Fig. 1 shows male body size for both P. chalceus ecotypes and for the P. littoralis populations in the Guérande region [see also additional file 1]. Mean body size for the pond populations is small (3.56; range: 2.9–4.2), intermediate for the canal populations (4.08; 3.4–4.6) and high for the P. littoralis populations (4.69; range: 3.9–5.1). Body size values for the females show a similar pattern but are always larger than male body sizes. Mean female body size is 3.91 for the canal populations (range: 3.4–4.3), 4.52 for the pond populations (range: 3.4–5.0) and 4.89 for the P. littoralis populations (range: 4.3–5.3).

Fig. 1 also shows male and female body sizes for the P. chalceus ecological groups and for the P. littoralis populations on a European scale [see also additional file 2]. Mean male body size is small for the stable (3.68; range: 3.2–4.3) and intermediate P. chalceus populations (3.68; range 3–4.3), somewhat higher for the temporary populations (mean: 3.92; range: 3.4–4.6) and high for the P. littoralis (mean: 4.32; range: 3.6–4.8) ones. Body size values of females show again a similar pattern and are always larger than male body sizes. Mean female body size is 4.04 for the stable P. chalceus populations (range: 3.2 to 4.6) compared to 4.11 for the intermediate populations (range: 3.1 to 4.7) and 4.3 for the temporary P. chalceus populations (range: 3.4 to 4.8) and 4.6 for the P. littoralis populations (4.1–5.2).

In the Guérande region and considering the two species (nested design ANOVA; six P. chalceus populations (canals and ponds pooled) versus three P. littoralis populations, the major part of variance (based on body size) is found among species (Table 1; 74.24% for males and 51.96% for females). If we consider three groups (three canal populations (P. chalceus), three pond populations (P. chalceus) and three P. littoralis populations, the major part of variance is even more pronouncedly found among groups (84.96% for males and 72.08% for females). Variance among populations within groups considering three groups instead of two drops from 17.5 to 2.35% for males and from 29.57 to 4.28% for females. This indicates that this variance was almost completely due to the differences in body size between populations of P. chalceus from different microhabitats. All variance components are statistically significant.

On a European scale and considering the two species (25 P. chalceus populations versus six P. littoralis), the major part of variance (based on body size) is found among species (Table 1; 68.37% for males and 48.37% for females). If we consider four ecological groups (14 temporary (P. chalceus), five intermediate (P. chalceus), six stable (P. chalceus) and five P. littoralis), the variance among groups drops (49.39% for males and 33.87% for females) and the variance within populations augments (45.62% for males and 58.53% for females). Variance among populations within groups considering four groups instead of two drops a little from 10.13 to 5% for males and from 12.29 to 7.61% for females. All variance components are statistically significant.

Fig. 2 shows male and female relative wing sizes for both P. chalceus microhabitats and for the P. littoralis populations in the Guérande region [see also additional file 1]. Mean male relative wing size for the canal populations is small (28.19; range: 20–35%), intermediate for the ponds (mean: 64.24; range: 25–82.5%) and high for the P. littoralis populations (mean: 103.59; range: 92.5–112.5%). Relative wing size values of females show a similar pattern and are not larger than male relative wing sizes. Mean female relative wing size for the canals is 26.93 (range: 17.5–32.5%), 62.31 for the ponds (range: 25–80%) and 103.04 for the P. littoralis populations (range: 92.5–110%).

Fig. 2 also shows male and female relative wing size in ecological groups of P. chalceus and of P. littoralis populations on a European scale [see also additional file 2]. Mean male relative wing size is small for the populations of the old, highly stable salt marsh areas (35.28; range: 22.5–62.5%), some higher for the populations of the salt marshes of intermediate stability (mean: 51.07; range: 25–85%), higher for the populations of the small, unstable areas (mean: 82.23, range: 27.5–105%) and very high for the P. littoralis populations (mean: 106.16; range: 90–112.5%). Relative wing size values of females show a similar pattern and are not larger or smaller than male relative wing sizes. Mean female relative wing size for the stable populations is 33.43 (range: 20–82.5%) compared to 49.91 for the populations of intermediate stability situations (range: 25–85%), 80.41 for the temporary populations of the highly unstable salt marshes (range: 40–97.5%) and 107 for the P. littoralis populations (range: 92.5–115%).

In the Guérande region and considering the two species (nested design ANOVA; six P. chalceus populations (canals and ponds pooled) and three P. littoralis populations), the major part of variance (based on relative wing size) is found among species (Table 2; 78.84% for males and 80.37% for females). If we consider three groups (three canal populations (P. chalceus), three pond populations (P. chalceus) and three P. littoralis populations), the major part of variance is even more clearly found among groups (95.48% for males and 93.92% for females). Variance among populations within groups considering three groups instead of two drops from 18.64 to 0.64% for males and from 16.21 to 0.8% for females. This indicates that this variance is almost completely due to the differences in relative wing size between populations of P. chalceus from different microhabitats. All variance components are statistically significant.

On a European scale and considering the two species (25 P. chalceus populations versus six P. littoralis populations), the major part of variance (based on relative wing size) is found among species (Table 2; 57.61% for males and 60.16% for females). If we consider four ecological groups (14 temporary (P. chalceus), five intermediate (P. chalceus), six stable (P. chalceus) and six P. littoralis populations), the major part of variance is again even more pronounced among groups (82.22% for males and 82.58% for females). Variance among populations within groups considering four groups instead of two drops from 36.11 to 6.34% for males and from 32.8 to 6.89% for females. This indicates that this variance is again almost completely due to the differences in relative wing size between populations of P. chalceus from different ecological or salt marsh area stability groups. All variance components are statistically significant.

In Guérande, both Idh1-2 and Idh1-4 alleles are frequent in ponds, whereas canals are nearly fixed at Idh1-4 (Fig. 3A) [see also additional file 1]. P. littoralis populations are fixed at the Idh1-6 allele. Allele Idh1-1, Idh1-3, and Idh1-5 are very rare in P. chalceus and therefore not shown in Figure 4. Considering the two species (AMOVA; six P. chalceus populations (canals and ponds pooled) and three P. littoralis), the major part of variance (based on IDH1) is found among groups (Table 3; 61.93%). If we consider three groups (three canal populations (P. chalceus), three pond populations (P. chalceus) and three P. littoralis populations, the major part of variance is still found among groups (64.25%). Variance among populations within groups considering three groups instead of two drops from 11.49 to 0.1%. This indicates that this variance is almost completely due to differences in IDH1 between populations of P. chalceus from different microhabitats. All variance components are statistically significant.

At a European scale, both Idh1-2 and Idh1-4 alleles are frequent in the intermediate stability populations, whereas the temporary populations are nearly fixed at the Idh1-2 allele and the stable populations at the Idh1-4 allele (Fig. 3B) [see also additional file 2]. P. littoralis populations are fixed at the Idh1-6 allele. Considering two groups (25 P. chalceus populations versus six P. littoralis populations, the major part of variance (based on IDH1) is found among groups (Table 3; 62.29%). If we consider four ecological groups (14 temporary (P. chalceus), five intermediately stable (P. chalceus), six highly stable (P. chalceus) and six P. littoralis populations, the major part of variance is somewhat lower but is still found among groups (53.26%). Variance among populations within groups, considering four groups instead of two, drops from 12.26 to 2.07%. This indicates that this variance is almost completely due to differences in IDH1 between populations of P. chalceus from different ecological groups, as P. littoralis is fixed in a different allele. All variance components are statistically significant.

The number of studied individuals and allozyme allele frequencies for each population in the Guérande region is given in additional file 3. In the Guérande and considering two groups (AMOVA; six P. chalceus populations (canals and ponds) versus three P. littoralis populations), the major part of variance (based on four neutral allozymes) is found between species (Table 3; 58.43%). If we consider three groups (three canal populations (P. chalceus), three pond populations (P. chalceus) and three P. littoralis populations, the variance among groups drops to 41.99% and the major part of variance is now found within populations (53.02%). Variance among populations within groups considering three groups instead of two remains in the same range (3.39% for two groups compared to 4.98% for three groups).

The number of studied individuals and allozyme allele frequencies for each population at a European scale is given in additional file 4. At a European scale and considering two groups (AMOVA; 25 P. chalceus populations and six P. littoralis populations), the major part of variance (based on four neutral allozymes) is found within populations (Table 3; 49.81%) and among groups (36.49%). If we consider four ecological groups (14 temporary (P. chalceus), five intermediately stable (P. chalceus), six highly stable (P. chalceus) and six P. littoralis populations, the variance among groups drops to 19.05% and the major part of variance is still found within populations (65.24%). Variance among populations within groups considering four groups instead of two remains in the same range (13.70% for two groups compared to 15.71% for four groups). All variance components are statistically significant.

The 459-bp COI mitochondrial sequences revealed two haplotypes in the Guérande region. Haplotype one was shared by individuals of both canal and pond ecotype. Haplotype two was exclusive to the canal ecotype (Table 4). The 497-bp 16S sequences revealed only one haplotype in the Guérande region (Table 5).

The 459-bp COI sequences included 32 variable sites on a European scale (29 parsimony informative) and revealed nine unique haplotypes (four for P. chalceus and five for P. littoralis, with no haplotype shared by the two species). Most haplotypes were exclusive to a particular sampling site, with the exception of haplotype one and three which appeared in eight different localities, and haplotype five, which was found in three localities (Table 4). Selective neutrality was confirmed for this gene (P > 0.1 for all test statistics with D* and F*).

The neighbour joining tree in Figure 5 shows that both species form clearly separated entities (high bootstrap values). Differences are found in 28 positions between haplotypes 5,6,8,9 (P. littoralis) and haplotypes 1 and 2 (P. chalceus; Table 4). 27 base differences are found between haplotypes 5,6,8,9 and haplotypes 3 and 4 (P. chalceus). 27 base differences are found between haplotype 7 (P. littoralis) and haplotypes 1,2 (P. chalceus) and 26 base differences with haplotypes 3,4 (P. chalceus). Intrapopulation differences in both P. chalceus and P. littoralis are very small (only a limited number of individuals studied) and between each haplotype there are only one to at most two bases different.

The 497-bp 16S sequences included 11 variable sites (9 parsimony informative) and revealed three unique haplotypes (Table 5; two for P. chalceus and one for P. littoralis, with no haplotype shared by the two species). Only one haplotype was exclusive to a particular sampling site (haplotype two). Haplotype one appeared in all 11 P. chalceus localities, and haplotype three appeared in all three P. littoralis sites. Selective neutrality was confirmed for this gene (P > 0.1 for all test statistics with D* and F*). Both species form, as in the case for COI, clearly separated entities (Table 5). 10 base differences are found between haplotype 3 (P. littoralis) and haplotypes 1 and 2 (P. chalceus). Interpopulation differences in both P. chalceus are very small (between the two haplotypes there is only one base different). There were no interpopulation differences found in P. littoralis.

P. chalceus populations from three different sites in the Guérande region are analysed (microscale; Fig. 5; see also 2). Each site consists of two drastically differing microhabitats, situated only 10–20 metres from each other and separated by one or two dikes. We compare P. chalceus populations from three canals (CANAL1; CANAL2; CANAL3: Fig. 5) to three adjacent pond populations (POND1; POND2, POND3; see also 2). Furthermore, we analyse Guérande P. littoralis populations from three different sites (GUE1, GUE2, GUE3; Fig. 5; see also 12) nearby the aforementioned P. chalceus population couples.

The two related species are also studied on a macroscale with completely independent population samples (Guérande populations not included). Data on P. chalceus populations from the Netherlands (FRI), Belgium (BRA, WAT, MOK, ZWC, HEI, LIS, OOS, NIE, MOE), England (SEA), France (CAN, AUT, SOM, MSM, VEY, GAC, GIR, TOU, CAM, ROU), and Spain (IBI, ALB, MUR, ALM; Fig. 5; see also 1122. For P. littoralis, six populations are analysed here, five of them from France (AUT, MSM, TOU, CAM, ROU; 12). From these sites in France, we also sampled P. chalceus populations (see above). In Belgium, P. littoralis is critically endangered and the previous record went back to 1956 and was from Ostend 23. Recently, a supposed new P. littoralis population has been discovered in Belgium and is also included here (Fig. 5; ZWC). Populations of P. chalceus on a European scale were assigned to belong to one of three different salt marsh area stability types: temporary (BRA, WAT, MOK, HEI, LIS, OOS, MOE, TOU, CAM, ROU, IBI, ALB, MUR, ALM), intermediate (ZWC, NIE, MSM, FRI, AUT) and stable (SEA, CAN, SOM, VEY, GAC, GIR; see also 211. Temporary populations of P. chalceus are situated in the Mediterranean part of Europe or occur in small (<4 ha) and young (<400 years) Atlantic salt marshes. Stable and intermediate populations live in larger marshes situated along the Atlantic coast. The only difference between both salt marsh areas is their estimated age (Stable: >1000 years; Intermediate: between 400–1000 years). The age of salt marshes was estimated using historical information 24252627.

Body size (elytral length) and wing size were measured by means of a calibrated ocular under a binocular microscope. Generally, carabid wing size follows an allometric relationship with body size. 28 developed an index that corrects for this allometry, i.e. percentage of maximal realisable relative wing size. Relative wing size is wing length × width divided by elytral length × width. Relative wing size is then expressed as a percentage of the maximal wing size for a beetle of a given size. This index was shown to be an unbiased estimator for comparing different individuals, populations and species of carabid beetles 28. In ground beetles, females are generally larger than males. Therefore, we analyse male and female body sizes separately. To be complete, we analyse female and male relative wing size also separately. Body size and relative wing size are compared between species and populations with ANOVA's. Total variance is partitioned among groups (species or species and ecotypes), among populations within groups, and within populations by carrying out a nested design ANOVA using STATISTICA (version 7.1; StatSoft Inc., Tulsa, UK) on both a micro- and macroscale.

Data are used from five polymorphic enzymes: aldehyde oxidase (AO, E.C. 1.2.3.1), glucose-6-phosphate isomerase (GPI, E.C. 5.3.1.9), isocitrate dehydrogenase 1 and 2 (IDH1, IDH2, E.C. 1.1.1.42), phosphoglucomutase (PGM, E.C. 2.7.5.1.). Protocols of electrophoresis are provided by 29. Earlier work showed that one locus (IDH1) was non-neutral and we will always analyse it separately 11.

Departures from Hardy-Weinberg equilibrium expectation were tested with an exact test using the GENEPOP software (Version 3.2; 30). Significance levels were adjusted by using sequential Bonferroni correction. Similarly as in the analyses for body and wing size, total genetic variance is partitioned among groups (species, ecotypes), among populations within groups, and within populations by carrying out a hierarchical analysis of molecular variance (AMOVA) using ARLEQUIN (version 3.000; 31) on both a micro- and macroscale.

PCRs for nucleotide sequencing of COI utilized primers C1-J-1718 and C1-N-2191 and for 16S we utilized primers LR-J-1307 and LR-N-13398 32. DNA amplification reactions were performed in a 25 μL final volume. Each reaction mix contained 5 μL of extract, 1× buffer (Sigma), 1.5 mM MgCl₂, 200 μM of each dNTP, 0.4 μM of each primer and 0.6 U RedTaq DNA polymerase (Sigma). Initial denaturation was for 2 min at 95°C, followed by 35 cycles of 1 min at 95°C, 1 min 30 s at 48°C (and 46°C for 16S), and 2 min at 72°C; 9 min at 72°C completed the program. The reaction was purified with columns following manufacturer's recommendations. Sequencing was done by BigDye Terminator version 3.1 kits on an ABI 3130 sequencer (Applied Biosystems). Sequences were aligned using BioEdit version 5.0.6 33. We tested for neutrality of mutations following Fu & Li's method with D* and F* test statistics using DNASP 4.0 3435. A phylogeny of unique haplotypes was constructed from the calculated Kimura two-parameter distances using the neighbour-joining approach within MEGA (36; 1000 bootstrap replicates).