INVITED REVIEW: Life on the margin: genetic isolation and diversity loss in a peripheral marine ecosystem, the Baltic Sea


K. Johannesson, Fax: +46 52 66 86 07; E-mail:


Marginal populations are often isolated and under extreme selection pressures resulting in anomalous genetics. Consequently, ecosystems that are geographically and ecologically marginal might have a large share of genetically atypical populations, in need of particular concern in management of these ecosystems. To test this prediction, we analysed genetic data from 29 species inhabiting the low saline Baltic Sea, a geographically and ecologically marginal ecosystem. On average Baltic populations had lost genetic diversity compared to Atlantic populations: a pattern unrelated to dispersal capacity, generation time of species and taxonomic group of organism, but strongly related to type of genetic marker (mitochondrial DNA loci had lost c. 50% diversity, and nuclear loci 10%). Analyses of genetic isolation by geographic distance revealed clinal patterns of differentiation between Baltic and Atlantic regions. For a majority of species, clines were sigmoid with a sharp slope around the Baltic Sea entrance, indicating impeded gene flows between Baltic and Atlantic populations. Some species showed signs of allele frequencies being perturbed at the edge of their distribution inside the Baltic Sea. Despite the short geological history of the Baltic Sea (8000 years), populations inhabiting the Baltic have evolved substantially different from Atlantic populations, probably as a consequence of isolation and bottlenecks, as well as selection on adaptive traits. In addition, the Baltic Sea also acts a refuge for unique evolutionary lineages. This marginal ecosystem is thus vulnerable but also exceedingly valuable, housing unique genes, genotypes and populations that constitute an important genetic resource for management and conservation.


Genetic variation in non-neutral loci is the essential base for population development and long-term survival (Templeton 1981; Templeton et al. 2001). Loss of genetic variation hampers adaptation to new selective regimes such as climate change and environmental contamination. For single-locus traits genetic variation is only very slowly accumulated through spontaneous mutations (Nei et al. 1975), if not regained through gene flow from other populations. On the other hand, quantitative genetic traits, being composed of a high number of individual loci are generally not restricted by lack of new mutations; but random accumulation of slightly deleterious spontaneous mutations may be a problem, particularly in small populations (Lynch et al. 1999). If genetically depauperate populations dominate an ecosystem, this might have pervasive effects on ecosystem development and impair ecosystem resilience. Recent research shows that genetic diversity is intimately linked to ecosystem function and evolution (Neuhauser et al. 2003; Whitham et al. 2003; Hughes & Stachowicz 2004). Eelgrass genotype diversity (as indicated by diversity at microsatellite loci) is, for example, positively correlated with biomass production, plant density and diversity of associated fauna, thus contributing to the maintenance of key ecosystem functions (Reusch et al. 2005). Likewise, larval settlement success increases with increased genetic diversity in a barnacle species (Gamfeldt et al. 2005). An extreme example is when genetic erosion leads to the complete loss of a population that has evolved separately over a long period of time; as with extinct species, such an evolutionary lineage will not re-appear (Moritz 2002).

Ecologically marginal environments (e.g. mountain tops, brackish-water estuaries and arctic habitats) house populations that experience extreme selection regimes. Quite often are these environments also geographically marginal, with populations that are peripheral to species’ main distributions. From this we expect that ecosystems that are both geographically and ecologically marginal will contain populations that are genetically impoverished and extreme. Such ecosystems would be vulnerable, deserving more careful management of genetic resources than central ecosystems. The degree of environmental and geographical marginality influences the genetic structure of populations. Effects of local selection in a peripheral habitat may be swamped by a large gene flow from core populations of the species, tending to shift peripheral populations away from their local ecological optimum (García-Ramos & Kirkpatrick 1997). If, however, a peripheral population becomes more isolated, evolution might be more strongly influenced by local selection pressures.

Empirical data mostly support the prediction that peripheral populations have less genetic variation compared to core populations of the same species (Lesica & Allendorf 1995; Schwartz et al. 2003). In addition, isolated populations that inhabit marginal environments tend to be more genetically differentiated from central populations than isolated populations that do not live in extreme environments (Bouza et al. 1999).

Population bottlenecks reduce genetic variation in proportion to the duration and size of the bottleneck (Chakraborty & Nei 1977; Booy et al. 2000) and peripheral populations are more likely to pass through severe bottlenecks than other populations being smaller than central populations and inhabiting a more demanding environment. The most serious effect of reduced genetic variation is inbreeding. Although considered less threatening in large populations (Ne > 500) (Nunney & Campbell 1993), recent investigations indicate inbreeding to be a larger problem in natural populations than earlier believed (Keller & Waller 2002). Notably, effective population size (Ne) is generally much less than population census size (e.g. Turner et al. 2002), and this effect is likely to be stronger in peripheral populations owing, for example, to high stochastic population fluctuations.

Furthermore, life history characteristics may influence population isolation and genetic structure of species (Lesica & Allendorf 1995). Species with a relatively poor dispersal capacity are expected to be more affected by a peripheral location than long-range dispersers, as the former have a less effective gene flow over geographic scales (Bohonak 1999). Generation time and type of reproduction will influence genetic structures of discrete populations. For example, asexual reproduction tends to be more common in marginal than in central populations (Eckert 2002).

Overall, populations at the edge of species’ distributions are likely to have a more extreme genetic set-up than central populations, and we therefore predicted that marginal ecosystems with a high share of peripheral populations will reveal general patterns of low genetic diversity and gene frequency perturbation over species inhabiting these ecosystems. To test this, we performed a meta-analysis of population genetic data from a truly marginal marine ecosystem, the Baltic Sea. Our analysis was based on molecular genetic variation, some of which, in particular microsatellite and mitochondrial DNA (mtDNA) loci, are mostly assumed to be neutral while, for example, allozyme and coding nuclear DNA loci might or might not be under selection (directly or through linkage to selected loci). Earlier studies show that general processes, such as drift, shaping genetic diversity and structure affect neutral and quantitative loci in a similar way, suggesting that molecular markers can be used as rough indicators of overall genetic changes (Reed & Frankham 2001; Gilligan et al. 2005).

Baltic Sea

The Baltic Sea is an extreme marine environment. It is geographically peripheral, constituting the innermost part of a large estuary connected to the North Sea by narrow and shallow sills in the Öresund and Danish Belts. It is also ecologically marginal by its harsh physical conditions for both marine and freshwater species with low winter temperatures and permanently low salinities ranging from 1 to 2 psu (practical salinity units) in the innermost parts to 25 psu at the entrance to the North Sea (HELCOM 1996). The most conspicuous result of this is low species-diversity: roughly a tenth or less of the species richness found elsewhere in the NE Atlantic.

The postglacial history of the Baltic Sea started quite rapidly as the freshwater lake that preceded the Baltic Sea opened to the North Sea about 8500 bp. Most of the marine species present in the Baltic Sea today are survivors of a more diverse postglacial flora and fauna established during the Littorina period, 8000–4000 bp, during which the Baltic Sea was more saline than today (Ignatius et al. 1981; Russell 1985). In addition to the postglacial flora and fauna, recent introductions of mostly fresh- and brackish-water species contribute to the present species diversity (Leppäkoski et al. 2002).

The populations living in the Baltic Sea are, with few exceptions, both geographically and ecologically marginal, and some show clear evidence of being genetically deviant from North Sea and Atlantic populations of the same species (e.g. Väinölä & Hvilsom 1991; Nilsson et al. 2001; Luttikhuizen et al. 2003). We here present a comprehensive review and meta-analysis of the genetic status of Baltic Sea populations, comparing them to North Sea/NE Atlantic populations of the same species. More specifically we address the following questions: (i) Are Baltic Sea populations in general genetically less diverse than North Sea/NE Atlantic populations? (ii) Are the Baltic Sea populations genetically more differentiated from nearby North Sea populations than are expected from a simple isolation-by-distance relationship? Finally, we discuss possible processes generating the observed patterns and the implications of the observed patterns for management and conservation.

Materials and methods

We have compiled published data sets on genetic variation and population structure that include populations from both inside and outside the Baltic Sea. In some cases, we re-analysed primary data kindly provided by the authors (see Acknowledgements), but in most cases allele and haplotype frequencies, etc., were taken directly from publications. Data sets that did not include both Baltic and North Sea/Atlantic populations were not used. The ‘entrance of the Baltic Sea’ was defined as the border between the Baltic proper and the Belt Sea and the Öresund, respectively (Fig. 1). North and west of this border, salinity increases rapidly through the Öresund, the Danish Belts and Kattegat.

Figure 1.

The Baltic Sea is connected with the North Sea and the Atlantic Ocean through the Skagerrak, Kattegat, Öresund and Belts seas. Sea surface salinities along this transect are indicated. The line Falsterbo (south Sweden)–Travemunde (Germany) is the ‘zero’ position indicated in Figs 4 and 5, marking the entrance of the Baltic Sea.

We included data representing both supposedly neutral and non-neutral molecular markers; although the majority of data sets were allozymes, microsatellites or mtDNA; single data sets of random amplified polymorphic DNA (RAPD) and haemoglobin (Hb) variation were also included. One study using internal transcribed spacer (ITS) sequence data was omitted due to small and unequal samples sizes (Leskinen et al. 2004). In species where more than one data set were available for the same type of marker, we used the largest data set. If data from different markers were used to estimate genetic diversity, we averaged values over markers. Finally, in comparisons among marker types we used several data sets of the same species if these represented different markers.

Diversity within populations

For allozyme, microsatellite and haemoglobin data we used average expected heterozygosity (HE) rather than observed heterozygosity as a measure of genetic diversity, since this parameter is less affected by departures from Hardy–Weinberg equilibrium and sample size (Nei 1987). For sequence data we used haplotype diversity and for RAPD, simply the number of bands. Expected heterozygosity and haplotype diversity are both equivalent to gene diversity (Nei 1987). Allelic richness was not used because it is very sensitive to sample size, and because very few publications included this parameter. We used published information on genetic diversity per sample, or, if missing, calculated levels of diversity from allele or haplotype frequencies. To compare levels of genetic diversity inside and outside the Baltic Sea, we averaged estimates from the different samples within each area for each species, and then related Baltic levels to Atlantic levels (vBaltic/vAtlantic) to obtain a unidimensional and relative estimate illustrating loss or gain of genetic diversity in Baltic as compared to Atlantic populations of each species. To illustrate the overall pattern across species, we plotted vBaltic against vAtlantic.

The data on genetic diversity was subsequently broken down to test specific hypotheses about the importance of species’ potential for dispersal, systematic grouping, generation time and genetic marker used. When comparing levels of genetic diversity between Baltic Sea and North Sea/Atlantic populations we excluded populations from the Belts, Öresund and the Kattegat (Fig. 1, but see cod Hb data), as in these areas Baltic Sea and North Sea gene pools hybridize or mix which may increase levels of gene diversity.

Differentiation between populations

For each species in which data from at least five populations in total inside and outside the Baltic Sea were available, we produced genetic trajectories from pairwise genetic differences (FST) to illustrate how genetic distance increases at successively greater geographic distances from the innermost Baltic Sea populations. Nielsen et al. (2004) used a similar approach to illustrate genetic clines at the entrance of the Baltic Sea in cod and turbot, but used North Sea samples as reference. We used the border between the Baltic and the Öresund/Belts (Fig. 1) as the zero position along the x-axis indicating geographic position, to assess species’ genetic structures at the entrance to the Baltic Sea. We predicted that the genetic trajectories would conform to either of four different general patterns (Fig. 2): (A) no genetic differentiation, (B) a simple isolation-by-distance cline, (C) genetic perturbations of peripheral populations inside the Baltic Sea, and (D) a sigmoid genetic cline with the steepest part around the entrance to the Baltic Sea, thus correlating with the strong environmental shift present here. Obviously, different patterns of differentiation could combine to more complex structures.

Figure 2.

Hypothetical genetic trajectories expected from various models of genetic structuring of populations along the Baltic Sea–Atlantic Ocean geographic transition. The model trajectories describe genetic differentiation between the innermost Baltic population and populations at increasing geographic distance along the Baltic–Atlantic transect. The broken line marks the Baltic Sea entrance. The first model assumes no genetic differentiation at all (A). The second model (B) describes a trajectory resulting from a linear isolation-by-distance model. In model C end-populations inside the Baltic are strongly deviating from all other populations, while in model D there is a steep clinal shift around the Baltic Sea entrance with populations either side of the entrance being relatively homogeneous.

Taxonomical considerations

Our aim was to study how a peripheral ecosystem affected genetic structures within species, but as we considered the same processes being important within species-complexes in which taxa have a substantial gene exchange, we also included the Mytilus edulis species-complex. M. edulis is considered by some authors as two separate taxa in this area: M. edulis in the North Sea and Mytilus trossulus in the Baltic (McDonald et al. 1991). These taxa hybridize, however, extensively at the Atlantic–Baltic border and there is considerable introgression of edulis-type mtDNA into the Baltic (Riginos & Cunningham 2005). We therefore treated these taxa as comparable to a subdivided species and used the name M. edulis. The polychaete Hediste diversicolor may include two sibling species (Röhner et al. 1997; Sikorski & Bick 2004). For this reason we excluded Hediste from the comparison of genetic differentiation over the Baltic–Atlantic transect, but included it in the comparison of levels of genetic diversity between Baltic and Atlantic populations, since even if these are separate species without substantial gene exchange, they would be so closely related that levels of diversity should be possible to compare in a similar way as if these were part of a subdivided species.

Statistical procedures

Differences in mean diversity between Baltic Sea and Atlantic populations were tested with one-tailed t-tests. The rationale for using one-tailed tests was to test the hypothesis that Baltic populations had lost diversity compared to Atlantic populations. Differences in relative diversity of Baltic populations (vBaltic/vAtlantic) among taxonomic groups, etc., were tested with analysis of variance (anova). Since ratios are usually log-normally distributed, the vBaltic/vAtlantic values were log-transformed prior to statistical treatment (Underwood 1997). Zero values were replaced by 0.01. Deviation from the null hypothesis of a mean ratio of unity between Baltic Sea and Atlantic populations was tested by calculating the 90% confidence interval (t * standard error) around the geometric mean. Confidence intervals not overlapping unity indicate that Baltic Sea populations are less diverse than Atlantic populations. Since the hypothesis of Baltic populations having less variation is one-tailed, the use of a 90% confidence interval (around the mean) encompasses 95% of the distribution space on any (one-tailed) side of the mean. The slope of the genetic trajectories was tested with linear regression. Difference in average slope between species with high and low dispersal potential was tested with a two-tailed t-test. Differences in shape among trajectories were tested using a chi-squared test.


Diversity of Baltic and Atlantic populations

In total, we found 41 data sets describing genetic diversity of populations in 29 different species (Table 1). Species-by-species analysis revealed the majority of the Baltic populations were less genetically diverse than populations of the same species from the Atlantic. Furthermore, different data sets representing the same species were consistent (Table 1), with just one exception, the bivalve Macoma balthica. This species was 30% more diverse inside the Baltic than outside when measured at allozyme loci (Väinölä & Varvio 1989a), while mtDNA haplotype diversity was significantly lower in the inner Baltic (Finland) than in Atlantic populations. However, populations from Poland and Latvia (SE Baltic) were among the most diverse in mtDNA loci in the entire NE Atlantic, owing to a high number of haplotypes found in the SE Baltic (Luttikhuizen et al. 2003). Thus, both markers in fact indicated an unexpectedly high genetic diversity in southern Baltic Sea while showing different patterns in the northern parts. This exceptional pattern might be explained by mixing of two lineages (see Discussion).

Table 1.  Genetic data for species represented by populations from both the Baltic Sea and the North Sea. Dispersal potential was classified considering both larval and adult dispersal. Genetic diversity was tested using expected heterozygosity (for allozyme and microsatellite data) or haplotype diversity (mtDNA). One-tailed t-tests (not corrected for multiple testing) were used to indicate differences in genetic diversity between Baltic and Atlantic populations. Transient populations, samples from Öresund and Kattegat, were not included except for the cod-Hb data (see below). Data sets consisting of less than four populations were not used for tests of difference in genetic diversity (‘No test’). Underline indicates the data set considered most informative for that marker and species, this data set being used in further analyses. ***P < 0.001; **0.001 < P < 0.01; *0.01 < P < 0.05; NS, nonsignificant
SpeciesDispersal potentialGenetic markerNo. of lociSample sizeNo. of populationsBaltic diversity < AtlanticReference
  • Number of bands used as estimator of gene diversity.

  • The SE Baltic populations include haplotypes of the Atlantic lineage and only the Finnish population was considered a true Baltic population.

  • §

    HO was used instead of HE as estimator of gene diversity.

  • Kattegat and Öresund populations were obviously different from Baltic populations and considered representative of non-Baltic populations.

Plants and algae
 Cladophora rupestrisLowAllozymes 1 20–53 5 5*Johansson et al. (2003)
 Ceramium tenuicorneLowRAPD  12–17 3 3NSGabrielsen et al. (2002)
 Fucus serratusLowMicrosat. 7 23–14919 5NSCoyer et al. (2003)
 Fucus vesiculosusLowMicrosat. 5 31–50 8 8***Tatarenkov et al. (pers. communication)
 Zostera marinaLowMicrosat. 3 25–54 8 5***Reusch (2002)
 Zostera marinaLowMicrosat. 4 29–80 8 3***Reusch et al. (2000)
 Zostera marinaLowMicrosat. 8 25–54 8 9**Olsen et al. (2004)
 Cerastoderma glaucumHighAllozymes13100 1 1No testNikula & Väinölä (2003)
 Cerastoderma glaucumHighmtDNA   4–10 4 3*Nikula & Väinölä (2003)
 Mytilus edulisHighAllozymes 3 57–315 2 2**Väinölä & Hvilsom (1991)
 Mytilus edulisHighAllozymes 4 19–112 4 1**Johannesson et al. (1990)
 Macoma balthicaHighAllozymes13 60–90 3 2NSVäinölä & Varvio (1989a)
 Macoma balthicaHighmtDNA   9–2512 1*Luttikhuizen et al. (2003)
 Mya arenariaHighAllozymes 9 48–111 2 2**Lasota et al. (2004)
 Cyprideis torosaLowAllozymes 5 81–148 2 5*Sywula et al. (1995)
 Balanus improvisusHighAllozymes11 16–140 1 2No testFurman (1989)
 Pontoporeia femorataLowAllozymes 6 14–122 1 1No testVäinölä & Varvio (1989b)
 Mysis mixtaLowAllozymes 3 65–133 1 5NSVäinölä (1992)
 Crangon crangonHighAllozymes 3 27–85 7 1NSBulnheim & de Schwenzer (1993)
 Gammarus salinusLowAllozymes 3 30–90 5 2NSBulnheim & Scholl (1981)
 Gammarus zaddachiLowAllozymes 3 25–87 9 3***Bulnheim & Scholl (1981)
Other invertebrates
 Hediste diversicolorLowAllozymes§10 13–48 3 4NSRöhner et al. (1997)
 Echinorhynchus gadiHighAllozymes 6  6–106 2 1No testVäinöläet al. (1994)
 Halicryptus spinulosusLowAllozymes 7  4–60 2 5***Schreiber et al. (1996)
 Flounder (Platichthys flesus)HighAllozymes 8  9–54 5 2NSBorsa et al. (1997)
 Turbot (Scophthalmus maximus)HighMicrosat. 8 75–135 3 4NSNielsen et al. (2004)
 Cod (Gadus morhua)HighAllozymes10 63–100 8 1**Mork et al. (1985)
 Cod (Gadus morhua)HighMicrosat. 9 50–158 4 2***Nielsen et al. (2003)
 Cod (Gadus morhua)HighmtDNA   5–40 9 6**Árnason & Pálsson (1996), Árnason et al. (1998)
 Cod (Gadus morhua)HighHemoglobin 1 40–333 619***Sick (1965)
 Salmon (Salmo salar)LowmtDNA   9–1162021***Nilsson et al. (2001)
 Salmon (Salmo salar)LowmtDNA  32–50 9 5***Verspoor et al. (1999)
 Salmon (Salmo salar)LowAllozymes 8  9–16210 8*Ståhl (1987)
 Herring (Clupea harengus)HighAllozymes17 30–100 9 8NSRyman et al. (1984)
 Herring (Clupea harengus)HighMicrosat. 5 50–59 2 2NSAndréet al. (2001)
 Eel pout (Zoarces viviparus)LowAllozymes 4> 100 4 6*Christiansen & Frydenberg (1974)
 Harbor porpoise (Phocoena phocoena)HighmtDNA  13–27 2 1No testWang & Berggren (1997)
 Harbor porpoise (Phocoena phocoena)HighMicrosat.12 49–169 5 1NSAndersen et al. (2001)
 Grey seal (Halichoerus grypus)HighmtDNA  16–20 1 1No testBoskovic et al. (1996)
 Harbor seal (Phoca vitulina)HighMicrosat.§ 7 30–21610 2*Goodman (1998)
 Ringed seal (Phoca hispida)HighMicrosat. 8 27–70 1 2No testPalo et al. (2001)

A meta-analysis, including all 29 species supported the hypothesis of generally lower genetic diversity in Baltic Sea populations compared to Atlantic populations (Fig. 3). Indeed, this trend was statistically significant as the distribution of Baltic:Atlantic diversity ratios did not overlap with unity (Table 2). The expectation that the degree of diversity loss of Baltic Sea populations would be related to the biology of the species, for example being biased to certain taxonomic groups, or to life history characteristics such as generation time and dispersal capacity was, however, not supported. Although bivalves and plants had diversity ratios significantly below unity, the ratios of these groups were not significantly different from the ratios of the other groups and neither dispersal potential nor generation time affected diversity ratios of species (Table 2). However, diversity loss in Baltic compared to Atlantic populations was significantly more pronounced in mitochondrial loci (mtDNA) compared to nuclear loci (allozyme and microsatellite). In fact, Baltic populations had on average 50% less mitochondrial diversity than Atlantic populations, while only 11–12% less nuclear diversity (Table 2).

Figure 3.

Genetic diversity of Baltic Sea populations compared to Atlantic populations. For species with equal diversity in both seas, points will fall on the line. Points below the line indicate lower diversity in Baltic populations compared to Atlantic populations and vice versa.

Table 2.  Gene diversity ratios (Baltic:Atlantic populations). In all comparisons, we averaged estimates over markers, except when comparing diversity of different markers. For replicate data sets of the same marker we used the most informative (highest number of individuals, populations or loci). See Table 1 for species included in each group. Bold denotes that geometric mean + 90% confidence interval do not span unity; that is, diversity is significantly lower in the Baltic populations (one-tailed test: alpha = 0.05). F and P with associated degrees of freedom (d.f.) indicate the results from anova, testing differences in geometric mean ratio among treatment levels. Arithmetic means are given for comparison
FactorLevelsnGene diversity arithmetic meanBaltic:Atlantic geometric meanCI0.90Mean + CIF, d.f.P
All species290.840.780.100.89
Dispersal abilityGood150.840.800.120.920.0550.82
Poor140.840.770.200.971, 27 
Generation timeLong150.880.860.090.951.520.23
Short140.800.710.200.911, 27 
Taxonomic groupPlant 50.770.760.130.890.950.47
Mammal 40.990.950.441.385, 23 
Fish 60.860.840.171.01  
Bivalve 40.610.610.100.71  
Crustacean 71.000.880.521.40  
Other invertebrates 30.660.600.871.47  
Genetic markerAllozyme170.890.820.160.976.320.005
Microsatellite100.880.870.070.942, 32 
mtDNA 60.500.350.490.83  

Genetic differentiation

Pairwise genetic distances (FST) increased significantly with geographic distances, from the innermost Baltic Sea population and outwards, in 21 of 28 data sets, representing 20 species (Table 3). Too few samples made a statistical test inappropriate in the cod parasite Echinorhyncus gadi and the bivalve Mya arenaria, while Fucus vesiculosus, Gammarus salinus, cod (allozyme marker), herring (microsatellite marker) and Balanus improvisus, revealed nonsignificant slopes. Of these, B. improvisus did not show much of differentiation at all over 1820 km, but F. vesiculosus, G. salinus, cod and herring did indeed show divergence between populations, but a combination of poor representation in some parts of the cline and a complicated pattern of differentiation not consistent with a simple linear isolation-by-distance model resulted in statistical insignificances in the present comparison (but see also Fig. 4). Thus, essentially only one species, B. improvisus, seemed to be panmictic over the Baltic Sea–North Sea salinity gradient, while all other species showed varying degrees of genetic substructure. Although we had expected that species with lower dispersal potential were more strongly genetically differentiated over increasingly larger geographic distances, we found no support for this expectation; mean slope of linear regressions did not differ between species with low and high dispersal potential (P = 0.49; Table 3).

Table 3.  Isolation-by-distance relationships (pairwise FST regressed on geographic distances). Data are sorted by decreasing slope of linear regression. ***P < 0.001, **0.001 < P < 0.01, *0.01 < P < 0.05; NS, means nonsignificant; no test was performed when n < 5. There was no difference in mean slope between species with low and high dispersal potential (P = 0.49)
SpeciesDispersal potentialMax distance (km)R2Slope*1000P of slope
Cod (Hb)High13260.470.311***
Ceramium tenuicorneLow14000.930.289***
Eel poutLow18100.690.216***
Mytilus edulisHigh17170.580.208*
Macoma balthica (allozyme)High35120.800.191***
Cladophora rupestrisLow18000.730.181**
Cyprideis torosaLow22000.980.157***
Gammarus zaddachiLow24220.630.144***
Fucus serratusLow22600.520.137**
Salmon (mtDNA)Low54750.440.108***
Zostera marinaLow30000.720.097***
Macoma balthica (mtDNA)High33150.700.081**
Harbor sealHigh20440.490.081*
Cod (mtDNA)High29000.960.062***
Cod (microsatellite)High10800.860.049***
Fucus vesiculosusLow12380.050.047NS
Echinorhynchus gadiHigh26000.870.028No test
Halicryptus spinulosusLow46000.890.027***
Salmon (allozyme)Low30250.530.016***
Gammarus salinusLow34000.050.007NS
Cod (allozyme)High28850.480.002NS
Herring (microsatellite)High24750.750.002NS
Mya arenariaHigh16600.860.002No test
Mysis mixtaLow49000.920.002***
Balanus improvisusHigh18200.100.001NS
Herring (allozyme)High32000.250.001*
Figure 4.

Genetic trajectories of 20 species based on pairwise FST estimates between the innermost Baltic population and populations at increasing geographic distances from this population. X-axis indicates geographic position in relation to the Baltic Sea entrance with negative values for populations inside the Baltic and positive values outside. The inserted figure (lower right corner) indicates the observed number of trajectories similar to the four hypothetical ones presented in Fig. 2. This observed distribution of trajectories is unlikely to occur by chance (P = 0.001).

The genetic trajectories provide a more detailed picture of the genetic differentiation over increasingly larger geographic distances, starting from the innermost Baltic populations, passing the entrance of the Baltic and proceeding into the Kattegat–Skagerrak–North Sea area (Fig. 4). Sigmoid trajectories were the most frequent pattern found (11 of 20 species), followed by linear isolation by distance and a few examples of end-population perturbations (Fig. 4). Gammarus zaddachi, Mytilus edulis, Macoma balthica, turbot, eel pout, salmon, cod, herring, and probably Cladophora rupestris and Zostera marina, showed sigmoid trajectories, while Fucus vesiculosus and Fucus serratus, in particular, revealed end-population perturbations. This observed distribution of trajectories deviated significantly from a random distribution (P = 0.001).

The fit of individual trajectories into one of the four hypothetical models is admittedly a simplification, and it seems, for example, quite possible that processes resulting in two or more of the model effects could combine in nature. However, we feel confident with using this approach when looking for general patterns, while interpretation of individual cases should be done with caution. Thus, the overall conclusion is that a major part of the species reveals a genetic shift between the Baltic and the Atlantic populations, often in the vicinity of the Baltic Sea entrance (Fig. 4)

Comparing trajectories for different genetic markers

Trajectory data from different genetic markers were available in four species: Macoma balthica, herring, salmon and cod (Fig. 5). In cod, microsatellite, haemoglobin and allozyme data showed congruent positions of genetic clines with the steepest parts located very close to the Baltic Sea entrance. Likewise, the steepest parts of the clines coincided in microsatellite and allozyme trajectories of herring. However, in all three cases where mtDNA trajectories were compared with those of other markers, the position of the mtDNA cline was different from the other trajectories (Fig. 5). While the mtDNA clines of both salmon and cod appeared at the North Sea side of the other clines, the Macoma mtDNA cline was clearly on the Baltic side of the allozyme cline, indicating heavy introgression of mtDNA North Sea genotypes into the Baltic Sea in this species.

Figure 5.

A comparison of genetic trajectories of different molecular markers analysed within the same species. To reduce sampling noise, moving averages (N = 1–5) have been used to smooth the lines.


Two general conclusions emerge from our comparison of Baltic–Atlantic population genetic structures over a wide range of taxa: (i) Baltic populations have generally lower genetic diversity than Atlantic populations, and (ii) a majority of taxa studied show strong differentiation over the Baltic–North Sea border. Moreover, the low diversity of Baltic populations is unrelated to taxonomic group, generation time and capacity of dispersal, while strongly related to type of genetic marker. Although extensive work would be needed to identify the specific mechanisms generating the observed patterns in each species, we discuss below some likely processes. We also review studies of some species for which more comprehensive work has been undertaken to explain Baltic–Atlantic population genetic structures.

Loss of diversity through genetic drift

Baltic Sea populations had in general less genetic diversity than North Sea/Atlantic populations. Assuming that the North Sea–Atlantic populations owing to the relative stability gained by large sizes have not increased their genetic diversity over the 8000 years since the opening of the Baltic Sea, the observed lower diversity of Baltic Sea populations must be explained by a loss of genetic variation in Baltic populations, before, during, or after the colonization. If lost before the Baltic Sea was invaded, the populations must have been separated from the Atlantic populations already at an earlier stage, which indeed is a likely scenario for some of the species (see below).

For populations that lost genetic variation during or after their establishment in the Baltic Sea, two observations suggest that genetic drift played a primary role. Primarily, populations have lost substantially more genetic diversity at mitochondrial genes compared to nuclear genes. As the effective population size of uniparentally inherited loci is only one-fourth that of nuclear loci (Ne), random loss of diversity should be correspondingly higher (4×) in mtDNA genes than in, for example, microsatellite and allozyme genes (Avise 1994). Indeed, we found the average loss of diversity at nuclear loci to be 11% (allozymes) and 12% (microsatellites) while at mitochondrial loci 50% was lost, which is almost exactly the relationship predicted by theory. Second, loss of diversity was at similar levels in coding (allozyme) and noncoding (microsatellite) loci. Variation in the coding allozyme loci would potentially be affected by natural selection, while the noncoding microsatellite loci would presumably not be affected. Thus, no difference in loss of diversity between these markers suggests that selection is not an important mechanism explaining loss of molecular genetic variation in Baltic Sea populations.

As drift eliminates variation per generation, populations with short generation times are expected to lose more diversity than long-lived organisms per unit time (Fig. 6). We did not find support for such a pattern, however, possibly because species with short generation time also tend to have larger effective population sizes, Ne (e.g. invertebrates and annual algae).

Figure 6.

Expected losses of gene diversity (HE) in populations isolated over 8000 years, and unaffected by mutation and migration. Losses of mtDNA and nuclear DNA diversity are calculated for various magnitudes of effective population sizes (Ne), for species with short generation time (1 year, nonoverlapping generations) and long generation time (5 years, nonoverlapping generations). Observed levels of gene diversity loss in Baltic populations (50% in mtDNA and 10% in nuclear DNA; Table 2) are indicated to show mean effective population sizes that will generate losses of these magnitudes.

Another prediction is that species with high dispersal capacity should lose less diversity owing to larger effective population sizes, but again we found no support for this expectation (Table 2). Perhaps this is explained by subdivision of populations of poor dispersing species into semi-isolated subpopulations may in fact increase Ne, and thereby promote maintenance of genetic variation (Nunney & Campbell 1993). Moreover, we used type of reproductive strategy as the criterion to characterize species as having poor or good dispersal, but strategy (planktonic vs. nonplanktonic) might not be a good enough estimator of gene flow (Knutsen et al. 2003; Colson & Hughes 2004). Indeed, species with a low dispersal potential did not show stronger isolation-by-distance effects than species with high dispersal capacity (Table 3).

With two exceptions (the barnacle Balanus improvisus and the bivalve Mya arenaria), all species included in the analyses have an early history of colonization in the Baltic, most probably dating back to the Littorina period (8000–4000 bp). Baltic populations may have lost genetic diversity compared to Atlantic populations due to varying degrees of isolation over 8000 years and relatively small population sizes, and occasional population bottlenecks. We calculated stochastic losses of haplotype (mtDNA) and allele (nuclear DNA) diversity in populations of various sizes assuming complete isolation and no new mutations during 8000 years (Fig. 6). The rate of loss under these circumstances is proportional to the effective population size (Ne) and the number of generations elapsed (T):

Ht = H0 (1 − 1/2Ne)T

Our observations showed that on average, Baltic populations had lost about 50% of Atlantic diversity in mtDNA loci (Table 2). Bringing this value into the theoretical relationship (Fig. 6) this corresponds to a mean effective size (Ne) of around 11 000 in a species with a short life cycle (1 year, nonoverlapping generations), and 5000 in a species with a long life cycle (5 years, nonoverlapping generations). For nuclear loci, the average loss was 11–12% (Table 2) corresponding to a mean effective size of 13 000 in ‘annual’ and 7000 in ‘multiyear’ species (Fig. 6). Effective population sizes (Ne) are probably magnitudes smaller than census sizes (N) of adult populations (Turner et al. 2002), and thus our calculations fits well with a scenario of genetic drift as a contributor to diversity loss in Baltic Sea populations that have been at least partly isolated since established. Occasional bottlenecks might have contributed to further reduction of the genetic variation (Chakraborty & Nei 1977), while intermittent periods of more effective genetic exchange with Atlantic populations would have opposed this by re-introducing Atlantic genes.

Data for Baltic Sea harbour seals (Phoca vitulina) support a scenario of diversity loss due to genetic drift and repeated bottlenecks (Härkönen et al. 2005). Baltic population sizes have been estimated from archaeological and hunting data since the seals were established in the Baltic Sea 8000 years ago. Moreover, historic data show that harbour seals were absent in the Kattegat and Skagerrak during most of the period 8000–200 bp, and thus the Baltic Sea population has been effectively isolated from Atlantic populations. Effective population size was inferred from behavioural studies (Härkönen et al. 2005), and estimated population sizes (including a recent population bottleneck) explain well the 20% loss of genetic diversity observed in microsatellite loci of the Baltic Sea population (Goodman 1998; Härkönen et al. 2005).

An example of a recently introduced species, the small planktonic crustacean Cercopagis pengoi, show that founding a new Baltic Sea population might involve severe population bottlenecks that substantially reduce genetic diversity. This species was unintentionally introduced to the Baltic Sea in 1992 from the Black Sea/Caspian Sea. Of seven mtDNA haplotypes present in the ancestral populations, only one is found in the Baltic Sea population (Cristescu et al. 2001).

Overall, genetic drift clearly seems the most important process eliminating genetic diversity in isolated and size-restricted Baltic Sea populations. Despite this, life-history characteristics, such as dispersal potential, generation time and taxonomic group seems marginally, if at all, important.

Loss of diversity through asexual reproduction

Increased frequency of asexual reproduction further reduces genetic variation. Noticeably, among facultative asexual species, populations living in marginal environments tend to reproduce asexually to a higher degree than central populations (Eckert 2002; Billingham et al. 2003; Kearney 2003) In the Baltic Sea, a number of plant and algal species reproduce asexually more frequently than elsewhere (Hällefors et al. 1981; Malm et al. 2001). Serrão et al. (1996, 1999) argue convincingly that it is the harsh brackish-water conditions that restrain sexual reproduction. Several species included in the present study show evidence of increased clonality inside the Baltic Sea. Zostera marina populations close to Åland (Fig. 1), at the border of the species’ distribution, for example, consist of 95% clonal plants, as compared to near zero clonality in the North Sea (Reusch et al. 2000). In Fucus vesiculosus, the northernmost population (Öregrund, close to Åland) has 30% cloned individuals: a unique finding since nowhere outside the Baltic Sea clonal reproduction of new attached plants has been found in this species (Tatarenkov et al. 2005). Also, the new species Fucus radicans so far only found in the northernmost part of the Baltic Sea (at salinities < 6) has populations with 80% asexually reproducing plants (Bergström et al. 2005; Tatarenkov et al. 2005). Baltic populations of the red alga Ceramium tenuicorne are dominated by individuals that reproduce vegetatively, while in the Skagerrak reproduction is to a large extent (> 95%) sexual (Gabrielsen et al. 2002; Bergström et al. 2003). Both Z. marina and F. vesiculosus show significantly less genetic diversity in Baltic populations compared to North Sea populations (Table 2), and in C. tenuicorne there is a tendency for loss of genetic diversity as estimated from the number of RAPD bands (P = 0.11).

Processes generating Atlantic–Baltic Sea differentiation

About half of the 20 species analysed showed a steepened genetic cline at or somewhat outside the Baltic Sea entrance. Such clines may form either as a consequence of gradient selection followed by a successively impeded gene flow (primary cline), or where genetically separated lineages of a species meet and hybridize (secondary clines) (Coyne & Orr 2004). Primary clines mainly evolve as a consequence of divergent natural selection acting on specific traits in a continuously distributed population (Endler 1977). It is expected that neutral loci will not, in the first place, be affected by divergent selection, but linkage to selected loci or as a consequence of impeded gene flow owing to partial reproductive barriers generated by selection on ecological traits, will result in differentiation also in neutral markers in primary clines (Johannesson 2001; Rolán-Alvarez et al. 2004). Primary clines are likely to be connected to strong environmental gradients (ecotones), being the ultimately cause of the cline (Endler 1977). Secondary clines form when and where evolutionary lineages that have evolved in allopatry come into secondary contact and hybridize. Secondary zones are initially characterized by congruent clines of selected and neutral markers, but neutral clines are expected to fade out over time by gene flow (Futuyma 2005). Both primary and secondary clines might be expected at the entrance of a marginal environment, which is ecologically aberrant and partly isolated.

Primary zones of hybridization

The steep clines in allozyme, haemoglobin and microsatellite loci in cod are likely examples of primary genetic clines over the Baltic-Atlantic transition. These three clines coincide at the Baltic entrance while the mtDNA cline seems less sharp and somewhat shifted towards the North Sea (Fig. 5). There are indications of selection acting directly on part of the allozyme variation, and on the haemoglobin variation (Sick 1965; Mork & Sundnes 1985). In addition, an efficient barrier to gene flow has evolved as a consequence of divergent selection on reproductive traits, such as egg buoyancy, sperm motility (Nissling & Westin 1997) and spawning season; the NE Atlantic cod spawn during winter while the Baltic cod spawn during summer (Wieland et al. 2000). These ecological differences have probably evolved as a consequence of adaptation to different environments, and are likely to act as barriers to gene flow, resulting in accumulated differences also in neutral markers such as microsatellite and mtDNA (Nielsen et al. 2003; Árnason 2004; see Fig. 5).

Herring also reveals a pattern of similar allozyme and microsatellite clines at the Baltic entrance (Fig. 5). As for cod, both congruent clines and the relatively low magnitude of differentiation suggest a primary hybrid zone being formed without a preceding stage of lineage separation.

Secondary zones of hybridization

The Baltic–Atlantic genetic shift of the bivalve species Macoma balthica is a clear-cut example of a secondary cline. The genetic differences between the two seas are very strong in both allozyme and mtDNA loci, and in between are sharp genetic clines (Fig. 5). Earlier studies indicate the presence of two evolutionary lineages, one present in the NE Atlantic and one in the Baltic Sea. The lineages were separated at least 10 million years ago, and have come into secondary contact after extensive isolation (Väinölä & Varvio 1989a; Luttikhuizen et al. 2003; Väinölä 2003). In fact, the central and northern Baltic Macoma balthica populations are more closely related to populations in the White Sea and Alaska than to North Sea populations in both allozyme and mtDNA loci. It is suggested that repeated invasions of this species from the Pacific to the Atlantic, gave rise to two separate lineages (Luttikhuizen et al. 2003; Väinölä 2003). These lineages are still interfertile and hybridize upon secondary contact. For some reasons, the Baltic and White Seas are the only areas in which the most recent invading lineage is found at present.

Similar patterns are seen in the blue mussel, Mytilus edulis, with steep clines in several allozyme loci around the entrance of the Baltic Sea (Väinölä & Hvilsom 1991). Here, again two distinct evolutionary lineages, M. edulis and Mytilus trossulus (Varvio et al. 1988), that separated 3.5 million years ago or earlier (Riginos & Cunningham 2005), are in secondary contact. The two lineages have introgressed heavily, as particularly evident in the structure of mtDNA variation with no uncontaminated individuals of trossulus genotype found even deep inside the Baltic Sea (reviewed by Riginos & Cunningham 2005). Also, nuclear DNA loci show evidence of substantial introgression of edulis alleles into the Baltic Sea (Riginos et al. 2002; Riginos & Cunningham 2005). On the other hand, allozymes of trossulus-type dominate completely inside and edulis-type outside the Baltic Sea. The explanation for the maintenance of these differences is likely differential selection acting directly on the allozyme loci or on loci strongly linked to these (Theisen 1978; Johannesson et al. 1990; Riginos & Cunningham 2005).

It is noteworthy that for both bivalve species, the most recently invading lineage from the Pacific occupies the innermost part of the North Sea–Baltic Sea region. A likely explanation is that these lineages had a much wider northern Atlantic distribution following the last glacial retreat but were outcompeted by older Atlantic lineages that expanded northwards following temperature increase after the last glacial period (Väinölä 2003). However, the Baltic became a blind alley on the route north and a particularly harsh environment for the southern lineages, while the northern lineages here found a refuge, probably by being better adapted to the extreme Baltic environment. Contrastingly, in Nova Scotia the situation is reversed with M. trossulus occupying the more saline open-coast areas and M. edulis the low-saline estuarine areas (Riginos & Cunningham 2005). This suggests that adaptation of M. trossulus to the low-saline Baltic environment is a secondary effect.

A similar scenario of repeated trans-Arctic invasions has been proposed for the red alga Phycodrys rubens (van Oppen et al. 1995), although a more extensive analysis is needed to reconstruct the phylogeography of the two European clades of this species. In contrast, very little DNA sequence differences has been found between Atlantic and Pacific individuals of the seagrass, Zostera marina, suggesting that Atlantic populations might be the result of a recent, or even currently open, trans-Arctic connection to the Pacific Sea (Olsen et al. 2004).

Genetic diversity of marginal marine ecosystems

To our knowledge, no similar study has previously reviewed genetic structures of multiple species inhabiting the same marginal ecosystem. However, there are examples of clinal genetic structures of marine species living in marginal marine environments, typically estuaries, characterized by strong salinity gradients. Most of these examples are bivalve species from the Mytilus group such as M. edulis/M. trossulus, but also M. galloprovincialis and Perna canaliculus show clinal patterns in estuarine habitats in Norway (Ridgway & Nævdal 2004), New Zealand (Gardner & Kathiravetpillai 1997; Gardner & Palmer 1998) and the US East Coast (Koehn et al. 1980). Moreover, estuarine sites in Holland reveal genetic alterations in species such as Macoma balthica (Luttikhuizen et al. 2003) and Littorina littorea (de Wolf et al. 2003). Notably, all these species have broadcasting larvae able to travel hundreds of kilometres between place of birth and place of settlement, and despite this, there are pronounced local-scale genetic structures. These observations corroborate our results from the Baltic Sea and suggest that species characteristics, such as dispersal potential and generation time, have no big impact on the genetic structuring of populations in a marginal habitat. Instead, physical properties of the environment, such as geographic or ecological periphery, is perhaps more important in promoting isolation and genetic perturbation.

Managing Baltic Sea biodiversity

Until recently, species diversity has been the main focus for the conservation of Baltic Sea biodiversity. Laikre et al. (2005), however, reviewed genetic diversity of Baltic Sea fish populations and stressed the importance of including information on genetic variation in stock management. Our review clearly supports this notion and pinpoints the necessity of a wide geographic reference (e.g. Baltic Sea–North Sea). Comparative studies can teach us much about evolutionary processes and phylogeographic relationships by combining genetic data with distributional data (e.g. Bernatchez & Wilson 1998). Moreover, observations of lost genetic diversity and genetic breaks across multiple species will have strong implications for ecosystem management.

The Baltic Sea ecosystem is fragile with much reduced species-richness and thus a low resistance to local extinctions. Here, the genetic dimension will be of particular concern, first because it has implications for ecosystem function (Hughes & Stachowicz 2004; Reusch et al. 2005), and second because genetic isolation of several populations implies restricted exchange of individuals (in this case between North Sea and Baltic Sea populations). Thus, in the event of local extinctions, we cannot expect a rapid recolonization of Baltic Sea areas. Furthermore, owing to different genetic histories, there is a risk that North Sea individuals, if introduced to the Baltic, would be genetically suboptimal and contribute maladaptive genes to surviving fragments of marginal Baltic populations (outbreeding depression, see, e.g. McKay et al. 2005). If, for example, individuals of North Sea cod were introduced into a diminishing Baltic stock, there is a risk that Baltic genotypes giving rise to locally adapted traits, such as summer instead of winter spawning (Wieland et al. 2000) and eggs with neutral buoyancy at lower salinities than elsewhere (Nissling & Westin 1997) would be broken down and important traits eroded.

The best strategy to promote long-term survival of Baltic Sea populations is therefore to protect and sustain local populations. Considering the relationship between census and effective population sizes (Turner et al. 2002), this means that actual population sizes must be orders of magnitude larger than 10 000–50 000 for most populations. It should also be noted that for some species there is not only one but several genetically distinct populations inside the Baltic Sea. This is certainly the case for salmon (Nilsson et al. 2001), herring (Jørgensen et al. 2005), Fucus vesiculosus (Tatarenkov et al. personal communication) and Zostera marina (Olsen et al. 2004), and probably for a range of other species as well. If needed, populations must be restored by individuals from as close as possible populations, and individuals from outside the Baltic Sea avoided.

For some species, the Baltic Sea contains unique lineages, separate from those currently dominating in the Atlantic. For example, the two bivalve species Mytilus edulis and Macoma balthica have established unique evolutionary lineages in the Baltic that are rarely found elsewhere in the NE Atlantic. Indeed, the closest relatives to these Baltic Sea populations are instead found in NE Canada and Alaska. Additionally, a cryptic species of brown alga, Fucus radicans, has recently been discovered in the northern Baltic Sea (Åland and northward), where it is the dominating macro-alga (Bergström et al. 2005; Tatarenkov et al. 2005). This species has hitherto not been found outside the Baltic, and it remains to be established if it is endemic to the Baltic Sea or a remnant Arctic species.

Evidently, a marginal ecosystem such as the Baltic Sea, might thus be of great concern in conservation and management, not only because of populations here being more vulnerable, but also because such an ecosystem might both produce and protect more or less extreme evolutionary lineages. Consequently, far greater focus than today has to be put on the management of intraspecific diversity of species living in marginal ecosystems, such as the Baltic Sea.


This work benefits from several colleagues making raw data and unpublished, or in press, manuscripts available to us. We thus thank Lena Bergström, Jim Coyer, Tove Gabrielsen, Rita Jönsson, Lena Kautsky, Björn Källström, Jeanine Olsen, Cynthia Riginos, Andrey Tatarenkov, Risto Väinölä and Thierry Wirth. In addition, we got very helpful comments on draft manuscripts from Risto Väinölä, Nils Ryman, Jon Havenhand and three anonymous referees. This work is part of the research project EUMAR, supported by the European Commission (grant EVK3-CT2001-00048, Additional funding was received from the Swedish Science Research Council, the Swedish Environmental Protection Agency through the research programme MARBIPP, and from the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning (FORMAS).

Kerstin Johannesson is an evolutionary biologist with a general interest in the evolution of marine species, in particular the evolution of intraspecific diversity. Carl André is a molecular ecologist interested in population biology issues of marine species such as biocomplexity, connectivity and metapopulation structure.