Me16 primers; Inoue et al. 1995). These advancements have not only increased the availability of genetic data but also enhanced the integration of genetic information with other data types (e.g., environmental variables) (Wenne et al. 2020, 2022) and increased the scope for investigating population structure, phylogeography and the adaptive potential in the context of seascape genomics.

Despite the growing body of genomic research on Mytilus spp., Ireland remains underrepresented in many large‐scale studies (Nascimento‐Schulze et al. 2023; Vendrami et al. 2020). This is notable given that Ireland includes an extensive hybrid zone and represents an ecologically and economically important region within the North‐east Atlantic. Although the Irish hybrid zone has been monitored since the 1970s, the increasing impacts of climate change, combined with a growing mussel aquaculture sector that has recently faced severe declines in output and revenue, call for a more fine‐scale and genomically informed investigation (Bord Iascaigh Mhara 2024; Gosling et al. 2008; Lynch et al. 2020; Seed 1974). While several Northern European countries have adopted SNP‐based and multi‐locus approaches, research in Irish waters still relies largely on single‐marker screening (Gosling et al. 2008; Lynch et al. 2020). As such, Ireland represents a relevant case study to provide additional baseline genomic data on blue mussel populations, supporting broader efforts to understand connectivity and hybridisation in the North‐east Atlantic.

Thus, the aims of this study are to use a panel of SNP markers originally sourced from selected publications addressing genetic structure at the inter‐ and intraspecific levels of the Mytilus spp. complex in European waters (Fraïsse et al. 2016; Hammel et al. 2021; Simon et al. 2018; Wilson et al. 2018), along with a geographically comprehensive sampling across the island of Ireland as a case study to: (i) elucidate the genetic composition and population structure of the Mytilus spp. complex in the Celtic and Irish Seas, including admixed and pure M. edulis populations; (ii) investigate the distribution of the different genotypes around the Irish coasts; and (iii) explore oceanographic and environmental factors that could drive the observed population structure, as insight into potential connectivity, here defined as the inferred potential for gene flow between sites, estimated through patterns of ocean current resistance and genetic differentiation.

2. Materials and Methods

2.1. Study Sites and Sampling

A total of 781 adult individuals of Mytilus spp. (shell length between 45 and 55 mm) were sampled in 26 locations around Ireland (Figure 1), from a variety of habitats and stock types (i.e., farmed and wild stocks). Samples were collected by local stakeholders either by hand, harvested from ropes by mussel workboat, or by dredging, and shipped fresh to the Atlantic Technological University (ATU)—Galway city and stored at −20°C. In addition, mussel tissues preserved in absolute ethanol from the Adriatic Sea (Italy, n = 11) and from the Baltic Sea (Sweden, n = 10) were included as reference/outlier samples, which represent other lineages present in Europe but not in Irish waters. Sampling details are summarised in Table 1.

FIGURE 1.

Mussel sampling locations. (A) Ireland within the European context, showing the Celtic Sea, Irish Sea, North Sea, the Atlantic Ocean, as well as the Adriatic and Baltic sample locations. (B) Zoomed‐in map of Irish sampling sites. Location codes correspond to Table 1.

TABLE 1.

Information on the sampling campaign including: Sampling locations, geographic coordinates, sampling month and year, location code, number of individuals screened per population (N), stock type, habitat type, type of culture (in case of farm stock) or type of substrate (in case of wild stock).

Country	Location	Region	Latitude	Longitude	Sampling date	Code	N	Stock type	Habitat	Type of culture/substrate
Republic of Ireland	Mulroy Bay	North	55.15488	−7.68000	November 2022	MB	30	Farm	Subtidal	Rope growth
	Lough Foyle	North	55.09047	−7.08548	November 2022	LF	30	Farm	Subtidal	Bottom growth—seabed
	Clew Bay	West	53.86222	−9.62820	April 2022	CB	29	Farm	Subtidal	Rope growth
	Killary Fjord	West	53.60487	−9.80624	June 2022	KF	29	Farm	Subtidal	Rope growth
	Bertraghboy Bay (South Connemara)	West	53.39689	−9.82947	February 2023	SCBB	22	Wild	Intertidal	Natural rock
	Aran islands (Inish Meáin)	West	53.08083	−9.57142	July 2022	AI	30	Wild	Intertidal	Artificial structure
	Cromane Wild	Southwest	52.09275	−9.96550	June 2022	CRW	30	Wild	Subtidal	Bottom growth—seabed
	Cromane Farm	Southwest	52.13825	−9.93817	June 2022	CRF	30	Farm	Intertidal	Bottom growth—seabed
	Snave Bantry Bay	Southwest	51.71223	−9.47268	November 2022	SBB	30	Farm	Subtidal	Rope growth
	Roaringwater Bay	South	51.53507	−9.41850	June 2022	RWB	28	Farm	Subtidal	Rope growth
	Cork Harbour	South	51.88000	−8.25000	March 2023	CKH	30	Wild	Intertidal	Bottom growth—seabed
	Waterford Estuary	South	52.24000	−6.97000	October 2022	WF	30	Wild	Intertidal	Bottom growth—seabed
	Kilmakilloge	South	51.77100	−9.82938	June 2022	KLM	30	Farm	Subtidal	Rope growth
	Dungarvan Harbour	South	52.10006	−7.58435	December 2022	DH	29	Wild	Intertidal	Bottom growth—seabed
	Rosslare	East	52.25820	−6.31087	July 2022	RSW	30	Wild	Subtidal	Bottom growth—seabed
	Wexford Harbour	East	52.33273	−6.43135	July 2022	WX	30	Farm	Subtidal	Bottom growth—seabed
	Arklow Wind Farm	East	52.78000	−5.95231	July 2023	AWWF	30	Wild	Intertidal	Artificial structure
	Wicklow	East	52.93878	−5.93775	May 2022	WK	30	Wild	Subtidal	Bottom growth—seabed
	Dún Laoghaire Marina	East	53.29000	−6.14000	September 2022	DLM	30	Wild	Intertidal	Artificial structure
	North Bull wall	East	53.35000	−6.16000	September 2022	NBW	30	Wild	Intertidal	Bottom growth—seabed
	Malahide	East	53.45000	−6.13000	September 2022	MH	30	Wild	Intertidal	Bottom growth—seabed
	Rogerstown	East	53.51256	−6.13003	November 2022	RT	30	Wild	Intertidal	Bottom growth—seabed
	Dunany Point	East	53.86600	−6.24622	November 2022	DP	30	Wild	Intertidal	Bottom growth—seabed
	Carlingford Lough	East	54.07942	−6.23708	May 2022	CL	30	Farm	Subtidal	Bottom growth—seabed
Northern Ireland (UK)	Dunseverick	North	55.23881	−6.43339	August 2023	NI‐12	35	Wild	Intertidal	Rocky shore
	Bangor	North	54.67210	−5.63544	September 2023	NI‐13	39	Wild	Intertidal	Rocky shore
Italy	Piattaforma acqua alta	Adriatic Sea	45.31667	12.50000	May 2022	A	11	Wild	Intertidal	Artificial structure
Sweden	Sundsbådan	Baltic Sea	58.78970	17.77921	September 2021	B	9 Wild		Subtidal	NA

Note: NA notation when information was not provided. See the map in Figure 1 for sampling locations.

2.2. DNA Extraction

Gill tissue was dissected from thawed Mytilus samples and preserved in absolute ethanol to establish a tissue bank (currently stored and available upon request at the Marine and Freshwater Research Centre in ATU‐Galway, Republic of Ireland). DNA extraction was carried out using the E.Z.N.A. Tissue kit (Omega BioTek, Norcross, GA, USA) and according to the manufacturer's instructions with adaptations: to homogenize the tissue, in a nuclease‐free 1.5 mL microcentrifuge tube, approximately 30 mg of gill tissue was minced by vortex at maximum speed for 15 min with three Zirconium Ceramic Oxide beads (diameter 1.4 mm, Fisherbrand). Then, incubation at 55°C was performed for 2 h with vortexing every 30 min for 15 s.

To ensure that the protocol yielded sufficient amounts of DNA for downstream genotyping procedures, genomic DNA of a subset of samples was quantified with the dsDNA Broad Range Assay kit using a Qubit 3.0 fluorometer (Thermofisher Scientific, Ireland).

2.3. Adhesive Protein Gene Single Marker Genotyping

To confirm successful DNA extraction and to obtain a preliminary genetic/taxonomic screen in line with previous studies carried out in Ireland (Gosling et al. 2008; Lynch et al. 2020), all samples were genotyped using a single marker approach developed by Inoue et al. (1995) that targets the foot adhesive protein gene with primers Me15 and Me16. PCR was carried out in a 20 μL reaction mixture containing: AccuStart II PCR ToughMix (1×) (Quantabio), GelTrack Loading Dye (1×) (Quantabio), 0.5 μM of primer Me15 (5′‐CCAGTATACAAACCTGTGAAGA‐3′), 0.5 μM of primer Me16 (5′‐TGTTGTCTTAATGGTTTGTAAGA‐3′), and 2 μL of template DNA (concentration between 2 and 40 ng/μL). The thermal conditions were as follows: 94°C for 3 min followed by 35 cycles of [95°C for 15 s, 56°C for 20 s, 72°C for 20 s]. The PCR was carried out using a MiniAmp Plus thermocycler (Applied Biosystem). The visualisation of the amplicons was conducted through gel electrophoresis in a 2% agarose gel stained with SYBR Safe DNA Gel Stain (Invitrogen, CA, USA), run at 100 V for 50 min. The size of PCR amplicons was established by comparison to a GeneRuler 50 bp DNA Ladder ready‐to‐use (Thermo Scientific, Ireland) upon visualisation by UV light in a Bio‐Rad Gel Doc EZ Imager (Bio‐Rad, Ireland).

2.4. SNPs Selection and Genotyping

The set of SNPs used in the present study was originally sourced from selected publications addressing genetic structure at the inter‐ and intraspecific levels of the Mytilus spp. complex in European waters (Fraïsse et al. 2016; Hammel et al. 2021; Simon et al. 2018; Wilson et al. 2018), aiming for discrimination between pure lineages and hybrids found in Irish waters, as well as cost‐effectiveness. To facilitate genotyping using a Biomark HD system (Standard BioTools, South San Francisco, CA, USA), primer pairs were designed using 1000 bp flanking sequences from the M. galloprovincialis genome assembly (LOLA—European Nucleotide Archive, project IDs PRJEB24883; Gene Bank GCA_900618805.1; Gerdol et al. 2020), whereby conserved DNA regions were identified by alignment against 17 additional Mytilus genomes (Corrochano‐Fraile et al. 2022; Gerdol et al. 2020; Murgarella et al. 2016; Regan et al. 2024; M. Gerdol, personal communication). Ultimately, a panel of 91 SNPs that showed a clear genotype distinction across reference genomes, and flanking regions suitable for primer design was retained for genotyping the study samples, after a pre‐amplification step: 70 SNPs from Simon et al. (2018) and 9 from Hammel et al. (2021) (all of which were originally published in Fraïsse et al. 2016), and 12 from Wilson et al. (2018). Data from each run were analysed using the Standard BioTools SNP Genotyping Analysis software v.1.0.2 (Standard BioTools, https://www.standardbio.com/), with an assay reference library. Within each run, the optimal cycle for each SNP was determined based on cluster segregation and amplification success. All base calls were manually verified and adjusted to ensure the accurate identification of homozygous and heterozygous clusters. SNPs that showed either undefined cluster patterns or amplification failure in more than 15% of the samples were discarded. Similarly, samples that failed to amplify more than 15% of the SNP markers were discarded from the dataset. We applied this 15% threshold as a more stringent criterion compared to other studies (Mathiesen et al. 2017; 25%), to reduce missing data and increase the reliability of downstream analyses.

Details on primer design, reference genomes and protocol optimisation are provided in Supporting Information.

2.5. Descriptive and Summary Statistics

An explorative and diagnostic analysis of the full dataset was performed, including 26 Irish sites, 1 Adriatic site and 1 Baltic site. The R package Adegenet (Jombart 2008; Jombart and Ahmed 2011; R Core Team 2023) was used to run a Discriminant Analysis of Principal Components (DAPC) and customisation from the Rscript ‘ggDAPC’ (Frantine 2023; GitHub—https://github.com/wilsonfrantine/ggDAPC) was used to produce all the DAPC plots of this study. The function sNMF of the package LEA (Frichot and François 2015) was used to run several admixture analyses with different numbers of predefined clusters (K), which ranged from two to five, to explore the population structure and admixture of the different populations. To assess the quality of the markers, allele frequencies were calculated for each locus within each population to identify monomorphic markers, as well as the inbreeding coefficient F _is per locus across populations, within populations and per population across loci using the R package Genepop v.1.2.2 (Rousset 2008). Loci that were either monomorphic or with minor allele frequencies < 0.01 were discarded after manual inspection (McDevitt et al. 2022). Loci with a global F _is > 0.3 or < −0.3, indicating departures from Hardy–Weinberg expectations, were further examined population by population, to determine whether high absolute values for F _is could be explained by admixture. Loci that showed a F _is > 0.3 or < −0.3 in more than 50% of the populations, excluding those that displayed a consistent level of admixture (based on the admixture preliminary analysis and population‐level F _is), were discarded from downstream analyses. The threshold of |0.3| was chosen empirically, based on the distribution of F _is values across loci in this dataset, and was used as a conservative quality control filter to minimise the risk of including loci with technical artefacts or extreme genotype frequency distortions (heterozygote deficiencies or excesses). After this step, 72 SNPs were retained for the subsequent analyses (list of SNPs loci dataset in Table S1).

2.6. Population Structure of Irish Mytilus spp.

A reduced dataset including only samples from the Irish coast was used to assess the genetic makeup and population structure of Irish mussels. A DAPC was run (as detailed above), and allele richness (A _r), observed heterozygosity (H _o) and expected heterozygosity (H _e) for each population were calculated with the function divBasic from the package diveRsity (Keenan et al. 2013). The R packages RLDNe (Do et al. 2014; Robinson 2024), dartRverse and dartR (Gruber et al. 2018; Mijangos et al. 2022) were used to calculate the effective population size (N _e) employing the function gl.LDNe. STRUCTURE v2.3.4 (Pritchard et al. 2000) was employed to analyse the genetic structure and the putative number of clusters of the Irish populations. The analysis parameters were as follows: 5000 length of burn‐in period, 50,000 MCMC, admixture model, allele frequencies correlated, K from 1 to 8, with 5 repetitions per K. The results were uploaded to the CLUMPAK online server (Kopelman et al. 2015; https://clumpak.tau.ac.il/bestK.html, accessed: 5th October 2024), and the best K was chosen following the Evanno method (Evanno et al. 2005). Furthermore, the software FSTAT V2.9.4 (Goudet 2001) was used to calculate populations' pairwise differentiation index F _st with significance p‐value corrected by Bonferroni multiple testing. Parameters were set as follows: global test of Hardy–Weinberg within and overall samples (500 iterations), ‘population differentiation’ test not assuming HW within samples, 5/100 nominal level for multiple tests and 1000 permutations. To minimise the influence of missing data on pairwise F _st with significance testing, we applied an additional filtering step. Loci with elevated missing genotypes (2%–8%) that interfered with F _st computation were iteratively inspected and discarded. Nine loci were therefore discarded, resulting in a final dataset of 63 loci used for the F _st analysis (Table S1).

All plots of this study were produced with the R packages ggplot2 (Wickham et al. 2016) and ggalluvial (Brunson and Read 2023).

2.7. Admixed and M. edulis Populations in Irish Mussels

A further investigation was conducted on the Irish populations that showed a ‘pure’ Atlantic M. edulis genotype. Based on the STRUCTURE Q values from the Irish Mytilus spp. analysis, populations for which samples showed admixture proportions > 0.2 (Mathiesen et al. 2017) were categorised as ‘Admixed Irish Populations’, while populations for which samples showed admixture proportions < 0.2 were categorised as ‘Irish M. galloprovincialis genotype’ or ‘Irish M. edulis genotype’, depending on the dominant cluster, respectively.

To assess the population structure of Irish M. edulis , the ‘Irish M. edulis genotype’ populations were analysed with a separate DAPC and STRUCTURE with the same parameters set as above. Two datasets were analysed: (i) all Irish populations with an average admixed proportion < 0.2 and (ii) only Irish populations in which all individuals showed admixture < 0.2 (i.e., excluding populations in which only some individuals showed a higher level of admixture).

2.8. Isolation by Distance and Correlation Between Genotype Composition, Pairwise F _st and Environmental Variables

To investigate potential drivers of the genetic differentiation and genotype composition of the Irish mussels, a two‐step approach was used. First, the relationship between genotype composition (extent of admixture) and environmental variables was examined at each site. We then investigated if genetic differentiation between sites was related to the least‐cost distance between them, taking into account sea currents and environmental differences between sites.

2.8.1. Environmental Variables Acquisition

The Regional Operational Model (ROMS) for the Northeast Atlantic was used to characterise the environmental conditions within the study area, which has a mean horizontal resolution of 1.9 km (Nagy et al. 2020; https://www.marine.ie/site‐area/data‐services/marine‐forecasts/ocean‐forecasts accessed on 11‐February‐2025). For each site, we obtained the long‐term minimum (min), maximum (max) and average (mean) of sea surface temperature (SST °C), salinity (PSU) for the period 2017–2023 and wave height (m) for the period 1994–2023. Some inshore sites fell outside of the area covered by the ROMS model. For these sites, data were extracted from the closest grid square for which data were available. The distance between the study sites and the geographic co‐ordinates of the modelled environmental data ranged from 144 m to 11.79 km, with a mean distance of 4.8 km (Table S4 and Figure S6).

2.8.2. Oceanographic and Geographic Resistance to Dispersal

Resistance to dispersal between sites during the main dispersal period for Mytilus spp. (March–May) (Doherty et al. 2009) was estimated using mean eastward and northward sea surface current velocities (u‐component velocity and v‐component velocity, respectively, in m s⁻¹) during those months for the period 2012–2023. Current data were obtained from the Northeast Atlantic ROMS model described above, and from the Global Ocean Physics Reanalysis, E.U. Copernicus Marine Service Information (CMEMS) Marine Data Store (MDS) (DOI: 10.48670/moi‐00021; accessed on 11‐February‐2025). Rasters of current data from the two data products were combined to create a resistance layer with the disaggregate and merge functions from the raster package in R (Hijmans 2025) preserving the higher spatial resolution of the Northeast Atlantic ROMS model.

In the gdistance package in R (van Etten 2017), a geographical correction was applied to the resulting resistance layer with the geoCorrection function and the least‐cost distance of dispersing between each pair of sites was calculated using the costDistance function. This produced an anisotropic matrix of least‐cost distance estimates (in arbitrary units) for each pair of sites, considering movement in both directions (i.e., from site ‘a’ to site ‘b’, and from site ‘b’ to site ‘a’).

For each pair of sites, a single least‐cost distance value was obtained using the minimum of the least‐cost distances from ‘a’ to ‘b’ and from ‘b’ to ‘a’. When dispersal between sites was not possible without moving against the direction of the currents or crossing land, the least‐cost distance estimate was equal to infinity. Of the 325 possible connections, 148 were calculable (i.e., the least‐cost distance was below infinity in at least one direction). For these 31 site pairs, the least‐cost paths did not reflect realistic transport scenarios (e.g., involved transport off the shelf edge for over 1000 km before deflection back towards the coast). We therefore grouped the least‐cost distance values into three categories representing the extent to which ocean currents presented a barrier to exchange between sites: cat0: least cost distance = infinity (strong current resistance), cat1: > 3,000,000 (moderate current resistance) and cat2: least cost distance < 1,100,000 (light current resistance).

To visualise the relative connectedness of sites based on ocean current resistance, the least‐cost distance values were plotted for site pairs in cat2 to produce a ‘sink‐source’ network map using the R packages ggplot2 and rnaturalearth (Massicotte and South 2023).

For a detailed description of the construction of the resistance layer, the calculation of least‐cost distances, and the handling of land cells, please refer to Methods S2.

2.8.3. Statistical Analyses

To investigate the potential contributions of environmental conditions to the genotype composition at each site, a beta regression model was run using the betareg R package (Cribari‐Neto and Zeileis 2010). The response variable was the STRUCTURE Q values for the M. edulis genotype for each site. A preliminary beta regression model was run including all the explanatory variables Q_valuesEdulis ~ mean_SST + max_SST + min_SST + mean_salinity + max_salinity + min_salinity + mean_significant_wave_height + max_significant_wave_height + min_significant_wave_height. Variance inflation factors were used to check for collinearity of the explanatory variables with the vif function from the car R package (Fox and Weisberg 2019). The best fitting beta regression model was selected with the dredge function from the MuMIn R package to perform automated model selection (Bartoń 2024), using subset to exclude collinear variable combinations (e.g., min, max and mean SST). To confirm that the selected model explained significantly more variance than an intercept‐only model, a likelihood ratio test (LRT) was performed by comparing the log‐likelihoods of the full and null models. Model diagnostics were performed by examining residuals and leverage values (hat values) using the base R function hatvalues (package stats). The model coefficients were expressed relative to the original scale of the response variable using the inverse logit transformation with the inverse.logit (Canty and Ripley 2022; Davison and Hinkley 1997). Model predictions were visualised using ggplot2 and ggeffects (Lüdecke 2018) R packages.

To investigate the relationship between genetic divergence and ocean current resistance, a beta regression model was used in which pairwise F _st was the response variable and resistance category (factor3 levels) was the explanatory variable, using betareg. LRT and model residuals were also examined. Negative F _st values were replaced by 0.001 to meet beta regression assumptions (Wenne et al. 2022).

A post hoc testing of the estimation of marginal means was performed using the R package emmeans (Lenth 2025).

Finally, a more comprehensive beta regression model was run with pairwise F _st as the response variable, and current resistance category (factor with 3 levels), absolute difference in min SST (Δ°C), max salinity (ΔPSU) and max wave height (Δm) between sites as explanatory variables. The analysis steps followed the same procedure described above, including selection of the best‐fitting model, LRT, diagnostic checks (residuals), and post hoc testing of estimation of marginal means.

3. Results

3.1. Descriptive and Summary Statistics

Preliminary single marker genotyping (Adhesive protein gene, Inoue et al. 1995) confirmed the expectations of the low‐resolution power compared to the multi‐marker approach (SNPs panel) by misidentifying 17% of M. edulis , 13.6% of M. galloprovincialis and 50.5% of mixed ancestry (Figure S1, data available upon request).

Explorative analyses based on DAPC and Admixture showed that samples from the Baltic and the Adriatic Seas were genetically distinct from the Irish populations, in line with the genetic lineages of blue mussels in those regions: M. galloprovincialis for the Adriatic Sea, and M. trossulus for the Baltic Sea around the Askö area (Stuckas et al. 2017; Vendrami et al. 2020). Furthermore, no evidence of M. trossulus genetic ancestry was detected along Irish coasts (Figures S2 and S3). However, the sample size of the Baltic population is small compared to the rest of this study, and further sampling in this region would be necessary to confidently assess the presence or absence of M. trossulus introgression in Ireland. Consequently, the Baltic and the Adriatic populations were excluded from the downstream analyses that focused exclusively on the Irish populations.

Allele frequencies calculated per locus within each population identified six SNP loci that were monomorphic for the Irish populations and the Adriatic population. Additionally, two loci that were not strictly monomorphic had a minor allele frequency < 0.01. These eight SNP loci were excluded from subsequent analysis.

Global F _is values calculated for each locus across all populations identified 16 loci with F _is values that were either > 0.3 or < −0.3, suggesting deviations from Hardy–Weinberg expectations. These 16 loci were then closely analysed at a population level; for 12 populations, more than 50% of this subset of loci had F _is values beyond the threshold. These 12 populations (Inish Meáin from the Aran Islands—hereafter Aran Islands—AI, Clew Bay—CB, Killary Fjord—KF, Mulroy Bay—MB, Snave Bantry Bay—SBB, Waterford Estuary—WF, Malahide—MH, Cork Harbour—CKH, Kilmakilloge—KLM, Dungarvan Harbour—DH, Bertraghboy Bay—SCBB and Roaringwater Bay—RWB) in fact showed a consistent level of admixture (see Figure S3 and Section 3.2).

To discard SNP loci with high or low F _is unrelated to admixture, global loci F _is were recalculated excluding these 12 admixed populations (i.e., mixed ancestry from both genetic backgrounds for > 20% Mathiesen et al. 2017). Ultimately, three loci for which the F _is values exceeded the |0.3| threshold for more than 50% of the non‐admixed populations were removed, resulting in a final panel of 72 loci (see Table S1).

Global population level F _is calculated across all 72 loci varied between 0.006 (Wicklow—WK) and 0.22 (Cork Harbour—CKH) (details in Section 3.2).

3.2. Population Structure of Irish Mytilus spp.

The DAPC was computed for all 26 Irish populations with the final panel of 72 loci (Figure 2 for DAPC), in which the first two eigenvalues accounted for 73.97% of the total variance, with linear discriminant 1 (LD1) accounting for almost 58.5% and LD2 for over 15%. The Irish populations clustered into four groups that overlapped to different degrees. The most distinct group separated along LD1 included two populations from the west of Ireland (Aran Islands—AI and Bertraghboy Bay—SCBB) and one from the North‐east (Dunseverick—NI12). Another distinct group consisted mainly of populations from the east coast but also included Cromane Farm (CRF) and Cromane Wild (CRW) from the southwest, which clustered with this group. Dún Laoghaire Marina (DLM), North Bull Wall (NBW) and Malahide (MH) on the east coast were slightly separated on the LD2 axis from the main east of Ireland cluster. The fourth group lay between the Aran Islands (AI), Bertraghboy Bay (SCBB) and Dunseverick (NI12) cluster and the east of Ireland and comprised mainly sites from the west and south coast of Ireland.

FIGURE 2.

DAPC based on 72 SNP loci. (A) Individual‐level scatterplot, where each point represents a mussel individual. Colours indicate sampling sites, grouped by Irish coasts: Grey/black (west), yellow/orange (south), blue (east) and green (north, Republic of Ireland and Northern Ireland, UK). The percentage of variation explained by each discriminant function is shown on the axes. (B) Population‐level centroids from the same analysis. Each dot represents the average position of a sampling site in the discriminant space, labelled with its location code (see Table 1).

The different grouping displayed by the DAPC showed a similar pattern in the STRUCTURE analysis, with the best K = 2 (Figure 3). The ancestry of individuals from Aran Islands (AI), Dunseverick (NI12) and Bertraghboy Bay (SCBB) was mainly composed of the M. galloprovincialis SNP genotypes, with an average Q value of 0.97 for both Aran Islands and Dunseverick, and 0.87 for Bertraghboy Bay. Populations from the east coast of Ireland, Cromane Farm and Cromane Wild showed a composition of mainly M. edulis SNP genotypes, with Q values > 0.9. Finally, samples from the west and south coast of Ireland showed a different proportion of both genotypes, with composition of M. galloprovincialis genotype ranging from 0.2 to 0.4. Therefore, four main groups can be identified from the DAPC and three groups with different ancestry composition can be identified from the STRUCTURE analyses: one group with a genetic make‐up of mainly M. galloprovincialis genotype, one group of exclusively M. edulis genotype and a third group that showed a composition of both genotypes, which indicates populations with different levels of admixture (from the west and south coast of Ireland—Clew Bay CB, Killary Fjord—KF, Kilmakilloge—KLM, Roaringwater Bay—RWB, Snave Bantry Bay—SBB, and one site from the north—Mulroy Bay MB).

FIGURE 3.

Results from STRUCTURE with best K = 2 to investigate the population structure, genetic admixture and ancestry inference of the Irish samples. Each column represents an individual, and individuals are grouped by populations on the X axis. The Y axis indicates the ancestry coefficient (Q value), and columns are coloured proportionally according to the composition of each of the two genotypes. Details on population codes are presented in Table 1.

Allele richness (A _r), observed heterozygosity (H _o), expected heterozygosity (H _e), and inbreeding coefficient (F _is) calculated per population and averaged by genotype categories as indicated previously (Section 2.7), are reported in Table S2. Among populations, A _r ranged from a minimum of 1.28 (Dún Laoghaire Marina—DLM), to a maximum of 1.93 (Clew Bay—CB), H _o ranged from 0.06 (Lough Foyle—LF) to 0.31 (Aran Islands—AI and Dunseverick—NI12), H _e spanned from 0.07 (most of East coast populations and Cromane Farm—CRF) to 0.32 (Aran Islands, Dunseverick and Bertraghboy Bay—SCBB), and F _is from −0.02 (Dún Laoghaire Marina—DLM) to 0.22 (Cork Harbour—CKH). When looking at those indexes among the different genotypes, H _e, H _o, and A _r were higher in admixed populations and M. galloprovincialis (0.23–0.32, 0.20–0.3 and 1.85–1.886, respectively), while M. edulis populations showed substantially lower values (0.08, 0.075 and 1.39, respectively), which indicates much less genetic diversity. F _is varied between genotypes, with M. edulis having 0.08, M. galloprovincialis 0.11 and admixed populations 0.16. Effective population size (N _e) calculated per population is reported in Table S3, and it ranged from a minimum of 0.9 (i.e., Cork Harbour) to infinite values (four populations; set at ‘1000’ to be log₁₀ transformed). Ne was lowest (< 10) in Waterford Estuary (WF), Cork Harbour (CKH), Bangor (NI13), Dungarvan Harbour (DH) and Roaringwater Bay (RWB). The highest values (> 50) were found in Lough Foyle (LF), Carlingford Lough (CL), North Bull Wall (NBW), Aran Islands (AI), Cromane Farm (CRF), Killary Fjord (KF), Mulroy Bay (MB), Rosslare (RSW), Wicklow (WK), Wexford Harbour (WX), Dún Laoghaire Marina (DLM), Dunany Point (DP) and Dunseverick (NI12). Collectively, populations dominated by M. edulis genotype showed greater N _e compared to the M. galloprovincialis and mixed ancestry ones. When comparing populations for H _o and N _e (Figure 4), lower H _o and higher N _e values were observed in mostly east coast populations (Lough Foyle—LF, Carlingford Lough—CL, North Bull Wall—NBW, Dún Laoghaire Marina—DLM, Dunany Point—DP, Arklow Wind Farm—AWWF, Rogerstown—RT, Rosslare—RSW, Wexford Harbour—WX and Wicklow—WK) and Cromane sites (CRW and CRF), which suggests large populations with limited gene flow. Higher H _o and lower N _e were detected mainly in the west coast and south populations with consistent levels of admixture (i.e., Aran Islands—AI, Clew Bay—CB, Killary Fjord—KF, Bertraghboy Bay—SCBB, Kilmakilloge—KLM, Roaringwater Bay—RWB and Cork Harbour—CKH), which may indicate a recent admixture. Higher H _o coupled with higher N _e was detected in Dunseverick (NI12), potentially indicating large and genetically diverse populations. Finally, lower H _o coupled with lower N _e, which suggests strong genetic drift or inbreeding was observed in Bangor (NI13), Dungarvan Harbour (DH) and Waterford Estuary (WF).

FIGURE 4.

Observed heterozygosity (H _o) (left Y axis, grey bars) and log₁₀ transformed effective population size (N _e) (right Y axis, red line) for the sampled Irish populations. N _e values that were ‘infinite’ were set at 1000 for the log₁₀ transformation. H _o for Malahide and North Bull Wall was not possible to calculate. Grey shade background colour indicates sites from the west coast of Ireland, yellow background colour from the south coast, blue from the east coast and green from the north coast (Republic of Ireland and Northern Ireland, UK).

The pairwise F _st ranged from 0 to 0.56; the adjusted nominal level (5%) for multiple comparison with Bonferroni correction was set at 0.000154 (Figure 5). Overall, both the F _st values and the significance of the p‐value supported the separation of Irish mussels into four major groups. Aran Islands (AI) and Dunseverick (NI12) were significantly different from the rest of the locations, excluding Bertraghboy Bay (SCBB). Most of the locations on the east coast of Ireland were significantly different from the locations on the west coast (except for Cromane sites) and the Aran Islands, Dunseverick and Bertraghboy Bay. Finally, Malahide (MH), North Bull Wall (NBW) and Dún Laoghaire Marina (DLM) were found to be significantly different from most of the east coast locations, suggesting a substructure within this ‘pure’ Irish M. edulis genotype of the east coast.

FIGURE 5.

Heatmap of pairwise F _st with p‐values that are significant highlighted in bold. F _st and p‐values were calculated in F _stat where p‐values adjusted nominal level (5%) for multiple comparisons (Bonferroni) is 0.000154.

3.3. Mytilus edulis Genotype Populations in Irish Mussels

The DAPC (Figure 2) and the pairwise F _st (Figure 5) analyses of Irish populations (Section 3.2) revealed further genetic structure among the Irish east coast samples of ‘pure’ M. edulis genotype (i.e., admixture proportion < 0.2). STRUCTURE analyses performed on two datasets (Figure S4 and Figure 6) identified the second dataset, which contained only ‘pure M. edulis ’ individuals, as the most appropriate to investigate the intraspecific population structure (12 populations). The DAPC (Figure S5) and STRUCTURE analyses (best K = 2; Figure 6) showed that mussels from Dún Laoghaire Marina (DLM) and North Bull Wall (NBW) grouped separately compared to the other populations, with an average ancestry coefficient for Cluster 1 of 0.1, while for the rest of the populations it ranged from 0.49 to 0.62.

FIGURE 6.

Results from STRUCTURE with best K = 2 to investigate the population structure of the ‘pure’ Irish Mytilus edulis genotype at a finer intraspecific level. Each column represents an individual, and individuals are grouped by populations on the X axis. The Y axis indicates the ancestry coefficient (Q value), and the columns are coloured proportionally according to the composition of each of the two intraspecific clusters.

3.4. Isolation by Distance and Correlation Between Genotype Composition, Pairwise F _st and Environmental Variables

The results of the beta regression showed that pairwise genetic differentiation (F _st) was significantly associated with ocean current resistance categories, as confirmed by the likelihood ratio test (LRT) (χ ² = 14.91, df = 4, p = 0.0006; pseudo R ² = 0.046) (Table 2). Post hoc tests confirmed that site pairs in the high‐resistance category (cat0) had higher F _st values (mean predicted F _st 0.197) than sites in the moderate current resistance category (cat1, mean predicted F _st 0.184), and site pairs in the low‐resistance category (cat2, mean predicted F _st 0.135). The only significant difference in F _st was detected between cat0 and cat2. These results are in line with the general assumption that population pairs with little connectivity via oceanographic pathways present a higher genetic differentiation. This pattern was observed in the IBD model (Figure 7), where the west and east coasts appeared to function as source regions, with east coast sites potentially receiving gene flow from the north, and the west coast sites contributing to the southwest, but not receiving from any other region. In contrast, the south region emerged as a sink, likely serving as the end point of connectivity from both the east and the southwest coast.

TABLE 2.

Model results from beta regression analyses showing the top‐ranked models identified by the dredge function.

Model	Best fitting model	Pseudo‐R‐squared	Predictors	Model estimates (on logit scale)	p
Environmental	Q( M. edulis ) ~ min SST + max salinity + max wave height	0.24	Max salinity	−0.05	0.03*
			Max wave height	−0.89	< 0.001***
			Min SST	−0.14	0.12
IBD	F st ~ connectivity	0.05	Cat0	−1.41	< 0.001***
			Cat1	−0.08	0.68
			Cat2	−0.45	< 0.001***
IBD + environmental	F st ~ connectivity + Δwave height	0.14	Cat0	−1.94	< 0.001***
			Cat1	−0.10	0.60
			Cat2	−0.38	< 0.001***
			ΔWave Height	0.70	< 0.001***

Note: For each model, the pseudo‐R ² is reported along with the parameter estimates (logit scale) and p‐values for predictors (* for significant p‐values). Response variables include ancestry coefficient for Mytilus edulis (Q values) and pairwise F _st. Predictors include minimum sea surface temperature (min SST, in °C), maximum salinity (max salinity, in PSU), maximum significant wave height (max wave height, in meters), connectivity (categorical variable describing site‐to‐site connections in the IBD model), and absolute difference of maximum wave height between site pairs (Δwave height). Environmental variables were retrieved from the Regional Operational Model (ROMS) for the Northeast Atlantic (horizontal spatial resolution 1.9 km; temporal resolution of min SST and max salinity: 2017–2023; max wave height: 1994–2023). Significance codes: ***p < 0.001; **p < 0.01; *p < 0.05.

FIGURE 7.

‘Sink‐source’ network map resulting from the IBD model showing the relative connectedness of the sites based on the mean current resistance between them. (A) map of connection to each region, (B) map of connection from each region. White dots are starting points (site a), red dots are end points (site b), and the thickness and transparency of the connection lines indicate the inverse of the seacost distance (i.e., the combination between currents and coastal distances): The thicker and darker the line, the better the connection is between site a and site b.

The results of the environmental factors beta regression showed that maximum wave height (max wave height), minimum sea surface temperature (min SST) and maximum salinity (max salinity) were the environmental variables that best explained the distribution of the M. edulis genotype across populations (Figure 8). All three variables together explained almost 24% of the total variability of genotype composition (pseudo‐R ²), and the LRT confirmed that the model including these predictors fit the data significantly better than the null model (χ ² = 137.24, df = 3, p < 0.001) (Table 2). Maximum wave height was highly significant, maximum salinity was moderately significant, while minimum SST was not significant (all three showed negative effect sizes). These results collectively showed a negative correlation with M. edulis genotype composition, meaning that M. edulis genotype was more frequent at sheltered, cooler and fresher (i.e., lower salinity) sites. As shown in Figure 8, Waterford Estuary (WF) had much lower salinity than all other sites. Therefore, the leverage of the model, and an additional model was run without that site to check if max salinity was still a good predictor (Figure S7); the results corroborated the original model.

FIGURE 8.

Scatterplots displaying the correlation between Mytilus edulis genotype proportion (Q values, Y axis) and environmental variables (X axis). (A) is maximum wave height (m), (B) is minimum sea surface temperature (°C), and (C) is maximum salinity (PSU). Red lines are beta regression predictors, with confidence interval shaded in pink. Grey dots are individuals' Q values, and coloured dots are the averaged populations' Q values. Colours indicate regions (blue east coast, yellow south coast, black west coast and green north coast), and details on population codes are presented in Table 1.

The results of the final beta regression model (Figure 9) showed that between‐site genetic differentiation was significantly related to ocean current resistance and absolute difference in maximum wave height (inverse logit of the model estimate = 0.22, Table 2). The model explained 14% of the variation (p < 0.001; pseudo R ² = 0.14). Post hoc tests confirmed that category 0 had the highest F _st, followed by cat1, cat2. Overall, genetic differentiation was highest between sites unconnected by ocean currents and most distinct in terms of maximum wave height.

FIGURE 9.

Scatterplots displaying the correlation between pairwise F _st (Y axis), and (A) connectivity category, and (B) absolute difference in maximum wave height between sites (Δm). Connectivity category is the categorical variable describing site‐to‐site connections in the IBD model, where category 0 is unconnected, sites in category 1 have moderate connection and sites in category 2 have good connection. Red dots in (A) and red lines in (B) are model‐predicted marginal effects from the beta regression, with confidence interval shaded in pink. Grey dots are pairwise F _st values from the data.

4. Discussion

4.1. Genetic Structure and Connectivity of the Mytilus Complex in the North‐East Atlantic: An Irish Case Study

This study provides the first SNP‐based population genomics investigation of the Mytilus species complex in Ireland, representing a key addition to regional efforts across the North‐east Atlantic. Because we used a set of multiple SNPs shared with other studies of hybrid zones in European waters (Hammel et al. 2021; Simon et al. 2018; Wilson et al. 2018) our results can be interpreted in the context of previously published patterns of Mytilus spp. genetic structure. Collectively, the results confirmed a distinct genetic structure among Irish coasts and highlighted how the use of a multi‐marker compared to a single‐marker approach can provide greater accuracy in resolving fine‐scale genetic structure, especially in admixed populations. In addition, this study explored how environmental and oceanographic features may shape genetic differentiation. While we refer to ‘connectivity’ throughout, we use this term to describe potential connectivity inferred from patterns of ocean current resistance and genetic differentiation.

Mussels from the east coast were composed almost exclusively of the M. edulis genotype, with evidence of intraspecific structuring in Dún Laoghaire Marina and North Bull Wall, and very limited admixture from M. galloprovincialis . In contrast, populations from the south and west coasts showed an increasing gradient of admixture: locations in the south‐east and south (i.e., Waterford Estuary, Cork Harbour and Dungarvan Harbour) presented few individuals with higher Q values of M. galloprovincialis genotype, but a higher and more consistent proportion of M. galloprovincialis ancestry was observed in populations in the south‐west and west. This pattern suggests that pure M. edulis populations dominate the Irish Sea, while more admixed populations are found along the southern and western coasts. This genetic structure is also supported by the pairwise F _st analysis and is in accordance with previous studies that used the single marker approach (i.e., Adhesive protein gene; Inoue et al. 1995) to show similar patterns that have remained relatively stable since the 1970s, with a clear distinction between the east coast of Ireland, dominated by M. edulis , and the M. galloprovincialis admixed populations prevailing in the south and west coasts (Gosling et al. 2008; Gosling and Wilkina 1981; Lynch et al. 2020). This distribution pattern that resembles a clinal gradient differs from the one observed in other M. edulis and M. galloprovincialis European hybrid zones: both in SW England and in the French Atlantic coast hybrid zones are characterised by a patchy composition often referred to as mosaic hybrid zones (Diz and Skibinski 2024; Fraïsse et al. 2016).

Beyond this broad clinal pattern, specific populations deviated from the expected structure gradient. The populations sampled in Cromane (southwest) were composed almost entirely of M. edulis despite their geographic location in a region where admixed populations are common. Similarly, populations from the Aran Islands (more specifically, Inis Meáin), Bertraghboy Bay (west coast) and Dunseverick (north coast) were dominated by M. galloprovincialis genotype individuals, in contrast to their neighbouring sites that showed a more admixed composition. These pockets of ‘pure’ M. edulis and M. galloprovincialis genotype populations within the admixture region along the west coast were not detected in previous studies (Coghlan and Gosling 2007; Gosling et al. 2008; Lynch et al. 2020). This discrepancy could be the result of patchy sampling coverage, the lack of overlap between the exact sites in this study and those in previous studies for these particular populations, or the higher resolution power of the SNP panel approach compared to the single marker method. As discussed by Larraín et al. (2019), multi‐locus approaches outperform single‐locus methods for characterising Mytilus spp. populations. In this study, a comparison between SNP panel genotyping and a single‐marker approach (adhesive protein gene for preliminary genotyping developed by Inoue et al. 1995, data available upon request) showed that while the single‐marker method is effective in populations with pure M. edulis ancestry (e.g., populations from the east coast), it produced significant discrepancies in populations with admixture, including those dominated by the M. galloprovincialis genotype. This highlights that while the single‐marker method is useful as a cost‐effective preliminary screening, this approach can misidentify species, especially for introgressed individuals. Thus, when resources are available, resolving genetic structure in admixed populations should be performed using a multi‐marker approach.

Oceanographic currents, their seasonal variability, and geographic distance are major drivers of genetic connectivity and structure in marine organisms with a pelagic larval stage (Coscia et al. 2020; Gilg and Hilbish 2003; Robins et al. 2013). Coastal currents around Ireland follow a broad clockwise direction from south to north‐west, and a gyre located in the Irish Sea often leads to self‐recruitment, that is, when the settlement of offspring is produced by the same population rather than by larvae dispersed from other populations (Brown et al. 2003; Emsley et al. 2005; Horsburgh and Hill 2003; Nolan et al. 2023; Sponaugle et al. 2014). The impact of the Irish Sea gyre on mussel population structure has been discussed by Gosling et al. (2008) and Robins et al. (2013), who hypothesised that the distinct M. edulis genotype population on the east coast could result from self‐recruitment due to the thermal fronts north and south of the Irish Sea. These fronts may act as soft barriers, limiting the influx of Mytilus spp. larvae, particularly during the spring spawning season.

In this study, the IBD model, which incorporates coastal distances weighted by oceanographic current resistance, revealed a clear pattern of relative isolation of the eastern and western sites, aligning with the Irish ocean currents system described above (Brown et al. 2003; Nolan et al. 2023). While the southern populations appeared well connected to both eastern and southwestern sites, the east coast primarily received input from the north, whereas west coast populations were largely isolated within the region, acting as sources only to south‐west locations. This pattern suggests potential local self‐recruitment driven by site‐specific hydrodynamic processes (Robins et al. 2013). This large‐scale pattern was significantly correlated with the pairwise F _st in the beta regression model, where greater distances and challenging dispersal paths corresponded to higher genetic differentiation. Similarly, a strong isolation by distance pattern was observed in the broadcast spawning species Cerastoderma edule in the Irish Sea (Coscia et al. 2020), supporting the idea that IBD is a primary driver of large‐scale genetic differentiation in broadcast spawners. However, at shorter distances and along lower resistance paths, the correlation between geographic distance and genetic differentiation was less pronounced. For example, the ‘Sink‐source’ network map (Figure 6) indicated potential connectivity between Cromane sites and the Aran Islands, a pattern not corroborated genetically. As reported by Demmer, Neill, et al. (2022), local oceanographic interactions, including wind‐driven and tidal dynamics, are essential for understanding finer‐scale larval dispersal and self‐recruitment in Irish mussel populations. Even in the absence of strong oceanographic barriers, site‐specific environmental conditions along with competition with local populations (i.e., M. edulis ) might prevent M. galloprovincialis from establishing in some areas, such as Cromane (Coscia et al. 2013; Knights et al. 2006). More in‐depth studies in these pure M. edulis and M. galloprovincialis sites among admixed regions could provide more insight into the processes that determine genetic composition.

The environmental drivers that shape the genomic composition of blue mussels, especially in the context of hybrid zone dynamics, have long been a focus of research, with fine‐scale environmental dynamics often playing a key role (Beaumont et al. 2004; Kijewski et al. 2019). In this study, environmental data for the spawning season (March–May) across 12 years identified maximum wave height and maximum salinity as the strongest predictors of M. edulis genotype composition. Wave exposure showed a significant negative association, followed by salinity, while sea surface temperature had a weaker, non‐significant effect. These results align with previous studies across European hybrid zones (including the Irish Atlantic coast), where the M. galloprovincialis genotype tends to dominate in wave‐exposed habitat, while M. edulis is more common in sheltered, freshwater‐influenced sites (Bierne, David, Langlade, and Bonhomme 2002; Gosling and Wilkina 1981). However, this pattern is not universally observed. For example, Hilbish et al. (2002) reported inconsistent spatial patterns in genotype distribution, and later studies in Ireland challenged the generality of these associations (Coghlan and Gosling 2007; Gosling et al. 2008).

The genotype‐environment association in Mytilus is more complex than previously assumed, likely influenced by local self‐recruitment patterns and the genetic signature of source adult populations. Coghlan and Gosling (2007) found no difference in the genetic composition between spat and adult populations, suggesting limited selective sorting at early life stages. Moreover, Lukić et al. (2024) reported no significant impact of wave exposure on M. edulis growth and survival under controlled experimental conditions. Additionally, other studies have linked fluctuations in M. galloprovincialis abundance along the Irish Atlantic coast to broader environmental shifts, such as rising sea surface temperatures (Gosling et al. 2008); however, colder spells and increased freshwater input seem to play a role in reducing the abundance of this genotype (Lynch et al. 2020), highlighting the role of episodic environmental variation. In this context, the results of this analysis reflect these observations: although wave height was the strongest predictor among the environmental variables investigated, other factors such as salinity and sea surface temperature might still influence shaping the local genomic composition of adult blue mussels.

Our model explained only 24% of the variation in genotype composition and was limited by a coarse resolution, especially in bays and fjords. As a result, the model reflects broader environmental patterns rather than capturing fine‐scale local dynamics. Moreover, other anthropogenic impacts can play a substantial role in the connectivity of mussel populations, shaping their genomic makeup (Simon et al. 2019). While the analysis of anthropogenic drivers (i.e., transplantation of mussel seed and spat, from wild seabed to farm bottom growth facilities; Bord Iascaigh Mhara 2024) is beyond the scope of this study, it remains an important avenue for future research.

To further explore how environmental factors interact with geographic isolation, we tested pairwise genetic differentiation (F _st) in relation to both IBD and environmental distances. Our model showed that sea current resistance has the greater impact on population genetic differentiation in Mytilus spp. around the coast of Ireland, followed by wave height. This is consistent with what has been observed with cockle species in the Irish and Celtic Seas (Coscia et al. 2020). As discussed in Coscia et al. (2020), neutral genetic structure can strongly link with geographic and hydrodynamic dispersal potential, while environmental variables (such as SST) often are linked with non‐neutral loci, giving insight into potential drivers of adaptive divergence. In the current study, we did not test explicitly for adaptive divergence at our loci, but future work should strive to distinguish between adaptive vs. neutral loci, as the former class may exhibit stronger associations with environmental variables.

4.2. Genetic Diversity of Irish Blue Mussels

In this study, a clear pattern of genetic diversity was observed across Irish Mytilus spp. populations, enabled by the resolution provided by multi‐marker SNPs data. Despite being native to Irish coasts, M. edulis populations exhibit reduced genetic diversity but greater effective population sizes, while M. galloprovincialis and admixed populations maintain greater genetic diversity, likely shaped by gene flow and hybridisation. M. galloprovincialis genotype populations exhibited the highest levels of allelic richness (A _r), observed heterozygosity (H _o) and moderate effective population size (N _e), followed by admixed populations, with M. edulis genotype populations showing the lowest values for A _r and H _o. The effective population size resulting from this study is similar to the one observed by Gurney‐Smith et al. (2017), in which farmed mussel populations composed of mixed ancestry between M. edulis and M. galloprovincialis showed lower Ne compared to wild M. edulis . Moreover, the allelic richness and observed heterozygosity results are consistent with the variation in genetic diversity observed across Mytilus spp. populations by Vendrami et al. (2020), who also reported a higher genetic diversity in M. galloprovincialis genotype populations compared to M. edulis . Furthermore, they observed that introgression of M. galloprovincialis ancestry in M. edulis populations had a weak but positive effect on H _o level, suggesting that genetic diversity in admixed populations can be influenced by both the genetic background and the patterns and magnitude of introgression. The genetic diversity observed in Irish blue mussels aligns in part with that reported by Mathiesen et al. (2017). In both studies, admixed populations of M. edulis and M. galloprovincialis genotypes exhibited similar values of A _r and H _o; however, a consistent difference emerged when comparing pure M. edulis and M. galloprovincialis genotype populations. Specifically, M. edulis populations in the Arctic region exhibited a higher diversity than the Irish M. edulis populations, while Irish M. galloprovincialis populations displayed higher genetic diversity compared to the reference Atlantic M. galloprovincialis included in the Mathiesen et al. (2017) study. These results suggest that Irish M. galloprovincialis and admixed populations maintain greater genetic diversity compared to pure Irish M. edulis populations, potentially due to differences in gene flow. The lower diversity in Irish M. edulis populations may reflect limited connectivity and high self‐recruitment as discussed in the previous section.

The inbreeding fixation index (F _is), calculated across all loci for each population, was positive for most populations, with higher values in admixed and M. galloprovincialis genotype populations. The elevated F _is in admixed populations may reflect substructuring (Wahlund effect) if populations with different levels of introgression are pooled together (De Meeûs 2018). A similar pattern was observed in the southeast English hybrid zone, where Diz and Skibinski (2024) reported elevated F _is in hybrid populations. They suggested that multiple backcrossing and asymmetrical introgression between the parental genotypes could contribute to positive F _is values. To better understand the observed combination of higher genetic diversity and positive F _is in Irish admixed populations, future studies should aim to disentangle fine‐scale patterns of hybridisation and their relationship with F _is. This could involve investigating individual ancestry proportions and F _is, temporal variation in hybridisation dynamics and potential selection pressures acting on hybrids.

Collectively, these results indicate that despite being historically native to the Irish coast, Irish M. edulis populations may experience increased genetic isolation and relatively lower genetic diversity within the overall Irish context. In contrast, M. galloprovincialis and admixed populations exhibit higher genetic diversity and larger effective population sizes. As reported by Fraïsse et al. (2014, 2016) and Simon et al. (2018), hybridisation in Mytilus spp. populations could lead to a variety of introgression patterns (i.e., detrimental or beneficial effects), which play a central role in genetic diversity and adaptive potential. Genetic diversity and adaptive potential are key aspects in the aquaculture management framework, especially in the context of changing climate conditions (Brauer et al. 2023). As observed in Crassostrea gigas (Pacific oyster), crossbreeding between pure selected lines could increase genetic diversity and produce phenotypically superior offspring (Liang et al. 2023), at least in F1 hybrids owing to heterosis. However, these benefits may be short‐lived: in subsequent generations, recombination between divergent genomic backgrounds could potentially result in outbreeding depression and reduced fitness. Moreover, it is important to highlight that hybridisation in Mytilus spp. should be investigated on a case‐by‐case basis by employing larger panels of SNPs or Genome‐Wide approaches (e.g., RAD‐seq, WGS), as introgression could lead to unfavourable outcomes in certain hybrid zones, which may be detrimental for aquaculture (Dias et al. 2009; Dias, Piertney, et al. 2011; Gurney‐Smith et al. 2017; Nascimento‐Schulze et al. 2021).

4.3. Conclusion and Future Perspective

This study presents the first application of a 72 SNPs panel to investigate Mytilus spp. population genomics across 26 Irish populations, providing valuable genetic data for the North‐east Atlantic Ocean, and confirming a clear distinct genetic structure among the Irish coasts of previous studies. The multi‐marker approach offers a higher resolution of genetic admixture, diversity, isolation‐by‐distance and environmental associations. While regionally focused, these results contribute to broader knowledge on blue mussel hybridisation dynamics, population structure and environmental adaptation in temperate coastal ecosystems, and aquaculture management under climate change.

While this study provides a significant step towards better understanding the population structure of blue mussels in Irish waters, some limitations remain, including uneven geographic coverage and relatively small sample sizes. Thus, future work should address the following: (i) expand spatial sampling to monitor changes in Mytilus spp. population structure, especially in unsampled areas (e.g., Galway Bay, Sligo coast); (ii) targeted sampling of Irish M. galloprovincialis genotype populations, including key locations such as the Aran Islands, Bertraghboy Bay and Dunseverick; (iii) investigating genotype composition across different aquaculture practices; (iv) investigate fine‐scale local environmental‐genetic interactions by collecting high‐resolution environmental data; (v) including high‐resolution environmental and ocean circulation models to investigate connectivity and gene flow patterns; (vi) conducting genome‐wide association analyses to identify putative non‐neutral loci and hence better assess adaptive potential and (vii) investigating how different genotypes respond to environmental stressors to anticipate future responses to climate change.

Funding

This work was supported by Bord Iascaigh Mhara (BIM) (RFT150321), European Commision, Marine Institute (Ireland), Department of Agriculture, Food and the Marine (DAFM, Ireland), Atlantic Technological University (Ireland).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1: eva70185‐sup‐0001‐supinfo.docx.

Cariolato, E. , Reed T., Brophy D., Graham C. T., Lucy F. E., and Mirimin L.. 2025. “Population Genomics and Connectivity of the Blue Mussel Species Complex: Insights From a North‐East Atlantic Hybrid Zone.” Evolutionary Applications 18, no. 12: e70185. 10.1111/eva.70185.

Data Availability Statement

The SNPs genotyping data that support the findings of this study are openly available in the Dryad Digital Repository at https://doi.org/10.5061/dryad.t1g1jwtgj.