Figures legends:
Figure 1: Locations of study sites of collection ofAmnicola limosus populations in the St. Lawrence River system and experimental design of reciprocal transplant experiment. A: The study sites are located near the island of Montreal, QC, Canada (right of the map). Sites are colored based on water calcium concentration (mg/L). Sites with gobies absent are HA-LCGA, OKA-LCGA, IB-LCGA, PB-LCGA, IPE-LCGA, and sites with gobies present are PG-HCGP, PST-HCGP, PON-HCGP, BEA-HCGP, GOY -HCGP. The two exceptions are RAF-LCGP (low calcium and gobies present) and PDC-HCGA (wetland, high calcium, and gobies absent). B: Experimental design showing the two origin waters (Ottawa River in grey or St. Lawrence River in blue) with six populations from each river (replicates), water treatment (OR-LC: Low calcium - Ottawa River; SL-HC: St. Lawrence River - high calcium) and goby cue treatment (\(+\)/\(-\): with or without).
Figure 2: Results of the environmental association analyses. (A) Biplot of the q-values obtained from the poolFreqDiff analysis testing for consistent differences in allele frequencies between populations from invaded and uninvaded environments, as a function of the q-values derived from the C2-statistic in the STD model in Baypass, assessing the association of SNP allele frequencies with goby presence/absence. The dashed vertical and horizontal grey lines indicate the FDR of α 0.01. ( B) Biplot of the q-values for the poolFreqDiff analysis comparing populations of low and high calcium habitats as a function of the Bayes Factor (BFis) from the Baypass STD model (median of three independent runs) testing the association with calcium concentration. The dashed vertical line (BFis > 20 dB) indicates outlier SNPs significantly associated with the calcium covariable and the horizontal line gives the q-value FDR of α 0.01.
Figure 3: Fecundity (total number of eggs produced) and survival as a function of water treatment, origin water, and goby cue treatment in the reciprocal transplant experiment. Each dot represents a measurement for one population (blue: St. Lawrence populations HCGP, grey: Ottawa river populations LCGA), summarized by the corresponding boxplots (showing the median, minimum and maximum, values, 1st and 3rd quartiles, and outliers as full dots). The black dots are the mean for each treatment, with squares and triangles for treatments with or without goby cues. Fecundity was analyzed with a GLM (negative binomial distribution), with only the origin water being significant (p = 0.015). Survival was analyzed with a GLMM (binomial distribution), with the origin and treatment water being significant (p = 0.020 and p = , respectively).
Figure 4: Genome-wide diversity indices. (A) violin plots of nucleotide diversity pi, (B) Watterson’s Theta, (C) and Tajima’s D according to habitats (Ottawa River LCGA; St. Lawrence River HCGP), with median and interquartile ranges shown in the box plot insert. (D) Observed heterozygosity per population, comparing habitat types (black dots: average per habitat, error bars show one standard error interval), except for PDC (HCGA, wetland refuge) and RAF (LCGP) which have inverted calcium concentration and goby presence characteristics.
Figure 5: Population genetic structure. (A) Heatmap of the scaled covariance matrix Ω (with ρij the correlation coefficient between pairs of populations) with hierarchical clustering tree (using the average agglomeration method), obtained from the core model of Baypass. (B) Pairwise FST matrix between the twelve study populations. The gray bars correspond to the Ottawa River populations (LCGA) and the blue bars to the St. Lawrence populations (HCGP), except for PDC (HCGA, wetland refuge) and RAF (LCGP).