Changes in estrogenicity and micropollutant concentrations across unit processes in a biological wastewater treatment system

A range of organic micropollutants, comprising synthetic and natural trace substances, are present in water at low concentrations (Luo et al. 2014). Of particular significance are ubiquitous detections of endocrine disrupting compounds (EDCs) and pharmaceutical and personal care products (PPCPs) in environmental media (Archer et al. 2017). Concern over EDCs has arisen due to their potential ability to cause effects on humans and wildlife at very low doses by interfering with the endocrine system (Clarke & Smith 2011), while many PPCPs are persistent and biologically active compounds with recognised endocrine disruption functions (Kasprzyk-Hordern et al. 2009). Domestic wastewater contains significant amounts of estrogen hormones and unmetabolised drugs (Archer et al. 2017). Conventional wastewater treatment plants (WWTPs) are not designed to treat organic micropollutants, and the effluent from such plants is now considered to be a major point source for EDCs and PPCPs in the receiving environment (Luo et al. 2014). The wide detection of pharmacologically and estrogenically active organic micropollutants in receiving waters world-wide has led to concerns over possible impairment of reproductive processes in exposed freshwater and marine organisms (Luo et al. 2014).

Previous studies have shown that several PPCPs and EDCs (e.g., carbamazepine (CBZ), triclosan, 17α-ethynylestradiol and bisphenol A) are poorly removed by conventional secondary sewage treatment (Clara et al. 2004; Nakada et al. 2006; 2007). While their removal can be achieved using advanced treatment technologies such as ozonation (Nakada et al. 2007), advanced oxidation (Chen et al. 2012), activated carbon (Snyder et al. 2007), and membrane filtration (in particular nanofiltration and reverse osmosis) (Yoon et al. 2007; Xue et al. 2010; Ho et al. 2011), most conventional WWTPs lack such advanced treatment technologies. Besides, sewage biosolids are an inevitable by-product of the treatment of municipal wastewater and have been widely used in agriculture for a long history. Recent studies have indicated that some PPCPs, including potential endocrine disruptors, which often adsorb to sludge during wastewater treatment, can persist in agricultural soils following biosolid application and can also travel through the soil column and leach into agricultural tile drainage at detectable levels (Bottoni & Caroli 2015). The impact of this persistence in soils is unknown, but the link to human and land animal health is likely tied to the capacity for plants to absorb and accumulate these chemicals in their consumed tissues. Obviously, the avoidance of this further contamination lies on the removal efficiency of these micropollutants by WWTPs or pre-treatment of biosolids.

Organic micropollutants differ in their effects on the endocrine system, and a holistic assessment of the effects requires that the concentrations be linked to potential toxicological effects using appropriate bioassays. Several in Vitro and in Vivo bioassays, such as E-Screen (Vanparys et al. 2010), Yeast estrogen screen (YES) (Balsiger et al. 2010), fish vitellogenin and ELISA (Le Fol et al. 2017), have been used for measuring estrogenic activities. While some studies have explored the biodegradation mechanisms of micropollutants such as estrogens (Yu et al. 2013) by either studying the transformation and fate of natural estrogens along the process of WWTPs (Ben et al. 2017), or investigating the change of estrogenicity (Zhao et al. 2015). Few researches coupled the removal of these micropollutants with the change of estrogenicity by each unit process in WWTPs.

In this present study, we link the decrease in micropollutant concentration to the change in estrogenicity across unit processes (including anaerobic digestion) in a typical WWTP by investigating the removal of 10 EDCs and PPCPs, and the associated change in estrogenicity. Specifically, we investigate the removal of four estrogens (estrone, 17β-estradiol, estriol, 17α-ethynylestradiol), an anti-epileptic drug (CBZ), an antibiotic (sulfamethoxazole (SMX)), an antibacterial chemical (triclosan), a cosmetic (oxybenzone), a surfactant (4-nonylphenol), and a plasticizer (bisphenol A) in the liquid stream by sedimentation, biological nitrification and denitrification, and ultraviolet (UV) oxidation, and anaerobic digestion of the micropollutants associated with sludge, along with the change in estrogenicity occurring across individual processes.

Reference standards of estrogens, estrone (E1), 17β-estradiol (E2), Estriol (E3), 17α-ethynylestradiol (EE2), were obtained from Sapphire Bioscience (New Zealand). Reference standards of CBZ, SMX, triclosan (TCS), oxybenzone (OBZ), 4-nonylphenol (NP), and bisphenol A (BPA) were purchased from Sigma–Aldrich (New Zealand). The internal standards (IS), atrazine-d5 (PESTANAL®, analytical standard), was purchased from Sigma–Aldrich (New Zealand). HPLC grade methanol, acetone and acetonitrile were obtained from Ajax FineChem (New Zealand).

A WWTP in an urban setting from northern New Zealand serving a population of >100,000 people was selected as being representative of conventional WWTPs. No big industries feed to the plant. The wastewater is pre-treated by fine screening followed by primary treatment involving grit removal and sedimentation (retention time 2–3 hours). This is followed by secondary treatment by activated sludge reactors – anoxic and aerobic reactors along with a clarifier (hydraulic retention time of 8–12 hours) and anaerobic digesters (hydraulic residence time 12–20 days). The tertiary treatment process involves filtration and UV disinfection. The details of the WWTP are presented in Figure 1. Samples (∼3 litre) were collected from: A) the influent (after pre-treatment), B) primary effluent, C) clarified effluent, D) final effluent, E) inlet and F) outlet of anaerobic digester (Figure 1). Methanol rinsed 4 L amber glass bottles were used to collect three litres of each sample. Following collection, the pH was adjusted to 2 and the samples were stored at 4 °C, with analysis performed within 24 hours.

One litre of samples A-D was filtered with 934-AH glass microfiber filter (Whatman, Global Science, New Zealand) in triplicate. Three mL of the filtrate from each sample was stored for the YES assay, while 500 mL of filtrate was spiked with 20 μL of 2.5 mg/L methanol solution of atrazine-d5 for solid-phase extraction (SPE) using 500 mg hydrophilic-lipophilic balance cartridges (Waters Corp., Global Science & Technology Ltd, New Zealand). The SPE cartridges were preconditioned with 5 mL methanol followed by 5 mL ultra pure deionized (DI) water at 5 mL/min, and the sample was loaded to the preconditioned cartridges at 10 mL/min. Cartridges were dried under high vacuum and then eluted with 5 mL methanol/acetone (1/1, v/v) at 4 mL/min. The eluate was collected and dried by evaporating the organic solvent under nitrogen. The residue was dissolved in 500 μL methanol, and ultrapure DI water added to make a 1 mL solution.

Fifty mL sludge were taken from samples collected at points E and F and frozen at −20 °C overnight. These were then freeze dried over 48 hours until the sample weight no longer changed. Five hundred mg freeze-dried samples were extracted with 4 and 2 mL methanol and then twice with 2 mL acetone; the sample was ultrasonicated for 10 min following solvent addition at each of these steps. After the first methanol extraction, atrazine-d5 was added as IS to the sample. Extraction involved centrifuging the sample at 4,000 rpm for 10 min, followed by supernatant collection and evaporation to around 0.2 mL under a gentle stream of nitrogen gas. The concentrated extracts were diluted with 500 mL ultrapure DI water, followed by pH adjustment to 2 and filtering with 934-AH glass microfiber filter. These samples were then subject to SPE as per the procedure described previously.

Liquid chromatography (LC) and analysis were performed on a Shimadzu Series LC-MS 2020 system equipped with a degasser, a binary pump, and a quadrupole mass spectrometer (MS) detector with electron spray ionization (ESI). Chemical separation was achieved using a 4 μm C12 polar reversed-phase column with a length of 150 mm and inner diameter of 2.0 mm (Synergi MAX-RP, Phenomenex), and a 4 × 3.0 mm C12 precolumn guard cartridge (Synergi MAX-RP, Phenomenex). The column temperature was set to 30 °C and a mixture of acetonitrile and methanol (v/v, 2/3) was used as the organic solvent (B). Optimal MS conditions were determined by performing ESI positive/negative scans for m/z of 50 to 500 for individual analytes directly infused into the MS at a concentration of 1 μg/mL, identifying the typical m/z ratio for each analyte, and fine tuning the desolvation line, direct current, and radio frequency voltages in the selected ion mode. The analysis of micropollutants was done using a binary gradient of DI water (A) and organic solvent (B) at 0.2 mL/min, as the following program: 30% B held for 2.0 min, increased linearly to 90% by 5 min and held for 5 min, and stepped back to 30% and held for 4.0 min. An 8-min equilibration step at 30% B was used at the beginning of each run to bring the total run time per sample to 24 min. An injection volume of 10 μL was used for all analyses.

The estrogenicities of samples A-D were assessed by a four-hour YES bioassay (Balsiger et al. 2010; Chen et al. 2012) and were assessed in two ways: (i) following filtration of the sample and (ii) after diluting 2 μL of the concentrated SPE extracts to 1 mL with DI water. The recombinant yeast strain DSY 219 was provided in media by Marc B. Cox (University of Texas at El Paso, USA). The strain was maintained in synthetic complete media lacking uracil and tryptophan (SC-UW) to select for plasmid retention. The 10X media consists of 6.7 g of Yeast Nitrogen Base, 5 g of Dextrose and 0.72 g of Tryptophan/Uracil Drop-Out supplement in 100 mL of DI water. The solution was sterilised by filtration and stored at 4 °C until used. For assay use, the yeast reporter strain was cultured overnight in SC-UW media at 30 °C in a shaking incubator, following which the cells were diluted to an optical density of 0.08 at 600 nm (OD600) and then shaken in an incubator at 30 °C until the culture reached an OD600 of 0.1. The yeast cells harvested by centrifuging 1 ml aliquots of the culture at 2,000 rpm for 2 min were resuspended in 250 μL 4× concentrated SC-UW. Each 750 μL of sample was mixed with 250 μL of yeast by vortex and incubated for 2 hours in a 30 °C incubator. After incubation, 100 μL of culture was transferred to an opaque 96 well plate, followed by addition of 100 μL freshly prepared Tropix Gal-Screen in Buffer B substrate (Applied Biosystems, CA, USA). The plate was incubated at room temperature for 30 minutes to 2 hours covered. The blank experiment was also carried out in the same manner as the test samples, but with deionised water. Hormone induced chemiluminescent signal was measured by Victor™X luminescence plate reader with 2 seconds' exposure time.

Estradiol equivalents (EEq, in ng/L) of the water samples were determined by: log EEq = log EC₅₀ - log ((top – bottom)/(y-bottom) −1)/slope, where Top and Bottom are the maximal and the basal responses respectively, EC₅₀ of the agonist is the concentration that provokes a response half way between the basal (Bottom) response and the maximal (Top) response, and y is the activity response of the samples.

Dose–response curves for standard E2, BPA and CBZ were developed using the five-parameter logistic equation (GraphPad Prism version 5.00 for Windows, GraphPad Software, San Diego, California, USA) by the same method described above. EC₅₀ of BPA and CBZ was calculated and compared to that of the natural estrogen, E2.

The limit of detection (LOD) and limit of quantification (LOQ) were determined using responses obtained under the optimal voltages for 10 μL injections containing 50 μg/L of an individual compound. These were calculated as α × C × N/S × 10 μL, where α is 3 for LOD and 10 for LOQ, C is the concentration of the compound, and N/S is the ratio of noise to signal provided by the LC-MS software. The LOD and LOQ for the 10 EDCs being tested are listed in Table 1. Calibration curves for the micropollutants were prepared using 0.5, 1.0, 10, 50, 100, 500 and 1,000 μg/L standards. The method recovery was determined according to the methods of Nie et al. (Nie et al. 2009) and the values for EDCs and PPCPs being tested are listed in Table 1.

Concentrations of the 10 EDCs and PPCPs in the influent, primary effluent following sedimentation, clarified effluent following biological nitrification/denitrification, and the final effluent after filtration and UV oxidation are shown in Figure 2, and for these compounds in sludge entering and exiting the anaerobic digester are shown in Figure 3. The removal of these substances by specific unit processes is discussed below.

The surfactant 4-NP was the only chemical to show significant removal via primary sedimentation. This chemical has been reported to be present at high concentrations (500–1,500 ng/L) in municipal wastewaters in Austria (Clara et al. 2005), Japan (Nakada et al. 2006), Greece (Stasinakis et al. 2008) and the UK (Kasprzyk-Hordern et al. 2009). However, 4-NP was only detected in the influent at 15 ng/L in the present study, a level comparable to concentration at a Japanese WWTP that received a mixture of municipal and industrial wastewater (Xue et al. 2010), and is perhaps a reflection of the absence of large industries in the region. Primary sedimentation had no effect on the concentration of the remaining nine EDCs and PPCPs, which were present at levels similar to those reported in raw wastewater or primary effluent (Archer et al. 2017).

During secondary treatment, the waste stream is split into liquid effluent and sludge. Based on the concentrations in the influent and the liquid effluent, a reduction of over 75% for E1, E2, E3 and EE2, 87.7% for TCS, and 100% for OBZ was observed. The non-removal of BPA is surprising as it has been reported that its removal by activated sludge is high (Nakada et al. 2007; Bertanza et al. 2011) and independent of plant operational parameters such as the sludge retention time (Stasinakis et al. 2010); however, inhibitors in wastewater such as allylthiourea or Hg₂SO₄ can lower BPA removal (Kim et al. 2007). While the observations do not clarify whether the removal is a result of biodegradation or due to sorption to sludge, the literature renders support to degradation being a major contributor to their removal. The degradation of natural estrogens occurs under both denitrifying and nitrifying conditions (Yu et al. 2013) and the synthetic hormone appreciably transforms under nitrifying conditions (Suarez et al. 2010). TCS removal by the activated sludge processes is high – removals of 75% in Denmark (Chen et al. 2011), 75% in UK (Kasprzyk-Hordern et al. 2009) and 50% in Japan (Nakada et al. 2007) have been reported. Biodegradation of TCS primarily occurs under aerobic conditions (Chen et al. 2011); however, despite its degradation, its high octanol-water partition coefficient implies that it also adsorbs to sludge (Sharipova et al. 2017), consistent with observations of a relatively high TCS concentration in the mixed sludge samples (Figure 3). The high removal of oxybenzone observed in this study is comparable to other reported values (Oppenheimer et al. 2007). Observations of poor SMX and CBZ removal in this study are consistent with reports of these compounds being highly persistent at other full scale sludge treatment plants (Clara et al. 2004; Göbel et al. 2007). The lack of alkyl side chains in the chemical structure of CBZ was suggested as the main factor impeding their degradation by ammonia monooxygenase, and the three fused aromatic rings provide stability to the molecule making CBZ one of the most persistent micropollutants in wastewater (Paredes et al. 2018). SMX removal, however, can vary greatly, with some studies reporting 0–84% (Castiglioni et al. 2006) or −138–60% (Göbel et al. 2007) elimination in plants, which can be explained by the presence of heterotrophic bacterial strains capable of biotransforming, and even mineralizing SMX (Paredes et al. 2018).

The tertiary stage treatment by filtration and UV oxidation was quite effective, resulting in complete removal of E1, E2, E3, EE2 and TCS. CBZ removal was 56.2%, while BPA and SMX were unaffected. Removal by filtration systems is poor (Nakada et al. 2007) and photochemical oxidation can significantly degrade natural and synthetic hormones (Sornalingam et al. 2016) and many of the pharmaceuticals. Therefore, the observed removals can largely be attributed to UV oxidation. While the poor removal of CBZ in DI water and the variable removal of CBZ in natural waters following UV, and to a greater extent UV/H₂O₂ treatment were reported (Paredes et al. 2018), suggesting that the water matrix plays an important role in compound degradation. BPA photodegradation by UV light is low (Doong & Liao 2017) while SMX removal by UV oxidation can be inhibited by NO³⁻ and HCO₃⁻ in wastewater (Mouamfon et al. 2011).

Of the selected organic micropollutants (see Table 2), SMX is hydrophilic (octanol-water partition coefficient (log K_ow) <1) and all other substances are hydrophobic with log K_ow > 2.5, indicating a strong tendency to sorb to sludge. Figure 3 presents the concentrations of target compounds in the mixed sludge from the primary sedimentation and secondary treatment. Despite its high hydrophobicity NP was not detected, possibly due to its low concentration in the influent. Anaerobic digestion is effective at lowering the concentrations of estrogens E1, E2, E3 and EE2 as well as pharmaceuticals OBZ and SMX to below LOQ. Some studies with anaerobic digesters have reported low removal efficiencies for natural estrogens (Muller et al. 2010), while others have reported their high removal (>85%) in mesophilic and thermophilic digesters, and such discrepancies can be attributed to differences in plant operation (e.g. the fraction of primary sludge mixed with secondary sludge) (Wang & Wang 2016). No removal of BPA, CBZ and TCS by anaerobic digestion was seen. Their resistance to biodegradation under anaerobic conditions is consistent with reports in the literature (Gonzalez-Gil et al. 2016; Wang & Wang 2016). It was proposed that the efficiency of the anaerobic digestion process was not dependent on operational parameters but compound-specific: some OMPs were highly biotransformed (e.g. SMX), while others were only slightly affected (e.g. CBZ and TCS) (Gonzalez-Gil et al. 2016).

The estrogenicities (as EEq) of the raw filtered samples and the extracts from SPE cleanup are shown in Figure 4 for different stages of the wastewater treatment system. Although the estrogenicity values reflect the effect of all compounds in wastewater and not just the 10 quantified in the study, estrogenicity reduction shows a similar pattern to removal of the 10 micropollutants across the unit processes. The influent (sample A) has an estrogenicity of 65 ± 10 ng/L, which increases to 80 ± 12 ng/L after primary treatment (sample B), then decreases to 48 ± 6 ng/L after integrated aerobic/anoxic treatment (sample C), and then further decreases to 5 ± 2 ng/L after filtration and UV oxidation (sample D), corresponding to absolute luminescence measurements that are about 10 times higher for the effluent compared to the DI water control. The filtered samples and the SPE extracts give similar values. The estrogenicity value for the raw influent is comparable to that for municipal sewage treatment plants abroad (Leusch et al. 2005) and the whole treatment processes in this present WWTP achieved the removal of estrogenicity to 92.3%. The small increase in EEq level after primary treatment is within the statistical variation in data; however, factors such as deconjugation may have contributed to the increase – estrogens are often excreted in inactive form as sulfonite or glucuronide conjugates, and deconjugation would enhance their estrogenic activity (Ternes et al. 1999). Activated sludge treatment resulted in a 40.0% reduction in estrogenicity, similar to Holbrook et al. (Holbrook et al. 2002) but lower than Leusch et al. (Leusch et al. 2005), possibly due to differences in hydraulic and solids retention times at the facilities (Holbrook et al. 2002). UV oxidation was the most effective process, giving a 89.6% reduction in estrogenicity. Comparing the degradation of the 10 selected organic micropollutants to the estrogenicity reduction, the high residue / poor removal of BPA, CBZ and SMX should have no significant contribution to changes of estrogenicity across unit processes in the WWTP.

Although many micropollutants have been reported to be detected in the sewage sludge/biosolids, in the present study, among 10 selected micropollutants, only three of them were found in the final digested sludge/biosolid with considerable concentrations, i.e. BPA (120.1 μg/kg dw), CBZ (16.8 μg/kg dw) and TCS (856.2 μg/kg dw). Table 3 shows that these concentrations are on the lower spectrum of concentrations reported in biosolids for these chemicals internationally. Quantification of biosolids' estrogenicity using the YES assay was not performed in this study, as biosolids can be toxic to yeast (McNamara et al. 2009). However, the estrogenicity values obtained using the YES assay for aqueous samples of BPA and CBZ are presented along with E2 in Figure 5 to demonstrate that BPA (EC₅₀ = 1.43 × 10⁻⁶ mol/L) and CBZ (EC₅₀ = 2.08 × 10⁻⁶ mol/L) had much lower estrogenicity than E2 (EC₅₀ = 2.87 × 10⁻⁹ mol/L).

Tracking the degradation of 10 EDCs and PPCPs across the different unit processes of a municipal WWTP showed that, among 10 selected micropollutants, (i) E1, E2, E3, EE2 and OBZ were degradable under aerobic and under anaerobic conditions; (ii) BPA had neither aerobic nor anaerobic degradability; (iii) CBZ and TCS were only aerobically degradable; (iv) SMX was only anaerobically degradable; (v) 4-NP was significantly removed via primary sedimentation. YES showed that primary settling had little effect on estrogenicity, biological anoxic/oxic transformations gave ∼40.0% reduction, but UV oxidation was the most effective and showed an 89.6% reduction in estrogenicity. Of the 10 chemicals monitored in the study, BPA, CBZ and SMX showed small or no degradability. Similarly, residuals of three micropollutants, BPA, TCS and CBZ, were observed in biosolids, indicating that some chemicals are not readily degraded by aerobic as well as anaerobic processes. However, YES assay values show a 92.3% estrogenicity reduction in the liquid stream across the plant, yielding 5 ng/L EEq in the effluent – a value that compares favourably with the observed 1 ng/L EEQ for DI control. That suggests YES assay is enough to evaluate the estrogenicity reduction efficiency of different unit processes in a WWTP, while chemical analysis could result in contradictory interpretation because some chemicals of interest do not contribute to changes of estrogenicity, such as BPA and CBZ in this present study.

This study was supported by a grant from the University of Auckland Faculty Research Development Fund. Yeast strains for the YES assay were kindly provided by Dr Marc B. Cox, University of Texas at El Paso.