RESULTS

Morphological analysis

Canonical variate analysis (CVA) shows that differences exist between females and males morphology of A. fasciatus (Fig. 3A). These differences are displayed along the second axis, which explains 30.39% of global variance. Specimen shape of Ayata Lake is discriminated from those of the three other localities along the first axis, which represents 45.73% of global variance (Fig. 3A). The variation highlighted along axis 1 is related to landmarks 3, 9, 11, 12 and 14 and 3, 4, 8, 10, 12, 13 for axis 2. For a better appreciation of the shape differences, we expose transformation grids based on procrustes coordinates covariance matrix (Fig. 3B). Because of the overlapping of Mellah lagoon, Mellah Marsh and Bizerte lagoon specimens, we used discriminate function to obtain a shape pair comparisons female/male. Transformation grids that discriminate sexes within populations were also described (Appendix 1). Based on transformations grids, Ayata Lake fish have a shorter head length than the other populations and the end of the mouth position litters higher (see points 2 and 14 in figure 3B). The base of opercule seems to be larger for Ayata Lake specimens. We noted variations between sexes in point 9, which corresponds to the insertion of the anal fin (smaller base in Ayata Lake) and more distant from pelvic fin in Ayata Lake. We see a difference between males and females along axis 2 (Cv2) in points 3, 4, 5, 7, 8 and 9. (Fig. 3B). Males are a bit wider than females, from the anal part to the caudal part of the body.

Genetic analysis

The seven-enzymatic systems (MDH, GOT, PGI, PGM, EST, IDH, LDH) gave clear zymograms. Among fifteen identified loci, eleven were polymorphic (GOT m, GOT f, PGI m, PGI f, PGM-2, EST f1, EST f2, EST m, LDH m, LDH f, IDH f). The mean number of alleles per locus (Am) varies between 1.93 for Ayata Lake and 1.33 for Mellah Marsh (Tab. II).

The highest value for both allelic richness (Ar = 1.45) and private allele richness (Ap = 0.32) are noted for Ayata Lake sample, while Bizerte lagoon is nearly close to Mellah lagoon in allelic richness (Ar = 1.21). However, the lowest is observed in Mellah marsh (1.11 for allelic richness and 0.06 for private allele richness) (Tab. II). For the expected heterozygosity, the highest value is 0.1413 ± 0.1803 in Ayata Lake and the lowest one in Mellah marsh (He = 0.0294 ± 0.0644).

The estimation of Wright fixation index Fis, according to Weir and Cockerham in each sample, revealed a significant deviation from panmixia with Fis = 0.282 (0.050 -0.505, 95% CI) in Bizerte lagoon and Fis = 0.524 (0.345 -0.653, 95% CI) in Ayata Lake. The estimation of global Fst shows a high significant value (Fst = 0.1246, P = 1). Removing PGM-2 locus, this value decreases and becomes not significant (Fst = 0.015, P = 0.053). We also note that after a jacknife analysis, when removing Ayata Lake sample, we obtain a lower and not significant value of Fst (0.0064, P = 0.186). Genetic distances calculated among pairwise samples according to Reynolds et al. (1983) showed that the highest distances were between Ayata Lake and the three other populations. The same finding was observed with Fst values as shown in table III.

Neighbour-joining tree was drawn basing on genetic distances of Reynold (Reynolds et al., 1983). The tree differentiates the populations in two clusters. The first group, supported by a bootstrap value of 100, contains Bizerte lagoon with Mellah lagoon and Mellah marsh as sub clade. This clade shows also high relationships between the two samples of Mellah region. The second clade only includes Ayata lake samples (Fig. 4).