In vitro bioanalytical evaluation of removal efficiency for bioactive chemicals in Swedish wastewater treatment plants

Chemical contamination of wastewater is a problem of great environmental concern, as it poses a hazard to both the ecosystem and to human health. In this study, we have performed a bioanalytical evaluation of the presence and removal efficiency for bioactive chemicals in wastewater treatment plants (WWTPs), using in vitro assays for toxicity endpoints of high relevance for human health. Water samples were collected at the inlet and outlet of five Swedish WWTPs, all adopting a treatment technology including pretreatment, primary treatment (sedimenation), seconday treatment (biological processes), post-sedimentation, and sludge handling. The water samples were analyzed for cytotoxicity, estrogenicity, androgenicity, aryl hydrocarbon receptor (AhR) activity, oxidative stress response (Nrf2) and the ability to activate NFĸB (nuclear factor kappa-light-chain-enhancer of activated B cells) signaling. We observed clear androgenic and estrogenic activities in all inlet samples. Androgenic and estrogenic activities were also observed in all outlet samples, but the activities were lower than the respective inlet sample. AhR activity was observed in all samples, with higher activities in the inlet samples compared to the outlet samples. The removal efficiency was found to be high for androgenic (>99% for two plants and 50–60% for two plants) and estrogenic (>90% for most plants) compounds, while the removal efficiency for AhR-inducing compounds was 50–60% for most plants and 16% for one plant.

Water sample extraction. The extraction of the water samples (2.5 L) was conducted using an automatic solid phase extraction system (SPE-DEX, Horizon Technology, Salem, NH, USA) using HLB extraction disks (Atlantic HLB-H Disks, diameter 47 mm; Horizon Technology, Salem, NH, USA). The disks were conditioned with 280 mL methanol and 420 mL Millipore water, and then the wastewater samples were loaded to the disks at a flow rate of 50 mL min −1 . After washing the disks twice with 24 mL 5% methanol in Millipore water, the disks were dried under vacuum for 30 min, and eluted 3 times with 25 mL methanol. The samples were then evaporated under a gentle nitrogen stream at 35 °C. Final extracts were in 0.5 mL ethanol. As an operational control blank sample, the same volume of Milli-Q water was concentrated by solid-phase extraction as described above.
The enrichment and dilution of the samples were expressed as relative enrichment factor (REF), calculated as described by Escher et al. 12 : The dilution and enrichment factors are calculated by the following equations: The initial water value was 2.5 L per sample which was extracted to a volume of 0.5 mL, leading to an enrichment factor SPE of 5000. When incubated with the cells, the concentrated water samples were diluted by 100 fold with the cell medium to get a final concentration of 1% ethanol and a REF of maximum 50. The concentrated water samples were then analyzed in a dilution series with lower REF values. A REF >1 means that the water sample is concentrated, while a REF <1 means that the water samples is diluted as compared to the initial water.
Bioanalysis. The concentrated water samples were analyzed in quadruplicate in seven bioassays, representing important toxicity endpoints. The applied bioassays are summarized in Table 2 and detailed information on the www.nature.com/scientificreports www.nature.com/scientificreports/ methods are provided in Supplemental Information (section SI-1). All cells used for this study were from already established cell lines and no animal experiments were conducted in this study. The operational control blank sample was analyzed in all bioassays and found to have no biological activity (data not shown).
Data evaluation. Bioactivities of water samples and positive controls were normalized to plate vehicle controls, set to 1. For nuclear receptor based assays, standard curves for positive controls were obtained by fitting data to a four parameter sigmoidal curve fit using GraphPad Prism 7. For Nrf2 and NFĸB, the standard curves for positive controls were based in linear regression using GraphPad Prism 7. Water samples were analyzed in dilution series to enable the calculation of effect concentration (EC) values, using linear regression in GraphPad Prism 7. Only data points within the linear range of the concentration-response curve was used for the calculation of EC values, as proposed by Escher et al. 26 . For nuclear receptor based assays, the effect concentration 20% or 50% (EC 20 and EC 50 ) were calculated. For Nrf2 and NFĸB activities, the effect concentration for induction ratio 1.5 (EC IR1.5 ) was calculated. The bioanalytical equivalent concentration (BEQ) for REF 1 for each sample was calculated using the linear range of the concentration-response curve and the dose-response relationship for the positive control, corrected for the dilution factor. Removal efficiency was calculated by comparing the BEQ value at REF 1 for WWTP inlet and outlet samples. For each bioassay, the limit of detection (LOD) was calculated as 1 plus 3 times the standard deviation (SD) of the normalized vehicle control values. For antagonistic assays, the LOD was calculated as 1 minus 3 times the SD. The LOD for each assay is presented in the Supplemental Information (Table SI-2). We could, however, observe concentration-response relationships in the reporter gene assay that could imply cytotoxic effects also at lower REFs. For example, a biphasic concentration-response relationship was observed for multiple samples with increasing concentrations leading to a decrease in the reporter activity at higher REFs ( Figure SI-2A). We hypothesize that the observed effect is due to cytotoxicity not detected by the MTS, and conclude that these concentrations should be omitted from the calculation of EC 20 values.

Results and Discussion
EC 20 values (Table 3) were calculated using the data-points in the linear range of the dilution series, at non-cytotoxic concentrations ( Figure SI-2B). The inlet wastewater samples for WWTPs 1, 4 and 5 showed an effect clearly above EC 20 also at the lowest analyzed concentration, and these samples were therefore analyzed at even lower concentrations ( Figure Leusch et al. 28 who observed high levels of androgenic compounds in untreated wastewater, while all the outlet water samples were below the limit of detection for androgenicity. These two papers report drastic decreases in androgenicity between WWTP inlet and outlet samples which is in line with our results, although we observe effects already at much lower REFs as well as a clear androgenic activity in the outlet water samples. König et al. 2 have reported an EC 10 value in the range REF 10-20 for a water sample collected a few hundred meters downstream of the effluent discharge from a Serbian WWTP. www.nature.com/scientificreports www.nature.com/scientificreports/ The effect at REF 1 was also expressed as DHT equivalent concentration (DHTEQ) ( Table 4). The DHTEQ in the inlet wastewater samples varied greatly between 0.6 and 60 nM, while the outlet water samples all had a DHTEQ of around 0.3 nM. The DHTEQ values reported by others also vary greatly; van der Linden et al. 29 report DHTEQ values of 2.6-2.8 pM in WWTP outlet samples, Bain et al. 30 observed DHTEQs in the range of 100 pM to 1 nM in the inlet wastewater sample to a WWTP but no androgenic activity in the outlet water, and Välitalo et al. 8 report DHTEQs between 48 and 230 pM in the inlet wastewater to five Finnish WWTPs, but no androgenic activity in the outlet water. When comparing the results it should, however, be noted that these studies have been carried out with a different bioassay for androgenicity, the AR-CALUX system, than our study. Both the AR-CALUX system and the AR-EcoScreen are based on transfected mammalian cells and show a very similar responsiveness to the positive control DHT with EC 50 values of 0.31 to 0.45 nM for the AR-CALUX 30 and 0.45 nM for the AR-EcoScreen (Table SI-3).
The removal efficiency, calculated based on the DHTEQ values at REF 1 (Table 5), for androgenic compounds was very high (>99%) for the WWTPs where a high level of androgenic compounds was observed in the inlet wastewater (WWTPs 1 and 4). For the WWTPs where relatively low levels of androgenic compounds was observed in the inlet wastewater (WWTPs 2 and 3), the removal efficiency was lower (50-60%), but the outlet water had DHTEQ values in the same range as the plants with high removal efficiency. Although these five WWTPs have very similar treatment techniques, the removal-efficiency for androgenic compounds varies greatly. The plants with very high removal-efficiency also has a higher DHTEQ level as compared to the plants with lover removal-efficiency and it could be speculated that this difference is due to the mixture of androgenic compounds present in the inlet waters (e.g. that the higher level of androgens could compose of compounds that are more easily to remove by the current treatment techniques).
Antiandrogenic effects were assayed in the AREcoScreen assay which was stimulated by DHT, with hydroxyflutamide (OHF) used as a positive control for antiandrogenicity. Only concentrations that were non-cytotoxic in both the MTS test and in the AREcoScreen agonism test (discussed above) were included in this assay. One sample showed an antiandrogenic reponse just above the LOD ( Figure SI-2E), while all other samples showed no antiandrogenic response at non-cytotoxic concentrations. Due to interference by cytotoxicity, we were unable to calculate an IC 20 value for this effect. Nivala et al. 27 have also reported that cytotoxicity is masking any antiandrogenic effects in WWTP inlet and outlet samples. estrogen receptor activity. The estrogenic activity of each sample was tested in the VM7Luc4E2 assay 31 with estradiol as the positive control. Samples were analyzed in dilution series to establish dose-response relationships and calculate effect concentration (EC) values. As expected, estrogen receptor activity was observed in all samples ( Figure SI-3A), with more pronounced activities in the inlet wastewater samples as compared to the outlet water samples. Similar to the androgen receptor activity assay, most samples showed a biphasic concentration-response relationship with decreasing reporter activity with increasing concentration at high REF values, even where the cell viability test did not indicate cytotoxicity. We hypothesize that this effect is caused by cytotoxicity not detected by the luminescent ATPase cell viability test, and excluded these high REF values from the analysis.
All inlet wastewater samples reached the maximum assay response already at very low REF values, and were therefore reanalyzed in further dilutions to enable the calculation of EC values ( Figure SI-3C (Table 3)  In similarity to our findings, Välitalo et al. 8 have reported estrogenic activities in both inlet and outlet waters from WWTPs, analyzed by a mammalian reporter gene assay similar to the one used in this study. Välitalo et al. also report that the estrogenicity was markedly decreased in the outlet water samples as compared to the inlet wastewater samples. Chou et al. 32 have investigated the removal efficiency of estrogenic compounds during wastewater treatment in a Taiwanese WWTP and reported strongly decreasing ER activity in the outlet water samples as compared to the inlet wastewater sample. The findings in both these studies are in concordance with the findings in the current study. Escher et al. 12  Most of these studies have analyzed samples from WWTPs with primary, secondary, and tertiary treatment steps, but the study by Chou et al. 32 have studied samples from a thin film transistor liquid crystal display WWTP and Escher et al. 12 report results for the secondary treated sewage effluent water which serves the as influent to water reclamation plants. At this stage, due to the limited number of studies, it is not possible to conclude that any of the studied wastewater treatment techniques is superior for removal of estrogenic compounds. In addition to difference in the treatment techniques, it can be hypothesized that the mixture of estrogens are varying greatly between the different study sites, also influencing the removal-efficiency.
A study by König et al. 2 have examined the estrogenicity in water samples collected upstream and downstream of a WWTP effluent discharge in Serbian River Danube. Upstream of the effluent discharge, the EC 10 value for ER activity was in the range of REF 5-100 depending on the assay used for detecting the effect. A few hundred meters downstream of the effluent discharge, the EC 10 was found to be in the range of REF 0.1-2, indicating that the discharged WWTP effluent water contains estrogenic compounds. The same was found for the Ammer River, Germany, where the EC 10  The estradiol equivalent concentration (E2EQ) was calculated at REF 1 (Table 4). The inlet wastewater samples had an E2EQ in the range of 100-930 pM while the outlet water samples had an E2EQ in the range 4-30 pM. The removal efficiency for estrogenic compounds was >75% for all WWTPs and ≥90% for four out of five studied WWTPs ( Table 5). The E2EQ values observed in this study are comparable with many other reports; Miege et al. 33 report E2EQ of 11 pM in the outlet water from a WWTP, Murk et al. 34  Antiestrogenic effects were assayed in the VM7Luc4E2 assay 31 which was stimulated by estradiol with raloxifen used as a positive control for antiestrogenicity. Only concentrations that were non-cytotoxic in both the luminescent ATPase cell viability test and in the VM7Luc4E2 agonism test were included in this assay. Antiestrogenic response were not observed in any inlet wastewater sample ( Figure SI-3E). For most outlet water samples (WWTPs 2, 3, 4 and 5) we observed ER activities below the LOD for the antiestrogenic response, but the effects were ambiguous and with no clear concentration-response relationships.
Oxidative stress response (Nrf2 activity). Oxidative stress response was assayed as the activity of Nrf2   Table 5. Removal-efficiencies bioassays where bioactivity was observed in both inlet and outlet water.
www.nature.com/scientificreports www.nature.com/scientificreports/ Aryl hydrocarbon receptor activity. The AhR activity was assayed in HepG2 cells transiently transfected with a luciferase plasmid responsive to ligand-activated AhR and a renilla plasmid to control for transfection efficiency. Generally, we observed relatively high AhR activities in the inlet wastewater for all five WWTPs with EC 20 values in the range REF 0.44-6.2 (Table 3 and Figure SI-5A and B). For all treatment facilities we observed a clear increase in the EC 20 in the outlet water as compared to the inlet wastewater, indicating that all treatment facilities removes AhR inducing chemicals from the water. For the outlet water samples, EC 20 values were in the range of  For two WWTPs (WWTP 1 and WWTP 4), the inlet wastewater sample showed a clear decrease in the renilla luciferase activity at the highest REF values, although the general cell viability test did not indicate any problem with cytotoxicity. A strong decrease in the renilla luciferase activity can cause false positive results in bioassays based on transiently transfected cells. Therefore, water samples for the inlet wastewater for WWTP 1 and WWTP 4 were only analyzed at lower REF values.
The EC values observed in this study are similar to other reports of AhR activities in wastewater. An Australian study including two WWTP effluent samples reported EC 10  The TCDD equivalent concentration (TCDDEQ) values were calculated for REF 1, and found to vary between 580 and 1200 pM for the inlet wastewater samples and 300-480 pM for the outlet water samples ( Table 4). The observed TCDDEQs are high compared to other studies, for example Nivala et al. 27 that report TCDDEQ values of around 0.8 pM in the inlet wastewater to a German WWTP and around 0.4 pM in the outlet water.
The removal efficiency for AhR activating compounds was 50-60% for most WWTPs studied, but WWPT 1 had a removal efficiency as low as 16% (Table 5).
NFĸB activation. The NFĸB activity was assayed in HepG2 cells stably transfected with an NFĸB responsive luciferase plasmid. For most samples, no NFĸB activity was observed ( Figure SI-6A). For some samples (outlet water from WWTP 2 and both samples from WWTP 3), we observed NFĸB activity above the threshold for induction of induction ratio 1.5, but with no clear concentration-response relationship ( Figure SI-6A). For two inlet wastewater samples (WWTP 2 and 4), we observed NFĸB activity above the threshold for induction of induction ration 1.5. The linear range of the concentration-response relationship was used to calculate the EC IR1.5 value, which was found to be REF 1.7 for WWTP 2 and REF 0.33 for WWTP 4 ( Figure SI-6B). The observed EC IR1.5 values for these two samples are similar to a previously published report studying NFĸB activity in WWTP inlet and outlet waters 27 . The TNFα equivalent concentration (TNFαEQ) for these two samples were calculated to be 0.05-0.2 ng mL −1 (Table 4).
NFĸB activity in water samples can be caused by either environmental pollutants or by the presence of endotoxins in the water. It has recently been demonstrated that bacterial lipopolysaccharides (LPS), and potentially other endotoxins, can be co-extracted from the water to the extract when using solid phase extraction methods 37 . In that study, Neale et al. 37 also report that the majority of the observed NFĸB effect in four Australian environmental water samples was caused by endotoxins and not environmental pollutants.

Comparing observed activities with effect-based trigger values.
To enable an evaluation of the hazards and risks associated with observed bioactivities in studies using in vitro methods, effect-based trigger values (EBTs) are currently being developed. The EBT is a threshold level for the bioactivity, and if a sample shows higher activity than the EBT it causes concern for the endpoint the EBT is developed for. There is currently not a scientific consensus regarding which actual EBT values that should be used, and the EBT value is also dependent on the assay used in the study, the positive control used to calculate the BEQ, and which endpoint the value is developed for (e.g. human health endpoints or ecotoxicological endpoints). As an example, Escher et al. 38 have proposed EBTs for a number of in vitro bioassays, based on a read-across approach from the guideline values in the Australian Guidelines for Water Recycling: Augmentation of Drinking Water Supplies. Of relevance for the in vitro bioassays used in this study, the authors proposed EBT values of 14 ng testosterone/L for androgenic effects, 0.2-1.8 ng estradiol/L (0.7-6.6 pM) for estrogenic effects. It is not possible to compare the proposed EBT value for androgenic activity or AhR activity with the activities observed in this study, as we have used a different reference compound. For estrogenic activity, we observed E2EQ values in the range 0.04-0.3 pM for the effluent waters, which is clearly below the proposed EBT values. It should, however, be noted that these EBT values have been developed with in vitro bioassays that are similar but not identical to the one used in this study. In a different study, van der Oost 39 has proposed EBT values derived from ecotoxicological endpoints. For estrogenic activities, the authors propose an EBT value of 0.5 ng estradiol/L (1.8 pM), which is clearly above the effects that we observe in the effluent water. For AhR activities, van der Oost propose an EBT of 50 pg TCDD/L (0.15 pM) based on observed effects of TCDD on the reproduction of rare minnow (Gobiocypris rarus) after exposure to TCDD. The TCDDEQs observed in this study is clearly above the proposed EBT which causes concern and prompts further research regarding which compounds that causes the effect and how these can be removed during wastewater treatment.
Using bioanalytical tools to evaluate toxicity removal efficiency in wastewater treatment techniques. WWTPs are challenged by the increasing risk of chemical pollutants, of which many are unknown, in the inlet wastewater 1,2,8,40 . Many WWTP operators are currently planning investments in new treatment technologies to target the chemical pollutants during the treatment process and are therefore in need of methodologies www.nature.com/scientificreports www.nature.com/scientificreports/ to evaluate the removal efficiency of bioactive chemicals both in the current treatment process and in potential new treatment techniques 9,25,27 , to make sound financial investments. Bioanalytical tools shows great promise to perform such evaluations, since they make it possible to evaluate the effects of all bioactive compounds in a water sample, without any a priori information on the chemical identity of the causative compounds. Aquatic effect-based monitoring tools, such as bioassays, has also been proposed as a part of the revised European Water Framework Directive 41 . In this study, we have demonstrated that in vitro bioassays based on mammalian cells, assaying initiating molecular events in key toxicity pathways, can be used for effect-based monitoring of wastewater treatment facilities and in our opinion these methods can be valuable both for monitoring purposes and to evaluate the effectiveness of new treatment techniques.  Figure 1. Radar plots showing bioanalytical equivalent concentrations (DHTEQ for AR activity, E2EQ for ER activity, and TCDDEQ for AhR activity) for influent and effluent water for the five studied WWTPs.