Our results confirmed a wide distribution and bioavailability of PFAS in the aquatic environment. The canal Brussel–Schelde (Fig. 1, Loc. 30) showed high accumulated concentrations in all biota. This canal is subject to direct influence from intensive industrial activities. The highest PFAS concentrations in perch were measured in the Zenne River (Fig. 1, Loc. 31), known for its very high background pollution and influences from Brussels . Furthermore, high concentrations were measured in mussels deployed in the canal Bocholt–Herentals (Fig. 1, Loc. 42). This canal is a connection between the Meuse and Scheldt basin, both known for large effects of industrial as well as household waste water as a source of PFAS . The Melsterbeek, however, showing high accumulated concentrations in eel, flows through a more agricultural region. Here, contamination with PFAS might be caused by agriculture, households or undefined point sources. However, our conclusions on this part are mere qualitative and based on personal interpretation and experiences of the general monitoring network of the Flanders Environment Agency. Further investigation using data on population density, area of industrial surfaces and emission indices could be used to investigate the relationship between accumulated concentrations in biota and possible sources with a more quantitative approach.
The use of PFOS is restricted since 2009 . This should eventually lead to a decrease of this substance in the environment. Previous studies on the water bodies used in the present study indeed showed higher concentrations of PFOS. A preliminary monitoring study from 2013 reported PFOS concentrations in muscle tissue of eel of 15, 33, 7.2 and 34 µg/kg ww in the upper-Scheldt, canal Ieper-IJzer, Kleine Nete and Demer, respectively . The present study showed concentrations of 5.6, 14.5, 12 and 11.3, respectively, at the same sampling locations in eel. These results showed a clear decrease and possibly revealed the effects of phasing-out the use of these compounds to the environment, with the exception of the Kleine Nete, which remained stable.
Compared to previous studies on yellow perch (Perca flavescens), the PFOS (118.6 ± 29 ng/g ww) and PFUnDA concentrations (3.8 ± 1.2 ng/g ww) in New Jersey, USA, were considerably higher than those reported in the present study, whilst, on the contrary, concentrations of PFDA (1.1 ± 0.4 ng/g ww) and PFDoDA (0.7 ± 0.2 ng/g ww) were lower in New Jersey . The concentrations of PFOS, PFUnDA, PFTrDA, PFDA, PFDoDA and PFOA were higher in the present study compared to those reported in perch collected in Finnish rivers (3.4 ng/g PFOS, 1.0 ng/g PFUnDA, 0.45 ng/g PFTrDA, 0.5 ng/g PFDA, 0.23 ng/g PFDoDA and 0.03 ng/g PFOA ), and to those reported in shad (Alosa agone), European whitefish (Coregonus lavaretus), burbot (Lota lota), rainbow trout (Oncorhynchus mykiss), perch, roach (Rutilus rutilus), brown trout (Salmo trutta) and Arctic char (Salvelinus alpinus) in glacial lakes from the Alps in France, Switzerland and Italy (6.0 ng/g PFOS, 0.3 ng/g PFUnDA, 0.5 ng/g PFDA and 0.3 ng/g PFDoDA ). A monitoring study in the Netherlands measured PFOS concentrations in bream (Abramis brama), roach, perch and pike-perch between 4.9 and 120 ng/g ww , which were higher than those reported in the present study. In the North Rhine–Westfalen basin in Germany, eel PFOS concentrations ranged between 8.3 and 49 ng/g ww . The PFOS concentrations in the Loire estuary ranged from 17.9 to 39.0 ng/g ww . Kwadijk et al.  examined the distribution of multiple PFAS in eel from The Netherlands and reported PFOS concentrations ranging from 7 to 58 ng/g ww. In Lake Möhne, Germany, PFOS concentrations of 37–83 ng/g ww have been reported in eel, whereas PFOA concentrations ranged up to 2.3 ng/g ww . These concentrations are comparable to those reported in eel in the present study. Compared to other studies investigating PFAS concentrations in mussels, the ΣPFAS concentrations were higher in the present study than reported in previous studies in fresh water mussels from Spain , and marine mussels from the Netherlands , Spain [35, 77], and Denmark , but lower than those reported in marine mussels from Portugal . Comparison to literature, revealed a large variation of PFAS concentrations measured in biota, both on a European and global scale. High concentrations might be due to the presence of different sources of PFAS contamination (e.g. point sources, diffusive emission sources). On the other hand, differences between species could be explained by their diet .
PFAS bioaccumulation, magnification and dilution
Longer chain PFAS preferentially partition to sediments, as the water solubility of PFAS is inversely proportional to the length of the carbon chain, while short-chain compounds remain dissolved in the water [51, 62]. The carbon chain length, as well as the identity of the anionic functional group of PFAS, is related to their bioaccumulative potential, with PFSAs being more bioaccumulative than PFCAs with the same fluorinated carbon chain length . Furthermore, PFCAs and PFSAs that contain at least eight fluorinated carbons (i.e. PFNA and PFOS and longer compounds) have a greater bioaccumulative potential . Although shorter chain PFAS can also bioaccumulate, they have a much smaller bioaccumulation potential and their bioaccumulation is mainly related to elevated concentrations in the water column [17, 34]. This might also explain the larger contribution of short-chain PFAS in mussels compared to fish, since both isotopic niche analysis and TL confirmed that the mussels occupied a lower position in the trophic food chain.
The differences in PFAS profiles between the mussels and fish species are likely also the result of different ways of exposure, caused by the sampling strategy and experimental design of this study. As the mussel cages were placed in the water column, without contact to the sediment, the mussels have been exposed solely to the water and suspended material, whilst the fish have been exposed to both the water and sediment. Although both species are considered to be spending time in close relation to the sediment compartment [7, 74], eel shows a more bottom-dwelling lifestyle. Consequentially, the dominance of hydrophilic PFCAs and PFSAs, with less than eight fluorinated carbons (i.e. PFOA, PFHxS and shorter PFAS), was expected in the mussels. Similarly, the fish species have also been exposed to sediments and hence to longer chain PFAS with a higher bioaccumulative potential. This also explains why PFOA concentrations were significantly higher in the mussels compared to both fish species, whilst the opposite pattern was often observed for longer chain PFAS. On the other hand, it has been shown that biotransformation (the degradation of PFAS precursors to, for example, PFOS) efficiency increased with increasing trophic level, from invertebrates to fish .
Dietary differences between perch and eel could explain the differences in PFAS concentrations and profiles between both fish species. Although on average, the TLs of both species did not differ, eels have a slightly broader range of TLs compared to perch. Despite that both are predatory species, feeding primarily on invertebrates and small fish species [63, 76], there are differences in the feeding ecology between both species in the studied populations. The broader isotopic niche (indicated by SEAc) indicates that, despite the overlap in isotopic niche area between both species, eels have a more diverse and flexible diet, which might consist of different invertebrate or fish species, compared to perch . For example, De Meyer et al.  showed that head morphology of eel (broad-headed vs. narrow-headed) could influence diet, trophic level and therefore pollutant accumulations. Furthermore, the diet of both species is known to depend on their size, as size-dependent diet segregation of both species has been reported before [29, 63, 76]. This segregation is also known to vary widely among populations , which could also explain dietary differences, and hence differences in exposure, between perch and eel. Additionally, spatial differences in diet may occur depending on local ecological variation in species composition and food availability. Finally, biotransformation of PFAS can be species-specific [4, 32], probably due to specific proteins involved in the process , resulting in different contamination profiles.
A positive correlation between accumulated concentrations in mussels and fish was found for PFTrDA, PFTeDA, and PFOS and PFDoDA in eel. This positive relationship reflects the possibility of mussels to predict the pollutant pressure at different locations, since high concentrations at a specific location are found in fish as well as mussels. With a negative correlation, as was the case for PFOA in perch, the relationship between species is contradictory, and mussels will not be able to predict high pollution levels in fish (and therefore a risk of secondary poisoning). Furthermore, all significant relationships between both fish species showed to be positive, which is a logical consequence since they occupied similar trophic levels of the same food web in each location.
The TMFs in the present study were compared to those of other studies on freshwater ecosystems, and were higher than those reported by Loi et al.  for PFOS (TMF = 1.3) and lower than those for PFUnDA (TMF = 1.7), although the general trend of biomagnification was comparable to the present study. Furthermore, the TMFPFOS calculated in the present study is in line with other European studies on lake and river food chains as reported by Rüdel et al. . On the other hand, Lescord et al.  reported a negative relation between accumulation and trophic level for PFOS, PFNA and PFUnDA in a foodweb from the high Arctic. In a study on alpine lakes in Northern Italy, a TMFPFOS of 3 was found . However, when they analysed a fish-only food web this value became not significant and lower than one. The presence of TMFs greater than 1 for PFOS, PFDA and PFUnDA was expected, as the biomagnification of these PFAS has been reported before [47, 58]. The observed biodilution for PFOA and PFTeDA could, in case of PFOA, be explained by differences in exposure between the mussels and fish species, as was described above. Regarding PFTeDA, its biodilution may be associated with its large molecular size, limiting the penetration of cell membranes [17, 45]. However, since species of different locations were compared, we might need to take into account the possible effect of location (ecological quality, physicochemical parameters) rather than just bare bio-magnification and -dilution effects on bioaccumulated concentrations. This is in agreement with Munoz et al. , who stated that the PFAS chemical structure is not be exclusively predictive of TMFs, since they are also influenced by trophic web characteristics.
Suitability of mussels in active biomonitoring of PFAS
The PFAS profiles in the mussels did not differ much among the three species (Additional file 1: Figure S1). Furthermore, isotopic niche determination showed an overlapping niche for both quagga mussels (D. bugensis) and Asian clams (C. fluminea). Therefore, it is appropriate to use both species in monitoring studies on locations with varying salinity and extrapolate the results. Although the PFAS profile of the blue mussels (M. edulis) differs slightly from those of the quagga mussels (D. bugensis) and the Asian clams (C. fluminea), this is likely the result of a smaller dataset for the first species. Therefore, more research, using a larger sample size, is necessary to fully confirm the suitability of using blue mussels simultaneously with the other two species.
Regarding their suitability in active biomonitoring, our results show that mussels can provide an overview of the contaminants present in the environment to which fish are exposed. The PFAS compounds that have been detected in the mussels were similar to those detected in the fish species, although concentrations differed, as was explained above.
Both ABM using mussels and PBM using indigenous fish, have their assets and liabilities. As the mussels provided a short time pollution profile (exposure during 6 weeks), fish allowed the integration of a lifetime exposure. Short time exposure might be influenced by seasonal variations in bioavailability, which is cancelled out using indigenous species. Furthermore, the numerous measurements below LOQ in mussels might give an underestimation of the situation. The bioavailability, however, is made clear through biomagnification in the accumulated concentrations of fish. Measuring in fish also may give additional information towards risk assessment for species of higher trophic levels feeding on fish, such as predatory birds or mammals (including humans). On the other hand, from an ethic perspective, the use of invertebrates might be encouraged. PFAS analysis can be done on a small amount of tissue and therefore on individual mussels, so no large numbers are needed. The results in the present study confirm the possibility of extrapolation between mussels and fish and between both fish species.
Human health risks
The maximum recommended amount of eel that could be consumed without posing health risks, according to the ATSDR guidelines , was lower than that of perch concerning PFOA contamination, but for PFOS the opposite was true. Nonetheless, for PFOS the differences between both species (ca. 1.4 times) were smaller than for PFOA (ca. 2 times). On the other hand, when using the EFSA value , the maximum recommended amounts of both fish were much lower. This sensitive value was determined with a decreased immunoresponse after consumption as a critical human health response and is to be tested against the sum of PFOS, PFOA, PFNA, PFHxS. However, as stated in the materials section, the sum of PFOS and PFOA was used in the present study. Although, this might have led to an underestimation of the actual risk, we believe it to be a good estimation since PFOS had the largest contribution to the total PFAS sum.
Due to consumption of their catch, in Flanders mainly recreational anglers and their families may be exposed to contaminated fish. A mean consumption of 2.7 g of perch per day and 18 g of eel per day was reported in an interview on anglers (ANB-VF/2015/4). The maximum recommended amounts of fish that could be consumed without posing health risks (Q-values; Table 3), calculated using the mean concentrations in the fish species, were higher than the mean consumption amounts in Flanders for both species using the ATSDR  MRL. The Q-values for eel were ca. 50 and 1.2 times higher than the reported consumption amount, for PFOA and PFOS, respectively. For perch, this was ca. 630 and 6 times higher. On the other hand, using the more strict EFSA  MRL, Q-values were below the reported consumption amount for eel and only 2 times higher for perch.
However, when using a worst-case scenario, based on the maximum concentrations in both fish species, health risks due to PFAS contamination are expected from the consumption of both fish species. In this worst-case scenario, the maximum edible amount of perch per day was 2.62 g/day and 0.82 g/day (Table 3), calculated using the ATSDR  PFOS MRL value and the EFSA  MRL values, respectively. For eel, these Q-values in the worst-case scenario are 2.17 g/day and 0.68 g/day, respectively, which are 8 to 26 times lower than the average eel consumption in Flanders.
Therefore, it is likely that the local recreational fishermen have a high chance of experiencing detrimental effects of accumulated PFAS concentrations. Evidentially, calculations were performed on mean consumption rates, indicating individuals exist that consume more. For these people even more locations might pose a health risk, since the Q-values for perch were very close to the mean consumption rate. PFOS accumulation in humans has been associated with multiple hepatotoxic, neurotoxic, reproductive, immunotoxic and thyroid disruptive effects, which could lead to severe diseases and even death (as reviewed by . Even at very low concentrations, PFAS can alter the lipodome, disrupting lipid and weight regulation . The Flemish government, however, already discourages consumption of eel and other predatory fish from Flemish waterbodies due to high concentrations of other pollutants (e.g. PCBs) .
Ecological health risk
Under the Water Framework Directive (WFD) the EU  defined Biota Quality Standards (EQSbiota) for freshwater, threshold concentrations for protection of the integrity of aquatic ecosystems and specifically for prevention of secondary poisoning and human health risk. For perfluoroalkyl substances, 9.1 µg PFOS/kg ww was set as the (human health based) threshold (EQSbiota, hh). For eel and perch, respectively, 44% and 58% of sampling locations exceeded the EQSbiota, hh, indicating potential health risks to the food web and to top predators (including humans) through fish consumption. In a German monitoring study, PFOS was above the EQSbiota,hh in 33% of the locations in perch muscle [64, 65]. However, the current EQSbiota, hh is based on the previous EFSA tolerable daily intake (TDI) value for PFOS of 150 ng/kg body weight, which can be converted to 10.5 µg per day considering a 70 kg person [26, 56]. This value is more than 200 times higher than the sensitive EFSA group TWI value  used in the human health risk determination in the present study. Furthermore, the EQSbiota was calculated considering a mean European daily fish consumption of 115 g , while Belgium is known to have a lower fish consumption compared to other European countries . All this leads to the conclusion that the current EQSbiota for PFOS might underestimate the risk for human health consequences through fish consumption, especially for Belgium, and needs to be revised.
On the other hand, the higher EQS of PFOS of 33 µg/kg ww  was determined specifically for protection of top predators against secondary poisoning (EQSbiota, secpois). Comparison to this standard resulted in an exceedance for 7% of the sampling locations for eel and 15% for perch. It was, however, stated that when determining the risk for secondary poisoning it is more appropriate to use whole fish measurements instead of fillet . An average conversion factor between both matrices of about 3 was determined for perch [64, 65, 81]. This would increase the exceedance in perch to 45% of all locations.
Furthermore, the setup of our study is in line with general recommendations for biota monitoring under de WFD . All biota used in the present study are considered good biomonitor species. However, in order to estimate the risk for secondary poisoning, taking into account biomagnification effects, the use of top predators (TL of 4 in freshwaters) is recommended. Both fish species included in the present study could be classified as such (TLperch: 4.97 ± 0.15; TLeel: 4.86 ± 0.14). Furthermore, their widespread (European) occurrence and limited home range allow for good monitoring practices [7, 81]. Although, within the present study a limited size range was targeted, differences in ranges and mean fish sizes between locations were detected . This might affect the mean PFAS concentrations per location and comparison between locations, since accumulated concentrations increase with size and age . As stated before, due to the high affinity of PFAS for proteins, liver tissue might have been a better matrix for sole monitoring purpose. However, since human health risk assessment was an important focus of the present study, muscle tissue was considered a more appropriate matrix. Finally, a standardization of hydrophobic compounds was proposed in the Guidance Document . For mercury and perfluorooctane sulfonate (PFOS), which do partition to proteins in contrast to the other lipophilic priority compounds, a standardization to a default dry weight fraction of 26% was recommended. This approach was not included in the present study. However, the standardization had a very limited effect on or even increased the variability of measured concentrations [81, 71]. Furthermore, Valsecchi et al.  reported that dry weight standardization, as a proxy for protein content, for PFOS is inappropriate because PFAS bind to specific proteins.