Insights from 180 years of mitochondrial variability in the endangered Mediterranean monk seal (
 
 Monachus monachus
 
 )

Mediterranean monk seals (MMS) are among the most endangered marine mammals on Earth. We screened mitochondrial variability (control region [CR1] and mitogenomes) of the species through a 180-yr timeframe and extended by 20% (n = 205) the number of samples from a previous investigation, including historical specimens from 1833 to 1975. Although we detected two new, rare CR1 haplotypes, genetic diversity remained extremely low. Fully resolved haplotype median network and rarefaction Corresponding author (e-mail: philippe.gaubert@univ-tlse3.fr). Current address: Laboratoire Evolution et Diversité Biologique (EDB) – UPS-CNRS-IRD, Université Paul Sabatier, 118 route de Narbonne, 31062 Toulouse, France. MARINE MAMMAL SCIENCE, 000(000): 1–23 (2019) © 2019 Society for Marine Mammalogy DOI: 10.1111/mms.12604

eastern Morocco, Algeria, and Italy (Mo 2011), but their exact status is unknown.
Studies on the genetics of MMS have concentrated on the extant, remaining populations. Seminal investigations by Pastor et al. (Pastor et al. 2004 exploring nuclear genetic diversity identified (1) a severe bottleneck in the Cabo Blanco population and the absence of structure among local pupping colonies, (2) low genetic variability and substantial differentiation between the two main breeding populations in Cabo Blanco and Greece, which are separated by more than 4,000 km, and (3) a signature of metapopulation dynamics across the species' range in the recent past, probably disrupted from the Roman period, a time during which the regular exploitation of monk seals in the western and central Mediterranean was widespread (see Johnson and Lavigne 1999). A more recent study investigating the variability in the mitochondrial control region showed that MMS were among the most genetically depauperate mammals ever reported, with the existence of only five, closely related control region haplotypes (Karamanlidis et al. 2016b). Coupled with the sensitivity of isolated colonies of MMS to stochastic events including toxic algal blooms and/or viral outbreaks (Hernández et al. 1998, Borrell et al. 2007, such low levels of genetic diversity have been considered potentially deleterious to the recovery of the species (Pastor et al. 2004(Pastor et al. , 2007. Because of the fragmented distribution of extant populations and their very low genetic diversity, scenarios for the demographic history of MMS and the time at which populations collapsed have remained speculative. In order to understand the ongoing process of a species' extinction, it is critical to take into account its past genetic diversity (e.g., Foote et al. 2013, Casas-Marce et al. 2017. To reach such an objective, access to historical specimens-notably from museum collections-is a critical step because such material often represents, especially in the case of endangered and/or elusive species, extinct or rare populations (Suarez andTsutsui 2004, Spurgin et al. 2014). In the case of the MMS, museum samples are the only source yet to be assessed for studying the impact of the historical, local extinctions that drastically fragmented the species' range (Sergeant et al. 1978).
Building on the recent study of Karamanlidis et al. (2016b), we reassess the genetic variability patterns in extant MMS by sequencing 40 new samples, including 25 historical specimens, from the historical and current species' range. More specifically, we analyze control region sequences and mitogenomes to address the following questions: (1) was the Mediterranean monk seal already a genetically depauperate species in the recent past? and 2) is there a signature of past metapopulation dynamics that would have connected the two main breeding colonies in the North Atlantic Ocean and the eastern Mediterranean Sea? We also discuss the applicability of mitochondrial markers in the conservation genetics of MMS.

Sampling
We obtained 16 recent   Western Sahara ("Cabo Blanco"; 9), and Morocco (1) (see Table S1). We also sampled 24 "historical"  specimens of MMS from 14 natural history museums that represent both the extant and historical species' range ( Fig. 1). The distinction between recent (from 1989 on) and historical (prior to 1975) samples reflects the fact that the last records of resident MMS groups from the central and western Mediterranean Sea were reported during the 1970s (Johnson 2004), suggesting that after this period the fragmentation of the species' range into North Atlantic and eastern Mediterranean populations as we know it today was effective. The historical samples originated from Bulgaria (Black Sea; 1), Croatia (2), France (1), Greece (Crete Island; 1), Italy (9; including Sardinia, Elbe Island, Capraia Island), Algeria (1), Portugal (Madeira; 1) and Western Sahara (4), plus four samples that had no precise origin. Among the historical samples, three had no exact dates of collection (but those were antecedent to 1975 according to the curators). The relative distribution of historical vs. recent samples was visualized using Kernel density plots reconstructed in RStudio v. 1. 1.383 (RStudio Team 2015) with the function density (Fig. 1).

DNA Extraction
Genomic DNA was extracted from fresh tissue and skin samples using the DNeasy Blood & Tissue Kit (QIAGEN, Courtaboeuf, France) following Figure 1. Distribution of samples of Mediterranean monk seals with known origin used in this study. In light blue, extant distribution of the species. Black and white circles correspond to recent and historical specimens, respectively, newly sampled as part of this study. Question mark means uncertain location. Between brackets, the respective contribution of recent/historical samples per locality; asterisks indicate new samples (relative to Karamanlidis et al. 2016b). At the bottom right, kernel density plots of the historical versus recent sample distribution through time as used in this study. The dotted box represents the timeframe (1990-2012) covered in Karamanlidis et al. (2016b). The red asterisk marks the mass die-off at Cabo Blanco, North Atlantic Ocean, in 1997. The photograph represents the sole Mediterranean monk seal ever collected from the Gulf of Naples, Italy in 1884 (Centro Musei delle Scienze Naturali e Fisiche of the University of Naples "Federico II" specimen Z245; not used in this study). MARINE MAMMAL SCIENCE, VOL. **, NO. *, **** the manufacturer's recommendations. To avoid cross-sample contamination, the historical samples including bones (postcranial and turbinal bones), tanned skins and connective tissues were processed in series of small batches (ca. 10 samples each) with negative controls through each extraction step in a laboratory dedicated to degraded DNA (http://www.labex-cemeb.org/en/degraded-dna), isolated from the other lab facilities where fresh tissues were processed. A subset of 10 historical samples was extracted 2-3 times independently and sequenced to control for potential cross-contamination and mismatch repair. The replicate sequences were identical for each of the 10 individuals (data not shown). Small pieces of tanned skin and connective tissue were extracted using the DNeasy Blood & Tissue Kit protocol, adding 4 μL of DTT (1M) during the lysis step and an additional 20 μL of proteinase K. Final elution volumes for fresh and tanned skin/connective tissue samples were 120 and 60 μL H 2 O, respectively.
The protocols for reducing bones into powder varied depending on sample type. Turbinal bones were reduced to powder with the Bullet Blender Homogenizer using 2-3 stainless steel balls (Next Advance, Troy, NY) in Safe-Lock 2 ml tubes (Eppendorf, Montesson, France) for three minutes at speed 10. Postcranial bones were first cleaned with a tissue soaked in bidistillate water, and their outer surface was removed using scalpel blades. Postcranial bones were then wrapped in aluminum foil and ground into coarse powder using a hammer cleaned with 6% diluted bleach between each sample.
Approximately 50-100 mg of bone powder was digested in 1-1.5 mL lysis buffer composed of 0.5M EDTA pH 8.0, 0.5% N-lauryl sarcosyl and 0.25 mg/mL proteinase K overnight at 56 C with gentle mixing. Purification and elution were performed with the MinElute PCR Purification Kit (QIAGEN) following the manufacturer's instructions. Final elution volume was 60 μL H 2 O.

PCR Amplification of Control Region and Mitogenome Shotgun Sequencing
We amplified 524 bp of the hypervariable region I of the mitochondrial control region (CR1) from the fresh samples using the primer pair of Karamanlidis et al. (2016b). For the historical samples, we designed four specific primer pairs with the Primer3 web platform (http:// primer3.ut.ee/) amplifying partially overlapping fragments ≤150 bp (Table 1), to obtain a 484 bp fragment encompassing all the variable sites of CR1 observed in MMS (Fig. 2). In this case, negative DNAextraction controls (see above) were included along with PCR blank controls to further assess potential contamination. PCR amplifications were carried out for both the large and small fragments in 20 μL final volume containing~5-50 ng of template DNA (as determined using Nanodrop), 0.1 mg/mL BSA, 4 × 0.25 mM dNTPs, 2 × 0.2 mM primers, 1× PCR direct loading buffer with MgCl 2 (1.5 mM) and 0.5 U GoTaq Flexi DNA Polymerase (Promega, Charbonnières-les-Bains, France). PCR cycling conditions included a first step of denaturation (94 C, 2 min), followed by 40-45 cycles of denaturation (92 C, 30 s), annealing (30 s; see Table 1 GAUBERT  (Hall 1999). The final CR1 alignment included 204 sequences (MS24 was not sequenced for CR1 because already available in GenBank; see Table S1). Five mitogenomes of MMS were obtained by shotgun sequencing, from four recent specimens in the eastern Mediterranean (Greece, Turkey, and Croatia) and one in the North Atlantic Ocean (Western Sahara) (see Table S2). Libraries were prepared after Tilak et al. (2015) and equimolarly pooled on different lanes, each combining ca. 20 vertebrate taxa. Lanes were sequenced using single-or paired-end 100 bp reads on an Illumina HiSeq2000 Analyzer at GATC Biotech, Konstanz, Germany. Quality of raw genomic data were checked with FastQC (https://www.bioinformatics. babraham.ac.uk/projects/fastqc/) and Geneious v.7.1.5 (Biomatters, Auckland, New Zealand) was used to trim Illumina adapters from raw reads before assembly. Given the absence of a complete mitogenome of MMS in the databanks, we used the annotated mitogenome of the Hawaiian monk seal Neomonacus schauinslandi (NC_008421) as a reference for mapping our sample with best coverage (MS24; see Table S2) in Geneious. Reads were first mapped using the "medium-low sensitivity" option, followed by "low sensitivity" iterative mapping repeated until no additional reads could be included. The final consensus sequence was obtained using the highest quality base assignment threshold; nucleotides were called for sites covered by ≥3 reads. The four other mitogenomes of MMS were assembled from the MS24 consensus sequence. A CR1 fragment was also generated by PCR (see above) from the five samples to validate the "quality" (i.e., 100% match) of the assembled mitogenomes. Final alignments of the 204 CR1 sequences and the five mitogenomes were performed in MMS_dlp564-R pos. 15,526 pos. 16,049 Figure 2. Primer design used in this study to amplify the hypervariable region I of the mitochondrial control region. Site positions (pos) refer to the mitogenome of Mediterranean monk seal with GenBank Accession number MG570472. Red and blue boxes on the control region sequence indicate new (this study) and previously known polymorphic sites, respectively. Primer details are given in Table 1. GAUBERT ET AL.: HISTORICAL GENETIC VARIABILITY IN MONACHUS MONACHUS Geneious using the Muscle algorithm (Edgar 2004) with default parameters and were manually adjusted in BioEdit.

Genetic Diversity and Variability
Genetic diversity within MMS was calculated for the CR1 in DnaSP v. 6.10.01 (Rozas et al. 2017), including number of haplotypes (h) and polymorphic sites (S), haplotype diversity (Hd) and nucleotide diversity (π). Genetic diversity estimates were also calculated for the historical (1833-1975; n = 24) and recent ; n = 180) samples, separately.
In order to assess haplotype geographic distribution, we built a medianjoining network (Bandelt et al. 1999) in Network v. 5.0.0.1 (http://www. fluxus-engineering.com) with ε fixed to 50 (maximal pairwise difference = 5; character weights = 10; tv/ts weight = 1:1) to obtain a full median network (Polzin and Daneshmand 2008). Because of the extremely low level of CR1 variability within MMS (see Karamanlidis et al. 2016b and this study), we refrained from using the calculation of F ST for estimation of population differentiation. Indeed, such relative measures of divergence can be drastically affected by levels of within-population genetic diversity (Meirmans andHedrick 2011, Jakobsson et al. 2013), making them potentially unfit to measure differentiation within a species where genetic diversity is low. Because of these limitations, we restricted ourselves to the description of the genetic variability among and within populations of the MMS. For this purpose, we counted the number of nucleotide differences among sequences using MEGA v. 7.0.18 (Kumar et al. 2016), delineating 12 marine subdivisions (Table 2) of the Mediterranean Sea (n = 9), Black Sea (n = 1) and the North Atlantic Ocean (n = 2) as defined by the International Hydrographic Organization (1953).
In order to assess the relationship between the number of haplotypes and sample size, we used EstimateS v. 9.1.0 (Colwell 2013) and the multinomial model (Colwell et al. 2012) to compute an individual-based rarefaction curve from CR1 sequences. We extrapolated the rarefaction curve by a factor of 2 (total n = 408) to evaluate the trend in haplotype diversity in the case of extended sampling. We also ran 100 randomizations to calculate the bias-corrected Chao 1 estimator, which estimates the "true" diversity of a sample set (Chao 1984, Chao et al. 2015. We used the five complete mitogenomes generated in this study to assess gene-specific polymorphism for the 13 coding genes (CDS), the two ribosomal genes (rRNA) and the control region. To do so, we calculated the total number of differences between individuals from the two main breeding populations of MMS (eastern Mediterranean Sea and North Atlantic Ocean).

RESULTS
Density curves showed that the temporal distribution of samples across the historical period was relatively uniform from 1833 to 1975 MARINE MAMMAL SCIENCE, VOL. **, NO. *, ****  compared to the recent period (1989 to 2014) where the distribution was clearly bimodal, notably because of an excess of samples coming from the massive die-off at Cabo Blanco in 1997 (Fig. 1). Mitogenomes produced from the five recent samples had a total length ranging from 16,739 to 16,747, their mean depth varying between 13× and 625×, and low levels of base ambiguities (4-11) ( Table S2). The CR1 fragments aligned with positions 15,526 (or 15,566 in the case of shorter fragments; see above) and 16,049 of the mitogenomes (Fig. 2). CR1 diversity indices for the total sample set (n = 204 sequences; length = 484 bp) were as follows: h = 7; S = 6; Hd = 0.654; π = 0.0018. Genetic diversity was slightly higher for historical samples (n = 24) compared to recent samples (n = 180): S his/rec = 5/4; Hd his/rec = 0.731/0.647; π his/rec = 0.0027/0.0017 (Fig. 3). Both sample sets were composed of six haplotypes, of which five were shared, meaning that both historical and recent samples also hold a unique, single haplotype each.
Two new haplotypes were identified, including MM06 (Cabo Blanco; one recent sample, but see Discussion) and MM07 (Adriatic Sea and France; two historical samples) (Tables 2, S3). The CR1 full median network did not include median vectors (i.e., missing haplotypes; Fig. 4). There was no apparent geographic structuring among the seven haplotypes, and all differed by 1-2 mutations (transitions). The most common haplotype (MM01; 50% of the samples) was distributed from the eastern Mediterranean Sea (Turkey; two samples) to Cabo Blanco (North Atlantic Ocean, Western Sahara; 1), also including Libya (1), Aegean Sea (Greece; Atlantic Ocean (Cabo Blanco and Madeira; 53) and in the Alboran Sea (Morocco and Algeria; 2). MM02 (ca. 14%) was present in the Black Sea (Bulgaria; 1) and the Aegean Sea (28). MM03 (ca. 5%) occurred in the Ionian Sea and Adriatic Sea. MM04 (1.5%) was present in the Aegean Sea, Ligurian Sea, and Sardinian Sea.
Mean numbers of differences among marine subdivisions were extremely low, ranging from 0 to 3 with minimum and maximum values between 0 and 5 (Table S4). Mean variability within marine subdivisions ranged between 0 and 2.67, with the highest values in the Adriatic Sea (2.67) and the Sardinian Sea (2). The individual-based rarefaction curve and its 95% intervals almost reached horizontal asymptotes after extrapolation by a factor of 2, with N haplotypes slightly above 7 (Fig. S1). Chao 1 estimator value was 7 (7-8.13 95% CI).

Mediterranean Monk Seals Have Been Genetically Depauperate for at Least 180 Yr
The use of historical samples allows for a direct assessment of genetic diversity loss over time (Rosenbaum et al. 2000). In the case of genetically depauperate species under extinction risk, the analysis of historical samples may allow investigation of whether human activities have had a direct impact on the loss of genetic diversity (Matocq and Villablanca 2001, Wandeler et al. 2007, Casas-Marce et al. 2017.
Screening the mitochondrial diversity of MMS over the last ca. 180 yr has shown that the species had already been genetically depauperate since at least the mid-19th century. Despite the detection of two new, rare haplotypes (h = 7; see below), CR1 diversity indices were among the lowest within pinnipeds and other endangered species, even though the percentage of variable sites almost doubled (from 0.6% to 1.1%) compared to Karamanlidis et al. (2016b). The ratio between haplotype numbers and the total number of samples (7/204 = 0.03) was the same as that reported in Karamanlidis et al. (2016b) for 165 recent samples covering the current species' range. Such low levels of mitochondrial diversity correspond to levels observed in other mammalian species at the brink of extinction (Rosel and Rojas-Bracho 1999, Rosenbaum et al. 2000, Weber et al. 2000, Rodriguez et al. 2011, Casas-Marce et al. 2017. Despite the analysis of 40 new samples (including 25 historical samples), we detected only two new CR1 haplotypes at very low frequencies (MM06-MM07: 0.5%-1.0%). This, coupled with the fact that (1) rarefaction analysis-including extrapolation-and Chao 1 estimator neared the total number of haplotypes (7) found in this study (Fig. S1) and (2) the MARINE MAMMAL SCIENCE, VOL. **, NO. *, **** full median network of CR1 haplotypes did not call for hypothetical (i.e. missing) haplotypes, suggests that the probability of detecting additional haplotypes in MMS through the last 180 yr is low, in turn reinforcing the current pattern of low genetic diversity in the species.
Our results also bear witness to the deleterious impact of local extinctions on the genetic variability of a patchily distributed species such as the Mediterranean monk seal. Indeed, haplotype MM07, which was recorded in two historical samples from France (collected before 1850) and Croatia ("Dalmatia"; probably collected before the First World War), is probably extinct because it could not be detected in the 180 recent samples that have been sequenced so far (Karamanlidis et al. 2016b;this study). Moreover, the sequencing of MMS specimens held at the Estacion Biologica de Doñana (Table S1) and originating from the Cabo Blanco 1997 die-off, enabled the identification of two haplotypes that were not described in this colony by Karamanlidis et al. (2016b). We speculate that the die-off, which reduced the size of the Cabo Blanco population by two-thirds (Hernández et al. 1998), could have led to the extinction of those two haplotypes, each represented by a single individual, including MM06 (new haplotype, endemic) and MM01 (shared with the Mediterranean). Therefore, it is of utmost importance to conduct genetic analyses on samples from Cabo Blanco collected after the massive dieoff in order to assess the impact of this dramatic demographic event on the genetic diversity of MMS.
The Mediterranean monk seal was already considered critically endangered when the first in situ research and conservation efforts were initiated in the late 1970s (Ronald and Duguy 1979). Although MMS have been exploited by humans since the Late Pleistocene (Steele and Alvarez-Fernández 2011), more recent exploitations are believed to have caused major local extinctions, notably by Romans in the Mediterranean Sea (until 300 AD) and by European sealing expeditions in the Atlantic Ocean (14-16th centuries) (Marchessaux 1989, Johnson and Lavigne 1999, González 2015. Following deliberate killing by fishermen and the development of mass tourism during the 20th century, breeding populations of MMS disappeared in the southern Black Sea, Albania, Croatia, Egypt, Israel, Italy, Lebanon, Libya, Tunisia, and Syria (Karamanlidis et al. 2016a). Earlier in the 20th century, MMS also went extinct in France (mainland coast and Corsica), Spain (mainland coast and Balearic Islands) and the northern Black Sea (Sergeant et al. 1978, Marchessaux 1989. Although such extinctions have most probably deeply impacted the distribution and demography of the species, our results indicate that the global level of mtDNA diversity in MMS remained roughly constant from the mid-19th century on. Indeed, we observed only a slight decrease in genetic diversity between historical and recent samples, and the number of haplotypes (six) remained the same between the two periods (Fig. 3). It seems reasonable to assume that in the timeframe of our study period, which covers ca. 180 yr, human activities did not significantly affect the mitochondrial diversity of the species.
Thus, it is not known whether human pressure exerted on MMS through historical times is responsible for the extremely low level of genetic variability observed in our study. In the Hawaiian monk seal, which shares with MMS one of the lowest levels of genetic diversity among mammals, there is evidence for a bottleneck due to hunting activities in the late 19th century, but also for the existence of low levels of heterozygosity before exploitation (Schultz et al. 2009(Schultz et al. , 2010. Schultz et al. (2010) speculated that a combination of long-term small population size and a bottleneck due to ancient-to-recent hunting activities depleted the genetic diversity of the Hawaiian monk seal. This could also be the case in MMS, although the exact impact of antique hunting activities such as those exerted during classical antiquity by Romans in the Mediterranean (Johnson and Lavigne 1999) is difficult to assess. Conversely, the dramatic decline of North Atlantic populations is more thoroughly documented, and seems to be linked to the heavy exploitation by Portuguese sailors for fur and oil during the 15th-16th centuries (Monod 1932, Marchessaux 1989, Pastor et al. 2004. Furthermore, the negative impact of sealing operations on the survival of North Atlantic MMS could have been further accentuated by unfavorable geoclimatic events also contributing to local extinctions, with the disappearance of (1) breeding sites in the Macaronesia Islands (Azores, Madeira, Canarias, Cabo Verde) following Late Pleistocene sea-level rise and (2) the whole Lobos Island (Bay of Dakhla, Western Sahara) after a tsunami caused by the "Lisbon earthquake" in 1755 (González 2015).
Given the very low levels of genetic diversity observed in the two surviving species of monk seals (see above) and the extinction proneness observed within the Monachinae in general (the third modern species, Neomonachus tropicalis, became extinct in the 1950s; McClenachan and Cooper 2008), low genetic diversity may actually constitute an intrinsic characteristic of monk seals since their early history ca. 6.3 MYA (Scheel et al. 2014). There are several, superimposable factors that may maintain low genetic diversity through time, including long-term small effective population sizes, recurrent population crashes, low juvenile survival, low reproductive success (Matocq andVillablanca 2001, Milot et al. 2007), and genome-wide selective sweep (Amos and Harwood 1998). Low birth rates and high pup mortality have been proposed as the factors maintaining small populations in MMS from Cabo Blanco (before the dieoff) and also Hawaiian monk seals (Gazo et al. 2000(Gazo et al. , 2001Schultz et al. 2009). The paucity of the data related to the population dynamics of MMS and the great influence of environmental and demographic conditions on those parameters (Karamanlidis et al. 2016a) renders low recruitment as a more reasonable hypothesis to explain slow population growth (Martínez-Jauregui et al. 2012), and by extension, long-term small effective population sizes. However, large-scale studies including both Atlantic and Mediterranean populations will have to be conducted to identify the main factors responsible for the low genetic diversity observed in MMS.

A New Picture of Mitochondrial Variability Across the Historical Range of the Mediterranean Monk Seal
Our study documented a less geographically confined mitochondrial haplotype distribution than previously assumed (Fig. 4, Table 2), suggesting that historical, local extinctions-but also the sampling bias towards extant MARINE MAMMAL SCIENCE, VOL. **, NO. *, **** populations of previous studies-yielded an "actualistic" interpretation of the clear-cut separation between the eastern Mediterranean Sea and the North Atlantic Ocean. All the five previously identified CR1 haplotypes (Karamanlidis et al. 2016b) were found to have wider, sometimes overlapping distributions. Notably, we showed that MM01, thought to be diagnostic of the eastern Mediterranean breeding population (Harwood et al. 1996, Karamanlidis et al. 2016b, was not restricted to the Aegean, Ionian, Adriatic, and Libyan Sea. Rather, it also characterized the first-ever sequenced samples from the Cilician Coast (East; Danyer et al. 2013Danyer et al. , 2014, historical samples from the Western Mediterranean, in the Thyrrenian and Sardinian Seas, and a new sample from the massive die-off at Cabo Blanco, ca. 4,000 km West (Table 2). Similarly, we showed that MM05, previously thought to be the unique haplotype in the North Atlantic, was also detected in the western Mediterranean (Alboran Sea) from one recent (Morocco) and one historical (Algeria) sample. MM02, distributed in the central and northern Aegean Sea (Karamanlidis et al. 2016b), was also found in the first-eversequenced representative of the extinct population of the Black Sea (Kiraç 2011). Despite uneven sampling in the Black Sea, our results suggest that dispersal probably occurred across the Turkish Straits System (TSS, consisting of the Marmara Sea, Istanbul and Çanakkale Straits), where individuals were observed in 2014 (southern Sea of Marmara; Inanmaz et al. 2014, Dede et al. 2016. MM04, a rare haplotype initially found in a single individual from the Aegean Sea (Karamanlidis et al. 2016b), was detected westward in two historical samples from the Ligurian and Sardinian seas. Given the low frequency of this haplotype in the Aegean Sea (<1.0%) and the relatively low number of samples from the Italian seas (n = 9), detecting MM04 in more than 20% of the Italian samples could be an indication of a haplotype once frequent in the Italian seas and now at the brink of extinction. Further sampling will have to be conducted to assess whether the small surviving colonies that might inhabit the Tyrrhenian-Sardinian-North African waters (Mo 2011) represent one of the last sources for this haplotype, or if MM04 originates from the Eastern Mediterranean recovering population.
The two main areas currently inhabited by the MMS are separated by ca. 4,000 km, and appear to be genetically isolated (Harwood et al. 1996;Pastor et al. 2004Pastor et al. , 2007Karamanlidis et al. 2016b). However, the sharing of haplotypes between the North Atlantic Ocean and the Mediterranean Sea populations supports the hypothesis that in the past, MMS constituted a metapopulation with gene flow connecting the Atlantic all the way to the Black Sea possibly via geographically intermediate populations that have now gone extinct (Pastor et al. 2007). The widespread distribution of two haplotypes found in our study, together with the greater levels of genetic variability observed in four marine subdivisions (Cabo Blanco, Aegean Sea, Adriatic Sea, and Sardinian Sea; Table S4), support the assumption of historical gene flow connecting the North Atlantic to the eastern Mediterranean, notably involving the passage through the Strait of Gibraltar in the west. Thus, one of the main conclusions from our analysis is that the current genetic separation between North Atlantic and eastern Mediterranean breeding populations might not have been effective in the past. The current pattern of low genetic diversity and population isolation observed in MMS is likely explained by a series of local, historical extinctions coupled with the effect of genetic drift on small breeding populations with already low levels of genetic diversity (as in the critically endangered Iberian lynx; Casas-Marce et al. 2017). Such a scenario concurs with the conclusions of Pastor et al. (2007) based on the analysis of microsatellites data, which support (1) recent, strong genetic drift resulting in marked genetic differentiation between the North Atlantic and the eastern Mediterranean populations, and at the same time (2) an imprint of past gene flow across the species' range as suggested by the distribution of allele frequencies.
Insights in the genetic variability of MMS should be used to guide conservation planning for the species, including the discussion on the benefits/risks of translocations and captive breeding programs that could be envisaged to save the species from extinction (Boulva 1979, Johnson et al. 2006, Pastor et al. 2007, Schultz 2011, Karamanlidis et al. 2016b). Our results supporting low genetic variability and past-possibly recentgene flow across the species' range could constitute a new baselinetogether with the conclusions of Pastor et al. (2007)-for the nonexclusive management of the two main extant populations in future conservation strategies. Further studies-notably targeting the nuclear genome-are necessary to give a time estimate to such past events of gene flow and assess the most critical parameters involved in the survival of MMS, notably regarding the issue of inbreeding vs. outbreeding depression (see Frankham et al. 2011).

Implications for the Conservation Genetics of the Mediterranean Monk Seal
The various attempts at assessing patterns of genetic diversity within MMS have highlighted limited genetic variation in the species and the difficulty to find informative, i.e., variable, markers to study/monitor the species (Harwood et al. 1996;Stanley and Harwood 1997;Pastor et al. 2004Pastor et al. , 2007Schultz 2011;Karamanlidis et al. 2016b). Our efforts to expand temporal and geographic coverages to an unprecedented level allowed us to characterize two new haplotypes and three new polymorphic sites on the CR1 fragment of MMS (Fig. 2, Table S3). However, genetic variability within the species remained very low (mean number of nucleotide substitutions = 0.5%; min-max = 0.2%-1.0%), and there was only one mutation (0.2%) differentiating the two main CR1 haplotypes mostly characterizing the North Atlantic (MM05) and the eastern Mediterranean (MM01) populations (similar observations apply also to mitogenomes, with only 0.3% substitutions).
The sequencing of five mitogenomes revealed very low levels of overall mtDNA variation in MMS. No variation was recorded within the four samples sharing the MM01 (CR1) haplotype in the Adriatic and Aegean Sea. Between MM05 and MM01, only seven genes out of 16 showed some level of polymorphism (Table S5), generally ranging between 0.1% and 0.4%. Genes commonly used in mammalian phylogeography, such as cox1 and cytb exhibited very low levels of polymorphism (3 and 4 variable sites, respectively). The control region had the highest level of polymorphism (2.4%; 31 variable sites), although most of the variability MARINE MAMMAL SCIENCE, VOL. **, NO. *, **** originated from repeated motifs and indels located in the hypervariable regions, as typically observed in mammals (e.g., Fumagalli et al. 1996).
Given the low levels of polymorphism observed in MMS despite extensive sampling coverage in time and space, our study highlights the limitations of mtDNA in assessing the population dynamics of the species. Moreover, mtDNA constitutes a single locus that may suggest a different scenario than nuclear, multilocus genotyping because of its maternal inheritance, smaller effective population size, and different sensitivity to demographic events such as bottlenecks (e.g., Fay and Wu 1999). However, attempts at generating multilocus nuclear data in MMS-in this case, microsatellites-have so far suffered from very low levels of detected polymorphism (Pastor et al. 2004(Pastor et al. , 2007(Pastor et al. , 2011.
Mediterranean monk seals continue to be exposed to a number of substantial threats, including habitat loss and deterioration, and deliberate killing (Karamanlidis and Dendrinos 2015), the latter being the most frequent cause of death in the Aegean Sea and the eastern Mediterranean coast of Turkey (Dede et al. 2015). Because populations are small and fragmented, they are also highly sensitive to diseases and stochastic events such as the mass-mortality event observed at Cabo Blanco in 1997, which killed more than 200 animals (Karamanlidis et al. 2016a). Although the ability of MMS to use marine caves for pupping and shelter may have temporarily saved the species from extinction (Bareham and Furreddu 1975), it will be crucial for its long-term survival to know whether the low genetic diversity observed in the species is associated with inbreeding depression, lower fitness and lower adaptability as predicted in population genetics theory (Lande 1998, Spielman et al. 2004, Frankham 2005. Following Karamanlidis et al. (2016b), we recommend pursuing the systematic collection of DNA samples throughout the species' range in order to establish a long-term DNA register that will help investigating temporal trends in the genetic diversity of MMS. In this context, additional mtDNA-sequencing from historical samples should allow a more accurate assessment of the species' genetic diversity through time and the refinement of the metapopulation scenario hypothesis posited in this study. In addition, the development of new nuclear markers via highthroughput sequencing technologies, but also investigations on more ancient time frames using archaeozoological samples are promising pathways to unravel the population dynamics of this emblematic species and develop efficient, DNA-assisted conservation strategies.
MM06 and MM07 are new haplotypes. Lines in bold mean that the information relative to the specimens could be double-checked with museum curators/data bank managers. The numbers in green cells refer to samples for which the mitogenomes were sequenced (see Table S2).
[Excel file] Table S2. Details of the mitogenomes of Mediterranean monk seals sequenced in this study. [Excel file] Table S3. Variable positions in the seven control region haplotypes found in this study. Site positions refer to the mitogenome of Mediterranean monk seal GenBank accession number MG570472. Alignment length was 16,747 bp. Table S4. Mean number of nucleotide differences (below diagonal) and range (above diagonal) in Mediterranean monk seals calculated among and within 12 marine subdivisions of the Mediterranean Sea and the North Atlantic Ocean. The yellow diagonal indicates the mean number of differences within each marine subdivision. MS = Mediterranean Sea; NAO = North Atlantic Ocean. Table S5. By-gene polymorphism across the mitogenome between the two main breeding populations of Mediterranean monk seals (eastern Mediterranean [n = 4] and Cabo Blanco, North Atlantic Ocean [n = 1]). Figure S1. Individual-based rarefaction curve of the expected number of haplotypes relative to sample size. The gray vertical line marks the sample size of this study (n = 204). Extrapolation was made for 408 samples (factor of 2).