**3. Results**

### *3.1. Community Composition and Isotopic Signatures*

A total of 8752 samples comprising basal resources, invertebrates and fish were collected from the three lakes, 8645 (148 taxa) of which were invertebrates and fish (Table 1, Table S1).

Malacostraca (Amphipoda, Decapoda and Isopoda), Gastropoda, Anthozoa, Bivalvia, Polychaeta and Ophiuroidea together made up 93.09 ± 3.86 % of invertebrates.

Invertebrate abundance was lower in IP than the other two lakes (Table 1, paired-χ<sup>2</sup> test, χ2 at least 20.46, *p*-value always <0.0001, Table S2). The composition of both the invertebrate and fish community also varied (contingency table, χ2 at least 170.2, *p* always <0.001, Cramer's V at least 0.46, Tables S2 and S3). The abundance of some taxa, such as Decapoda and Anthozoa, decreased, while that of others (such as Amphipoda) increased with the pollution level of the lake (paired-χ<sup>2</sup> test, χ2 at least 48.36, *p*-value always <0.0001). The number of fish taxa varied, i.e., 23 in LP, 17 in IP and 7 in HP. The relative abundance of fish differed between lakes (paired-χ<sup>2</sup> test, χ2 at least 34.21 *p*-value always <0.0001) and was lowest in HP. The standard length of *Anguilla anguilla* was lower in HP (35.10 ± 3.63 cm) than the other two lakes (48.81 ± 6.07 cm in LP, 47.66 ± 2.12 cm in IP) (Mann–Whitney test with Bonferroni correction in cases of multiple comparisons, U = 13.0, *p*-value always <0.05). Similarly, *Diplodus annularis* had an average standard length of 6.25 ± 0.14 cm in HP, which was lower than LP

(8.76 ± 0.31 cm) and IP (9.43 ± 0.41 cm) (Mann–Whitney test with Bonferroni correction in cases of multiple comparisons, U = 2.0, *p*-value <0.05).

**Table 1.** Parameters describing the communities in each Lake. LP, IP and HP: low, intermediate and high eutrophication. N◦ indicates the sample size. Numbers in parentheses indicate the number of samples analysed. Community indicates both fish and benthic invertebrates. Stable isotopes of δ<sup>13</sup> C and δ15N are reported as mean (%) ± s.e. For each parameter, di fferent superscript letters (a,b,c) indicate di fferences between lakes (one-way ANOVA or Mann–Whitney test; *p* < 0.05).


Among the basal resources, detritus showed depleted δ13C values, while primary producers were δ13C-enriched (Figure 2). δ13C-enrichment was also observed in pelagic fish with specialist diets such as *Atherina boyeri* (δ<sup>13</sup> C = −15.34 ± 0.08% in LP, −17.45 ± 1.35% in IP and −16.14 ± 0.90% in HP). Since neither the mean nor the variance (σ2) of δ13C in the primary producers di ffered significantly either within each lake or between lakes (one-way ANOVA and associated Levene's test for homogeneity of variances, F at least 0.1412, *p*-value always >0.05), we concluded that the presence of a salinity gradient within a lake could not have an e ffect on the isotopic variability of the baseline. δ15N values of primary producers increased with eutrophication (one-way ANOVA, F: 5.80, *p* <0.01).

The isotopic di fferences observed in basal resources reflected those observed in the whole community (Table 1, Figure 3; Kruskal–Wallis, Hc at least 127.1, *p*-value < 0.001; for δ13C Mann–Whitney with Bonferroni correction for multiple comparisons, U at least 45,590, *p*-value always <0.001 and for δ15N Mann–Whitney, U at least 421.57 with Bonferroni correction for multiple comparisons, *p*-value always <0.001, Figure 3) and in *A. anguilla* and *D. annularis*.

The δ13C and δ15N isotopic signatures of *Anguilla anguilla* and *Diplodus annularis* di ffered between lakes (Kruskal–Wallis, Hc at least 35.5, *p*-value <0.001). δ13C values were higher in the least polluted lake (Table 2; Mann–Whitney with Bonferroni correction for multiple comparisons, U at least 3392.5, *p*-value always <0.001), while δ15N values increased with eutrophication (Table 2; Mann–Whitney with Bonferroni correction for multiple comparisons, U at least 555, *p*-value always <0.001).

Specifically, in *A. anguilla,* the more generalist of the two species, δ13C values reflected the shift of inputs from marine to terrestrial origin passing from the least to the most eutrophic lake (Figure 2 and Tables 1 and 2).

**Figure 2.** Isotopic standard ellipse areas (SEAcs) of *Anguilla anguilla* (continuous line) and *Diplodus annularis* (dashed line) in lakes with low (LP), intermediate (IP) and high (HP) eutrophication. Isotopic signatures (Mean ± s.e.) of primary producers (empty symbols) and detritus (full symbols) in lakes with low (circle), intermediate (square) and high (triangle) eutrophication. The greyscale reflects the origin of the main organic matter inputs from terrestrial (dark grey, left), to marine (light grey, right).

**Figure 3.** δ13C **(a)** and δ15N **(b)** values (%) of basal resources, invertebrates, fish and the whole animal community in each lake. LP, IP and HP: low, intermediate and high eutrophication. Isotopic values are reported as mean ± s.e. Greyscale indicate degrees of eutrophication: LP (white), IP (grey), HP (black). Arrow indicates increasingly eutrophic conditions. Different letters (a, b, c) within panels indicate differences between lakes (Mann–Whitney test with Bonferroni correction for multiple comparisons; *p*-value <0.05).

**Table 2.** Isotopic niche and food web metrics of the eel *Anguilla anguilla* and the seabream *Diplodus annularis* in each lake. LP: low, IP: intermediate, HP: high eutrophication. N: sample size, ITUs: Isotopic Trophic Units, δ13C (%) and δ15N (%) (mean ± s.e.), CR: Carbon Range, NR: Nitrogen Range, L: number of feeding links, S: number of ITUs in the diet, L/S: Linkage density, SEAc: Standard Ellipse Area "corrected" (SEAc) by degree of freedom, TNW: Trophic Niche Width. For details of metrics, please refer to the materials and methods section. For each parameter, different superscript letters (a,b,c) indicate differences between lakes (Mann–Whitney test with Bonferroni correction for multiple comparisons; *p*-value <0.05).



**Table 2.** *Cont*.

### *3.2. Niche Metrics and Diet of Anguilla anguilla*

The isotopic signatures and niche metrics of *Anguilla anguilla* varied among lakes (Table 2, Figures 2–4; Kruskal–Wallis, Hc least 12.06, *p*-value <0.001). The highest δ15N values were observed in HP (Table 2; Mann–Whitney with Bonferroni correction in cases of multiple comparisons, U at least 0.1, *p*-value always <0.001). The Carbon Range increased with eutrophication (Figure 4, Table 2) and the largest Nitrogen Range was observed in the eutrophic lake (Table 2).

**Figure 4.** Boxplot of the distribution of δ13C (**a**) and δ15N (**b**) isotopic signature of *Anguilla anguilla* and *Diplodus annularis* in each lake: LP, IP and HP: low, intermediate and high eutrophication. For each lake, the thick horizontal line represents the median of the distribution, the box includes 50% of the data, the symbol (x) represents the mean and the whiskers reach the highest and lowest value within 95% of the distribution. Different letter (a, b, c) within panels indicates differences among lakes (Mann–Whitney test with Bonferroni correction in cases of multiple comparisons; *p* <0.05).

Overall, no correlation between the body length and δ13C (%) of *Anguilla anguilla* was observed in any lake (Pearson correlation, *p* >0.05).

*Anguilla anguilla* had 5 ITUs in LP and HP and 12 ITUs in IP, where the eel-resource ITU linkage density was highest (Table 2). ITU-based mixing models showed no differences between lakes in terms of the overall contribution of invertebrates to the eels' diet (Figure 5a).

By contrast, the consumption of basal resources increased and piscivory decreased with increasing levels of pollution (i.e., from LP to HP; Table 3 Figure 5a).

*A. anguilla* showed a generalist diet including 20 different categories of food source (Table 3, Figures 6 and 7). Some of these were common to the three lake populations (e.g., Actinopterygii, Bivalvia, Gastropoda, Decapoda and Polychaeta) but their consumption varied. The Bray–Curtis index (BC) applied to diet showed a lower similarity between the HP population and the others (76% similarity between LP and IP vs. 41% between LP and HP, and 54% between IP and HP). Specifically, in LP the diet of *A. anguilla* was mostly based on Actinopterygii (34.76% ± 1.90), Decapoda (27.84% ± 4.60) and Gastropoda (13.06% ± 0.40), in IP on Actinopterygii (30.65% ± 0.50) and Decapoda (26.13% ±

1.90) and in HP on Polychaeta (30.41% ± 0.60), Actinopterygii (12.18% ± 1.50), Bivalvia (10.35% ± 0.90), detritus (9.03% ± 0.50) and Decapoda (9.11% ± 1.90) (Table 3, Figure 6).

**Figure 5.** Contribution to the diet of *Anguilla anguilla* (**a**) and *Diplodus annularis* (**b**) of basal food sources (white), invertebrates (black) and fish (grey) in the lakes with low (LP), intermediate (IP) and high (HP) eutrophication. The overall contribution of basal resources, invertebrates and fish is reported as the mean (%) ± s.e. Different letters (a,b,c) within panels indicate differences between lakes in the contribution of food sources to the diet (χ2-test, *p*-value <0.001).

**Table 3.** Proportional contribution (in %) of food sources to the diet of *A. anguilla* in each lake, obtained from ITU-based mixing models. LP: low anthropogenic pressure, IP: intermediate anthropogenic pressure, HP: high anthropogenic pressure. The contribution of each food source is reported as the mean (±s.e.). "Taxa" indicates the number of taxa belonging to the respective group in the diet of *A. anguilla.* The overall contribution of basal resources, invertebrates and fish is reported as the mean (%) ± s.e. Different superscript letters (a,b,c) indicate differences between lakes in the contribution of categories of food sources to the diet (χ2-test, *p*-value <0.05). For details please refer to the methods section.



**Table 3.** *Cont*.

**Figure 6.** *Anguilla anguilla* and *Diplodus annularis* food webs in lakes with low (LP), intermediate (IP) and high (HP) eutrophication. Each node at the base of the food web represents a food source (in terms of class and respective families). Arrows point from each food item to its consumer: *Anguilla anguilla* (black arrows) and *Diplodus annularis* (grey arrows). The arrows' thickness is proportional to the trophic interaction strength. TNW indicates trophic niche width. *O* indicates the niche overlap between *D. annularis* and *A. anguilla*. For details of metrics, please refer to the results section.

**Figure 7.** Invertebrate abundance in the environment (histograms), and selectivity values for each invertebrate group in the diet of *Anguilla anguilla* (double line) and *Diplodus annularis* (single thick line) in LP, IP and HP (low, intermediate and high eutrophication). Selectivity values greater than 1/m (dotted line) indicate preference.

The difference in resource use was associated with a difference in trophic niche width (Figures 2–6), which increased with eutrophication (TNW: 1.81, 2.06 and 2.32 in LP, IP and HP respectively), with significant differences between HP and LP (bootstrap comparison among populations, *p* <0.0001).

### *3.3. Niche Metrics and Diet of Diplodus annularis*

The δ13C and δ15N of *Diplodus annularis* varied across lakes (Figures 2–4, Kruskal–Wallis for both δ13C and δ15N, Hc at least 35.5, *p*-value <0.001). In addition, no significant correlation between the body length and δ13C (%) of *Diplodus annularis* was observed in any lake (Pearson correlation, *p* >0.05).

δ15N increased with the level of pollution (Table 2, Figure 4). The Carbon Range was highest in LP while no differences in Nitrogen Range were observed between lakes (Table 2). The seabreams had

four ITUs in LP and two ITUs in both IP and HP (Table 2). No differences in the linkage density (L/S) of the ITUs or in SEAc were observed between lakes (Table 2).

Mixing models based on the single ITU values showed similar average contributions of basal resources, invertebrates and fish in the lakes (Figure 5b; paired-χ<sup>2</sup> test, *p*-value always >0.05).

However, when taxa in each category were distinguished, diet similarity between LP and HP was 55%, while between IP and both LP and HP it was 61% (Bray–Curtis index, BC).

Overall, the diet of *D. annularis* was based on 17 different taxa, and invertebrates represented more than 70% of it in all lakes (Table 4, Figures 5 and 6). Among these, Decapoda (22.03% ± 1.37), Gastropoda (20.65% ± 1.06) and Actinopterygii (12.69% ± 0.49) contributed most to the diet of *D. annularis* in LP (Figure 6); Decapoda (27.86% ± 1.69), Actinopterygii (10.84% ± 0.28) and Amphipoda (9.86% ± 1.54) in IP (Figure 5); and Polychaeta (34.76% ± 1.12), Actinopterygii (18.56% ± 0.78) and Decapoda (14.67% ± 1.03) in HP.

**Table 4.** Diet composition of *Diplodus annularis* in each lake. Proportional contribution (in %) of food sources to the diet of *D. annularis* obtained from ITU-based mixing models. LP: low anthropogenic pressure, IP: intermediate anthropogenic pressure, HP: high anthropogenic pressure. The contribution of each food source is reported as the mean (± s.e.). "Taxa" indicates the number of taxa belonging to the respective group in the diet of *D. annularis.* The overall contribution of basal resources, invertebrates and fish is reported as the mean (%) ± s.e.


The trophic niche was significantly narrower in HP than LP and IP (Figure 6; TNW: 1.98 vs. 2.15 and 2.28 respectively; bootstrap comparison among populations, *p* <0.05). The niche overlap between *D. annularis* and *A. anguilla* decreased from the eutrophic to the unpolluted lake (O = 0.95, 0.87 and 0.84 in HP, IP and LP respectively).
