Marine Biology

, Volume 159, Issue 9, pp 1871–1884

Recent history of the European Nassarius nitidus (Gastropoda): phylogeographic evidence of glacial refugia and colonization pathways

Authors

    • Department of Ecology, Faculty of SciencesUniversidade da Coruña
  • Jeanine L. Olsen
    • Marine Benthic Ecology and Evolution Group, Centre for Ecological and Evolutionary StudiesUniversity of Groningen
  • Lucía Couceiro
    • Department of Ecology, Faculty of SciencesUniversidade da Coruña
  • José Miguel Ruiz
    • Department of Ecology, Faculty of SciencesUniversidade da Coruña
  • Rodolfo Barreiro
    • Department of Ecology, Faculty of SciencesUniversidade da Coruña
Original Paper

DOI: 10.1007/s00227-012-1975-9

Cite this article as:
Albaina, N., Olsen, J.L., Couceiro, L. et al. Mar Biol (2012) 159: 1871. doi:10.1007/s00227-012-1975-9

Abstract

Because marine species respond differentially to factors governing survival and gene flow, closely related taxa may display dissimilar phylogeographic histories. New data for the patchily distributed gastropod Nassarius nitidus throughout its Atlantic–Mediterranean range (collected during 2008 and 2009) were used to investigate its phylogeography and recent demography. Results based on mitochondrial COI sequences of 422 N. nitidus individuals from 15 localities revealed contrasting phylogeographic and demographic patterns among N. nitidus populations from each basin. Data suggest the existence of two glacial refugia, one in the Atlantic, around the Iberian Peninsula, and the other in the Paleo-Mediterranean Sea (Adriatic). Bayesian skyline reconstructions suggest that the Adriatic population of N. nitidus remained largely unaffected by the Last Glacial Maximum (LGM), whereas the Iberian Atlantic region experienced dramatic exponential growth after its conclusion. Contemporary North Sea populations of N. nitidus are the endpoint of a leading-edge recolonization process from a southern position. Additionally, a reanalysis of pre-existing material for the continuously distributed close congener N. reticulatus was used to compare both species in the late histories. In contrast to N. nitidus, N. reticulatus prospered during the LGM and experienced an earlier Atlantic expansion during the previous interglacial period. Despite similar life history and dispersal potential, the results here presented suggest that subtle differences in microhabitat requirements between the two species have had important consequences for their particular distribution in response to glacial events.

Introduction

Understanding the interplay of historical, ecological, geological and climatic factors governing the distribution of genealogical lineages within and among closely related species is the primary objective of phylogeography (Avise et al. 1987). Events that occurred during the Tertiary and Quaternary periods have played an important role in species origin and evolution towards their present status. In particular, the last 2.5 million years (MY) have been a time of intense climatic fluctuations in the North Atlantic marine intertidal (Cunningham and Collins 1998; Wares and Cunningham 2001). Following the extension of the northern ice sheets during glacial periods and the drop of sea level, the range of distribution of intertidal organisms was generally reduced and moved southwards. Although the fast pace of habitat alteration often prevented benthic organisms from adapting and triggered local and regional extinctions (Cunningham and Collins 1998; Williams et al. 1998; Thomas et al. 2004), it is also known that glacial refugia provided survival of a fraction of individuals, while the same type of organisms disappeared from the surrounding areas (Andersen and Borns 1994; Hewitt 1996; Maggs et al. 2008; Provan and Bennett 2008).

The phylogeographic history of a species is the sum of abiotic processes related to coastal topography (Wares et al. 2001; Marko 2002; Duran et al. 2004; Waters et al. 2005), prevailing currents (Galarza et al. 2009), nutrients and chance events (Cowen and Sponaugle 2009), as well as to biotic interactions related to competition (Kinlan and Gaines 2003) and species-specific traits such as larval development (Kyle and Boulding 2000) or dispersal capacity (Shanks et al. 2003; Gysels et al. 2004a; Roman and Palumbi 2004). As a consequence of such complex interaction, the fact that two species have gone through the same paleoclimatic episodes, have a close phylogenetic relationship or have similar larval dispersal capacities does not necessarily guarantee that they share phylogeographic patterns (Taberlet et al. 1998; Avise 2004).

The estuarine dog-whelk Nassarius nitidus (Jeffreys, 1867) is a benthic dweller on muddy-sandy bottoms of the Black Sea, the Mediterranean and the NE Atlantic, the latter extending to southern Norway (Hansson 1998). Although mainly restricted to sheltered bays and estuaries along the northern Atlantic coast of Spain (Fig. 1a), it is found through a wider range of habitats in the Wadden Sea and North Sea (Rolán and Luque 1994; Craeymeersch and Rietveld 2005). Due to morphological similarities, N. nitidus is still commonly misidentified as the netted dog-whelk N. reticulatus (Linnaeus 1758). N. reticulatus, however, has a much broader Atlantic distribution (Fig. 1a) including some enclaves in the Spanish Mediterranean (Rolán and Luque 1994; Couceiro et al. 2007). N. reticulatus has fewer habitat requirements with respect to sediment, salinity or embayment type (Fretter and Graham 1994). Although the two species display subtle differences in habitat preferences (upper estuaries for N. nitidus; wider range of salinity and exposure conditions for N. reticulatus), they are occasionally found in sympatry, particularly along the northern coast of Spain. Both species have similar reproductive cycles (Barroso and Moreira 1998; Valli et al. 2003), a relatively long planktotrophic larval stage (1–2 months), and share a scavenging-based diet (Tallmark 1980; Fretter and Graham 1994). Although our knowledge about N. reticulatus is relatively extensive (Barreiro et al. 2001; Barroso et al. 2005; Chatzinikolaou and Richardson 2008; Couceiro et al. 2009; Ruiz et al. 2010), studies on N. nitidus are scarce and, to a great extent, have focused on taxonomic issues (Collyer 1961; Rolán and Luque 1994; Sanjuán et al. 1997; Carvajal-Rodríguez et al. 2006; Couceiro et al. 2011). With respect to past and present phylogeographic status of N. reticulatus, Couceiro et al. (2007) analysed mtDNA from samples along 1,700-km of the NE Atlantic coastline. Their results suggested the existence of a single genetic population that experienced a demographic expansion from a refuge in the Bay of Biscay between 160 and 61.5 ky BP.
https://static-content.springer.com/image/art%3A10.1007%2Fs00227-012-1975-9/MediaObjects/227_2012_1975_Fig1_HTML.gif
Fig. 1

Map of the European coasts showing sampling sites for N. nitidus. a Diversity plot following Petit et al. (2003) scheme in which red and blue circles represent diversities above and below the mean, respectively. Diameter is proportional to distance from the average value. Beside each the North Sea, North Iberian Peninsula (NIP) and Adriatic regions, the proportion of private haplotypes with respect to the total number of haplotypes is indicated. Surveyed sites lacking N. nitidus are marked with black stars. From N to S and from W to E: Norway, Germany, UK (2 sites), France (Atlantic, 2), Portugal (2), Spain (3), France (Mediterranean, 2), Italy, Greece, Bulgary, Turkey and Israel. Distribution ranges for N. reticulatus and for N. nitidus are indicated by green and orange lines, respectively. b Haplotype frequencies. Dashed lines indicate the three clusters detected in the parsimony network (Fig. 2). The LGM coastline is marked by a solid black line; the British and Scandinavian ice sheet extension is depicted in grey. Redrawn and modified from several sources (colour figure online)

The aims of the present study were to assess the phylogeography and historical demography of the patchily distributed N. nitidus. We used mitochondrial cytochrome c oxidase subunit I (COI) sequences from Atlantic and Mediterranean samples to: (1) investigate N. nitidus population structure along its distribution range; (2) explore the demographic history of N. nitidus populations with a focus on the Last Glacial Maximum (LGM; 26.5–19 years ago) (Clark et al. 2009); and (3) ascertain whether N. nitidus demographic processes were comparable to those experienced by the continuously distributed N. reticulatus. This study completes those preliminary insights of Couceiro et al. (2011) to N. nitidus population structure and constitutes the first approach to recent demography of both N. nitidus and N. reticulatus species.

Materials and methods

Sampling

An extensive search of N. nitidus from Norway to Turkey was conducted in 2008–2009. The survey covered almost entirely the geographic range of the species (which is also present in the Black Sea). The first author (N. A.) personally visited locations for which N. nitidus had been reported and/or consulted with local, place-based colleagues (see Fig. 1a; Table given in Online Resource 1, OR1).The species was not found along large stretches of potentially suitable coast on the Atlantic basin. Rather, specimens were found in only a few sheltered bays, which were hundreds or thousands of km distant from the next nearest area of presence. In the Mediterranean, N. nitidus was found only in the Adriatic despite several efforts to collect it in the western Mediterranean, Tyrrhenian and Aegean Seas (Fig. 1a; OR1).

At each of the 15 sampling sites (Fig. 1), individuals were attracted by mussel bait and brought into the laboratory for immediate processing. When transport to the laboratory was delayed, specimens were preserved in 96 % ethanol.

DNA extraction and sequencing

DNA was extracted from foot tissue using Chelex-100 resin (Estoup et al. 1996) or a spin column kit (DNeasy Blood and Tissue Kit, Qiagen International). Universal primers were used as described by Folmer et al. (1994). A 710-bp fragment of the mitochondrial COI gene was amplified, from which a complete 477-bp alignment was obtained. Reactions were performed in a 25 μl solution containing 5 μM of each primer (Bonsai Technologies Group), 0.2 mM of each dNTP (Fermentas), 2 μM MgCl2, 1× Buffer, 0.4 U AmpliTaq DNA Polymerase (Applied Biosystems) and 1 μl of template DNA. The thermal cycling protocol consisted of an initial denaturation of 95ºC for 2 min followed by 40 cycles of 95 °C for 30 s, 45 °C for 30 s and 72 °C for 1 min. PCR products were purified with exonuclease I and shrimp alkaline phosphatase (Fermentas) and sequenced on ABI 3730XL (Applied Biosystems) using BigDyeTM (Applied Biosystems) terminator cycling conditions.

Data analysis

COI sequences were unambiguously aligned and edited in CODONCODE ALIGNER v.2.0 (CodonCode Corporation, Dedham, MA, USA). Species identification was verified by following Couceiro et al. (2011). Haplotype and nucleotide diversities were determined with DNASP v.4.5 (Rozas et al. 2003). Gene genealogies were constructed using the statistical parsimony method described by Templeton et al. (1992) with TCS (Clement et al. 2000). To provide statistical support to genealogical groups detected in the parsimony network, a Bayesian phylogenetic analysis was carried out in MRBAYES3 (Ronquist and Huelsenbeck 2003), using two N. reticulatus sequences Couceiro et al. 2007) as outgroup. Since the N. reticulatus COI fragments were shorter than those from N. nitidus, our sequences were trimmed and the phylogenetic analysis was performed using only the 395-bp common segment. Searches were run for 10 million MCMC generations and sampled every 100. First 50,000 generations were discarded. The best-fit model of evolution for the data was determined using hierarchical likelihood ratio tests in MRMODELTEST v.2.3 (Nylander 2004) through PAUP v.4.0 (Swofford 2002).

Genetic structure within and between groups of samples (based on haplotypes and diversity) was tested using ARLEQUIN v.3.1 (Excoffier et al. 2005). Analyses of molecular variance (AMOVA, Excoffier et al. 1992) and also pairwise comparisons (Reynolds et al. 1983; Slatkin 1995) were performed based on F (Fst, Fct and Fsc, Wright 1951) and Φ (Φst, Φct and Φsc, Excoffier et al. 1992) diversity indexes. For Ф-statistic calculations, Tamura–Nei + Γ distance (arlequin best alternative to the HKY + Γ model followed by our data) was used. Significance was assessed by permuting data for 16,000 replicates and by applying Hochberg’s correction (1988).

To test for isolation by distance (IBD) over the entire sampling range, as well as for the Atlantic data, we used the software ISOLDE as implemented in GENEPOP v.4.0 (GENEPOP on the Web; Raymond and Rousset 1995; Rousset 2008), where a Mantel test (Mantel 1967) between genetic (Φst/1 − Φst) and geographic pairwise distances under a two-dimensional stepping stone model (Rousset 1997) was performed via 16,000 permutations.

The demographic history of each phylogeographic group was explored in two ways. First, neutrality statistics based on the distribution of the mutation frequencies (Tajima’s D; Tajima 1989) and on the haplotype distribution (Fu’s Fs; Fu 1997) were estimated (H0: constant effective population size), and the mismatch distribution profile obtained (Rogers and Harpending 1992). Significance was assessed by 16 000 replicates in ARLEQUIN v.3.1 (Excoffier et al. 2005). The fit of our results to a model of demographic expansion (H0: population expansion; Rogers and Harpending 1992) was further evaluated by considering both the sum of square deviations (SSD) and the raggedness index (rg, Harpending 1994). Whenever both the neutrality tests and the mismatch-based distribution tests fit an expansion model, the relationship τ = 2ut (Rogers and Harpending 1992) was used to approach the time since expansion, where τ is the empirical estimate of the time of expansion, u is the mutation rate per sequence per year, and t is the time to present since the beginning of the expansion. It is important to recall that a non-rejection of the null hypothesis of expansion should not necessarily be treated as proof that the hypothesis is true. In particular, the SSD test is known to be very conservative and rarely leads to rejection of the expansion model (Schneider and Excoffier 1999; Ramos-Onsins and Rozas 2002).

The posterior distribution of the effective population size through time and the moment of haplotype coalescence (Most Recent Common Ancestor, MRCA) was inferred using the Bayesian Skyline Plot model (BSP; Drummond et al. 2005) implemented in BEAST v.1.5.3 (Drummond and Rambaut 2007). The BSP framework is a valuable tool for tracking changes in effective population size because all of the genealogy, demographic history and substitution-model parameters are co-estimated in a single analysis. Therefore, confidence intervals represent the combined phylogenetic and coalescent uncertainty (Ho and Shapiro 2011). As the model assumes no historical gene exchange, populations are considered to be independent. Although an analysis of isolation with migration using ima (Hey and Nielsen 2007) proved impossible due to repeated crashes of the program, the other analyses provided enough evidence to consider the presence of two separate, isolated populations of N. nitidus, supporting the compatibility of our data under an isolation model in BEAST. Consequently, the Adriatic and North Iberian Peninsula (NIP) groups were treated separately, while the SW, NE and BE samples were excluded from the simulation. Each subset was run twice for 50 million generations and sampled every 1,000. A strict molecular clock was assumed under an HKY + I + Γ substitution model, starting from an UPGMA tree and applying default priors after a burn-in period of 5 million steps. The fit of data to a strict molecular clock rate was verified by calculating the standard deviation of the uncorrelated lognormal relaxed clock. Log files for each run were pooled with logcombiner to check for effective sample size values (ESS). We assumed Wilke’s “trait-specific” substitution rate estimated for the COI gene of small, aquatic, dioecious, ectothermic and one-year generation time Protostomia (1.57 ± 0.45 % My−1, under the HKY + I + Γ model; Wilke et al. 2009) to obtain results in relevant time units (errors associated with the substitution model were incorporated in data estimations). The clock was calibrated using substitution rates calculated for 12 pairs of sister taxa from five groups of Protostomia separated by discrete biogeographic events. As always, molecular clock calculations must be interpreted with caution because the clock rate is not corrected for ancestral polymorphisms and is likely to overestimate divergence times.

The same BSP procedure in BEAST was also applied to the close relative N. reticulatus using 156 (395-bp) COI sequences previously published (for which only one single but widely spread Atlantic population has been reported; Couceiro et al. 2007). The N. reticulatus data fit an HKY model (modeltest best-fit model of evolution), and consequently, a 1.24 ± 0.22 % My−1 substitution rate (Wilke’s clock rate under the HKY model) was used to transform estimates in appropriate time units. To rule out that any difference between N. reticulatus and N. nitidus could be due to the use of different substitution rates for each species, the demographic history of N. reticulatus was also estimated using the same substitution model and rate employed with N. nitidus.

Results

Genetic diversity and haplotype genealogy

The 477-bp COI fragment obtained for 422 individuals of N. nitidus produced a total of 68 polymorphic sites (14.3 %) of which 33 were parsimony informative (6.9 %). Of 81 haplotypes identified, 64 (79 %) were unique to a single location (Table 1). Haplotypes can be found in GenBank (accession numbers: GQ219744–GQ219751, GQ219753–GQ219765 and GU270587–GU270648).
Table 1

N. nitidus sampling localities, identifiers (ID), geographic coordinates, sample size (n), number of segregating sites (S), number of total (Ht) and private (Hp) haplotypes, haplotype (h) and nucleotide (π × 102) diversities (with SD)

Locality

ID

Position

n

S

Ht

Hp

h

π × 102

Lysekil, Sweden

SW

58°16′ N, 11°25′ W

41

1

2

1

0.049 (0.046)

0.01 (0.01)

Grevelingen, Netherlands

NE

51°42′ N, 03°59′ W

30

3

2

0

0.186 (0.088)

0.12 (0.06)

Oostende, Belgium

BE

51°13′ N, 02°57′ W

23

4

3

1

0.312 (0.115)

0.23 (0.13)

Fuenterrabía*, Spain

SP (1)

43°22′ N, 01°47′ W

25

13

6

1

0.727 (0.061)

0.77 (0.13)

Santoña*, Spain

SP (2)

43°26′ N, 03°27′ W

21

13

7

2

0.733 (0.082)

0.76 (0.32)

Neda*, Spain

SP (3)

43°29′′ N, 08°10′ W

26

10

4

1

0.618 (0.058)

0.77 (0.08)

P. porco*, Spain

SP (4)

43°20′ N, 08°12′ W

29

14

8

3

0.702 (0.079)

0.63 (0.12)

Sada*, Spain

SP (5)

43°21′ N, 08°14′ W

34

13

8

4

0.783 (0.039)

0.78 (0.09)

Rianxo, Spain

SP (6)

42°38′ N, 08°48′ W

47

15

7

3

0.698 (0.038)

0.82 (0.06)

Poio, Spain

SP (7)

42°25′ N, 08°41′ W

31

16

7

3

0.662 (0.056)

0.80 (0.07)

S. Simón*, Spain

SP (8)

42°18′ N, 08°37′ W

26

14

8

4

0.711 (0.079)

0.80 (0.08)

Civitanova, Italy

IT (1)

43°18′ N, 13°43′ W

25

27

21

14

0.983 (0.017)

0.82 (0.08)

Cesenatico, Italy

IT (2)

44°11′ N, 12°24′ W

30

33

25

14

0.982 (0.016)

0.85 (0.07)

Venice, Italy

IT (3)

45°25′ N, 12°20′ W

20

27

17

9

0.984 (0.020)

0.96 (0.08)

Grado, Italy

IT (4)

45°40′ N, 13°24′ W

14

12

8

4

0.890 (0.060)

0.57 (0.10)

Regional total

 North Sea

  

94

5

4

2

0.160 (0.050)

0.09 (0.03)

 NIP

  

239

37

27

18

0.722 (0.018)

0.81 (0.03)

 Adriatic

  

89

48

54

47

0.976 (0.007)

0.82 (0.04)

Total

  

422

68

81

67

0.759 (0.019)

0.75 (0.03)

Asterisks denote sites where N. reticulatus was found in sympatry with N. nitidus

Locations in the North Sea region of the Atlantic basin (henceforth North Sea) exhibited the lowest haplotype (mean: 0.16) and nucleotide (mean: 0.09 × 10−2) diversity estimates. Diversity gradually increased southwards in NIP sites (0.72 and 0.81 × 10−2) and reached their maxima in the Adriatic Sea (0.976 and 0.82 × 10−2) (Table 1; Fig. 1). The most common haplotype (h20) was present at every locality with the exception of Venice (Fig. 1). Its relative abundance gradually decreased from an overwhelming dominance in North Sea sites, to 23–52 % in NIP, finally reaching minimum values in the Adriatic (7.9 % on average). Haplotypes h16 and h21 were likewise abundant, but they occurred exclusively in the Atlantic (average frequency of 12 and 23 %, respectively). H16 was found along the entire Atlantic coastline, while h21 was only detected in NIP sites. The predominance of just three haplotypes in the Atlantic contrasted with the high number of low-frequency haplotypes found in the Adriatic, where only h53, h59 and the aforementioned h20 were locally abundant.

The haplotype network (Fig. 2) revealed three secondary lineages, each one derived from the common haplotypes h16, h21, and h59 and interconnected through the central, highly abundant h20. For these secondary lineages, posterior probabilities were ≥ 0.92 (see the Bayesian phylogenetic tree OR2). Clade 1 (around haplotype h21) was endemic to the NIP (posterior probability value 1.00) and considerably divergent (six mutational steps) from the ancestral h20. Clade 2 (posterior probability value 0.92), although mostly found at the NIP sites, also had representatives in the North Sea. Interestingly, both clades 1 and 2 emerged from h21 through haplotypes found only in the Adriatic (for clade 1, this observation was supported by a 0.98 posterior probability of the phylogenetic analysis). Mean sequence divergence between clades 1 and 2 was 2.22 % (Tamura–Nei distance) and included 19 fixed mutational positions. Finally, clade 3 (posterior probability value 0.99) consisted of 19 haplotypes recently derived from h59 and was confined to the Adriatic Sea (phylogenetic posterior probability value 0.99; this analysis included in clade 3 two single types sampled elsewhere, probably as a result of the shortening of sequences necessary for tree construction). The remainder of the network (81 %) consisted of predominantly single-mutation haplotypes always < 3 mutations away from the dominant h16, h21, h59 and h20. This star-like pattern is often regarded as a signature of a recent population expansion. It is worth noting that the haplotypes found on each side of the Strait of Gibraltar were mutually exclusive, with the exception of the ancestral, highly abundant in the Atlantic h20 and the phylogenetically close h42, for which nine individuals (seven and two, respectively) were collected in the Mediterranean. At least two of the terminal haplotypes collected in the NIP (h29 and h40, and probably h41) appear as direct descendants of a series of Adriatic types (2–3 mutational steps not sampled in the NIP). In addition, the h41 position was supported by the Bayesian phylogenetic inference (see OR2).
https://static-content.springer.com/image/art%3A10.1007%2Fs00227-012-1975-9/MediaObjects/227_2012_1975_Fig2_HTML.gif
Fig. 2

Statistical parsimony network for 81 N. nitidus COI haplotypes. Each circle represents a haplotype with its name (outside) and sampling sites (inside). Circle size is proportional to strain frequency. Coloured dashed lines delineate the three main lineage clusters. Lines between haplotypes are single mutational steps; small black dots indicate unrecorded haplotypes (colour figure online)

Population structure

Estimates of population differentiation based on haplotype and nucleotide distances were nearly identical; therefore, only the results for nucleotide distances are presented. The AMOVA revealed a highly significant structure (Φct = 0.196, P < 0.001) among the three sets of samples (North Sea, NIP and Adriatic). Differences within groups were not significant and accounted for only 3.4 % of the total variance. Except for a few cases involving sites NE, BE and IT (3), most inter-regional pairwise comparisons were significant after Hochberg’s correction. Pairwise Φst values ranged from 0.067 to 0.563 for the North Sea–NIP contrasts, from 0.073 to 0.435 for the NIP–Adriatic combinations, and from 0.063 to 0.491 for the Adriatic–North Sea ones (see Table in OR3).

Genetic divergence estimates were positively correlated with geographic distance for both the whole set of samples, as well as for the more restricted Atlantic set (P < 0.001, r = 0.494 and P < 0.001, r = 0.518, respectively; not shown).

Historical demography

Under the assumption of selective neutrality, Tajima’s D and Fu’s Fs regional estimates for the North Sea and NIP failed to detect any significant departure from the null hypothesis of equilibrium (Table 2). For the Adriatic, significant D and Fs estimations indicated an excess of low-frequency haplotypes. Subsequent mismatch distribution analyses revealed different patterns among regions (Fig. 3). In the North Sea, a strongly skewed, unimodal profile was obtained, and neither the SSD nor the rg index rejected the null hypothesis of demographic growth. The NIP graph showed a multimodal distribution. In this case, the statistical tests for population expansion produced conflicting results. Whereas the rg statistic rejected an event of population growth, the SSD did not. Finally, the Adriatic was characterized by a smooth unimodal curve with high statistical support. As for this region both the neutrality and mismatch distribution tests were consistent with an episode of population proliferation, we used the ARLEQUIN inference of τ to estimate the time since expansion. According to a τ = 4.084 estimation, expansion would have started some 273 ky BP (95 % confidence interval: 182–315 ky BP). It must be noted that the molecular clock rate employed in this study resulted in wide posterior probability intervals for date estimations. For this reason, we should be circumspect when interpreting results. Nevertheless, even taking the previous fact into account, BEAST coalescent reconstruction of past demographic events again showed contrasting patterns between the NIP and the Adriatic (Fig. 4). According to the beast inference, the Adriatic N. nitidus population underwent a slow exponential growth approximately from 220 (150–320) ky BP and then reached a plateau after the LGM (0–15 ky BP; Fig. 4). Effective population size increased by 60-fold in the course of this 200-ky expansion. The MRCA is estimated to have occurred 347 (156–584) ky BP. For the NIP haplotypes, the MRCA was estimated to have arisen some 522 ky BP (95 % highest posterior density interval: 203–899 ky BP), due to an early split between the lineages leading to clades 1 and 2. Only after the end of the LGM (15–30 ky BP) did a demographic burst take place of approximately 40-fold.
Table 2

N. nitidus. Analyses of population stability (Tajima’s D and Fu’s Fs tests) and population expansion (sum of squared deviations SSD and raggedness rg mismatch distribution tests)

Region

Tajima’s D

Fu’s Fs

Mismatch SSD, rg

North Sea

−1.260

(0.088)

−0.285

(0.285)

0.020, 0.670

(0.101, 0.700)

Northern Iberian Peninsula

−1.040

(0.143)

−5.538

(0.087)

0.072, 0.175

(0.076, 0.017)

Adriatic Sea

1.887

(0.007)

25.951

(<0.001)

0.001, 0.019

(0.421, 0.548)

Parameters and probabilities (in brackets) are shown for each regional group. Significant P values are indicated in bold and italics

https://static-content.springer.com/image/art%3A10.1007%2Fs00227-012-1975-9/MediaObjects/227_2012_1975_Fig3_HTML.gif
Fig. 3

Mismatch distribution for the three regional groups of N. nitidus populations, North Sea, North Iberian Peninsula (NIP) and Adriatic. Bars represent the observed distribution of pairwise differences, while solid lines show the expected distribution under a pure demographic expansion model (colour figure online)

https://static-content.springer.com/image/art%3A10.1007%2Fs00227-012-1975-9/MediaObjects/227_2012_1975_Fig4_HTML.gif
Fig. 4

Bayesian skyline plot (BSP) for N. nitidus in Atlantic (NIP) and Mediterranean (Adriatic) refugia and for N. reticulatus in the Atlantic. Curves represent population sizes (effective population size per generation time; Ne.τ) progress to their most recent common ancestor (MRCA). Dashed vertical lines indicate the lower end of the 95 % highest posterior density (HPD) interval for the MRCA estimate, and grey fill represents the 95 % HPD of the BSP. The LGM extent according to Clark et al. (2009) is shown on each chart. Above and below the graphs, two horizontal diagrams indicate warm (white) and cold (grey) Marine Isotopic Stages (MIS; Gibbard) (colour figure online)

According to the genealogical trees generated for the BSP reconstructions of N. nitidus (figures not shown), the point estimate for the coalescence time of clade 3 haplotypes (220; 95 % HPD: 110–350 ky BP) is older than the upper confidence interval for those in both clades 1 (76; 95 % HPD: 18–164 ky BP) and 2 (62; 95 % HPD: 16–127 ky BP) (posterior probabilities ≥0.93). These estimates suggest an earlier persistence of the Adriatic clade 3 and a more recent origin of both the Atlantic clades 1 and 2, consistent with events surrounding the LGM.

The historical demographic analysis of the congener N. reticulatus (from Belgium to southern Portugal) indicated that the MRCA may have existed at least 353 (75–847) ky BP. This species seems to have increased its effective population size around 75-fold in an exponential expansion that started at 100 ky BP (88–113 ky BP) and ended at the time of the LGM (Fig. 4). According to these models, contemporary population sizes for both the NIP and Adriatic N. nitidus groups would be similar but smaller as compared to those of N. reticulatus.

Discussion

The impact of a major historical event is expected to be manifested by phylogeographic concordances across unrelated taxa (Avise 2004). While there is agreement about the homogenizing effects that glaciations had in the NW Atlantic biota (Wares and Cunningham 2001; Addison and Hart 2005), current evidence for the NE Atlantic is more difficult to interpret, and a single generalized pattern across taxa cannot be inferred (Maggs et al. 2008). Reasons for this are related to sea surface temperature isotherms, the complexity of the coastline and the much greater variation in suitable habitat types along the European coasts. Nevertheless, some of the large-scale patterns are fairly consistent among taxa, and N. nitidus is no exception.

The Mediterranean–Atlantic split

The Strait of Gibraltar, together with the Almeria-Oran oceanographic front (Tintore et al. 1988), is a well-documented barrier to gene flow between the Atlantic and Mediterranean, as illustrated for seaweeds (Andreakis et al. 2004), seagrasses (Coyer et al. 2004; Olsen et al. 2004), invertebrates (Roman and Palumbi 2004) and fishes (Gysels et al. 2004a, b) (revised by Patarnello et al. 2007). Our analyses of N. nitidus also support this phylogeographic break, with each region comprising distinct clades and radically different diversities, that is, a few abundant haplotypes in the Atlantic and many low frequency ones in the Adriatic. Indeed, only two haplotypes were shared between the two basins.

Within the Atlantic, two groups were identified (Table 2). Although the ancestral h20 and the more recent h16 occurred along the entire Atlantic coastline, their dominance reached maximum values only in the North Sea. In addition, the mismatch distribution indicates different demographic histories for each of these subregions, admixture at the NIP (see below) and a very recent northward expansion into the North Sea. A division between the North Sea and the NE Atlantic has been repeatedly observed for other species (Luttikhuizen et al. 2003; Jolly et al. 2006; Spalding et al. 2007; Strasser and Barber 2009), and it is believed that a transition area between the Northern European and the Lusitanian marine provinces is situated in the Hurd Deep within the present-day English Channel (Provan et al. 2005; Maggs et al. 2008).

Refugia

The high haplotype diversity, deep divergence and abundance of private haplotypes in the Adriatic and the NIP are consistent with at least two glacial refugia (Hoarau et al. 2007; Maggs et al. 2008). One refugium, located in the low-stand, Paleo-Mediterranean Sea (Bianchi and Morri 2000) would account for the current Adriatic haplotypes, while the other, situated on the Atlantic coast of the Iberian Peninsula, would account for the Atlantic lineages.

The high haplotype diversity and phylogenetic depth of the Adriatic branch are consistent with long residency in a Mediterranean refugium. Population growth and expansion are evident, and no sign of admixture was detected. Our estimate places the start of this steady proliferation in the Saalian glaciation, probably during the interstadials 180–240 ky BP (Marine Isotopic Stages, MIS, 7a, c, e). The slight slowdown detected during the LGM indicates that this population was only weakly affected by cooling events.

The Atlantic Iberian coast was the second refugium based on the dominance of three distant lineages, one of them the ancestral lineage. Although the results obtained for the tests based on the multimodal mismatch distribution for the NIP may be indicative of a large stable population (Rogers and Harpending 1992), they are also consistent with the admixture of at least two different expanding units (Alvarado Bremer et al. 2005). Indeed, we identified two highly divergent Atlantic clades in the Iberian samples, each of them with a star-like topology (Figs. 2, OR2). Moreover, the Bayesian reconstruction showed a minimum-sized population that underwent rapid expansion after the LGM. The North Sea region comprises a subsample of clade 1 haplotypes that undertook a process of sudden population expansion. These patterns are consistent with Pleistocene marine refugia found in several other marine intertidal organisms including a seaweed (Hoarau et al. 2007), a crustacean (Campo et al. 2010) and a vertebrate (Chevolot et al. 2006).

Mediterranean origin

Lineages surviving in the Atlantic refugium (clade 1 at 76 ky BP and clade 2 at 62 ky BP) had a Mediterranean origin given the MRCA estimated for the Adriatic clade 3 at 220 ky BP. Previous studies of the benthic vertebrate, Pomatoschistus microps (Gysels et al. 2004b) and the gastropod Hydrobia acuta (Wilke and Pfenninger 2002) also support a Mediterranean origin for lineages surviving in two different Mediterranean and Atlantic refugia during the Pleistocene. Sea level fluctuations between 76 and 220 ky BP (Gibbard and Cohen 2008) are believed to have facilitated the movement of organisms through the Strait of Gibraltar (Kooistra et al. 1992). The early Mediterranean N. nitidus population expansion (220 ky BP) and the very recent NIP expansion (19 ky BP) estimates are in good agreement with this hypothesis.

A scenario in which clades 1 and 2 arose in the Mediterranean and migrated into the Atlantic after the LGM is unlikely. First, each Atlantic clade occupies a peripheral position in the genealogical network and only shares one haplotype with the Mediterranean cluster. Second, the star-like pattern of each Atlantic clade is typical of a pre-established population undergoing expansion rather than experiencing a leading-edge colonization process (Hewitt 1996). Third, if a glacial episode had removed the less frequent haplotypes (including ancestors to modern clades derived from h20) after migration of the clade to the Atlantic basin, the ancestral genetic signature for the Atlantic populations (but not the Mediterranean) would be expected to have disappeared (as it is in this basin where climatic pressure reached extreme conditions). In contrast, a high number of Adriatic haplotypes in the branch leading to the Atlantic clade 1 are present. In this light, the ancestral h20 would be absent, or at least scarce, in the Atlantic coast; yet it is abundant. Thus, even taking into account the limitations inherent to the interpretation of haplotype networks, the requirement that all of these events would have co-occurred makes support for this scenario weak.

Pre- and post-LGM expansions

Although the LGM affected both the Atlantic and Mediterranean, it did so in different ways. In the Atlantic basin, the Scandinavian and English ice sheets covered the Baltic Sea (totally) and the North Sea (partially). Coastal areas such as the Hurd Deep and the English Channel were left exposed, as the sea level reached a low stand of −100 m (Lambeck 1997; Clark and Mix 2002). In the Mediterranean, conditions were milder, with sea level changes shifting benthic taxa to the deeper central basins (Bianchi and Morri 2000; Patarnello et al. 2007). Despite the southern and eastern displacement of the Adriatic population of N. nitidus, demographic expansion continued, whereas in the Atlantic, this snail survived in the NIP refugium but did not expand until the end of the LGM 20 ky BP. Subsequent colonization of the North Sea area from the NIP then proceeded. Our results match the post-LGM patterns obtained for other marine invertebrates (Gysels et al. 2004a; Maggs et al. 2008) and seaweeds (Coyer et al. 2003; Provan et al. 2005; Hoarau et al. 2007; Olsen et al. 2010) inhabiting the Atlanto-Mediterranean territories and are also comparable to those obtained for seaweeds that experienced a post-LGM expansion in distant, temperate-sub-Antartic waters (Fraser et al. 2009).

Although survival in glacial refugia during the LGM was widespread, the demographic impact that we can infer depends upon whether the refugia were large or small. Interestingly, the Atlantic congener N. reticulatus (scarcely present in the Mediterranean; Fig. 4) appears to have not been significantly affected by the LGM. Its population expansion started during the penultimate interglacial period and has remained relatively stable since the LGM. Our results are consistent with the estimation of Couceiro et al. (2007) of a population growth starting 62–160 ky BP. These authors also found little evidence for any population structure in N. reticulatus and concluded that the high genetic diversity in the mid-Atlantic was consistent with long-term presence of the species in the Bay of Biscay. Similarly, examples of other taxa whose status was little affected by the LGM include the brown seaweed Ascophyllum nodosum (Olsen et al. 2010) and the barnacle Pollicipes pollicipes (Campo et al. 2010).

Patchy versus continuous distributions

Our intertidal and subtidal surveys of potential habitat for N. nitidus suggest a naturally patchy distribution. While we cannot claim to have searched every kilometre of coastline, a major effort was made to visit sites recorded in the literature, those reported by colleagues with local knowledge and many new ones through our own newly initiated efforts. Dependency on soft, sheltered, brackish habitats appears to be more important than traditionally thought, especially recalling the fact that the congener N. reticulatus is comfortable in a broader range of muddy-sandy habitats that overlaps with those of N. nitidus. While acknowledging that further field explorations should be accomplished to confirm the evidences presented here, the observed differences between N. nitidus and N. reticulatus phylogeographies are comparable to those reported in previous studies of marine gastropods where small differences in the life history of congenerics have resulted in contrasting distributions (Wilke and Pfenninger 2002; Crandall et al. 2008; Marko 2004). Likewise, and despite the fact that N. nitidus and N. reticulatus are closely related and share similar life histories and dispersal capacities, differences in microhabitat requirements may have shaped their specific response to climatic shifts and thus their respective post-LGM distributions—one patchy and one continuous.

Genetic drift versus artificial population connectivity

The presence of three Atlantic haplotypes at the tips of the Adriatic clade, one of them supported by the phylogenetic analysis, remains difficult to explain in evolutionary terms. Two hypotheses may account for our observations. First, both regions have a long tradition of shellfish culture and commercial exchange. The Adriatic is a typical source for clam transfers to the Iberian Peninsula, and it has been identified as the source of several nonindigenous gastropods (Gibbula adansonii, Gibbula albida, Rapana venosa and Hexaplex trunculus) in sites close to our locations SP (6) and SP (7) (Rolán 1992; Rolán and Bañón 2007). In the case this theory was true, gene flow and genetic diversity estimations here presented would be hardly reliable. However, our results are in proper agreement with geoclimatic history and with studies on other coastal benthic organisms. A second possibility would be strong genetic drift acting on the Atlantic populations. As mentioned before, it is assumed the NE Atlantic was a hostile environment for intertidal species during glacial intervals, especially for brackish habitat species. Brackish biota is characterized by a high degree of population differentiation resulting from the discrete nature of habitat and their relative confinement. Thus, for both Nassarius species, the effects of genetic drift acting on local settlements are expected (reviewed by Bilton et al. 2002; Johannesson and André 2006). To our opinion, in order to avoid stochastic events acting on small size mitochondrial populations, further investigations based on non-mitochondrial genetic information are required to resolve this issue.

Conclusion

Pleistocene climatic shifts affected population structure and recent demography of the Atlanto-Mediterranean estuarine dog-whelk N. nitidus. A sharp division between Adriatic and Atlantic populations was found, with the Atlantic populations originally derived from the Adriatic. Two refugia were identified: the Adriatic experienced a steady population growth over 200 ky, while the NIP experienced rapid expansion only during the last 20 ky. Our results are consistent with numerous other studies, most of which are also exclusively based on mtDNA. However, it is increasingly recognized that the addition of nuclear loci (especially those under selection) may provide a more complete and more nuanced view of the phylogeographic history of a species and the superimposed ecological pressures that have further shaped distribution. In the present study, for example, the Atlantic congener N. reticulatus seems to have been scarcely affected by the LGM despite similar morphology and life history features. Such results highlight the fact that seemingly insignificant intertidal habitat differences may have allowed some species to avoid the ice by going slightly lower and deeper down the shore.

Acknowledgments

We thank C. Bernárdez, J. Craeymeersch, L. Grassia, L. Kellner, F. Kerchov, B. Lundve and C. Mazziotti for providing samples, R. Bao for information and advice about Quaternary chronostatigraphy and paleoceanography and J. Coyer and G. Hoarau for discussions about data analysis and interpretation while N. A. hosting in the Olsen lab. Financial support for this work was provided by the Spanish Ministerio de Educación y Ciencia (MEC) grant CTM2004-04496/MAR (partially co-funded by FEDER, Fondo Europeo de Desarrollo Regional) and the Xunta de Galicia grant PGIDT05PXIC10302PN. N. A. acknowledges her postgraduate fellowship from the Ministerio de Educacion (FPU-MEC, AP2006-03231).

Supplementary material

227_2012_1975_MOESM1_ESM.pdf (96 kb)
Supplementary material 1 (PDF 96 kb)
227_2012_1975_MOESM2_ESM.pdf (72 kb)
Supplementary material 2 (PDF 72 kb)
227_2012_1975_MOESM3_ESM.pdf (75 kb)
Supplementary material 3 (PDF 74 kb)

Copyright information

© Springer-Verlag 2012