To investigate the single and combined effects of Ag and P concentrations on biofilm quality, a diatom-dominated biofilm (diatoms representing 85–90% of algal biomass, Leflaive et al., 2015) was collected in the field. This biofilm was then exposed for 3 weeks in a full factorial design to a gradient of silver concentration at three distinct P concentrations. Impacts on microbial community structures were investigated independently from the present experiment, results being fully available in Leflaive et al. (2015). The present study specifically investigates the impacts of P and Ag stressors on biofilm fatty acids profiles, elemental composition, and silver concentration, giving an evaluation of biofilm potential quality for consumers. To measure the real/effective quality of biofilms, we fed a model consumer, Gammarus fossarum, with the distinct freeze dried biofilms, and followed organisms survival and growth throughout a 42-days experiment.

All details of the biofilm biomass production can be found in Leflaive et al. (2015). Briefly, the diatom-dominated biofilm was collected in a mesotrophic reservoir (Lake Saint-Ferréol, South-West France) using polyethylene plates (10 cm × 5 cm) placed vertically, 60-cm deep in the lake. Substrates were immerged during 3 weeks (May–June), then brought back to the laboratory in a cool box and placed two by two in beakers filled with 500 mL of modified COMBO medium (Kilham et al., 1998). The experiment was carried out in a controlled culture chamber (18°C, light intensity between 50 and 60 μE m^-2 s^-1, 16 h/8 h light/dark cycles). Twelve conditions were tested: three P levels (20, 100, or 500 μg/L) and four Ag concentrations (0, 5, 50, or 150 μg/L), each treatment being replicated three times (36 microcosms in total). Phosphorus concentrations were chosen as representative of what can be found along a gradient of P pollution in aquatic ecosystems (EEA, 2014). Concerning silver concentrations, total Ag concentrations have been shown to vary in natural waters between 0.03 and 500 ng L^-1 (Luoma, 2008), but these concentrations are expected to increase in the future (Geranio et al., 2009; Gottschalk et al., 2009). To maintain levels of silver close to what could be found in contaminated rivers in our semi-continuous exposure conditions, we chose to expose biofilms to a 5 μg Ag L^-1 concentration. The 50 and 100 μg Ag L^-1 concentrations were retained as acute contamination levels, to evaluate the tolerance limits of biofilms to silver exposure. To limit too strong changes in our exposure conditions, media were entirely renewed each week. For that purpose, we emptied the beakers and re-filled them with newly prepared medium with the metal at the appropriate concentration. To limit silver adsorption on beakers surface, glass beakers were pre-saturated during 24 h with Ag solutions at concentrations similar to those of the diverse treatments.

Just after the introduction of biofilms in beakers (t₀) and at the end of the experiment (after 3 weeks of biofilm growth, t_final), 5 mL of medium from each beaker (n = 3 replicates per treatment, 36 samples in total per sampling date) were sampled, acidified with 15 μL of 70% HNO₃ and stored at 4°C for later Ag quantification. At t_final, biofilms present on polyethylene substrates were scrapped, then homogenized in 20 mL of incubation medium. Some subsamples were taken for measuring bacterial and micro-eukaryotic community structures. In particular, 1 mL was fixed (2% formaldehyde) and kept at 4°C for algal determination and counts (see Leflaive et al., 2015 for community structure results). The remaining volume of suspension was concentrated (centrifugation 7,000 × g, 10 min), supernatant was eliminated and the biofilm was stored at -20°C until being freeze-dried.

Silver concentrations in culture media was measured on t₀ and t_final samples by atomic absorption spectrophotometry (graphite furnace Varian SpectrAA 300, detection limit = 0.1 μg Ag/L). The concentration of silver in the biofilm was measured after acidic digestion with 1 mL bidistilled HNO₃ (15 N) and 1 mL H₂O₂ (30%, Sigma) for 24 h at 70°C. After full digestion, the samples were evaporated at 70°C and redissolved by 2% bidistilled HNO₃ for analysis by ICP-MS (Agilent 7500 ce) with an uncertainty of 5% and a detection limit of 0.001 μg L^-1. A reference standard solution SRM1646a (NIST, United States) was used to certify the accuracy and precision of the analytical procedure. The data quality was assessed by comparing the certified and the reference standard solution in terms of recovery (%), and by checking the precision of the ICP-MS analysis by the relative percentage differences (RPD) and the relative standard deviation (RSD) among the reference material replicates. The recovery for Ag was higher than 84.5% for all the samples. The precision of the instrument was within the 10% of RSD, thus was acceptable. All the materials used for silver sampling and measurements were cleaned with 1 N HCl for 24 h. Due to the cost of Ag analyses, measurements were done for each treatment on a pooled sample of biofilm corresponding to a mix of equal quantities of the three replicates (12 samples, i.e., 3 P-levels × 4 Ag concentrations). Initial PO₄^3- concentrations were measured spectrophotometrically (ammonium molybdate method, AFNOR, 1990) in culture media used for microcosms filling.

Freeze-dried biofilm was first gently homogenized in a plastic centrifugation tube using a glass pestle. The C and N content of biofilm was measured on the 36 samples (3 P-levels × 4 Ag concentrations × 3 replicates) on ca. 1 mg samples using a CHN elementary analyzer (Carlo Erba NA2100, Thermo Quest CE International, Milan, Italy). Biofilm P content was quantified on the same number of samples after alkaline digestion with persulfate, and mineral P was then measured spectrophotometrically following the AFNOR (1990) procedure. Results are expressed as the mass percentage of the element in different resources, and C:N:P ratios correspond to molar ratios.

Fatty acids analyses were performed on the 36 dried biofilm samples (3 P-levels × 4 Ag concentrations × 3 replicates). Analyses followed the procedure described in Crenier et al. (2017). Briefly, lipids were extracted twice with a chloroform/methanol solution according to the method proposed by Folch et al. (1957). Fatty acids were then converted into fatty acid methyl-esters (FAME) by acid catalyzed transesterification and analyzed on an Agilent Technologies^TM 6850 gas chromatograph. FAME were identified after a comparison of retention times with those obtained from Supelco^® and laboratory standards, and were quantified using internal standards (13:0).

All parameters (Ag concentration in water and biofilm, biofilm %C, %N, %P, C:N, C:P and N:P ratios, fatty acid contents, Gammarus fossarum growth rates) except survival curves were analyzed using two-way ANOVAs, considering P level and Ag concentrations as categorical predictors. Data were log transformed when necessary in order to meet the variance homoscedasticity condition for using ANOVAs. When ANOVA indicated significant differences between treatments, multiple comparisons were conducted using Tukey’s HSD test. Due to the elevated costs of analyses and the quantity of biological material required, Ag concentrations in biofilms were not replicated, rather measured on a pooled sample coming from the three replicates. The interactive effects of P and Ag were thus impossible to calculate, and only the main effects were analyzed. Survival were analyzed using Kaplan–Meier survival curves and log-rank tests for comparing curves. By accounting for censored data (i.e., when organisms were still alive at the end of the experiment), this approach enabled to investigate the effects of P and Ag during the whole experiment. Yet, since mortality remained null in several treatments of this experiment, it was statistically impossible to compare the 12 treatments. To make the analysis feasible, survival values were pooled by P levels, then by Ag levels. This permitted to compare the effects of P and Ag concentrations on gammarids survival, the analysis of the interactive effects of P and Ag being not possible. Survival curves analyses showing significant differences for both Ag and P treatments, the three curves obtained for P effect and the four curves obtained for Ag effects were compared pairwise. To take into account the multiple comparisons, p-values considered as significant followed a Bonferroni adjustment. Finally, impacts of resource quality parameters (biofilm C/P ratios, PUFA contents, and Ag concentrations) were analyzed using linear regressions. All statistical analyses were computed with STATISTICA (SAS institute). Statistical significance was inferred at p ≤ 0.05.

Despite significant initial differences in Ag concentrations in water between the four Ag treatments, Ag concentrations measured at t0 were systematically lower than nominal concentrations (differences with expected concentrations ranging from 1% in the Ag50 treatment to 37% and 48% in the Ag150 and the Ag5 treatments, respectively). This observation suggests that a part of the metal might have been very quickly adsorbed to the biofilm and/or to the microcosms walls at the microcosm filling. The lack of interaction with P level as well as the high recovery of dissolved silver in the culture medium (e.g., >80% for the 150 μg Ag L^-1, P500 treatment) suggests that a fast complexation and a precipitation in presence of phosphates, if it occurred, might have been reduced.

After 1 week of biofilm exposure to silver (t_final), silver concentrations were largely reduced when compared to what was initially introduced. These reductions could be partly attributed to silver uptake by biofilm microorganisms (algae, bacteria, fungi…), but also to the great potential of growing algal cells and biofilms to adsorb ionic metals on their surfaces. Indeed, adsorption processes are generally more common in algae than active uptake into the cells (Ratte, 1999). Adsorption has already been shown to increase with the presence of extracellular polymers (EPS), as already observed in previous studies (e.g., González et al., 2015, 2016), such substances presenting numerous potential binding sites (van Hullebusch et al., 2003). Since nutrient depletions are well known to stimulate the production of EPS by phototrophic biofilms (Lyon and Ziegler, 2009), an increase in Ag adsorption was expected in the P20 treatment when compared to the P500 treatment. However, in our study, lower P levels did not change significantly silver concentrations remaining in water. Similarly, even if we were unable to test for the interactive effects of P and Ag on silver concentration in biofilms (absence of replication for these chemical analyses), higher P concentrations in water were expected to lead to reduced Ag concentrations in biofilms. Again, this general trend was not clearly visible, and complementary studies will be required to understand the mechanisms explaining the changes in silver concentrations, both in the water column and in the biofilm.

Elemental compositions of resources are generally considered as good proxies of resources quality for consumers (Sterner and Elser, 2002). Indeed, life history traits of several species have been shown to be controlled by the imbalance between their elemental requirements and what they can effectively find in their resources (e.g., Elser et al., 2001; Frost and Elser, 2002). Since C proportions in resources is rarely limiting for consumers, resources presenting low C:N and/or low C:P ratios are generally considered as potentially high quality resources for their consumers (Sterner and Elser, 2002). Our study revealed that both biofilm C:P and N:P ratios were significantly reduced by the P concentration of the culture medium. This effect might be simply explained by the fact that algae have the potential to immobilize nutrients in excess in their biomass, this process being generally called luxury consumption (Droop, 1974). This effect occurs concomitantly with a reduction of the C-rich EPS production (Lyon and Ziegler, 2009), these compounds generally largely contributing to increasing algal communities C:nutrient ratios (Pannard et al., 2016). In addition, bacterial communities have the potential to quickly change, selecting for species able to use optimally nutrients available and adjusting the stoichiometry of the whole community to that of their resources (Danger et al., 2008). Unexpectedly, silver contamination also led to significant reductions of biofilm C:N and C:P ratios, even if these reductions remained low when compared to those observed along the P gradient. Data on the effects of contaminants on resources elemental compositions remain extremely scarce in the literature. Some previous studies showed that biofilms tend to increase their production of C-rich EPS as a response to contaminant exposure (García-Meza et al., 2005; Serra and Guasch, 2009; González et al., 2016). This physiological response, aimed at increasing the metal-binding sites of the biofilm, would thus be expected to increase biofilm C:nutrient ratios (Pannard et al., 2016). In our study, the opposite effects were observed. This result might thus alternatively be explained by the large Ag-induced shifts in prokaryotic and microeukayotic community structures observed in our study (see Leflaive et al., 2015). One could expect that different species with different luxury consumption capabilities might be selected by Ag contamination. Another potential, non-exclusive, explanation could be that organisms stressed by the Ag contamination tend to increase their antitoxic defenses, these defenses generally relying on N-rich enzymes and leading to apparent lower biofilm C:N ratios.

Some fatty acids, and especially long-chain PUFAs have been described as pertinent indicators of resources quality. Indeed, most metazoans are unable to (or, at least, not in sufficient amounts) synthesize these essential compounds that must be found in their diet. These compounds are notably involved in the synthesis of key hormones and represent important molecules in cell membranes, controlling in particular membrane fluidity (Arts et al., 2009). In algal communities, diatoms are known to produce high amounts of long chain PUFAs (20:5ω3, Dunstan et al., 1994) in comparison with green algae and cyanobacteria that mainly produce shorter chain PUFAs (Masclaux et al., 2009) or are unable to synthesize highly unsaturated fatty-acids (Muller-Navarra et al., 2004), respectively. In the present study, both Ag contamination and P concentrations greatly altered algal communities, reducing diatoms proportion and increasing green algae proportions (see Leflaive et al., 2015, Figure 1). Both P and Ag stressors led to similar effects, without acting interactively. Thus, reductions in biofilm PUFA and 20:5ω3 content were expected for both stressors. In the present study, only the P increase led to significant reductions in these compounds. Surprisingly, we did not observe any significant effect of silver on biofilm PUFA and 20:5ω3 contents. In contrast, silver contamination significantly modified biofilm SAFA and the total amount of fatty acids, these parameters reaching their minimal values for an exposure to 5 μg Ag L^-1. The effect of P level on biofilm fatty acid profiles can certainly be explained by the replacement of 20:5ω3-rich diatoms by green algae, as already observed in diverse studies dealing with natural communities (e.g., Muller-Navarra et al., 2004; Bec et al., 2010). The absence of silver effect on biofilm PUFA and 20:5ω3 contents, despite the replacement of diatoms by green algae, could be explained either by the replacement of some diatom species by other diatom species containing more PUFAs or by an increase of diatoms’ PUFA content as a response to Ag contamination. However, Ag and P stressors led to similar changes in biofilm community compositions (Leflaive et al., 2015), both communities showing similar reductions in the abundance of the dominant diatom species (Achnanthidium minutissima, Kützing and Cymbella excisa, Kützing). The second explanation thus appears as the most probable. Even if data remain scarce in the literature, such an effect of fatty acid synthesis deregulation was already found in the macroalgae Fucus sp. exposed to copper (Smith et al., 1985). Similarly, some studies showed that fatty acid profiles of unicellular organisms (Green algae: McLarnon-Riches et al., 1998; Diatom: Jones et al., 1987; Euglenophyceae: Rocchetta et al., 2006) were susceptible to change after an exposure to metals. Yet, there is still no consensus on the toxic metals effects on algae fatty acids profiles, some studies showing reductions in long chain PUFAs due to reductions in their abundance or synthesis caused by oxidative stress (Rocchetta et al., 2006), while other studies showed increases in algal PUFAs after alterations of enzymatic activities (McLarnon-Riches et al., 1998). The precise understanding of silver impacts on biofilm fatty acid contents would greatly benefit from specific investigations of fatty acid synthesis on algal communities.

While biofilm Ag, nutrients, and PUFA concentrations remain only potential quality parameters for biofilm consumers, measuring the impacts of biofilms consumption on metazoans is the only way for evaluating the effective quality of biofilms, and anticipate the indirect effects of the multiple stressors selected (Ag and P) on higher trophic levels. Consumers’ growth measurements have been shown as a powerful mean for evaluating resources quality. In particular, effects of resources stoichiometry and/or highly unsaturated fatty acids on consumers’ growth have already been successfully tested on planktonic and terrestrial herbivores (e.g., Elser et al., 2001; Schade et al., 2003; Masclaux et al., 2009). More recently, such effects have been revealed in the crustacean species, Gammarus fossarum, both for the effect of resources P (Danger et al., 2013) and highly unsaturated fatty acid contents (Crenier et al., 2017). In the present study, biofilm C:P ratios were significantly related to G. fossarum growth, growth being reduced when organisms were fed with the highest C:P ratio resources. Organisms P content being directly related to organisms’ nucleic acid production and cell proliferation (Elser et al., 2003), eating low C:P resources can help organisms to overcome P limitations of their growth. Changes in microbial biomass PUFA and 20:5ω3 contents induced by biofilm exposure to high P concentrations, while significant, were too small for detecting any significant effect on consumers’ growth. Finally, contrary to expectations, accumulation of Ag in biofilm biomass had strictly no influence on consumers’ growth. Yet, some studies showed that toxicity of Ag on zooplankton species could be higher via trophic transfer than by direct uptake from water (Hook and Fisher, 2001; Bielmyer et al., 2006). In contrast, other studies suggested that Ag toxicity was very variable depending on water chemistry, and that free ionic silver was the most toxic form to invertebrates (Ratte, 1999). In our study, elemental quality of resources seemed thus to be more important than biofilm silver concentration for G. fossarum. However, it must be noted that responses could have been different if other life history traits had been considered. For example, PUFA concentrations have been regularly shown to control organisms’ reproduction (e.g., Masclaux et al., 2009), and multiple stressor effects might also impact organisms reproduction.

Finally, it must be noted that in contrast to the observed stimulation of G. fossarum growth, our results also showed a low but significant negative effect of P on organisms’ survival. Such an effect has already been observed in another study (Arce-Funck et al., 2018). It was proposed that higher growth rates generated by higher resource quality increases molting frequency. Yet, molting is by far the most sensitive stage of molting organisms’ development, especially when exposed to contaminants (McCahon and Pascoe, 1988). Stimulation of organisms’ growth thus generally co-occur with an increase in organisms’ mortality. In contrast, the highest survival found for Ag-fed gammarids is more difficult to explain. One could imagine a stimulation of immune defenses of gammarids or a reduction of potentially pathogenic bacteria growing in biofilms. Such effects yet remains to be tested.