Multiple stressors in periphyton – comparison of observed and predicted tolerance responses to high ionic loads and herbicide exposure

Authors


Summary

  1. As a result of the increasing human impact on aquatic ecosystems, freshwater organisms are often exposed to multiple stressors simultaneously. The joint actions between stressors can result in combined effects and unexpected ecological effects. Therefore, a better understanding of the interactive effects on ecosystems is required.
  2. This study aimed to identify potential interactions between high ionic loads and herbicides. A microcosm study, using periphyton as model community, was conducted with a factorial design. Two levels of ionic loads were used as single stressor and in combination with prometryn. Structural (biomass, algal class and diatom composition) and functional parameters (tolerance development) were determined over a growth period of 6 weeks. The concept of pollution-induced community tolerance (PICT) was applied to quantify integrated community responses. Long-term community responses to the combined exposure were predicted using the model of independent action.
  3. No co-tolerance of high ionic loads and prometryn or vice versa was found. Stress-induced succession resulted in a distinct community structure for each stressor and combination of stressors. Multiple stressors led to the selection of opportunistic species and higher tolerances to prometryn than predicted by the model of independent action. However, joint effects for high ionic loads and prometryn were concentration and time-dependent. The PICT concept enabled the quantification of community-level effects in systems receiving multiple stresses.
  4. Synthesis and applications. Multiple stressors might explain the failure to achieve good ecological status for many European water bodies within the context of the EU-Water Framework Directive (WFD). We propose PICT as a diagnostic tool for investigative monitoring to clarify stressor conditions by testing the tolerances of local communities to preselected site-specific compounds.

Introduction

Aquatic organisms and biological networks are often exposed to diverse natural and anthropogenic stressors simultaneously. Interactions among multiple stressors might be more commonly non-additive than simple additive, which might cause ‘ecological surprises’ (Darling & Côté 2008). The type of interactive effect is dependent on stressor pairs, trophic level and biological response levels, for example, species interactions within communities might dampen or exacerbate the impacts of multiple stressors (Crain, Kroeker & Halpern 2008). Thus, there is a clear need for concepts and appropriate approaches to analyse, quantify and predict community-level responses to multiple stressors (Breitburg et al. 1998).

For chemical mixtures, two basic concepts have been developed for calculating the expected combined effect based on known responses of the individual toxicants: (i) concentration addition for compounds with a similar mode of action and (ii) independent action for dissimilar-acting compounds (Faust et al. 2003). While concentration addition formalizes the dilution principle, independent action is based on statistically independent effects. Several studies successfully applied these concepts to chemical mixtures (Altenburger & Greco 2009) and combinations of different stressors (Coors & De Meester 2008), but were mainly applied in single species tests.

Backhaus, Arrhenius & Blanck (2004) have taken a first step by applying the principles of mixture toxicity at a community level, using a short-term community assay. However, there is a lack of long-term studies considering integrated community parameters, such as community tolerance based on species shifts, for prediction. One reason for this might be the absence of suitable community approaches that allow for a quantification and prediction of long-term community-level responses.

The concept of pollution-induced community tolerance (PICT) was introduced by Blanck, Wängberg & Molander (1988) and enables quantitative measures of community-level responses by causally linking exposure to community-level effects. Tolerance is the result of adaptation processes, which maintain overall functions within organisms (physiological acclimatization) and communities (species shifts). The PICT approach is based on the assumption that chronic exposure to toxicants entails changes in species composition, known as toxicant-induced succession (Blanck 2002). The reliability of PICT has been demonstrated for several compounds in microcosm and field studies using periphyton (e.g. McClellan, Altenburger & Schmitt-Jansen 2008; Rotter et al. 2011), but few studies address the effect of simultaneous stressors on periphyton (e.g. Muñoz et al. 2001; Tlili et al. 2010). However, under multiple stressor conditions one stressor might change the vulnerability of a biological system to a second stressor (De Lange et al. 2010). One single stressor might induce tolerance mechanisms triggering co-tolerances to similar-acting stressors or the exposure to several selecting stressors simultaneously might lead to the development of multiple tolerance mechanisms for the entire set of stressors (Blanck, Wängberg & Molander 1988).

Apart from pesticide contamination, the increasing secondary salinization of freshwaters has been recently described as one of the most important stressors for freshwater ecosystems, causing environmental risks world-wide (Cañedo-Argüelles et al. 2013). Furthermore, high ionic loads and pesticides often occur simultaneously in freshwater (Hart et al. 1991; Coring & Bäthe 2011; Von der Ohe et al. 2011) and revealed joint effects on macroinvertebrates (Schäfer et al. 2011). However, approaches quantifying community responses to multiple stressors, in order to enable the prediction of joint effects, are still lacking.

In this study, we address the question of individual and combined effects of salt and toxic stress on long-term community-level responses. Stressors were chosen according to a former field study of the Elbe River basin with the simultaneous presence of prometryn and high conductivity (Rotter et al. 2011). The herbicide prometryn inhibits the photosynthetic electron transport by competing with the plastoquinone QB for its binding sites at the D1 protein (Huppatz 1996). Therefore, algae exposed to prometryn have a diminished photosynthetic rate (Brown & Lean 1995), which reduces the growth rate. In contrast, environments with increased ion concentrations lead to disturbed ionic steady states of Na+, Cl, K+ and Ca2+ in organisms (Hasegawa et al. 2000). The loss of water activates several secondary acclimation processes, which increase the demand for energy supplying reactions (Fodorpataki & Bartha 2004). Based on these considerations, we assume dissimilar modes of action of prometryn and high ionic loads.

We hypothesize that (i) exposure to high ionic loads, herbicides and the combination of both triggers a stressor-specific replacement of species resulting in (ii) enhanced tolerance to prometryn in single and combined prometryn exposures. However, due to dissimilar modes of action we (iii) do not expect co-tolerance between high ionic loads and prometryn. The objective of this study was to quantify the single and joint effects of both stressors on changes in community tolerance in order to enable the prediction of joint effects using the mixture toxicity model of independent action. For a mechanistic understanding of tolerance development in multiple stressor environments, a microcosm study in a factorial design was conducted, and structural changes of periphyton (biomass, algal class and diatom composition) were related to community tolerance development under control conditions as well as under single and combined stressors. Periphyton pre-exposed to one single stressor was used to investigate co-tolerance. While the tolerance to prometryn was tested in short-term toxicity tests, the index of salinity was used as measure of tolerance to high ionic loads.

Materials and methods

Periphyton Cultivation

Microcosm experiments were conducted over a period of 6 weeks. Periphyton was used as model community and was grown on glass discs (1·5 cm diameter) according to Blanck (1985). Unfiltered freshwater from the River Parthe (Germany) was used as inoculum to obtain a naturally sourced community of algae. River monitoring showed no prometryn contamination, conductivity values around 800 μS cm−1 and low pesticide levels within the nanogram range (Sächsisches Landesamt für Umwelt, Landwirtschaft und Geologie 2013). The light regime was set to a 14:10 h light/dark cycle using neon lamps (36 W; photosynthetic photon flux density above water surface = 200 μmol m−2 s−1), cultivation temperature was approximately 22 °C, and water was stirred continuously. In total, 14 microcosms each containing 14 L of water and 200 vertically exposed glass discs were used. Water was replaced weekly, and pH, oxygen and conductivity were measured in the fresh river water and in aquaria directly before water exchange.

Two aquaria without additives served as controls (Co) and two further containing 0·01% (v/v) of dimethyl sulfoxide (DMSO, CAS RN: 67-68-5; Merck, Darmstadt, Germany) were solvent controls (DMSO-Co). Two aquaria were exposed to a nominal concentration of 15 μg L−1 prometryn (Prom) (CAS RN: 7287-19-6; Riedel-de Haën, Seelze, Germany), which is known to cause an increase in tolerance (Schmitt-Jansen & Altenburger 2005). The stock solution of prometryn was dissolved in DMSO and was freshly spiked each week subsequent to the changing of water. Two levels of ionic loads were selected: the first corresponding to the former field study with a mean conductivity of 1919 μS cm−1 (Rotter et al. 2011), and the second was taken to mimic mean conductivities of the lower River Werra, which has been heavily polluted by direct discharges of salt wastewater from the potash industry since 1950 (Bäthe & Coring 2011). In order to increase ionic loads, sodium sulphate (CAS RN: 7757-82-6; Merck, Seelze, Germany) and calcium chloride (CAS RN: 10043-52-4; Merck, Seelze, Germany) were added in equal weight ratios. The salt concentrations were elevated to a final conductivity of about 2000 μS cm−1 (C2000) and 5000 μS cm−1 (C5000) in two aquaria each. Finally, four aquaria were exposed to mixtures of 15 μg L−1 prometryn and elevated conductivity, two microcosms to 2000 μS cm−1 (P2000) and two to 5000 μS cm−1 (P5000), providing a factorial design. The sodium and chloride concentrations in C5000 treatments were in the range of 24·6–31·8 mM NaCl (see Table S1, Supporting Information).

In the following, we refer to increased conductivity as high ionic load, because salinity is based on chloride concentrations, and no limiting nutrients such as ammonium, nitrate or phosphate were introduced to the system.

Chemical Analyses

A preliminary experiment was conducted to analyse the stability of prometryn over the exposure period of 1 week (see Appendix S1, Supporting Information for details). Due to minimal deviations from applied concentrations, we refer to nominal concentrations.

Furthermore, the physicochemical water parameters (temperature, pH, oxygen concentration, conductivity) of one control microcosm were measured directly after water renewal and after 7 days. Analyses for dissolved ions (Cl, math formula: ion chromatography; Ca2+, Na+: ICP-OES) were conducted following standard protocols.

Biological Analyses

Short-term toxicity tests for quantifying PICT

Community tolerance was determined by short-term inhibition tests of photosynthesis measured by a pulse-amplitude modulation (PAM) – fluorescence-based method. After 3, 4, 5 and 6 weeks of growth, 24 colonized glass discs of each microcosm were sampled. Tests were performed in 24-well plates, each containing 2 mL of freshwater. Periphyton was exposed to six concentrations of prometryn in triplicate, ranging from 3 × 10−5 to 30 mg L−1 dissolved in DMSO. Samples without contamination and samples containing 0·1% DMSO served as controls. Short-term tests were incubated on a rotary shaker under a photosynthetic photon flux density of 110 μmol m−2 s−1 at 20 °C. After 1 h of exposure to prometryn, chlorophyll a fluorescence was measured using a MAXI-Imaging-PAM (Walz, Effeltrich, Germany).

Prometryn tolerance was quantified as EC50 (effect concentration at median efficacy) for inhibition of photosynthesis. Relative inhibition of fluorescence in relation to controls was calculated to model concentration–response relationships, according to Seefeldt, Jensen & Fuerst (1995).

display math

where Amin and Amax denote the minimal and maximum responses, x the concentration and p stands for the slope. After data inspection, Amax was fixed at 100% and Amin at zero. Analyses were performed using nonlinear regression least squares curve fitting with the software OriginPro 8G (Microcol Software Inc., Northhampton, MA, USA). EC50 data were tested for outliers using Grubbs’ test (Grubbs 1969).

For prediction of joint effects, the concept of independent action was used. The equation was formulated following Backhaus, Arrhenius & Blanck (2004)

display math

where EMix represents the expected effect (scaled between 0–1) of an n-compound mixture and E(ci) the individual effect of each stressor i. The transformation of the observed effects to proportional data and the rescaling of predicted EMix to absolute values were conducted according to Coors & De Meester (2008)

display math

whereas ei is the observed effect in absolute units, econtrol is the mean of all controls (DMSO-controls and controls) per week. The maximum possible effect emax was defined based on the maximal observed EC50 of this experiment (41·13 mg L−1). More details and an example calculation of the prediction are shown in Appendix S2 (Supporting Information). We consider interactions between stressors present where the predicted EC50 lies outside of the standard error for observed effects of combined stressors.

Algal class composition and biomass

Fluorescence-based measurements of photosynthesis pigments were conducted to analyse algal class composition and biomass of periphyton. The PHYTO-pulse-amplitude-modulated fluorometer (Heinz Walz GmbH, Effeltrich, Germany) allows discrimination by multiwavelength-excitation and was used according to Schmitt-Jansen & Altenburger (2008). Minimal fluorescence (F0) was measured after 3, 4, 5 and 6 weeks for three spots per glass disc with three replicates taken from each microcosm.

Taxonomic analyses of diatoms

For identification of diatom species, seven discs per microcosm were preserved in 5% formaldehyde. Samples were taken after 3, 4, 5 and 6 weeks of cultivation. Preparation of the samples was conducted according to Krammer & Lange-Bertalot (1986). Clean diatom samples were used to count and to identify at least 400 valves per sample using light microscopy and taxonomic literature of Krammer & Lange-Bertalot (1986).

The index of pollution sensitivity (IPS) of each sample was calculated using omnidia software v8.1 (Lecointe, Coste & Prygiel 1993). Furthermore, Index of Salinity (Ziemann 1999), Shannon Diversity Index and Evenness (Mühlenberg 1989) were calculated.

Statistical Analyses

A repeated measures anova was conducted to analyse the effects of high ionic loads, prometryn (between-subject factors) and time (within-subject factor) and their interaction on community tolerance, biomass and algal class composition (Von Ende 2001). Data of controls and DMSO-controls were pooled into one control group after statistical testing indicated that there were no significant differences between the two control groups (Table S2, Supporting Information). Normal distribution had to be assumed due to few replicates. Mauchly's test of sphericity was used to test for equal variances, Greenhouse Geisser correction was applied if assumption of sphericity was violated. For significant anova results (< 0·05), the source of the differences was located using the Tukey's post hoc test, comparing all possible pairs of groups. The Bonferroni-adjusted P-value for multiple testing was applied for responses of cyanobacteria, green algae and diatoms. Statistical analyses were conducted using the software spss v19 (IBM, Chicago, IL, USA) and SigmaPlot v11 (Systat Software Inc., San Jose, CA).

Canonical ordination analyses were conducted to relate the effects of prometryn, high ionic load and time to the species composition. Diatom species with a relative abundance below 1% were excluded from analyses. Species abundances were transformed using Hellinger transformation (Legendre & Gallagher 2001). Furthermore, Cocconeis pediculus had to be excluded, because it was either present in high numbers or absent, which led to an imbalanced analysis. A Detrended Correspondence Analysis on the diatom data revealed a linear gradient, which requires a Redundancy Analysis (RDA). RDA was carried out on the reduced diatom data set of 31 species against the variables conductivity, prometryn and time. Statistical significance for RDA axes and environmental parameters were assessed using permutation test with 999 random permutations. Multivariate analyses were carried out using vegan 1.17-9 of the R statistical environment (R Development Core Team 2010).

Results

Physicochemical Water Parameters

The actual prometryn concentration in the microcosm was constant during 1 week (mean = 14·0 ± 1·3 μg L−1) and matched well with the nominal concentration of 15 μg L−1. Water temperature in microcosms was stable (22·1 ± 0·9 °C), pH increased from 8·03 ( ± 0·43) to 8·38 ( ± 0·16), oxygen concentration (8·76 ± 0·49 mg L−1) decreased by 9% and mean conductivity of 650 ( ± 191) μS cm−1 decreased by 7·5% to 601 ( ± 120) μS cm−1 within 1 week in control microcosms.

Community Tolerance

According to the Grubbs' test, nine of 56 EC50 values were identified as outliers. A total of five of these values were excluded from statistical analysis, because maximum inhibition of the corresponding concentration–response curve was <60%, which prevents the accurate modelling of EC50 values.

All factors (ionic load, prometryn and time) significantly increased community tolerance and showed significant interactions among each other (Table S3, Supporting Information). The controls and DMSO-controls showed similar sensitivities to prometryn (mean EC50 = 2·2 ± 1·5 mg L−1), which were stable for the duration of the experiment (Fig. 1; exact values are shown in Table S4, Supporting Information). Sensitivities of treatments with high ionic loads (C2000 and C5000) were in the same range as controls (mean EC50 = 2·6 mg L−1). In contrast, all prometryn-exposed communities showed an increase in EC50 over time. Thus, Prom treatments were significantly more tolerant than controls from the fifth week (Tukey's Test, P < 0·001). The community tolerance of the combined treatments P2000 was in the range of single Prom treatments from the fourth week and was significantly higher than controls in week 5 (Tukey's Test, P = 0·01). Apart from the sixth week, the predicted EC50 for P2000 was in the range of the observed values.

Figure 1.

Effective concentrations (EC50) as a measure of community tolerance to prometryn after 3 - 6 weeks of exposure. Two replicate microcosms were conducted per treatment (Co = control; DMSO-Co = dimethyl sulfoxide (solvent) control; C2000 = 2000 μS cm−1, C5000 = 5000 μS cm−1, Prom = 15 μg L−1 prometryn; P2000 = prometryn + 2000 μS cm−1; P5000 = prometryn + 5000 μS cm−1). EC50s were calculated from concentration–response relationships (log-logistic data analysis) based on photosynthesis inhibition. Error bars represent standard errors of modelled EC50. Different lowercase letters indicate significant differences between treatments. Triangles are predicted EC50, based on the concept of independent action.

The combined treatments with high ionic loads (P5000) continuously increased EC50 values over time and developed tolerances beyond all other treatments (13 times higher EC50 than controls after 6 weeks). EC50s of P5000 were significantly higher than controls, C2000, C5000 and Prom from the fourth week and were significantly higher than all other treatments after 5 weeks (Tukey's Test, P < 0·001). The tolerances of P5000 treatments were clearly higher than predicted values (independent action model) from the fourth week, also reflected in a significant interaction of high ionic loads, prometryn and time (P = 0·009).

Structural Parameters

Biomass

The minimal fluorescence (F0) was measured as indicator for algal biomass. The biomass of each treatment is shown in Fig. 2 for the third and sixth weeks. The whole data set was used for statistical analyses and is shown in Table S5 (Supporting Information). Whereas P5000 and C5000 significantly increased in biomass from the third to the sixth week (significant interaction between high ionic loads and time, P < 0·001), neither the effect of prometryn nor the combined effect of high ionic load and prometryn changed significantly with time (Table S3, Supporting Information). After 6 weeks of exposure, biomass of C2000 was in the range of controls, P2000 and Prom showed lowest level of biomass, and C5000 and P5000 doubled in biomass compared with controls.

Figure 2.

Biomass development of periphyton communities per treatment from the third to the sixth week. Data were derived from minimal fluorescence (F0 in relative units) measured by multiwavelength-excitation pulse-amplitude modulation fluorometry. Error bars represent standard deviations of three measurements of each three glass discs per microcosm. Different lowercase letters indicate significant differences between treatments after 6 weeks, no significant differences were found after 3 weeks. Co = control; DMSO-Co = dimethyl sulfoxide (solvent) control; C2000 =2000 μS cm−1, C5000 = 5000 μS cm−1, Prom = 15 μg L−1 prometryn; P2000 = prometryn + 2000 μS cm−1; P5000 = prometryn +5000 μS cm−1.

Algal classes

A detailed description of algal class composition is presented in Appendix S3 (Supporting Information). All communities significantly changed in their algal class composition over time, (Bonferroni-adjusted P-value < 0·017, see Table S3, Supporting Information). All controls and C2000 had similar compositions with equal relative proportions of diatoms, chlorophytes and cyanobacteria. After 3 weeks of growth C5000, Prom, P2000 and P5000 had slightly more diatoms and less chlorophytes than the control group. However, these treatments clearly developed in different directions within 6 weeks of colonization. After 6 weeks, P2000 and C2000 were similar to the control group and had about 70% diatoms and <5% cyanobacteria, whereas the high ionic load treatments (C5000 and P5000) were highly dominated by diatoms (>85%). However, Prom communities were dominated by diatoms (55%), chlorophytes were in the range of controls, but the relative abundance of cyanobacteria was significantly higher (23%) than in all other treatments (Tukey's test, P < 0·01).

Diatom community

A total of 104 diatom species were classified from 14 different microcosms covering 4 weeks. Values for community metrics (Shannon Weaver Index, species richness, evenness, salinity index and IPS) are shown in Table S6 (Supporting Information). The Shannon Weaver diversity, species richness and evenness tended to decrease from the third to the sixth week but showed no consistent differences between treatments. According to the salinity index, periphyton exposed to high ionic loads (C5000 and P5000) was α-oligohalobic and β-mesohalobic indicating low and moderate salt effects, respectively, whereas all other treatments were classified as typical freshwater (β-oligohalobic) after 6 weeks of colonization. Control communities had on average significantly higher IPS values than treated communities (unpaired t-test, d.f. = 54, t = 5·29, P < 0·001), indicating good ecological quality of controls. In contrast, C2000 and P2000 belong to the moderate and C5000, Prom and P5000 to the poor ecological quality class.

The RDA was carried out on three variables (conductivity, prometryn and time) and 31 common diatom species (species are listed in Table S8, Supporting Information). The first three axes of the RDA explained 45·1% of the variance in the diatom data. The result of the RDA is illustrated in Fig. 3 as a biplot of species data and treatment factors. RDA1 explains 23·6% of the variance and leads to a separation of prometryn-exposed communities (Prom, P2000, P5000) and non-exposed communities (Co, DMSO-Co, C2000 and C5000). The second axis of the RDA contributed an additional 14·9% to the explained variance and is characterized by the conductivity gradient, discriminating treatments with elevated ionic loads from other treatments. Six of the 31 analysed diatom species were clearly related to treatment conditions in terms of control conditions, prometryn exposure and increased conductivity. The species Cocconeis placentula (Ehrenberg), Cymbella silesiaca (Bleisch) and Achnanthes minutissima var. minutissima (Kützing) are clustered in the ‘control-area’. Fragilaria capucina var. gracilis (Hustedt) is related to prometryn as single stressor, whereas Nitzschia palea (Kützing) appears more with high ionic load/prometryn mixtures. Navicula halophila (Grunow) is related to RDA2 and correlates with high conductivity. According to the permutation test, RDA axes and treatment factors (conductivity, prometryn and time) were significant at P = 0·001 (Table S7, Supporting Information).

Figure 3.

Redundancy analyses (RDA) based on most abundant (>1%) diatom species of periphyton communities colonized in microcosms with different treatments and sampled after 3, 4, 5 and 6 weeks. For a better visualization, species in the centre are suppressed and species supporting the separation of the treatments are highlighted. For illustration, treatment groups are framed. Co = control; DMSO-Co = dimethyl sulfoxide (solvent) control; C2000 = 2000 μS cm−1, C5000 = 5000 μS cm−1, Prom = 15 μg L−1 prometryn; P2000 = prometryn + 2000 μS cm−1; P5000 = prometryn + 5000 μS cm−1.

Discussion

Response to Toxicant-Induced Stress

Our results support the first and the second hypothesis that prometryn exposure triggers a replacement of species resulting in an increased tolerance to prometryn.

Prometryn-exposed communities showed 50% less biomass than controls, reflecting the diminished photosynthetic rate. To survive in a polluted environment, species must develop tolerance mechanisms such as increased detoxification of compounds or alterations of the target site, for example the D1 protein (Eriksson et al. 2009). We found shifts not only in the species composition, but also in an enhanced abundance of cyanobacteria on the level of algal classes, which might indicate sensitivity differences in both structural levels. Previous studies on other PSII inhibitors found comparable results, for example Bérard, Leboulanger & Pelte (1999) reported that the cyanobacterium Oscillatoria limnetica was stimulated by atrazine exposure in the summer months. Also, Schmitt-Jansen & Altenburger (2005) found a 20% increase in relative abundance of cyanobacteria for isoproturon (PSII inhibitor) concentrations of 20–40 μg L−1.

On the species level, Fragilaria capucina var. gracilis showed the highest relative abundances in Prom treatments (43% after 6 weeks). Therefore, we assume F. capucina var. gracilis is a prometryn tolerating species. Also Guasch et al. (1998) found F. capucina var. vaucheriae as one of the dominating species in an atrazine polluted stream. According to our second hypothesis, prometryn-treated communities had significantly higher tolerance (four to seven times) compared with controls from the fifth week on. This indicates a time dependency in the development of community tolerance, which goes along with the succession process. The detected tolerance increase was in the range of other studies, which found an increase by a factor of three to six in prometryn-exposed communities in microcosms (Schmitt-Jansen & Altenburger 2005), and a seven times higher tolerance in pre-exposed field communities (Rotter et al. 2011). By impacting the physiology of sensitive species, toxicants such as prometryn clearly exert selective pressure in communities, resulting in toxicant-induced succession and an increase in the overall community tolerance.

Response to Ionic Stress

The responses to high ionic loads were concentration-dependent. The increase in conductivity to 2000 μS cm−1 did not affect the community at the level of algal classes (Fig. S1, Supporting Information), also the species composition of diatoms differed just slightly from controls (Fig. 3). Furthermore, community biomass was in the range of controls. These findings are in accordance with the gradient study from 200 to 3000 μS cm−1 of Leung, MacKinnon & Smith (2003), who found effects on the community composition of phytoplankton, but not on total biomass, indicating functional redundancy. C5000 communities were clearly affected by the high ionic load. After 6 weeks of colonization, the overall biomass of the C5000 community was almost twice that of controls. The increase in ions in our study stimulated periphyton, whereas the toxic input of prometryn reduced the photosynthetic rate and thus also biomass. This finding supports the perturbation theory of Odum, Finn & Franz (1979), considering a system as subsidized if ecosystem functions (e.g. biomass) are enhanced due to an input into a system, even if certain species may be stressed. Several studies analysing salt effects reported decreased biomass with increasing levels of NaCl (100–800 mM) (e.g. Fodorpataki & Bartha 2004). However, lower concentrations of NaCl (17–85 mM), which are comparable to C5000 in our study, increased the biomass of the green algae Botryococcus braunii (Rao et al. 2007). Due to equal or higher biomass than that of controls, we assume that C2000 and C5000 communities were not limited by osmotic or ionic stress.

Given the decrease in chlorophytes in C5000 treatments, we assume that chlorophytes of the community were less tolerant to high ionic loads than diatoms. The high relative abundance of N. halophila in C5000 communities and the associated correlation with C5000 treatments in the RDA indicate an elevated tolerance to high ionic loads. This is in accordance with Lange-Bertalot (2001) characterizing N. halophila as salt tolerant species, preferring waters with high electrolyte contents of various ion concentrations. According to our first hypothesis, enhanced ionic loads inhibit sensitive species (e.g. C. silesiaca) or algal classes, giving salt tolerant species the opportunity to increase their growth extensively, resulting in a specific community composition.

Apart from the measurement of tolerance to prometryn by using short-term tests, the index of salinity can be used as a metrics-based measure of tolerance to high ionic loads. Both approaches, based on species shifts due to the exposure to either toxicants (PICT) or salt (index of salinity), enable a ‘translation’ of structural changes into a nominal scale, which simplifies comparisons to other treatments or time points. The index of salinity indicated a clear development of salt tolerance in C5000 communities. Yang et al. (2003) found for a salinity gradient of lakes the highest changes in diatom assemblages between 1·5% and 1·9‰, which is the transition point from slightly oligosaline to eusaline lakes. For our study, 2000 μS cm−1 equates to 1‰ and 5000 μS cm−1 to about 2·8‰, which supports the findings of Yang et al. (2003) and might explain the clear difference between C2000 and C5000 diatom compositions. There might be a threshold of resistance for high ionic loads from which changes within communities become obvious at all structural levels.

High ionic loads as a single stressor did not induce an increase in tolerance to prometryn, and thus, no co-tolerance was found, supporting our third hypothesis. Also, prometryn exposure did not trigger an increase in salt tolerance, as measured by the index of salinity. As no co-tolerance was found for high ionic loads and prometryn, we confirm the expectation of dissimilar tolerance mechanisms for both stressors. This further indicates different qualities of selection pressure for ionic and toxic stress that might shift communities to different alternative stable states (Clements & Rohr 2009).

Combined Effects of Toxic and Ionic Stress

Our most important finding is that enhanced ionic loads in P5000 seem to stimulate productivity of tolerant species, compensating for the inhibitory effect of prometryn on biomass and inducing tolerances higher than predicted by the model of independent action. Interestingly, this was not found for P2000 indicating a concentration-dependent effect. Similar results were found for the simultaneous presence of 150 mM NaCl and copper, leading to higher biomass production of Scenedesmus opoliensis as compared to single treatments of copper (Fodorpataki & Bartha 2008). Crain, Kroeker & Halpern (2008) stated that nutrients and toxicants as well as salinity and toxicants have opposing effects and interact antagonistically, pointing out that the positive effect of nutrients or high ionic load can mitigate or overcompensate for the negative effect of toxins.

Compositional changes in communities revealed different mechanisms of species selection of single and combined stressors. Whereas single stressors led to a higher occurrence of specialists, combined stressors caused increasing numbers of omnipresent and opportunistic species, such as Nitzschia palea and Gomphonema parvulum. Furthermore, Fragilaria capucina var. gracilis was less abundant than in Prom treatments, and Fragilaria ulna acus almost disappeared in comparison with C2000 and C5000. Dissimilar-acting stressors may target a wider range of species than single stressors with similar mode of actions. Therefore, communities developed distinct species compositions for each specific stressor or stressor combination. Furthermore, the strong influence of elevated ionic loads was indicated by the diatom index IPS and the index of salinity categorizing C2000 and P2000 communities (moderate ecological class; β-oligohalobic) as well as C5000 and P5000 (poor ecological quality class; α-oligohalobic to β-mesohalobic) in the same functional classifications. According to the index of salinity, the high ionic load as single stressor and in combination with toxic stress led to an increase in salt tolerance.

Whereas the single stress of prometryn increased the tolerance to prometryn up to a factor of eight, tolerance levels of combined stress (P5000) were up to 16 times higher than that of controls and were clearly higher than predicted by the independent action model. This finding was supported by the statistical significant interaction of high ionic load and prometryn (P < 0·002). These elevated tolerances might be due to a higher biomass (limits the transfer into the biofilm), the species composition itself or physiological adaptations of species such as modifications at the receptor site of the stressor. Since both stressors were acting in parallel, one stressor might influence the tolerance of the other simply by selecting specific species. Functional redundancy within communities might also have led to the selection of opportunistic species with a high growth rate. This is in accordance with Howarth (1991), who postulated that ecosystems dominated by opportunistic species are more resistant to stress than those dominated by specialists. In comparison with controls, we did not find a clear loss in biodiversity, rather species characterizing control communities were no longer abundant in other treatments. This means that sensitive species were replaced by pollution tolerant opportunistic species (Howarth 1991).

An increase in tolerance clearly reflects serious stress responses and might indicate the loss of ecological resistance accompanied by the shift of communities to alternative stable states. Our results confirm the hypotheses that single and combined stressors trigger stressor-specific succession processes resulting in distinct species compositions and changes in community tolerance.

The determination of community tolerance causally links structural and functional changes to toxicant exposure and enables quantification of an integrated community response. This study revealed that the combined stresses of a toxicant and salt foster clear changes in the community composition accompanied by an increase in community tolerance, which exceeded tolerance levels of single stressors and model predictions. Using the PICT approach, this study conducted a first step in assessing the applicability of the independent action concept to long-term community responses in a multiple stressor environment. Our findings on combined effects of high ionic loads and pesticides also provide a first insight into effects of secondary salinization in multiple stress environments.

Management Implications

Multiple stressors might be one of the factors explaining the critical ecological status of many European water bodies. Therefore, a better understanding of stressor interactions is crucial for water quality assessment and the success of remediation activities. In unspecific stressor conditions, investigative approaches are required, enabling diagnosis and prioritization of impacts of stressors by causally linking exposure to effects. Although, several taxonomy based metrics, for example for eutrophication (trophic diatom index) or high salinity (index of salinity) exist, no suitable metrics exists for toxicants. Thus, diagnostic tools based on other criteria such as toxicant-induced community tolerance may support decision-making within the Water Framework Directive (WFD). If biological quality elements (e.g. diatom metrics) implemented in EU-WFD refer to a poor or bad ecological status () and chemical analyses point to various contaminations and stressors, PICT might be a tool to identify the worst triggering compounds. The tolerance of local communities towards site-specific stressors could be tested for a selection of candidates, in order to find the most adverse compounds. This might enable an adaptation of remediation processes and an increase in remediation success. We propose PICT due to its specificity, causality and the ability to quantify community-level responses as an advanced tool for ecological effect assessment in multiple stressor environments.

Acknowledgements

We would like to thank Judith Römer for analysing diatom samples. Thanks are also to Isabell von Rein and Iris Christmann for technical assistance. This study was performed within the Helmholtz Impulse and Networking Fund through the Helmholtz Interdisciplinary Graduate School for Environmental Research (HIGRADE) and CITE.

Ancillary