**4. Discussion**

Our results indicate that anthropogenic inputs a ffected the composition and abundance of the lake animal community. Specifically, the diversity of fish decreased, and trophic choices of the eel *Anguilla anguilla* and the seabream *Diplodus annularis* changed, with increasing eutrophication. *A. anguilla*, which was more piscivorous than the seabream at low and intermediate eutrophication, increased its consumption of invertebrates and basal resources. On the other hand, the seabream, which fed more on invertebrates, increased its preference for polychaetes. The high selectivity for polychaetes in the highly eutrophicated lake could be due to the facilitated capture of these preys and their good contribution to the energy supply of their predators compared to other aquatic invertebrates [66–68]. Changes in the feeding choices of the two fish species resulted in increased interspecific niche overlap, suggesting that eutrophication may have strong bottom-up e ffects on interspecific interactions [69,70].

Previous research demonstrates that the δ15N values of the aquatic biota reflect anthropogenic nitrogen inputs from the surrounding terrestrial areas [7,10,71,72]. In our study, increased eutrophication was associated with higher δ15N in the fish community, as previously observed by Santoro et al. [2] and Jona-Lasinio et al. [3] for invertebrates and primary producers respectively. Our results indicate that in transitional waters, individual fish δ15N could be a useful indicator of anthropogenic N transfer along food webs [10,73], while the range of δ13C in the fish population could reflect the diversity of C inputs, emphasising the need for ecological monitoring in these productive ecosystems.

The distinct carbon isotopic signatures (δ13C) of primary producers (e.g., terrestrial vs. aquatic vegetation) allowed us to discern the origin of the organic matter contributing to the nutrient pool of water bodies [7,33,34,74]. Specifically, depleted δ13C organic matter values indicated the contribution to the organic matter pool of allochthonous (terrestrial) carbon, while enriched δ<sup>13</sup> C indicated autochthonous primary production, as also observed in other aquatic ecosystems [7,34,35,74]. Although it was not in the remit of this study to investigate the cause, carbon enrichment was also visible in pelagic and strictly specialist species such as the sand smelt, *A. boyeri*. In addition, neither the mean nor the variance of δ13C of primary producers di ffered significantly between sites within the same lake or between di fferent lakes. This allowed us to exclude possible interferences in the isotopic baseline arising from environmental parameters such as the salinity and oxygenation levels of the respective area. The larger contribution of terrestrial organic matter in HP may be due to the large input of fresh water that this lake receives from the hinterland, as indicated by the low salinity generally observed in this lake compared to the other two. These results are consistent with what has been observed in similar environments [7,74,75].

Large Carbon Ranges sugges<sup>t</sup> multiple carbon sources at the base of the food web [7,35,62–64]. The supply of organic matter from multiple sources in the three lakes was evident in the δ13C values of *A. anguilla*, which shifted from marine to terrestrial input with the increasing pollution. Its larger Carbon Range in the eutrophic lake indicates that this species integrated both autochthonous and allochthonous carbon pathways, while in the unpolluted lake it relied mainly on the autochthonous one.

It is acknowledged that increased N-loads promote significant changes in aquatic productivity [76] that could potentially a ffect the composition of the prey community [77,78]. This in turn might be reflected in the feeding behaviour of consumers at all trophic levels [2,7,79].

In our study, increased N pollution resulted in altered community composition, with decreasing diversity, which seemed to a ffect the feeding preference and niche width of *Anguilla anguilla* and *Diplodus annularis*. The feeding regime of the two species is known to be characterised by marked generalism and trophic plasticity [19,22,42,43,80]. However, while the eel enlarged its trophic niche, feeding o ff multiple resources at various trophic levels in eutrophic conditions, the seabream concentrated on a small number of invertebrates. Although dietary changes are known to depend primarily on size and growth stage [81–83], in our study neither *A. anguilla* nor *D. annularis* showed a relationship between body size and δ13C signatures. The greater trophic generalism and omnivory of eels with eutrophication may be due to the di fferent density, accessibility and availability of the prey at higher trophic levels [19,22,24,42,43] as well as to an altered presence of potential competitors for the same

food sources [78]. This was also confirmed by the results of the Bayesian mixing models, which showed a shift with eutrophication in the diet of *Anguilla anguilla* mostly from pelagic (fish) to benthic (invertebrates) prey, regardless of their abundance. Bouchereau et al. [24,25] reported Teleosts, amphipods and decapods as the predominant prey in the diet of *A. anguilla* in two North-Mediterranean lagoons, and that prey selection could be linked to the activity and accessibility of the prey itself. Rosecchi [42], Pita et al. [23] and Chaouch et al. [19] indicated molluscs (bivalves and gastropods), crustaceans, polychaetes and Teleosts as the main items in the diet of *D. annularis* in lagoons and coastal waters. Lammens et al. [84] and Dörner et al. [80] identified *A. anguilla* as belonging to the piscivorous community in many European lakes. In our study, the adoption of piscivory in the least eutrophic lake allowed the eel to reduce niche overlap [80,84,85] and therefore potential competition with other fish.

However, the trophic behaviour of the two species can be expressed differently by individuals within populations [18,20,21], enabling *A. anguilla* and *D. annularis* to include several food sources in their diet even in a single area [19,25]. In this context, the individual isotopic characterisations of carbon (δ13C) and nitrogen (δ15N), coupled with diet reconstruction at the isotopic trophic unit level (ITU, as recently proposed by Rossi et al. [31]) were crucial to the detailed estimation of the diets of the two fish species with high trophic generalism and omnivory. This allowed us to consider the possibility that each individual consumer could draw on the whole spectrum of potential food sources available within each lake [31]. In this way, we are able to describe the trophic plasticity and generalism of two ecologically and economic important fish species, and hence the real variation of the diet within the same populations under a range of eutrophication conditions.
