We sampled 11 populations of Agile frog, Rana dalmatina in western France. They were considered as belonging to environments with different degrees of road influence.

• (i) seven populations were investigated in ponds far from important roads or highways, in meadows between Angers and Cholet, Maine & Loire, France (considered as undisturbed populations).
• (ii) four other populations were located close to the A11 highway, between Angers, Maine & Loire and Nantes, Loire Atlantique, France (considered as disturbed (“highway”) populations).

Population size was estimated by numbering clutches and by estimating the number of calling males. Ponds below 40 adults frogs or 20 clutches were considered as small populations.

In May 1999, genetic sampling was made by collecting a mean of 14.6 spawns for each population. About 10% of the eggs of each clutch of the same population were collected and reared in plastic pans of 5 litres. Then, tadpoles represent a strong sampling of the genetic pool of the population. After emergence, 24 to 36 froglets (mean=30.4) from each population were preserved at −23°C until genetic analysis and others were released. Crude protein extracts were used for horizontal starch-gel electrophoresis. Samples were homogenised in equal volume of distilled water and centrifuged at 12 000g for 5 minutes à 4°C. Migration was performed in 12% continuous Tris citrate starch gels, for 3–5 hours at 300 V and 4°C. Slices were stained for revealing specific enzymes using standard procedures [18] and nine enzymes encoded by 12 polymorphic loci were investigated (AK, EC 2.7.4.3; CK-1 and CK-2, EC 2.7.3.2; 6-PGDH, EC 1.1.1.44; GPI, EC 5.3.1.9; αGDH, EC 1.1.1.8; LDH-1 and LDH-2, EC 1.1.1.27; MDH-1 and MDH-2, EC 1.1.1.37; MPI, EC 5.3.1.8; PGM, EC 5.4.2.2).

Mean distance between undisturbed ponds and high traffic road was 2173 m $\pm 1092$. It was significantly higher than mean distance between “highway” populations and the roadway (mean=94.3 m; SD=62.2; t_Welch=4.645; df=6; p=0.003). All populations were small; the difference between size of undisturbed populations and highway populations was not significant (U_{Mann–Whitney}=7.5; p=0.23).

Firstly, Rana dalmatina populations of western France present an important allozymic variation. It is related with a high polymorphism and an average of 2.5 alleles per enzyme in undisturbed populations (Table 1). However, in highway populations the number of alleles per enzyme averages 2.23 and was significantly lower (t_Welch=4.339; df=3; p=0.023).

No significant difference in heterozygosity level was recorded among the seven undisturbed populations (F=0.139; df=6; p=0.991). In these populations, H_obs was significantly lower than $H_{\exp }({\mathrm{t}}_{\mathrm{Welch}}=14.3$; df=8; p<0.0001). On the other hand, no significant difference in heterozygosity level was recorded among the four highway populations (F=1.124; df=3; p=0.324) while H_obs was significantly lower than $H_{\exp }({\mathrm{t}}_{\mathrm{Welch}}=6.504$; df=3; p=0.007).

Moreover, H_obs was significantly lower in highway populations than in undisturbed ones (t_Welch=14.18; df=8; p<0.0001). Consequently, undisturbed populations present a significantly lower F_IS index than highway populations (F_IS=0.315; 0.544, respectively; t_Welch=5.605; df=3; p=0.011). The difference between H_obs and H_exp have been also analysed for each locus (Table 2).

Agile frog populations of western France evidence a high deficiency in heterozygotes. Indeed, F_IS index is high, suggesting an inbreeding process. Several explanations may be invoked for this peculiar result: (1) Rana dalmatina populations are small compared with Rana temporaria ones that exhibit population sizes superior to hundred individuals [19]. (2) It can reflect a complex reproductive system of this territorial species where all the adult males do not access to reproduction [20]. (3) The high breeding pond fidelity could also result in reducing the genetic variability of populations, especially in amphibians [21]. (4) A selection on a particular stage of life, as eggs for example, may also be regarded as a source of genetic depletion. (5) The heterozygosity deficiency revealed in Rana dalmatina populations could be induced by a reduction of reproductive adults number because of road mortality. As heterozygosity deficiency is more pronounced in highway populations than in undisturbed ones, hypothesis 1, 2 and 3 may not be sufficient to explain such a difference, because they rely on species characteristics whatever the population. A selection on eggs could not be evoked too because all of the clutches were sampled in the same generation, and adult heterozygosity analysis confirmed our results [22]. These hypothesese might only result in the heterozygotes deficiency. In contrast, we propose that highways consist in a threat for frogs during their reproductive movements, with a negative influence on dispersal [23], reducing population size and involving an inbreeding process.

In general, Agile frog populations size is not considerable. Highway populations seem to be slightly smaller than undisturbed ones (10 adults versus 26 on average, n.s.). Whatever the cause of such a difference (pollution, noise, road mortality...) the low number of reproductive adults may have genetic consequences. Indeed, population size is a factor involved in genetic depletion and small populations loose more rapidly their genetic diversity. In regard to our results, roadway proximity appears as one major cause of this lower genetic variability, thus strongly suggesting that such linear structures increase extinction risk for species. While genetic flows and biological diversity trace broad patterns across the landscape, transportation planning for human mobility traditionally focuses on a narrow strip close to a road or highway. Forman and Deblinger [24] evidenced that populations close to a high traffic road (distance<600 m) exhibited several deficiencies (biological diversity, genetic depletion) arguing they are located in a “road effect zone”. In the four highway populations of our study, the mean distance between pond and high traffic road is significantly lower than in undisturbed populations. Moreover, highway ponds are close to the roadside, in a distance less than 100 meters and the results on genetic depletion in highway populations, that are within the “road effect zone”, partly validate this model. Forman & Deblinger [24] advised a protected zone of about 600 m to prevent perturbation of mammal and amphibian roadside populations. This distance has been chosen according to noise reduction and gene flow preservation. Indeed, many species are injured by roadway, especially during migration, like snails [25], amphibians [14], birds [26,27] and mammals [28]. Thus, by mapping such a zone, we might learn to predict how future roads will influence environments and where conservation projects (underpasses, noise barriers, ...) have to be built [29,30].

The majority of natural populations extinction results from environment modification, especially in the urbanised landscapes [9]. Extension of human activities and roads construction are ineluctable issues in the twenty-first century. Within the general decline of amphibian populations, it raises considerable issues for species conservation. Then, preserving genetically distinct populations is a fundamental concern for biodiversity [31], considering that small populations are more exposed to extinction than others [3,32]. Thus, we conclude that busy roads and natural areas, like breeding ponds, should be well separated, and that future transportation systems across landscapes can provide for ecological flows and biological diversity in addition to safe and efficient human mobility.