Introduction

Phylogeography considers the temporal and spatial genetic structure within a species. However, most studies focus solely on the spatial dimension not including the dynamics of the reproductive output which could be revealed by signatures in the temporal variation of the genetic structure of populations (Berger et al. 2006; Barshis et al. 2011; Haye et al. 2021). Because populations with variable reproductive output show genetic structure over time (Flowers et al. 2002; Siegel et al. 2008; Marshall et al. 2010), the temporal genetic structure of populations could inform about past population reproductive dynamics modulated by reproductive output and provide an indirect assessment of the variability in reproductive success of populations. This is particularly relevant in marine species inhabiting variable environments.

Many marine benthic species broadcast-spawn their gametes to the water column and have a planktonic larval developmental phase of variable duration. Long reproductive larval development has been associated with high variance in reproductive output (Li and Hedgecock 1998; Pineda et al. 2010; Hedgecock and Pudovkin 2011), as a response to environmental variations linked to differential survival, and patchy and stochastic larval recruitment (Jackson et al. 2017; McKeown et al. 2017; Villacorta-Rath et al. 2018). These species are prone to displaying large variance in reproductive success and stochasticity in reproductive events (Li and Hedgecock 1998; Hedgecock and Pudovkin 2011; Pérez-Portela et al. 2018; Vendrami et al. 2021), which can generate greater population genetic differentiation between year classes and body size than between spatially discrete populations (Hilbish et al. 2002). This dispersal and connectivity pattern (Lee and Boulding 2009), known as the chaotic genetic patchiness paradigm (Johnson and Black 1984; David et al. 1997; Eldon et al. 2016), is associated with patchy and stochastic larval and recruitment survival, collective dispersal, and sweepstakes reproductive success (Hedgecock 1994; Flowers et al. 2002; Bierne et al. 2003; Vadopalas et al. 2012; Owen and Rawson 2013; Kesäniemi et al. 2014; Eldon et al. 2016; Plough et al. 2016; Jackson et al. 2017; Villacorta-Rath et al. 2018). Consistently, studies have shown that species with high dispersal potential (planktonic larval dispersal >2 weeks) usually display high temporal genetic instability of populations, because of variations in the reproductive output of local populations temporarily changing the local gene pool, according to the chaotic genetic patchiness paradigm (e.g., Ng et al. 2010; Jackson et al. 2017; Villacorta-Rath et al. 2018; Salloum et al. 2019; Vendrami et al. 2021).

Mytilus chilensis Hupé 1854 is a Chilean endemic marine blue mussel found in protected areas in the southeast Pacific on the Chilean coast (Barria et al. 2012; Molinet et al. 2015; Larraín et al. 2017) with a geographical distribution from the Tirúa River (38°S) to Punta Arenas (53°S) in the Magellan province (Toro et al. 2004, 2006; Astorga et al. 2020; Gardner et al. 2021), in which individuals form discrete dense beds of sessile adults at 10 to 25 meters of depth (Brattström and Johanssen 1983). The species’ natural populations have been extensively exploited, leading to the collapse of natural populations, as has been suggested for other commercial resources in the area (Gelcich et al. 2010).

Although there are other species of the genus in the geographic area (i.e., M. edulis and M. galloprovincialis), and other mussels in sympatry (Choromytilus chorus, Aulacomya atra, Semimytilus algosus, Perumytilus purpuratus), M. chilensis is the preferred species for aquaculture in the southeast Pacific because it provides better growth outcomes (Díaz et al. 2019). Currently, natural populations of M. chilensis contribute with less than ~1% of the harvested biomass, the 99.66% remaining being provided by the mussel aquaculture (390,000 tons on average) (Sernapesca 2020, 2021). In consequence, a significant socio-ecological system has been structured around the mussel aquaculture in the inner sea of Chiloé and Reloncaví Sound in northern Patagonia (Molinet et al. 2017; Fernández et al. 2018; Gonzalez-Poblete et al. 2018; San Martin et al. 2020). This system strongly depends on natural beds for the collection of spat required for aquaculture. Understanding temporal variability is particularly relevant, given that a few sites have been used for the collection of spat, such as southern Chiloé and different areas of the Reloncaví Sound (Fig. 1). The geographic area where more than 99% of the mussel aquaculture facilities are placed coincides with the occurrence of most natural beds of M. chilensis in northern Patagonia, between 41°30’S and 43°10°S in the Reloncaví Sound and the Inner Sea of Chiloé (Sernapesca 2020, 2021). Areas of spat collection vary in time according to spat availability (Winter et al. 1984; Uriarte 2008; Molinet et al. 2017, 2021); from ~15 years ago, larval abundance and recruitment of M. chilensis have shown a strong temporal and spatial variability (Avendaño et al. 2011; Barria et al. 2012; Campos and Landaeta 2016; Lara et al. 2016), becoming a main concern for the sustainability of the mytiliculture social-ecological system.

Fig. 1: Map of sampling sites of natural beds of Mytilus spp. in the southeast Pacific coast of Chile (northern Patagonia area).
figure 1

Sampling in these natural beds was done by local fishers and were replicated during four consecutive Austral summers (2014-2017).

Full size image

The reproductive cycle of M. chilensis includes broadcast spawning of gametes and external fertilization, followed by several free-living larval stages with a total duration of 30–45 days in the water column (Toro and Sastre 1995; Toro et al. 2004), providing the species with a high dispersal potential by larval connectivity (Bohonak 1999). Most of the population genetic structure studies of M. chilensis have described moderate to low genetic differentiation among mussel beds in northern Patagonia, including the Reloncaví Sound and the Inner Sea of Chiloé (Toro et al. 2004, 2006; Larraín et al. 2015; Araneda et al. 2016; Astorga et al. 2018, 2020). When considering the range of distribution including southern Patagonia (south of 51°S), significant genetic differentiation has been broadly reported between northern and southern Patagonia (e.g., Araneda et al. 2016; Astorga et al. 2020; Díaz-Puente et al. 2020), showing also that northern Patagonia is likely a single gene pool.

Taxonomical identification in the genus Mytilus can be challenging, particularly using traditional morphological and molecular DNA barcoding tools. For example, the mitochondrial marker (mtDNA) Cytochrome Oxidase I (COI), a standard and highly validated tool for DNA barcoding (Valentini et al. 2009), could not be a proper tool for taxonomical validation in the Mytilus species complex due to a lack of resolution among close species of Mytilus (Tinacci et al. 2018), the double uniparental inheritance (Garrido-Ramos et al. 1998) and the hybridization and introgression among Mytilus species (Beaumont et al. 2008; Toro et al. 2012). Altogether evidence suggests that a single mtDNA marker could not be enough information for taxonomical identification and validation of Mytilus spp. (Larraín et al. 2019). Furthermore, Guisti et al. (2020) questioned the validation of the endemic species in the Southeast pacific coast (Mytilus chilensis), suggesting that this is a lineage of M. galloprovincialis not significantly distinct from the northern hemisphere Mediterranean mussel. However, recent studies have provided robust evidence through single and multi-locus approaches of a distinct native Chilean mussel, M. chilensis, as a valid taxonomical unit, which is likely to dominate in the southeast Pacific coast (Larraín et al. 2019; Oyarzún et al. 2021; Araneda et al. 2021).

The goal of this study was to evaluate the interannual and spatial genetic structure of Mytilus spp. in northern Patagonia from six natural beds between 39°25’S, 73°13’W and 43°07’S, 73°42’W (Fig. 1), as an indirect measure of past variations in reproductive success, for 4 consecutive years. Considering the presence of syntopic species, hybridization, and introgression among species, we used COI as a phylogeographic tool to evaluate temporal and spatial genetic diversity and structure in the most abundant, widespread, and monophyletic haplogroup (lineage), putatively assigned as M. chilensis (M. cf. chilensis). Given that in northern Patagonia the spat availability is variable in space and time (Barria et al. 2012), temporal genetic population structure serves as a proxy of past variations in reproductive success and as a complementary approach to studies of variation in larval availability, reproductive output, and recruitment (Avendaño et al. 2011; Díaz et al. 2011; Barria et al. 2012), for understanding temporal reproductive variations.

Methods

Samples of Mytilus spp. were collected during four consecutive austral summers (March) from 2014 to 2017. To avoid overlapping generations, only juvenile-size individuals of ~2–3 cm of shell length were collected in each of the sites and years. Samples were obtained in situ by local fishers in six natural mussel beds located between Mehuín (ME; 39°25’S, 73°13’W) and Yaldad (YA; 43°07’S, 73°42’W) (Fig. 1 and Table 1).

Table 1 Sample size (N), COI and H1 genetic diversity of Mytilus cf. chilensis for sites/years including number of haplotypes (H), number of segregating sites (S), haplotype diversity (h), mean number of pairwise nucleotide differences (k), and nucleotide diversity (π).
Full size table

Mytilus spp. is considered as a species complex, with hybridization processes among sympatric and parapatric species (Beaumont et al. 2008; Toro et al. 2012). The overlapping morphological traits among species/lineages present in north Patagonia (M. edulis and M. galloprovincialis), particularly among juveniles, impedes robust morphological species identification.

Due to the biparental inheritance of mitochondrial DNA in Mytilus spp. (Rawson et al. 1996; Garrido-Ramos et al. 1998; Burzyński et al. 2006), although supposedly low in M. chilensis (Larraín et al. 2019) and given that the M-type (Male type) mitochondrial lineage can theoretically only be detected in gonadal and mantle tissues in adult Mytilus individuals (Gérard et al. 2008), only abductor muscle tissue from juveniles was collected to ensure obtaining only matrilineal mitochondrial lineages. Moreover, using small spat drastically reduces the probability of amplifying M-type mtDNA in males, as the latter is overexpressed in ripe adult male gonadal tissue. In juvenile somatic tissue, F-type is the predominantly overrepresented mtDNA type. In addition, the M-type and F-type mitochondrial lineages have an average divergence of at least 20% (Garrido-Ramos et al. 1998), which can be detected through simple visual inspection of DNA sequence alignment afterward.

Small pieces of 2–3 mm3 of fresh posterior abductor muscle were dissected from each individual and preserved in absolute ethanol at −20 °C until DNA extractions. Genomic DNA was extracted from 25 mg of tissue using the EZNA DNA extraction kit (Omega, USA) following the manufacturer’s protocol. To amplify a portion of the mtDNA COI, PCRs were performed using the universal primers HCO and LCO (Folmer et al. 1994).

We focused our phylogeographic analyses on the population genetic structure of the single and most widespread lineage or haplogroup in the study area. For this, the presence of different lineages was determined using genetic distance among COI sequence data for all sampled Mytilus spp. individuals (Fig. 2). For further population genetic analyses, only the single haplogroup that was frequent and widespread was included, which was putatively assigned to the Chilean endemic lineage, M. cf. chilensis. This approach was not designed as a taxonomical validation method but rather as a phylogeographic criterion for the analysis of populations of a monophyletic group. With this phylogeographic approach, issues with mislabeling in public databases such as GenBank (Tinacci et al. 2018; Giusti et al. 2020), lack of resolution of COI gene sequences to distinguish in certain cases between M. chilensis and M. galloprovincialis (Tinacci et al. 2018), and the potential introgressions among Mytilus species are avoided.

Fig. 2: Maximum likelihood unrooted phylogram of 281 COI haplotypes of the Mytilus spp.
figure 2

Complex, displaying three haplogroups putatively assigned to Mytilus cf. chilensis, M. cf. galloprovincialis and M. cf. edulis.

Full size image

To complement the mtDNA sequence data in subsequent analyses of temporal genetic diversity and spatial structure and to incorporate the information of another cellular compartment, sequence data of the nuclear gene H1 was included for individuals of the frequent and widespread COI haplogroup. To choose a nuclear sequence with variation at the population level, PCRs for several nuclear genes were tested with primers from the literature (data not shown) and with newly designed primers for H1 (H1-238F 5’-AAG GGT GTC TTC GTC GAG GT-3’ and H1-998R 5’-AGG GCT TTT GGT GGC TGA AT-3’) from aligned H1 sequences available from several species of the genus Mytilus in GenBank. Even though H1 sequences are not a commonly used marker for phylogeographic studies, we chose H1 because, in our assessment, it showed greater polymorphism than other nuclear sequence markers and great PCR and sequencing success.

PCRs for COI and H1 were carried out with the following conditions: 1× Buffer, 1.3 mM MgCl2, 0.2 μM of each primer, 0.8 mM dNTPs, 0.03 U/μL of Invitrogen Platinum Taq polymerase enzyme, 0.075 mg/mL BSA and 1 μL of DNA. The cycling conditions consisted of an initial denaturation at 94 °C for 10 min, 35 cycles of denaturation at 94 °C for 1 min, annealing for 1 min, and extension at 72 °C for 2 min, ending with a final extension at 72 °C for 13 min. PCR annealing temperatures were 40 and 57 °C, for COI and H1, respectively. Sequencing of purified amplicons was performed on an automated ABI 3730XL sequencer (Applied Biosystems). Obtained sequences were aligned using Geneious R10 (www.geneious.com).

To apply phylogeographic analytical methods designed for haplotypes to the H1 data (diploid), the nuclear H1 sequences were transformed into haplotypes using the Bayesian framework implemented PHASE 2.1.1 (Stephens et al. 2001) in the software DnaSP 6.0 (Rozas et al. 2017). The settings used for running PHASE were determined after several run tests with a different number of iterations and chains consisting of five chains of 10,000 iterations, a thinning interval of 10, and 10% Burn-in. Only haplotypes with posterior probability greater than 0.9 were used for analyses (e.g., Haye and Muñoz-Herrera 2013).

The following analyses were carried out separately for each marker (COI and H1). DnaSP 6.0 was used to obtain the genetic diversity measures for sites and years, including number of haplotypes, number of segregating sites, haplotype diversity, mean number of pairwise nucleotide differences, and nucleotide diversity). Sample sizes were evaluated to determine if they were enough to recover most of the genetic diversity within the sampled populations and years. For this, haplotype accumulation curves were constructed for each year and marker using the function haploAccum with a total of 10,000 permutations in the R 4.2.0 (R Development Core Team 2022) package SPIDER (Brown et al. 2012).

Genetic differentiation by sampling site and year was assessed with FST (Weir and Cockerham 1984), and the significance was assessed after false discovery rate correction (Q value < 0.05) (Benjamini et al. 2005). Pairwise FST among years was estimated both using haplotype frequencies and unique haplotypes in Arlequin 3.5 (Excoffier and Lischer 2010). Hierarchical evaluations of FST based on haplotype frequencies were used to compare and evaluate the effect of sampling site and sampling year in the population genetic structure, considering variation among sites within years and among years within sites, using analysis of molecular variance (AMOVA) in Arlequin 3.5.

Permutation-based multivariate analyses (PERMANOVA) were performed using the adonis function in the Vegan R package (Oksanen et al. 2020) to evaluate the decomposed variance of genetic distance matrices, with sampling site and years as crossed factors (instead of a hierarchical model such as AMOVA). For this, Bray–Curtis distance matrices based on haplotype frequencies were calculated in Vegan with the vegdist function for each marker. Obtained matrices were corrected using Hellinger’s Method with the decostand function, and significant values of PERMANOVA were assessed using 10,000 permutations.

Tests of congruence among distance matrices (CADM) were explored to evaluate the temporal stability of genetic differentiation (Legendre and Lapointe 2004) using site pairwise FST matrices for each year with the CADM.global and CADM.post functions implemented in the APE package (Paradis and Schliep 2019) in R. CADM was implemented using the pairwise FST matrices between sites for each year for both markers, with 10,000 permutations, to evaluate the congruence of the genetic distance matrix of each year against all the matrices of the other years through a Kendall coefficient of matrix concordance (W). A high and significant Kendall coefficient suggests a high concordance between year matrices (therefore temporal stability). In addition, extended Mantel tests were performed using a post hoc Holm’s correction using the CADM.post function. Considering that the null hypothesis was complete incongruence among matrices, higher Mantel values indicate greater concordance/correlation between matrices and, thus, temporal stability.

Results

We sequenced a total of 803 individuals of Mytilus spp. from 4 consecutive years in six natural beads along the Southeast Pacific coast. COI data revealed three haplogroups (Fig. 2 and Supplementary Table S1). One is highly frequent and widespread, putatively Mytilus cf. chilensis, that included 751 of the sampled individuals. The rest of the individuals corresponded to highly divergent and less frequent lineages assigned according to BLAST search to M. edulis (46 individuals, 106 mutational steps) and M. galloprovincialis (6, 21 mutational steps) (Fig. 2 and Supplementary Table S1), which displayed only haplotypes of a single frequency (with only one exception) and not randomly distributed among years and sites, being most frequent in the northern sites of Metri and Los Molinos, and particularly in years 2015 and 2016. All natural beds displayed more than one species of Mytilus in syntopy. Due to the low frequency of the two divergent haplogroups, phylogeographic analyses focused on the most frequent, widespread, and monophyletic linage, putatively M. cf. chilensis. Albeit this, all the posterior analyses based on COI data, were performed with all the sampled individuals, showing that the general patterns described in this study remain congruent when considering all Mytilus spp. collected (Supplementary Tables S1 and S5–S7).

The 751 individuals of the COI haplogroup identified as M. cf. chilensis consisted of an average of 30 individuals per sampling site and year, with 180–200 individuals per year, and were used to obtain partial sequences of the nuclear gene H1. The final COI and H1 datasets for M. cf. chilensis haplogroup contained 751 and 747 sequences of 610 and 691 base pairs of length, respectively (GenBank Accession numbers, COI MZ773651–MZ773880, and H1 MZ901916–MZ902024) (Supplementary Table S1). The COI sequences included 230 haplotypes, with a total of 121 polymorphic sites, of which 74 were informative for parsimony and 47 singletons. For H1, the 747 diploid sequences resulted in 1012 haploid sequences that corresponded to 111 haplotypes, with 77 polymorphic sites, 43 sites informative for parsimony and 34 singletons. The number of shared haplotypes between sites was 42 and 47 for the COI and H1 data, respectively, while the number of unique haplotypes for a single sampling site was 188 and 64, respectively. Overall, genetic diversity was high (Table 1); haplotype diversity values were, in all the cases, in the upper 15% (>0.85) of possible values. Simulations using haplotype accumulation curves suggest that the number of individuals sampled by year was not enough to account for the observed diversity, being this pattern is more pronounced in COI than H1 (Supplementary Figs. S1–S4). Extrapolated plateau of COI was reached with a sample size approximately three times greater than the one we used, and two times greater for H1. Number of segregating nucleotide sites was higher for COI as well as the nucleotide diversity that indicates less genetic distance within COI than H1 haplotypes. In general, diversity indices were higher in sites from the Inner Sea of Chiloé (i.e., PI, PU and YA).

The COI haplotype network had a star-like pattern with the most frequent haplotypes homogenously distributed among years and sites (Fig. 3). The seven most frequent haplotypes were present in all sites and years, and the four most common haplotypes always represented more than 50% of the individuals (Fig. 3). There was some variability of the haplotype frequencies among sites/years (Figs. 3 and 4). The year 2015, for example, showed a higher relative frequency of the less common haplotypes (still within the 14 most common ones) (Fig. 4). COI genetic differentiation between pairs of sites for each year was low (Supplementary Table S2), with FST values ranging from 0 to 0.176 between ME and MT in 2017. In 2014, the only significant FST values were those that included comparisons with the site LM; in 2015, the two significant values were restricted to population pairwise comparisons including the site PI; in 2016, most pairwise comparisons including ME and all the ones including YA were significant. Finally, the year 2017 showed the highest number of significant values across sites (Supplementary Table S2).

Fig. 3: Temporal and spatial networks of COI and H1 haplotypes in natural beds of Mytilus cf. chilensis.
figure 3

Median-joining networks of 230 COI (A, B) and 111 H1 (C, D) haplotypes of Mytilus cf. chilensis. In A and C, each color represents a sampling year (time) according, and in B and D, colors represent sampling sites (space). Each circle represents a haplotype, and the size is proportional to the number of individuals that display each haplotype. Each link between haplotypes corresponds to a mutational step and crossed lines in the links represent additional mutational steps. Acronyms for sample sites according to Fig. 1.

Full size image
Fig. 4: Temporal and spatial frequencies of COI and H1 haplotypes in natural beds in Mytilus cf. chilensis.
figure 4

Relative frequencies per sampling site and year of the 14 most frequent haplotypes of A COI and B H1 of Mytilus cf. chilensis. Acronyms for sample sites according to Fig. 1.

Full size image

The H1 haplotype network was characterized by one most frequent haplotype with the highest representation in all sites and years, and several additional shared haplotypes among sites and years which emerged subsequently one to each other from the most common one (Fig. 3). Excluding the most common haplotype, the next 13 most frequent haplotypes varied in presence and frequency between sites and years (Fig. 4). The pairwise FST values of H1 were mostly non-significant, the two significant values were from the year 2017 in pairwise values that included ME (Supplementary Table S2).

In general, there was a low genetic differentiation between years, with FST values from 0 to 0.047 (Supplementary Table S3). For COI, the sampling sites PU, and YA, showed significant temporal FST differentiation values in comparisons in LM between the years 2014 and 2015 and between 2015 and 2017, and in PU between 2016 and 2017. Differentiation using unique haplotypes UST showed significant values in ME between 2014 and 2015, in three comparisons in LM (2014–2015, 2015–2016 and 2015–2017), in PU between 2014 and 2017 and in YA between 2016 and 2017. For H1 no FST value was significant, and four values of UST were significant: in PI between the years 2014–2016 and 2014–2017, and in PU between 2014–2015 and 2015–2017.

AMOVA using year and site as a priori groupings, respectively, showed contrasting results among markers (Table 2). Most of the COI genetic variance was explained by differences within populations in both sites and years and genetic differentiation among years was non-significant (FCT = −0.004, p = 0.99), i.e., no significant temporal variation of the COI genetic variance. For H1, although only 0.33% of the genetic variance was explained by differences among years, the descriptive statistic FCT was significant (FCT = 0.003, p = 0.022), indicating that variation among years had a significant contribution to the H1 genetic variance. On the contrary, AMOVA grouping by sites instead of years in H1 led to non-significant values (FCT = 0.002, p = 0.240), suggesting that sampling sites did not contribute to the distribution of the genetic variance in H1.

Table 2 Analyses of molecular variance (AMOVA) for COI and H1 data of Mytilus cf. chilensis grouping by year and by site.
Full size table

PERMANOVA analyses using site and year as crossed factors were consistent with AMOVA results. For COI, there was significance when sites were considered as the fixed factor (Site F value = 1.299; p = 0.01; Year F value = 0.05, p = 0.820), while for H1 this result was inverted, i.e., the year was the most important factor to explain genetic variability (Year F value = 2.313; p < 0.001; Site F value = 0.128; p = 0.99) (Table 3). Similarly, CADM results for COI showed marginally significant Kendall coefficients of concordance (W) among distance matrices for analyses performed with the full data set of M. cf. chilensis (W = 0.39; χ2 = 22.150; p = 0.049) and using only shared haplotypes (W = 0.410; χ2 = 22.950; p = 0.042) (Table 4). The congruence of the genetic distance matrices among years suggests temporal stability for the COI data. In contrast, for H1, Kendall coefficient values of concordance were low and non-significant for the full data set (W = 0.266, χ2 = 14.888, p = 0.394) and shared haplotypes (W = 0.250, χ2 = 15.000, p = 0.452), suggesting temporal genetic variation of the H1 data (Table 4). A posteriori Mantel correlation values (CADM.post) for H1 were also incongruent, suggesting temporal genetic differentiation. Values ranged from −0.600 to 0.434 and were significant only for the comparison of 2015–2017 using shared haplotypes (Fig. 5). The year 2014 showed the lowest Mantel correlation values, ranging from −0.600 to −0.029, suggesting that in this particular year, haplotype frequencies were consistently incongruent with respect to all other years (Fig. 5).

Table 3 Permutation‐based multivariate analysis of variance (PERMANOVA) of COI and H1 data of Mytilus cf. chilensis.
Full size table
Table 4 Results of congruence among distance matrices analyses (CADM) comparing the pairwise FST matrices of each year with the rest of the years of COI and H1 data of Mytilus cf. chilensis.
Full size table
Fig. 5: Heatmap of a posteriori Mantel values of pairwise congruence among distance matrices (CADM) in Mytilus cf. chilensis for H1 data.
figure 5

Values and colors indicate the Mantel test statistic. A CADM with all haplotypes. B CADM with shared haplotypes. Significant p value with an asterisk.

Full size image

Discussion

With more than 800 individuals sampled from six natural beds through four consecutive years in the southeast Pacific, there was one most abundant lineage of Mytilus spp., putatively Mytilus cf. chilensis, which had high genetic diversity in comparison to natural populations of Mytilus worldwide (Zardi et al. 2007; Riginos and Henzler 2008; Han et al. 2016; Pickett and David 2018). Reduced genetic diversity can be expected with overfishing and leads to increased population vulnerability (Hauser et al. 2002; Pinsky and Palumbi 2014). The absence of signals of reduced diversity in M. cf. chilensis stands out in that a highly exploited species has a high degree of genetic diversity, a main requirement for the adaptation and resilience of populations (Bitter et al. 2019).

M. cf. chilensis showed to have homogeneous and well-connected natural beds and slight but significantly contrasting differentiation in both spatial and temporal genetic structure. The mtDNA COI and the nuclear H1 markers displayed different spatial-temporal genetic structure signals. Overall, COI data had a greater spatial genetic structure than the nuclear H1, and vice-versa, H1 showed a greater temporal genetic structure than COI.

The low but significant temporal genetic structure of M. cf. chilensis based on H1 could explain the described temporal variations in the availability of spat from natural beds (Avendaño et al. 2011; Barria et al. 2012; Molinet et al. 2021). Different temporal genetic structure signals between markers could be due to their different effective population size and degree of genetic diversity. Possibly, as shown by the haplotype accumulation curves, the COI haplotype diversity was not well-captured with the studied sample size. The effects of under-sampling such a high diversity, as the one detected for COI, albeit using a relatively high sample size, could also explain detected differences in the temporal genetic structure among studied genetic markers. Mito-nuclear discordance, i.e., different phylogenetic or phylogeographic signals among mitochondrial and nuclear markers, is a widespread and relevant phenomenon (Larmuseau et al. 2010; Toews and Brelsford 2012), normally attributed to differences in the linage sorting process between markers with different effective population sizes (e.g., Debiasse et al. 2014; Lee et al. 2021), and to life history traits such as dispersal, mating, and reproductive success (Toews and Brelsford 2012). However, it is improbable that linage sorting can explain the temporal population genetic variation of M. cf. chilensis. It is more likely a consequence of the high genetic diversity of COI when compared with H1 and enhanced by the different patterns of heredity of nuclear and mitochondrial genomes.

Another explanation for the slight temporal structure detected with the H1 nuclear marker could be habitat dynamics that affect variability in post-settlement processes, causing temporal reproductive segregation (Bierne et al. 2003). Analyses of the population abundance of M. chilensis indicate temporal variation and rapid changes in local abundance in the Reloncaví fjord (Molinet et al. 2015). In addition, local environmental conditions in the study area are highly variable in time, with seasonal fluctuations and intromission of warm sub-superficial waters (>10 °C) (González et al. 2011). The Northern Patagonia area, for example (from ~41.5°S to 43.9°S), has been described mainly as an estuary zone, with the influence of the interaction of freshwater from rivers and ice melting with oceanic water masses from the south Pacific (Sievers and Silva 2006), causing significant dynamics in salinity, chlorophyll (González et al. 2011; Narvaez et al. 2019), temperature, pH and oxygen (Sievers and Silva 2006; Narvaez et al. 2019). These temporal fluctuations may impose physiological challenges to populations of benthic marine organisms. In mollusks, variation in salinity and pH (acidification) affects respiration, growth, calcification, and reproduction rates (Navarro 1988; Gazeau et al. 2007; Duarte et al. 2014, 2018). In specific, for M. chilensis the capability and magnitude of trait responses depend on the environmental history of their native habitats to which individuals and ancestors have been exposed (Duarte et al. 2014).

Early studies and several following ones that aimed at detecting temporal genetic structure in marine species were performed with conservative allozyme loci and often were able to detect temporal genetic structure (David et al. 1997; Moberg and Burton 2000; Robainas-Barcia et al. 2008), and studies performed with sequence markers such as COI have also evaluated temporal genetic structure (Flowers et al. 2002; Haye et al. 2021). Given the temporal genetic population structure detected with H1, using an interannual scale evaluating persisting natural beds of Mytilus chilensis, this study prompts further analyses involving higher resolution methodologies that could more accurately depict population structure at a fine spatial scale (e.g., Ng et al. 2010; Villacorta-Rath et al. 2018; Quintero-Galvis et al. 2020). Although studies of temporal population genetic structure are often not comparable because they use different methodological approaches (i.e., focus species, geographic location, years of sampling, and molecular markers), available studies of marine species on the southeast Pacific provide somewhat contrasting results (Rojas-Hernandez et al. 2016; Quintero-Galvis et al. 2020; Haye et al. 2021). On the one hand, the brachyuran crab Metacarcinus edwardsii showed a pattern of temporal stability attributed to high dispersal coupled with high and temporally stable reproductive output (Rojas-Hernandez et al. 2016). On the other hand, the muricid gastropod Concholepas concholepas, also with high larval dispersal potential (>3 months), displayed high temporal genetic variation, likely linked to variations in reproductive success in response to the environment (Quintero-Galvis et al. 2020). Finally, the ascidian Pyura chilensis, which has low intrinsic dispersal potential and high potential for anthropogenic transport in ships and boat hulls, harbors three well-differentiated intraspecific lineages that have disparate patterns of COI temporal genetic structure. For this species, one of the lineages had a stronger temporal than spatial genetic structure, indicating reproductive and connectivity differences among species’ lineages (Haye et al. 2021).

In the genus Mytilus, temporal population genetic variation (e.g., among year classes) has been reported in coastal areas for M. edulis (Koehn et al. 1976; Pedersen et al. 2000; Hilbish et al. 2002; Gilg and Hilbish 2003; Simon et al. 2020), M. galloprovincialis (Gardner and Palmer 1998; Gilg and Hilbish 2003; Gilg et al. 2016), and M. trossulus (Pedersen et al. 2000). In general, like for M. cf. chilensis, Mytilus species from different geographic areas display low to moderate levels of temporal genetic structure, suggesting that the presence of a moderate reproductive success variation could be a widespread pattern for mytilids. Similarly, at a broader taxonomic scope, changes in the temporal genetic structure of populations have been widely reported, to different degrees, in different species of marine invertebrates with high dispersal potential, high fecundity, and in populations without geographic differentiation, including species of crustaceans (Robainas-Barcia et al. 2008; Jackson et al. 2017), gastropod mollusks (Salloum et al. 2019), and echinoids (Flowers et al. 2002; Calderón et al. 2012; Couvray and Coupé 2018).

The human-mediated spat movement likely potentiates the autonomous intrinsic high dispersal potential of M. chilensis given by its long-lived larvae, contributing to population connectivity, and maintaining highly diverse natural beds, as has been described for M. edulis in offshore installations in the North Sea (Coolen et al. 2020). A similar pattern was suggested by Norrie et al. (2020) for the mussel Perna canaliculus in New Zealand, in which larval spillover from farmed populations provided larval subsidy for the restoration of natural beds. Simon et al. (2020) also showed that for both M. edulis and M. galloprovincialis, there was a significant influence of human-mediated transport in population genetic connectivity. Mussel aquaculture could also promote temporal genetic structure by transporting spat from different source populations in time to an extensive geographic area where mariculture is developed. The enhancement of population connectivity driven by the mussel aquaculture similes the model of spillover subsidy of aquaculture farms and provides an opportunity for a network management approach of Mytilus spp. (Ludford et al. 2012; Norrie et al. 2020).

Geographic distribution of the genetic diversity has been widely studied in M. chilensis using different markers. Overall, studies show that the study area on the southeast Pacific coast is composed of connected and diverse populations (Toro et al. 2004, 2006; Araneda et al. 2016; Astorga et al. 2020; Díaz-Puente et al. 2020), although there is variation in the degree of structure detected among studies. Differences can be attributed to the different markers used (e.g., Araneda et al. 2016; Astorga et al. 2020), to the large temporal variation in reproductive success and environmental and anthropogenic local processes (e.g., Astorga et al. 2020), and to the aquaculture industry enhancing connectivity by translocating spat (Díaz-Puente et al. 2020). Temporal population genetic variation could contribute to the differences among phylogeographic and population genetic studies performed with samples obtained in different time periods and locations.

The low yet significant temporal genetic variation among years detected between distance matrices shows a greater differentiation of the year 2014 with respect to the other years. From 2015 on, haplotype frequencies of M. cf. chilensis using both markers significantly changed in the study area with respect to 2014. This change could be attributed to the strong El Niño Southern Oscillation (ENSO) event of 2015 (Blunden and Arndt 2016; Santoso et al. 2017; Molina et al. 2021); the pre- and post-ENSO change in haplotype frequencies could be linked to the environmental variability associated with ENSO events that can impact the reproductive success and connectivity dynamics of marine populations. A similar pattern of changes in the temporal genetic structure associated with ENSO was reported in a broad geographic area along the southeast Pacific for the ascidian Pyura chilensis, which also showed greater temporal genetic differentiation in the years 2015–2016, likely associated with the ENSO event of 2015 (Haye et al. 2021). Environmental variables, such as temperature and nutrients, display great variations during ENSO events (e.g., LaVigne et al. 2013; Santoso et al. 2017), affecting species’ reproductive traits (e.g., Cantillánez et al. 2005; Manríquez et al. 2018; Avila-Poveda et al. 2021). Particularly, and in agreement with interannual genetic structure, Narváez et al. (2019) showed that temperature anomalies of 2014 differed from 2015, 2016, and 2017 in the western and the eastern coasts of Chiloé Island (41–43°S). During ENSO events, the genetic pool of broadcasting species with reproductive output may get shuffled due to changes in reproductive success, larval development (e.g., Ruiz et al. 2008), and larval settlement (e.g., Fuentes-Santos and Labarta 2015). These results highlight the importance of interannual comparisons in genetic structure to reveal signatures in genetic diversity and population structure of strong stochastic events such as ENSO.

Despite the limitations of the markers used in terms of spatial resolution (in comparison with hypervariable markers), and taxonomical validation, our phylogeographic approach appears to be relevant for the study of the temporal and spatial genetic structure of a species highly demanded for human consumption, which represents a relevant social-ecological system, particularly for the northern Patagonian area in the southeast Pacific Coast.

Conclusion and final remarks

In phylogeography, studies often focus on the spatial structure using only one temporal scale, rarely incorporating the temporal dynamics of population genetic variation in nature, which is a proxy of past variations in reproductive success. Herein, temporal interannual comparisons of natural populations of M. cf. chilensis in northern Patagonia showed that the main haplogroup conforms to a connected gene pool with high genetic diversity and with some interannual variations in haplotype frequencies, which could explain fluctuations in natural spat availability. Results also highlight the need to analyze the spatial and temporal genetic variation of populations to shed light on population dynamics resulting from reproductive success, particularly on marine broadcasting species inhabiting highly variable environments prone to variable reproductive output and reproductive success. Special attention to population structure should be given to species, such as M. chilensis, that provide relevant social-ecological services. For adequate management of persistent mussel beds and sustainability of the mussel socio-ecological system, it is necessary to include both temporal and spatial monitoring of local populations of M. chilensis and their genetic diversity and differentiation, considering environmental heterogeneity and temporal variations, such as ENSO that could impact on reproductive success.