The Endangered Limpet Patella ferruginea Integrates a Metapopulation across the Species’ Range

Abstract

The population genetics of Patella ferruginea Gmelin, 1791, an endangered limpet endemic to the western Mediterranean, has been analysed using 11 polymorphic microsatellite markers on 533 individuals from 18 localities throughout its distribution area. The results showed a deficit of heterozygotes, denoting a certain degree of inbreeding, and, with an overall F_ST of 0.004, a low level of genetic variability among localities. These data indicate that the species is distributed as a metapopulation (an assemblage of discrete local populations with migration among them) along most of the species’ range. Moreover, 99% of the variability observed was within populations, with only 0.41% accounting for between-population variability. No pattern of isolation-by-distance was found, and 35.5% of the individuals were recognised as migrants. Altogether, the findings indicate that most of the populations studied are connected to each other to some extent and that larvae of the species show a higher dispersal capacity than previously assumed. The exchange network does not follow a clear direction but rather shows a chaotic pattern attributed to stochastic factors resulting from the complex interaction of biotic and abiotic factors. This pattern indicates the lack of strong barriers to dispersal in the study area and permeable barriers that do not limit population connectivity. A relatively high level of self-recruitment and occasional stochastic dispersal events at variable distances are also evidenced by the analyses. Currently, marine protected areas (MPAs) safeguard the benthic adults but not the larval phase of the species. Considering our results, the conservation of P. ferruginea should be based on a holistic approach in which the protection of its habitats extends from the benthic to the pelagic zones, which will help maintain the larval pool and promote larval dispersal and settlement and, ultimately, gene flow. Lastly, conservation efforts must prioritise the survival of the extant populations of P. ferruginea, both within and outside MPAs, over measures that require the manipulation or translocation of specimens.

1. Introduction

A huge body of evidence shows the widespread loss of biodiversity in the marine realm, particularly in coastal areas ([1,2], among many others). Anthropogenic disturbances, such as habitat destruction, fragmentation or deterioration, unsustainable overexploitation, pollution, introduction of exotic species and diseases, and global change are the main causes of “marine defaunation” [3]. The Mediterranean Sea is one of the marine areas where biodiversity is most quickly changing under the combined pressure of these factors [4,5].

In the face of such widespread biodiversity decline, understanding the genetic structure of populations and the underlying processes involved in connectivity is of critical importance for conservation management [6]. Advances in genetic and genomic tools have been providing new insight into the extent of genomic admixture among populations, which can aid the management of species, especially threatened ones, and, in turn, the conservation of biodiversity [7,8]. Genetic diversity and gene flow among populations ensure the evolutionary potential and viability of species by helping them withstand the effects of negative impacts on their long-term persistence and survival [9]. Preserving genetic diversity and gene flow requires knowledge of both the genetics and the specific life history traits of the species, as well as of oceanographic patterns, coastal topography, and habitat availability [10]. The interplay of these features can lead to complex patterns of genetic variation over space and time [11].

Here, we study the population genetics of the ferruginous limpet Patella ferruginea Gmelin, 1791, a gastropod endemic to the Mediterranean Sea that is among one of the most threatened marine invertebrates in this sea. It has been highlighted as a “species of Community interest requiring strict protection” in Annex IV of the Habitats Directive [12]. It is also listed as a “strictly protected species” in Appendix II of the Bern Convention [13], “endangered or threatened” in Annex II of the Barcelona Convention [14] and “in danger of extinction” in the Spanish Catalogue of Threatened Species [15]. In March of 2008, the “Strategy for the Conservation of Patella ferruginea in Spain” was approved, and it was updated in November of 2023 [16].

Patella ferruginea inhabits rocky coastal substrates (both natural and artificial) of the narrow upper midlittoral fringe in areas with medium to high wave exposure, high oxygen levels and low pollution levels [17,18,19]. This limpet was common and widely distributed in the western Mediterranean basin from at least the Pleistocene onward [20,21] and is known to have been collected since the Neolithic [22]. Since then, intense harvesting of the species by humans, facilitated by its large size and the accessibility of its habitat, has led to declines in local abundances and range fragmentation [20,21]. The species’ decline has worsened in the 20th century owing to this sustained harvesting, as well as habitat degradation, loss, and marine pollution [23,24,25,26]. The current distribution of the species is highly fragmented and formed mainly by a few relatively isolated and relict populations (compiled by [19]). Healthy or successful populations (with high densities and regular recruitment) are currently only found along the North African coast between the Strait of Gibraltar and Tunisia. Relict populations, which are characterised by low and highly variable densities and few specimens, have been recorded in Corsica and Sardinia, the south-eastern coasts of the Iberian Peninsula, Alboran and Pantelleria islands and in a few localities in Sicily and continental France and Italy [19,27,28,29,30,31,32].

Patella ferruginea has a slow growth rate, a long lifespan that can exceed 12 years and may reach lengths of up to 10 cm [21], although growth parameters show high variability [19]. It is a free-spawner with a bi-phasic life cycle comprising planktonic larvae and benthic juveniles and adults. This life cycle implies that, following fertilisation in the water column, dispersal occurs during the pelagic larval phase. Fertilisation success depends on the local concentration of sperm and eggs and hence on population density (degree of aggregation of conspecific adults). Under very low-density conditions, a local population or deme of a free-spawning species may become effectively reproductively sterile in that area (the so-called “Allee effect” [33]), becoming a ‘pseudopopulation’ that is maintained solely by larval input from other populations [34].

Several studies have described the reproductive biology of P. ferruginea [35,36,37,38,39,40]. According to these authors, it first matures as a male from ∼30 mm in maximum diameter, hence individuals less than 30 mm are usually considered juveniles. Guallart et al. [31] and Ferranti et al. [41] recently described the larval development of the species under laboratory conditions. According to Ferranti et al. [41], pediveliger larvae reach competence 3 to 4 days post fertilisation (dpf) (depending on water temperature); however, the effective settlement period is from 7 to 32 dpf but can last up to 40 dpf, delaying metamorphosis until the larvae reach an appropriate microhabitat for settlement. Although the length of the planktonic phase in nature remains unknown, based on these data, the distance that larvae can travel, potential barriers to dispersal and the possible origin and destination of migrants can all be investigated. According to Henriques et al. [42], limpet populations cannot be considered fully open because some local larval retention likely occurs. Estimating the degree to which a local population is open/closed or donor/recipient requires knowing the location of possible source populations, the factors involved in larval transport and detailed genetic studies [31]. We attempt here to address these aspects and the extent of the connectivity in P. ferruginea.

Several barriers influence the connectivity and distribution of Mediterranean marine species [43]. Monitoring larvae in the water column is highly challenging; therefore, detailed observations have not been reported for most benthic invertebrates. Population genetic studies based on molecular marker analyses are used as a proxy to understand processes related to connectivity ([44,45,46,47], among many others).

In previous studies of P. ferruginea, genetic diversity patterns were analysed using both mitochondrial and nuclear markers at different geographic scales (local and global) and yielded some contrasting results. Espinosa and Ozawa [48], based on their analysis of mitochondrial COI, showed a high level of genetic homogeneity among specimens from southern Spain and North Africa. However, local genetic studies carried out in Sardinia using inter-simple sequence repeat (ISSR) markers revealed genetic differentiation between populations from different zones and even between nearby populations [49,50,51]. In an attempt to resolve this contradiction, Casu et al. [52] analysed both ISSR markers and partial sequences of the mitochondrial COI, 12S and 16S genes from specimens distributed throughout the species’ range. Their results with COI showed high genetic homogeneity among all localities, in line with Espinosa and Ozawa [48], but the ISSR analyses showed a clear discontinuity between two large groups: the Sardinia and Corsica localities and the rest of the western Mediterranean localities. Within the first group, differentiation was detected between localities in NW Sardinia and those in NE Sardinia and Corsica. However, no differentiation was found among the populations of southern Spain and North Africa, including specimens from Pantelleria (in the Siculo-Tunisian Channel) and the Egadi Islands (SW Sicily). In a more recent study focusing on populations from the Algerian coast [53], genetic divergence, based on the analysis of a short fragment of COI, was detected between the eastern Algerian localities and the remaining W Mediterranean localities, which all resolved as closely related populations.

Machordom et al. [54] isolated and characterised 11 microsatellite markers for P. ferruginea from a set of specimens from the Chafarinas Islands (N Africa). Microsatellite markers are a useful tool for clarifying small-scale species connectivity patterns and identifying donor and recipient populations [55]. Cossu et al. [56], in a study analysing eight of the microsatellite loci previously developed [54], found some level of genetic diversity, a high percentage of self-recruitment, and dispersal barely exceeding 10 km for populations from Sardinia. Analysing the same microsatellite loci on more specimens from the Sardinian populations, Cossu et al. [56] subsequently showed that the power of detecting genetic differentiation among populations increases with sample size, with 30 or more samples being optimal. Additionally, in contrast to their previous study, the authors did not find evidence of a genetic structure along the Sardinian coasts.

Despite the species’ relatively short planktonic phase, and therefore assumed limited dispersal capacity and low connectivity among populations, the available information indicates that the larval dispersal capacity of P. ferruginea may be greater than previous thought, probably due to sporadic long-distance dispersal events [57]. Our main objective is to test this hypothesis: if the species has a higher dispersal capacity than previously assumed [21], we expect to observe high connectivity and gene flow among populations and a low level of genetic differentiation among them. Using the 11 microsatellites developed by Machordom et al. [54], we analysed the population genetics of samples from 18 localities throughout the species’ area of distribution and also estimated the migration rate to assess the degree that populations are close or open. Finally, we discuss source-sink patterns and make some suggestions for the conservation of the species.

2. Materials and Methods

2.1. Sample Collection

Over the course of nearly a decade (2005–2013), tissue samples were obtained from 533 mature individuals of P. ferruginea, each with a diameter greater than 30 mm. Sampling was conducted at 18 different localities along the western Mediterranean coasts and from islands such as Corsica or several near Spain, Morocco, Algeria, and Tunisia (Figure 1). Samples were collected only when a moderate to high number of specimens were found in each area, and sample sizes ranged from 14 (Cape Bon, Tunisia: BON) to 42 individuals (Ceuta, Strait of Gibraltar: CEU) (

Table 1. Location, code, coordinates and number of individuals (N) of the sampled populations of Patella ferruginea in the western Mediterranean Sea.

Locality	Code		Coordinates	N
Alborán Island (Spain)	ALB	1	35°56′26.89″ N 3°1′56.20″ W	35
Ceuta Port, S Gibraltar Strait (Spain)	CEU	2	35°53′43.76″ N 5°18′17.58″ W	42
Congreso Island northeast, Chafarinas (Spain)	LAR	3	35°10′39.90″ N 2°26′17.28″ W	24
Congreso Island southeast, Chafarinas (Spain)	LEV	4	35°10′46.75″ N 2°26′21.75″ W	22
Rey Island, Chafarinas (Spain)	FRA	5	35°10′50.11″ N 2°25′12.70″ W	34
Isabel II Island east, Chafarinas (Spain)	CIP	6	35°10′54.60″ N 2°25′37.71″ W	32
Isabel II Island west, Chafarinas (Spain)	CIW	7	35°10′53.41″ N 2°25′52.93″ W	24
Commune de Galéria (E Corsica, France)	COR	8	42°24′33.25″ N 8°36′5.86″ E	31
Motril Port, Granada (Spain)	GRA	9	36°43′8.24″ N 3°31′31.30″ W	30
Las Palomas Island, Algeciras Bay (Spain)	ISP	10	36°3′54.27″ N 5°26′4.78″ W	35
Kristel, Oran (Algeria)	KRI	11	35°50′13.79″ N 0°29′7.93″ W	31
Cape Tres Forcas (Morocco)	FOR	12	35°23′43.11″ N 3°0′17.32″ W	30
Sidi El Bachir (Morocco)	BAC	13	35°5′21.00″ N 2°31′46.64″ W	30
Cala Iris (Morocco)	IRI	14	35°9′5.72″ N 4°21′56.78″ W	32
Melilla Port (Spain)	MEL	15	35°17′35.98″ N 2°55′42.86″ W	35
Oran Port (Algeria)	ORA	16	35°42′45.93″ N 0°38′57.57″ W	35
Cape Bon (Tunisia)	BON	17	37°4′58.82″ N 11°2′31.31″ E	14
Kelibia Port (Tunisia)	KEL	18	36°50′2.70″ N 11°6′59.10″ E	17

). When possible, samples were taken without sacrificing the specimens following the protocol by Guallart et al. [58].

2.2. DNA Extraction and Microsatellite Amplification

DNA was extracted and purified from the sampled individuals using the QIAGEN Biosprint 15 DNA Blood Kit (Qiagen, Hilden, Germany). Each individual was genotyped with the 11 polymorphic microsatellite markers previously isolated by Machordom et al. [54]. Each PCR included template DNA (0.3 ng/µL), 1× buffer (2 mM MgCl₂), 0.5 µM of each primer, 0.2 mM of dNTPs and 0.25 U of Taq polymerase in a final reaction volume of 10 µL. The cycle was as follows: 94 °C for 5 min and 35 cycles at 94 °C for 30 s, 54 °C for 30 s and 72 °C for 30 s, with a final elongation at 60 °C for 30 min. To facilitate genotyping, the PCR products were fluorescently labelled using three techniques: direct forward tagging with Pf-31AH8 (VIC), Pf-D11A (6-FAM), Pf-G1M (PET), Pf-31ME8 (PET) and Pf-31IB1 (PET), direct reverse labelling with Pf-31IF2 (NED) and Pf-C10 (VIC) and nested PCR with Pf-31IFI (NED), Pf-G6A (6FAM), Pf-31IB2 (6FAM) and Pf-31IA5 (6FAM). The labelled products were sequenced with an ABI PRISM 3.730 DNA sequencer (Applied Biosystems, Waltham, MA, USA); the size standard used was GeneScan 500 LIZ (Applied Biosystems). Finally, GENEMAPPER v4.0 (Applied Biosystems) was used to analyse the electropherograms and obtain a matrix of the alleles corresponding to the 11 loci.

2.3. Analysis of Genotypes and Clonal Structure

The presence of possible null alleles in the data matrix was checked using MICRO-CHECKER v.2.2.3 [59]. To assess whether populations adhere to Hardy–Weinberg equilibrium (HWE), we used GenAlEx 6.5 [60] with both the original data matrix and the corrected matrix considering the null alleles. Statistical significance was corrected using the sequential Bonferroni method [61]. GenAlEx 6.5, with the multilocus matches option, was also used to analyse the clonal structure of the populations.

2.4. Genetic Variability

Four pairs of individuals with an identical genotype (possible clones) were detected (see results). These individuals were eliminated from the matrix to prevent them from interfering with the reliability of our results. GenAlEx 6.5 was used to calculate the observed (Ho) and expected (He) heterozygosity. Allelic richness, which can be affected by sample size, was estimated for a standardised sample size of 14 individuals per population (the lowest number of individuals sampled in the populations analysed) using the R package standArich_v1.00 [62]. Using GENEPOP v4.0 [63], we tested, for each population, HWE, obtaining the mean values of the inbreeding coefficient (F_IS) and the presence of linkage disequilibrium (LD).

2.5. Population Differentiation

To assess genetic differentiation between populations, we calculated Wright’s fixation index (pairwise F_ST) using Weir and Cockerham’s estimators in GENETIX v.4.03 [64]. We used this F_ST to perform principal component analysis (PCA) in GenAlEx to represent the amount of variation between the populations in our study and reveal a possible pattern of genetic structuring. Using the R package adegenet v2.1.1 [65], the F_ST values were also used to perform the discriminant analysis of principal components (DAPC), which evaluates the genetic variability between and within groups, optimising the variance between groups and minimising it within groups [66]. In addition, we obtained standardised F_ST values (F′_ST), assuming that each population has different alleles for each locus, and linearised F_ST (F_ST/1 − F_ST).

We next determined whether the observed genetic structure is conditioned by the geographical distance between populations, i.e., whether genetic isolation by distance (IBD) has occurred. For this, we performed a Mantel test [67] in GenAlEx with 10,000 permutations to identify any correlation between linearised F_ST values and logarithmically transformed geographical distances in kilometres. Geographical distances were calculated with GOOGLE EARTH PRO 7.3 by the shortest path across the sea, both along the coastline and by Euclidean distance.

The population genetic structure of P. ferruginea was assessed using a Bayesian approach in STRUCTURE 2.3.4 [68]. The model used considers correlated allele frequencies, locality as a priori information and HWE estimates, defines a series of genetic clusters (K) and determines the probability that an individual belongs to one of these clusters. The analysis, performed with K = 20 (two clusters more than the number of populations studied) and 20 replicates for each K, was run with 10,000 Markov chain Monte Carlo and a burn-in of 10,000 iterations. These results were used to determine the best value of K (Δk), using STRUCTURE Harvester (Earl and von Holdt, 2012), CLUMPAK [69] and Evanno’s method [70]. CLUMPAK was also used to identify the K for which Pr(K = k) is the largest using ln(Pr(X/K)) and to obtain the plots corresponding to the best value of K. According to [71], Evanno’s method may underestimate the number of clusters (K); therefore, we also used STRUCTURE SELECTOR [72] and Puechmaille′s method [71] to check the number of clusters. In addition, the molecular variance (AMOVA) of all the populations and groups inferred in the STRUCTURE analysis was quantified in ARLEQUIN v.3.11 [73], and the global F_ST value was obtained.

The probability that an individual in a population comes from that population (self-recruitment) or came from another was estimated using a Bayesian assignment analysis and a first-generation migrant detection analysis in GENECLASS2 [74,75]. The analyses were run with 10,000 permutations and an error of 0.05. For the assignment test, the criteria for computation was conducted according to the Paetkau et al. [76] simulation algorithm with a 0.05 score threshold. The default frequency for a missing allele was set to 0.1, and the type I error was established at 0.05. The number of simulated individuals was 10,000. These parameters were used for the Monte Carlo resampling probability computation methodology of Paetkau et al. [77]. The first-generation migrants were inferred using the “Detection of first-generation migrants” Bayesian method [75] under the following conditions: L = L_home/L_max; 10,000 simulated individuals; Type 1 error of 0.05 and the simulation algorithm of Paetkau et al. [77].

Finally, we used BOTTLENECK v1.2.02 [78] to evaluate possible recent reductions in effective population size using allele frequency data and three mutation models: the stepwise mutation model (SMM), the infinite allele model (IAM) and the two-phase mutation model (TPM) [79]. Analyses were performed with 10,000 iterations and a descriptor of the allele frequency distribution (“mode-shift”), which determines the type of distribution that each population presents. Two statistical tests, the Sign test [78] and the Wilcoxon signed-rank test [80], were used to assess each mutation model.

3. Results

3.1. Analysis of Genotypes and Clonal Structure

All loci analysed were polymorphic, except 31AH8B, which was monomorphic for the populations ALB, CIP, CIW, IRI, MEL and KEL, and 31IA5, which was monomorphic for GRA and ORA. According to the MICRO-CHECKER analysis, null alleles were detected in the populations. However, after the Bonferroni correction, no substantial differences were found between the original matrix and the one corrected for null alleles. Given this, and as null alleles can cause overestimates of the F_ST values [81], the corrected matrix was not used in further analyses. Regarding clones, four pairs were detected in the MEL (Melilla), FRA (Chafarinas), KRI (Oran) and ISP (Algeciras) populations. These were removed from the analyses as they may interfere with the reliability of the results.

3.2. Genetic Variability

Allelic richness values ranged between 5.896 (ISP) and 6.669 (CIW), and the mean value among the populations was 6.175 (

Table 2. Parameters of genetic variability by population. Na, standardised allelic richness; Ho, observed heterozygosity; He, expected heterozygosity; F_IS, inbreeding coefficient; * Populations deviating from Hardy–Weinberg equilibrium (p < 0.05). For population codes, see Table 1.

Population	Na	Ho	He	FIS
ALB	6.169	0.582	0.571	−0.007 *
CEU	6.039	0.568	0.588	0.046 *
LAR	6.325	0.511	0.576	0.131 *
LEV	6.260	0.566	0.565	0.021
FRA	6.240	0.526	0.575	0.088 *
CIP	6.448	0.531	0.573	0.087 *
CIW	6.669	0.574	0.588	0.045 *
COR	6.468	0.598	0.613	0.040
GRA	5.903	0.554	0.568	0.034 *
ISP	5.896	0.552	0.576	0.055 *
KRI	6.058	0.574	0.572	0.008
FOR	5.916	0.543	0.563	0.052 *
BAC	6.130	0.548	0.574	0.062 *
IRI	5.994	0.591	0.620	0.063 *
MEL	6.091	0.524	0.585	0.120 *
ORA	6.104	0.546	0.555	0.029 *
BON	6.455	0.653	0.607	−0.037
KEL	5.994	0.507	0.552	0.110 *
Mean	6.175	0.558	0.579	0.053

). Expected heterozygosity values ranged between 0.552 (KEL) and 0.620 (IRI), and those of observed heterozygosity between 0.507 (KEL) and 0.653 (BON). The mean values of expected and observed heterozygosity were 0.579 and 0.558, respectively. Only four populations (ALB, LEV, KRI, BON) had negative or very low F_IS values, indicating an excess of heterozygotes. The other 14 populations had positive F_IS values, indicating heterozygote deficits (or homozygote excesses) (Table 2). Thus, in general, populations of P. ferruginea have a deficiency of heterozygosity and, thus, a certain degree of inbreeding. Furthermore, most populations showed significant deviation from HWE but for only a single locus in most cases. Only FOR (Cape Tres Forcas) and CIW (in Chafarinas Islands) presented multiple loci that deviated from HWE, with three and two loci, respectively. Linkage disequilibrium (LD) was detected between two pairs of loci at the global level (31IB1 and 31IF2; and 31IB2 and 31IF2). However, this LD was not observed in all populations, likely owing to an inconsistency in the data; therefore, LD was discarded, and all loci were considered independent of one another.

3.3. Population Differentiation

The overall F_ST value indicated a low level of genetic differentiation (F_ST = 0.004) among populations, with many population pairs presenting negative F_ST values. Given the complexity of interpreting negative F_ST values, lacking clear biological meaning, and the potential indication of more genetic diversity within populations than between them, negative values were conventionally replaced with zeros (

Table 3. Values of F_ST (lower diagonal) and F’_ST (upper diagonal). Significant F_ST values are in bold (p < 0.05). See Table 1 for the population codes.

	ALB	CEU	LAR	LEV	FRA	CIP	CIW	COR	GRA	ISP	KRI	FOR	BAC	IRI	MEL	ORA	BON	KEL
ALB	0	0.010	0.001	0.025	0.046	0.000	0.005	0.039	0.006	0.000	0.002	0.012	0.000	0.060	0.005	0.003	0.023	0.038
CEU	0.004	0	0.000	0.015	0.041	0.000	0.000	0.031	0.000	0.000	0.004	0.000	0.000	0.031	0.000	0.000	0.035	0.000
LAR	0.000	0.000	0	0.013	0.029	0.000	0.000	0.043	0.000	0.000	0.000	0.008	0.000	0.019	0.000	0.000	0.033	0.000
LEV	0.011	0.007	0.007	0	0.057	0.017	0.010	0.034	0.000	0.016	0.036	0.011	0.002	0.049	0.007	0.017	0.010	0.030
FRA	0.003	0.000	0.001	0.000	0	0.033	0.040	0.090	0.021	0.027	0.006	0.051	0.049	0.106	0.048	0.033	0.041	0.083
CIP	0.000	0.000	0.000	0.009	0.000	0	0.000	0.041	0.000	0.000	0.000	0.000	0.000	0.035	0.000	0.000	0.020	0.013
CIW	0.001	0.000	0.000	0.006	0.000	0.000	0	0.018	0.005	0.000	0.000	0.000	0.000	0.033	0.000	0.000	0.017	0.010
COR	0.015	0.012	0.016	0.015	0.014	0.016	0.007	0	0.048	0.026	0.060	0.031	0.035	0.059	0.031	0.047	0.054	0.045
GRA	0.002	0.000	0.000	0.000	0.000	0.000	0.000	0.016	0	0.000	0.000	0.000	0.000	0.048	0.000	0.000	0.007	0.026
ISP	0.000	0.000	0.000	0.007	0.000	0.000	0.000	0.009	0.000	0	0.000	0.000	0.000	0.039	0.000	0.000	0.024	0.017
KRI	0.000	0.000	0.000	0.013	0.000	0.000	0.000	0.019	0.000	0.000	0	0.008	0.000	0.068	0.001	0.000	0.016	0.030
FOR	0.006	0.000	0.004	0.006	0.000	0.000	0.000	0.013	0.000	0.000	0.001	0	0.006	0.027	0.007	0.004	0.09	0.003
BAC	0.000	0.000	0.000	0.002	0.000	0.000	0.000	0.014	0.000	0.000	0.000	0.003	0	0.040	0.000	0.000	0.024	0.012
IRI	0.024	0.012	0.008	0.022	0.022	0.015	0.014	0.022	0.016	0.014	0.022	0.011	0.016	0	0.042	0.060	0.069	0.019
MEL	0.002	0.000	0.000	0.004	0.000	0.000	0.000	0.012	0.000	0.000	0.000	0.003	0.000	0.016	0	0.001	0.009	0.026
ORA	0.002	0.001	0.000	0.007	0.000	0.000	0.000	0.018	0.000	0.000	0.000	0.002	0.000	0.024	0.000	0	0.029	0.028
BON	0.009	0.014	0.014	0.005	0.004	0.009	0.006	0.019	0.003	0.010	0.007	0.011	0.008	0.025	0.003	0.012	0	0.071
KEL	0.015	0.000	0.000	0.014	0.013	0.005	0.004	0.018	0.007	0.005	0.007	0.000	0.004	0.009	0.009	0.010	0.027	0

), i.e., it is assumed that there is no genetic differentiation. Of the populations showing positive F_ST values, indicating some degree of genetic differentiation between them, the two Tunisian populations, BON and KEL (0.027), showed the highest level of differentiation. Given the relatively low values observed in the pairwise comparisons (<0.03), genetic differences among populations are overall low.

The PCA results showed that 37.20% of the F_ST variation was explained by the first axis and 24.92% by the second axis (62.12% considering the two axes). As shown in Figure 2, most of the populations clustered at the intersection of the axes; therefore, we concluded that they are genetically very similar to each other and included them in the same group. The other localities distinguishable from this group included BON and KEL (Cape Bon and Kelibia, Tunisia), COR (Corsica), IRI (Cala Iris, Morocco) and LEV (in Chafarinas Islands). The variation observed for these groups was not consistent with their geographical distribution. For example, BON and KEL were distant points on the graph but were only 30 km apart, whereas BON and LEV were relatively close points despite these populations being 1300 km apart.

The results of the DAPC, in general, support those of the PCA, with axis 1 and 2 accounting for 16.7% and 12.4% of the variation, respectively (Figure 3). Almost all the individuals were included in a core group, and somewhat differentiated from this group were the individuals from COR in Corsica (number 8 in Figure 3), IRI in Morocco (14) and BON in Tunisia (17). In this case, neither the Congreso Island north (LEV) nor the Kelibia (KEL) populations differed from the rest of the populations.

The results of the STRUCTURE analysis using Evanno’s Δk method showed three differentiated genetic groups (K = 3) (Figure 4): most of the populations belonged to the first group, except for some individuals from the three localities previously mentioned (IRI, COR and BON), the second group represented part of the ancestry of mainly IRI individuals, and the third was present mainly in individuals from the Corsica (COR) and Tunisian (BON) populations (Figure 4A). Using the lnProb and Puechmaille method, the most probable K value was two: almost all of the specimens belonged to the first group, with the other group’s contribution mainly found in the IRI population (Figure 4B).

To analyse the molecular variance (AMOVA), we used the genetic groups obtained in the STRUCTURE analysis, that is, K = 3 and K = 2. We also used K = 1, with all populations forming a group. For K = 1, 99.59% of the total variance was observed within populations, and the remaining 0.41% was between populations (

Table 4. Analysis of molecular variance (AMOVA) of the 18 populations of Patella ferruginea considering one (A), two (B) and three (C) genetic groups.

Genetic Groups	Source of Variation	df	Sum of Squares	Components of Variance	Percentage of Total Variation
A	Among groups	-	-	-	-
(K = 1)	Among populations	17	60.14	0.0118	0.41
	Within populations	1040	2962.04	2.8481	99.59
	Total	1057	3022.18	2.8599
B	Among groups	1	9.70	0.0542	1.86
(K = 2)	Among populations	16	50.44	0.0052	0.18
	Within populations	1040	2962.04	2.8481	97.96
	Total	1057	3022.18	2.9075
C	Among groups	2	18.43	0.0561	1.93
(K = 3)	Among populations	15	41.71	−0.0012	−0.04 (0)
	Within populations	1040	2962.04	2.8481	98.11
	Total	1057	3022.18	2.9031

A). For K = 2, similar values were observed. In this case, 1.86% of the variance was between the two delimited groups, which was more than the variance between populations of the same group (0.18%) (Table 4B). For K = 3, values similar to those for K = 1 and K = 2 were observed, with 98.11% of the variation found within populations, 1.93% between groups and no variance between populations within groups (Table 4C). Therefore, for all the K values considered within the population, genetic variation accounted for nearly all of the observed variation, with differences between populations (or subpopulations) contributing to a very small proportion.

The results of the Mantel test showed no evidence of isolation by distance (IBD). The value for the two measures of distance, Euclidean and coastline, was R² = 0.12 (p = 0.02) and R² = 0.11 (p = 0.03), respectively. Although the probabilities were significant (p < 0.05), these R² values were very low, indicating no correlation between genetic differentiation (F_ST) and geographical distance.

According to the GENECLASS assignment analysis, 341 individuals (64.5%) were assigned to their own population (self-recruitment), and 188 (35.5%) were assigned to other source populations, of which 59 were of unknown origin (

Table 5. Results of the GENECLASS assignment analysis. Individuals are arranged in rows according to their sampling location, and in columns according to their assigned origin. See Table 1 for the population codes. Own and unknown represent the origin of the population analysed or from populations not included here, respectively. In the columns, the first value corresponds to the number of individuals assigned to other localities, and the second represents the number of first-generation migrants.

Popula-tion	Own	Origin																			Total
		ALB	CEU	LAR	LEV	FRA	CIP	CIW	COR	GRA	ISP	KRI	FOR	BAC	IRI	MEL	ORA	BON	KEL	Unknown
ALB	24					1/0	2/0	3/0	2/0		1/1			1/1	1/1					0	11/3
CEU	34					1/2	1/1			0/1	0/1			1/1	1/1					4	8/7
LAR	14		1/0		0/1		1/1	1/0						1/0	1/0		1/0			4	10/2
LEV	12					0/2	1/1	1/0	1/0						1/0	1/0		2/0	0/1	3	10/4
FRA	20							2/0			2/1				1/0	2/0		2/1		4	13/2
CIP	20			1/1				1/1	1/1		1/1			0/1		2/1	2/0			4	12/6
CIW	15				1/1	1/0				0/1	1/0				0/1	1/0	0/1			5	9/4
COR	18				0/1		1/0	1/0		0/1		1/1			1/0			0/1		9	13/4
GRA	20						1/0	2/1	1/0					3/0					0/1	3	10/2
ISP	23						1/0		4/1				0/2	1/0	1/0					4	11/3
KRI	21	1/0	1/1				2/0		2/0				1/1			2/0				0	9/2
FOR	18		2/1	2/1			2/0	2/0	1/0		1/1			1/0	0/1					1	12/4
BAC	20		1/0			1/0		3/1			1/0				1/1	1/0				2	10/2
IRI	22		2/2							1/0	0/1		0/1	1/0			0/1	1/1		5	10/6
MEL	23		1/1				1/0	2/0	1/0		1/0						1/0	1/1		3	11/2
ORA	27		1/0				2/0			1/1	0/1	0/1	0/1		1/0	0/1				3	8/5
BON	3					0/1		1/0	1/0				1/1			3/0			0/1	5	11/3
KEL	7						1/0	1/0	1/1	0/1		1/2		1/0	1/0	4/0				0	10/4
TOTAL	341	1/0	9/5	3/2	1/3	4/5	16/3	20/3	15/3	2/5	8/7	2/4	2/6	10/3	10/5	16/2	4/2	6/4	0/3	59	188/65

). Particularly noteworthy populations included COR (in Corsica), which had nine individuals of unknown origin out of a total of 31 sampled, and BON (in Tunisia), which had 11 of the 14 individuals sampled belonging to the other or unknown populations. The populations to which most individuals were assigned from others were FRA and CIP (in Chafarinas Islands), COR (in Corsica), and FOR (Tres Forcas Cape, in Morocco). KEL (in eastern Tunisia) was the only population in which no individuals coming from other populations were found; however, the analysis of first-generation migrants recognised four individuals in KEL from three different populations. In any case, the numbers of individuals assigned were generally higher than those of first-generation migrants (Table 5), indicating a potential overestimation due to the existence of genetically close populations and a high level of gene flow and allele transfer between different populations.

Almost all the populations received first-generation migrants (Table 5). This is graphically represented in Figure 5, which shows the assorted exchange of individuals between populations, with CEU (along the southern coast of the Strait of Gibraltar), CIP (in Chafarinas Islands) and IRI (in Morocco) receiving the most, and ISP (along the northern coast of the Strait of Gibraltar) and FOR (in Morocco) donating the most. Moreover, the data indicate the lack of first-generation migrants from the ALB (Alboran Island) population.

Finally, the results of the bottleneck analysis revealed that most of the populations experienced a relatively recent reduction in population size (

Table 6. Results of the BOTTLENECK analysis, using the Sign and Wilcoxon tests, for three mutation models: IAM (infinite allele model), TPM (two-phase mutation model) and SMM (stepwise mutation model). The type of distribution of each population is listed under mode-shift, and significant values (p < 0.05) are in bold. See Table 1 for the population codes.

	Sign Test			Wilcoxon Test			Mode-Shift
	IAM	TPM	SMM	IAM	TPM	SMM
ALB	0.434	0.210	0.017	0.846	0.275	0.005	Normal
CEU	0.544	0.041	0.001	0.831	0.083	0.001	Normal
LAR	0.343	0.053	0.012	0.765	0.054	0.007	Normal
LEV	0.340	0.047	0.001	0.240	0.016	0.001	Normal
FRA	0.321	0.009	0.008	0.700	0.067	0.002	Normal
CIP	0.183	0.060	0.002	0.432	0.084	0.005	Normal
CIW	0.403	0.170	0.000	0.492	0.084	0.001	Normal
COR	0.445	0.125	0.000	0.465	0.067	0.005	Normal
GRA	0.308	0.073	0.000	0.432	0.084	0.001	Normal
ISP	0.452	0.125	0.008	0.898	0.175	0.005	Normal
KRI	0.590	0.144	0.010	0.966	0.147	0.007	Normal
FOR	0.309	0.039	0.008	0.638	0.021	0.005	Normal
BAC	0.335	0.010	0.000	0.577	0.009	0.000	Normal
IRI	0.150	0.180	0.015	0.160	0.557	0.010	Normal
MEL	0.411	0.375	0.060	0.846	0.432	0.014	Normal
ORA	0.432	0.066	0.002	0.846	0.105	0.003	Normal
BON	0.217	0.421	0.451	0.123	0.765	0.700	Normal
KEL	0.428	0.202	0.002	0.770	0.232	0.010	Normal

). In both statistical tests of the SMM, which is the best-fit model for the data analysed here, all the populations, except MEL in Melilla Port (only significant difference for the Wilcoxon test) and BON in Tunisia (for which no evidence of bottleneck was found), showed values significantly different than those expected for populations in mutation/drift equilibrium.

4. Discussion

The results presented here, particularly the low level of genetic differentiation among populations (F_ST = 0.004), indicate that P. ferruginea is distributed as a metapopulation (defined as an ‘assemblage of discrete local populations with migration between them’ [82]) in the study area. Our analysis of highly variable markers in 533 specimens from 18 localities across most of the range of distribution of the species, with the exception of the Tyrrhenian Sea, reveals a relatively high level of connectivity among local populations, with most of the populations having exchanged larvae and, consequently, genes. The pattern of this exchange network is chaotic, indicating non-directionality and, at least for the larvae of P. ferruginea, the lack of strong barriers to dispersal, but rather permeable ones that do not limit connectivity between populations.

Although many models of larval dispersal have been published (e.g., [83] and references therein), the stochastic nature of larval dispersal and connectivity among coastal marine populations may not be the exception but the rule [84]. This stochasticity may be driven by the complex interaction between coastal circulation, shoreline configuration and life-history traits of species. This source of uncertainty has often been overlooked in larval connectivity models of coastal marine populations [85]. Indeed, the east–west pattern of surface circulation along the western Mediterranean coast in December (when P. ferruginea larvae disperse), as described by Martínez et al. [86], could allow connectivity among most of the known populations of the species. However, the exchange network observed here does not show any clear direction of dispersal that would be consistent with the surface circulation pattern. It is possible that coastal configuration plays a major role in the dispersal process than previously considered; however, this hypothesis remains to be tested.

With regard to the life-history traits of P. ferruginea, Laborel-Deguen and Laborel [20,87], in their pioneering works, suggested the species has a short larval phase and limited dispersal capacity owing to the relatively large size of the eggs. Nevertheless, according to Guallart et al. [31], the main reproductive traits of this limpet (egg size and larval development) hardly differ from those of other common Mediterranean limpet species (i.e., P. rustica Linnaeus, 1758 and P. caerulea Linnaeus, 1758) [88,89]. This suggests that relevant differences in larval development and duration of pelagic phase among these species are not expected, and as such, these traits cannot be considered a biological constraint for the ferruginous limpet. Considering a planktonic larval phase of 3 to 7 days in P. ferruginea [31], Guallart et al. [57] suggest that the main “hotspots” in the SE Alboran Sea, separated by a distance of about 100 km or less, could facilitate connectivity among the North African populations, depending on the environmental conditions, which may vary from year to year. Interestingly, Ferranti et al. [41] found that, in the laboratory, settlement may occur up to 40 days after fertilisation in P. ferruginea. The latter findings suppose that the veliger larvae of P. ferruginea can disperse more widely and over longer distances (at least occasionally) than previously assumed. Thus, a genetic exchange may occur between populations separated by hundreds of kilometres, generating a homogenisation of genetic material. Our results corroborate this hypothesis, as do those of studies of other related species. For instance, Ribeiro [90], in a study on dispersal and connectivity of some European limpet species in relation to the general system of currents, estimated that larval dispersal can exceed 200 km, although the number of larvae decreased sharply with distance, and Sá-Pinto et al. [91] pointed out that the duration of the larval stage of related Patella species P. ulyssiponensis and P. rustica can be up to 35.5 days. Furthermore, as Guallart et al. [57] noted, some isolated specimens of P. ferruginea have been found more than 200 km away from established populations, such as in the Hormigas Islands (eastern Spain) [28,92] and Liguria (north-western Italy) [30], evidence potential sporadically long-distance dispersal events in this species. Although we report a substantial percentage of self-recruitment in the analysed populations, as also observed by Cossu et al. [56], 35.5% of the individuals were recognised as migrants. Therefore, eventual stochastic and asymmetrical dispersive events at variable distances may also happen. A similar pattern of dispersal has been proposed for other littoral Mediterranean gastropods with a short planktonic larval stage (e.g., Gibbula divaricata Linnaeus, 1758) or direct development (e.g., Dendropoma lebeche Templado, Richter and Calvo, 2016) [93].

The observed mean allelic richness was 6.175, which is almost double the value obtained by Cossu et al. [56] (3.35 and 2.85) in their study of two populations in Sardinia. Those authors analysed 8 of the 11 microsatellites used here but in fewer specimens. Therefore, more allelic variability has been found in the populations studied with the same markers. Most loci are polymorphic, which explains the higher level of allelic richness. However, in general, a deficit of heterozygotes was observed, which could indicate a certain degree of inbreeding, heterozygote counterselection, or homozygote superiority, leading to deviations from HWE. Setting aside the possibility of technical issues related to genotyping errors, other conventional explanations for this overall heterozygote deficit seem less plausible. For instance, this deficit was found in almost all the analysed populations, which greatly differ in their actual sample size. Some, like those from the Chafarinas Islands, count dozens of thousands of individuals, while others, like those from southern Spain, barely reach a hundred [19]. Thus, genetic drift would be an unexpected general explanation. Limited gene flow or population structure is also improbable, given the observed rate of migration and the lack of differentiation. In any case, such deviations are common among benthic marine invertebrates whose fertilisation occurs in the water column in the absence of sexual selection, further increasing the probability of outcrossing between related individuals (level of inbreeding) [94].

The observed levels of inbreeding are consistent with the low level of overall genetic differentiation (F_ST = 0.004). The low pairwise F_ST values observed among most populations are reflected in the genetic structure, which shows most of the populations comprising a central cluster with some populations being slightly divergent, such as those from Corsica (COR), Cala Iris in Morocco (IRI) and Cape Bon in Tunisia (BON). These data suggest some weak genetic structuring. The genetic differences found between some populations (although non-significant, such as the two Tunisian ones or those from Congreso Island) may be due to shoreline features, but not distance. Indeed, geographic location and the surrounding hydrodynamic and topographic environment could have a large impact on connectivity in coastal species [95].

To confirm the above, we performed a STRUCTURE analysis, which showed possible structuring of the populations in two (K = 2) or three (K = 3) genetic groups (Figure 4). For K = 2, the first genetic group characterises all the populations, with the second group representing the ancestry of predominantly certain IRI individuals. For K = 3, the addition of a new genetic group reveals another origin of some individuals, primarily from Corsica (COR) and Cape Bon (BON). In both cases and despite the group differentiation, the genetic group to which all populations belong is clearly predominant. The AMOVA results showed that, for K = 3, within the population differences explained 98.11% of the variability, with among group differences accounting for only 1.93%. Similar values were obtained when considering K = 2 and K = 1. Therefore, this pattern of genetic structuring is not significant enough to support any structure, as genetic variability was higher at the level of individuals than between the defined groups.

The genetic homogeneity of P. ferruginea was previously demonstrated by Espinosa and Ozawa [48] for populations from southern Spain and northern Africa based on an analysis of the mitochondrial COI gene. Similarly, Casu et al. [52] found no significant genetic differentiation across the entire distribution range of P. ferruginea based on mtDNA sequences. However, their ISSR analysis revealed two distinct groups: one encompassing localities in the Alboran Sea (Spain, Morocco and Algeria) and the other in the Sardinian-Corsican region. Casu et al. [52] attributed these differences to a potential barrier to gene flow in the Sardinian Channel, a stretch of approximately 180 km that separates North Africa from Sardinia. Unfortunately, we have not been able to include samples from the Tyrrhenian Sea in our study, including Sardinia, preventing us from assessing the potential impact of this barrier on our results.

Other studies conducted with populations from Sardinia have highlighted a marked genetic structuring [49,51,56]. This pattern has been attributed to limited larval dispersal, leading to restricted gene flow at a local scale, isolation by distance and finally, genetic differentiation among populations. For instance, Coppa et al. [96] and Cossu et al. [97] (following [87]) suggested that the larval stage of P. ferruginea lasts a maximum of 10 days, and Cossu et al. [56] indicated that dispersal in this species barely exceeds 10 km. However, as mentioned previously, more recent studies under laboratory conditions [31,41] show that the larval phase of P. ferruginea is similar in duration to other Mediterranean limpet species and can facilitate wider dispersal. The observed genetic divergences among Sardinian populations may be the result of both historical and contemporary processes. During Pleistocene periods of low sea levels, the Strait of Bonifacio could have acted as a significant barrier, and the region’s present-day irregular topography and dynamics might impede gene flow between localities along Sardinia’s north-western and north-eastern coasts.

In our F_ST analysis, null values of differentiation predominated among the western localities, except for one population from Morocco (IRI), one from the Chafarinas Islands (LEV), two eastern localities from Tunisia (BON and KEL) and one from Corsica (COR). Given these data, a pattern of isolation by distance can be ruled out, and the chaotic dispersal pattern observed may be due to occasional stochastic factors, as mentioned above. The migration analyses showed that 12.3% of the individuals were first-generation migrants, i.e., they migrated to another population, settled and bred with individuals of other origins, contributing to the result showing the assignment of 35.5% of the specimens to other populations.

Espinosa and Ozawa [48] hypothesised that the lack of genetic differentiation in P. ferruginea was the result of a bottleneck, which reduced the effective population size of the species. This bottleneck would have reduced the species’ haplotype diversity compared with the other species of Patella, which show greater diversity [98]. Our genetic structure results show no structure to support this hypothesis and suggest this process seems to have occurred recently. Our results also evidenced a bottleneck under the best-fit model, SMM, and both statistical tests (see Table 6). All the populations, except Melilla (MEL) (only significant for the Wilcoxon test) and Cape Bon (BON) (no support for bottleneck), presented values significantly different from those expected for populations in mutation/drift equilibrium. In other words, the results suggest most of the populations suffered a bottleneck. Taken together, our results indicate that extant populations should be considered as sub-populations of a large metapopulation distributed throughout the studied area.

Bouzaza et al. [53] analysed a fragment of COI in 51 individuals of P. ferruginea sampled from seven stations along the Algerian coast and found genetic differentiation between eastern and western populations. These results, which partially contradict ours, could be attributed to the different resolution of the genetic markers used: COI can reveal past historical events, whereas microsatellites provide evidence of current processes driving population structure [45]. It is plausible that significant bottlenecks in the past restricted the distribution of P. ferruginea to scattered sub-populations or small demes with limited connectivity between the eastern and western areas of Algeria as a consequence of past geological conformation or environmental conditions, with connectivity being re-established later.

Sá-Pinto et al. [91,99] and Cossu et al. [97] conducted analyses on genetic variability and gene flow in two common Atlanto-Mediterranean limpet species, P. rustica and P. ulyssiponensis. Their studies identified shared barriers to gene flow within the species’ respective ranges, specifically in the Atlantic-Mediterranean transition and across southern Italian shores [90], but within the western Mediterranean basin, the species exhibited genetic homogeneity. Villamor et al. [45] investigated the Mediterranean endemic limpet P. caerulea and observed a gradual genetic transition between the western and eastern Mediterranean basins, but in neither P. ulyssiponensis [97] nor P. caerulea [45] was detected a pattern of isolation by distance. Additionally, Cossu et al. [97] noted that P. ulyssiponensis displayed a genetic structure pattern indicative of a chaotic patchiness scenario. Therefore, the barriers in the Atlanto-Mediterranean and in the region dividing the western and eastern Mediterranean basin appear to be permeable for P. rustica and P. ulyssiponensis (and the latter for P. caerulea). However, these barriers currently impede the dispersion of P. ferruginea, as its distribution range is restricted by both limits, with the easternmost barrier likely located between east Sicily and the Calabrian Peninsula.

A striking finding of the present study was the identification of four clone pairs (individuals with identical genotypes) in the MEL (Melilla), FRA (Chafarinas), KRI (Oran) and ISP (Algeciras) populations. These clone pairs were considered outliers and were excluded from the analysis. Initially, we attributed these clones to potential errors; however, the repetition of the same error in four different populations is unlikely, given the careful sampling procedures and the fact that the cloned specimens were re-extracted to rule out laboratory errors. It is challenging to provide a straightforward explanation for the presence of these clone pairs. Molluscs are not known to exhibit asexual reproduction other than parthenogenesis [100], and larval cloning, though described in some phyla such as echinoderms [101], remains unknown in molluscs, at least in natural conditions [102]. However, exceptions exist, such as the marine brooding, hermaphroditic clam genus Lasaea, which contains clonal lineages apparently derived via allopolyploidy [103]. Certain species of the pelagic gastropod family Cavolinidae exhibit a mode of asexual reproduction in unfavourable environmental conditions, including fission in mature females producing two individuals, with one reverting to the male stage of the protandric cycle and the other being hermaphroditic, likely capable of self-fertilisation [100].

Implications for Conservation and Management

Research on larval dispersal and metapopulation connectivity has increased over the past few decades due to the importance of these processes for the implementation of spatial management strategies, particularly in relation to population persistence and the viability of marine species (e.g., [95,104,105,106], among many others).

Measures for the conservation of the marine environment focus mainly (almost exclusively) on the establishment of networks of marine protected areas (MPAs) ([107,108,109], among many others). The benefits of MPAs are indisputable, but they are very static and insufficient tools, especially for species with planktonic larvae that constitute metapopulations. These protected areas safeguard the benthic adult phases of these and other species but not their larval phases. Moreover, larvae are generally more sensitive to stressors than adults, making them even more vulnerable [110,111]. Therefore, for species with a bi-phasic life cycle, conservation measures should aim to protect the dynamic pelagic environment in order to maintain the larval pool and promote larval dispersal and settlement in favourable places and, ultimately, gene flow. Such measures will also ensure that well-preserved benthic adult populations not only produce larvae that can supply other areas in addition to their own but also receive larvae from outside. Some authors (e.g., [112,113]) have highlighted the need to move marine conservation beyond traditional 2-dimensional coastal approaches to the 3-dimensional pelagic environments incorporating dynamic oceanographic features. In the marine environment, everything is interconnected and water masses surrounding MPAs must also have favourable conditions. Furthermore, suitable habitats must be preserved outside these areas; otherwise, a good part of the larvae produced in the MPAs will be wasted. If on-going coastal physical destruction, fragmentation and transformation is further neglected in management planning, the few sectors of native or semi-native habitats that remain may ultimately be compromised [114]. Due to habitat loss and fragmentation, connectivity between populations may be reduced or interrupted. Although occasional larval exchange across variable distances may have been sufficient to maintain genetic panmixia, these events had little significance for ecological or demographic connectivity [8].

As mentioned in the introduction, over-harvesting, on one hand, and habitat destruction and fragmentation, on the other, are the two main causes of extinction of populations of P. ferruginea. If these causes persist, the number of populations will continue to decline progressively. Patella ferruginea is currently considered at risk of extinction and is under strict legal protection in the countries it still inhabits, meaning its harvesting and sale are prohibited. The first conservation measure must be strict compliance with this legislation, encouraged by increased surveillance along the coast and close monitoring of shell fishing, both inside and outside of protected areas. The second major measure is a comprehensive coastal management plan to halt the destruction, fragmentation and pollution of coastal areas in order to maintain favourable habitat conditions.

The metapopulation of P. ferruginea is made up of populations that contrast in abundance and density [19]. According to Hawkins et al. [8] and Kurland et al. [115], large populations contribute most to the connectivity and persistence of the metapopulation by reducing inbreeding and enabling the recolonisation of small sub-populations that are unable to persist without external input. Dispersal between sub-populations with a patchy distribution can have a stabilising effect on metapopulation size [116], and the effect of subpopulation extinction on the maintenance of genetic variation within metapopulations depends, to a great extent, on the degree of migration among subpopulations. In this context, the successful and abundant populations of P. ferruginea that inhabit the northern African coasts and serve as exporters of larvae play an important role. Yet, this migration is useless if the exported larvae do not find suitable places to settle or do so in areas where threats to their survival, such as harvesting, continue to operate. At the same time, these successful populations will suffer progressive genetic impoverishment if they only act as donors and do not recruit larvae from other areas. We agree with the assertion made by Cossu et al. [7] that the conservation of P. ferruginea must mainly focus on the strict safeguarding of current, well-preserved populations and the connectivity among them. To do this, effective protection management of the MPAs the species inhabit, as well as nearby unprotected areas, must be prioritised, as should creating new MPAs in areas with thriving populations.

The genetic homogeneity of P. ferruginea or the lack of structure found here and in other studies should be interpreted with caution, especially regarding its application in conservation measures other than the aforementioned effective protection of the current major populations both inside and outside the MPAs. Evidence of genetic homogeneity may encourage the use of other commonly used conservation tools, such as translocations [117]. In the case of P. ferruginea, two types of translocations based on the origin of the translocated specimens have been used as a measure to recover populations of this species. The first type involves the transfer of specimens from a natural donor population to another location, and the second is the transfer of young specimens produced by aquaculture or recruits collected in nature from artificial structures installed in locations with successful populations.

The earliest attempts were adult translocations from natural donors; however, these were not successful due to the low survival rates of the transferred specimens (e.g., [88], 25% after one year, and 10% after two years; [118], 20% and 10%, respectively). Subsequent experimental translocations for reintroduction purposes (e.g., [119]) used cages to protect transferred individuals against predation and wave action. These experiments resulted in higher survival rates than previous attempts; however, the rates were still low for both non-caged and caged specimens (30–61% after one year and 25–58% after two years, respectively) compared with that of non-translocated control specimens (85%). Moreover, these authors overestimated survival rates, as they did not include in their estimations the mortalities that occurred while handling and transporting specimens.

In summary, the high mid-term mortality rate of translocations of both adult and young specimens from wild populations renders this measure ineffective, particularly given the endangered status of the species. Moreover, success at the scale of the receiving population, if any, has not yet been proven. This type of translocation is problematic for several other reasons. First, translocations may have a negative effect on donor populations (e.g., a potentially significant decrease in the population), an aspect not yet studied in detail. Second, they imply habitat manipulation and alteration in donor and receiving areas with unknown consequences. Third, translocations are extremely costly, even when moving only a few tens of specimens, and unfeasible on a larger scale. Finally, translocations are often argued as compensatory measures to justify coastal infrastructure expansion policies in areas inhabited by this species under the spurious pretext that the affected specimens can be safely moved to other areas. In the updated “Strategy for the conservation of the ferruginous limpet (Patella ferruginea) in Spain” (SCS, [16]), translocation of specimens from natural populations is discouraged as a conservation or compensatory measure, given the negative impact it has on the populations of origin.

As an alternative to translocations of wild individuals, the SCS [16] recommends using young specimens obtained via aquaculture to reinforce populations or expand the distribution of the species. In recent years, substantial advances have been made in the laboratory-controlled reproduction of P. ferruginea [31,41]. However, its use as an effective conservation tool for restocking, stock enhancement or reintroduction still requires much research and development of a methodology that ensures a semi-industrial production of juveniles (“seeds”), taking into account the appropriate quantities and level of genetic diversity needed for such purposes. Thus, research efforts should focus on improving the aquaculture of the species, as well as how to respond quickly and adequately to local or regional threats to current populations, such as extensive contamination events (e.g., oil spills) or diseases. Despite its potential, restocking from aquaculture specimens should not be used on a large scale without more knowledge of the biology and genetics of P. ferruginea. As asserted by Luque et al. [19], restocking must always be performed after prior effective protection of suitable receiving areas where the factors causing its decline or disappearance do not exist or have been eliminated. Moreover, as restocking projects based on aquaculture production eventually involve the translocation of juveniles to a receiving locality, we recommend prioritising the advancement of techniques that improve the medium and long-term survival of translocated juveniles. Finally, the conservation effort of P. ferruginea must be coordinated at the international level to involve all countries where the limpet is found. Therefore, reinforcement of this synergy is necessary to advance conservation efforts rapidly.