A membrane bioreactor (MBR) was assessed for the removal of estrogens, androgens, and a selection of pharmaceuticals and personal care products. The biomass and aqueous components of the MBR were investigated to determine whether removal was by biodegradation or by adsorption to the biomass. Removal was monitored using chemical analysis by gas chromatography/mass spectrometry (GC-MS) as well as biological analysis using estrogenic and androgenic yeast assays. Results showed that the MBR was effective in removing the compounds of concern from raw influent with removal rates between 78 and 99%. Removal efficiencies were comparable or better than those reported for conventional activated sludge systems, which was attributed to the relatively high sludge retention time of the MBR. The biomass component showed significant concentrations of salicylic acid, triclosan, and 4-tert-octylphenol. Estrogenic and androgenic activity was also measured in the biomass. Estrone was identified as the main compound responsible for the estrogenic activity. It was concluded that the main removal pathway was biodegradation, but sorption to biomass may also be important, particularly for triclosan and 4-tert-octylphenol.
Much concern currently exists about the release of endocrine-disrupting chemicals (EDCs) and pharmaceuticals and personal care products (PPCPs) into the aquatic environment. These chemicals enter the sewer system via disposal or excretion and are not completely removed during wastewater treatment. Several studies have detected PPCPs [1,2] and EDCs, such as natural steroidal hormones [2,3] and alkyl phenols , in environmental samples. The long-term effects of the release of these chemicals are not well known . Research has shown adverse affects on aquatic wildlife [6,7] that have been linked to the presence of EDCs and PPCPs. The human health implications of these environmental contaminants are unknown.
Membrane bioreactor (MBR) technology for wastewater treatment is advancing rapidly, particularly as a decentralized system. Decentralized wastewater treatment systems (or package plants, as they are sometimes called) are becoming the preferred option for sewage treatment in areas where connection to a centralized public sewer network is not possible or economically feasible. Membrane bioreactors have several advantages over conventional activated sludge (CAS) systems, such as having a small footprint, lower sludge production, and a largely disinfected effluent . Membrane bioreactors are comprised of a combination of an activated sludge process and microfiltration/ultrafiltration membrane separation. The application of package plant systems as either single or cluster formations has been shown to produce treated effluent of sufficient quality, in terms of biological oxygen demand and suspended solids, to meet environmental regulations for direct discharge into water courses . However, like centralized wastewater treatment systems, there remains some concern as to the fate and removal of trace contaminants by package plant treatment processes. This is particularly important in regions with no suitable receiving waters or where the treated wastewater is for reuse. Membrane bioreactors commonly perform better than CAS processes in terms of effluent water quality and can often achieve good nitrogen removal through extended solids retention times (SRT) and high mixed liquor suspended solids concentrations . If decentralized MBRs can be demonstrated to remove organic contaminants of concern such as EDCs and PPCPs, this will increase their attractiveness for application in wastewater treatment for water reuse in rural and regional communities.
The ability of MBR technology to remove EDCs and PPCPs has not been well studied, particularly in terms of sorption to MBR biomass. A few comparative studies on the fate of EDCs and PPCPs in the aqueous phase of MBRs are available [2,11] and have shown promising results. Clara et al.  reported 90% removal of the EDC bisphenol A and the pharmaceutical ibuprofen, and Kim et al.  showed that MBRs were effective in eliminating hormones (estriol, testosterone, and androstenedione) and some pharmaceuticals (acetaminophen, ibuprofen, and caffeine) with approximately 99% removal. Kimura et al.  investigated the use of an MBR for the removal of six acidic PPCPs and found comparable removal rates to CAS. Spring et al.  studied a pilot scale MBR and found that it was more effective at removing the estrogens estrone and ethynylestradiol than a full-scale CAS treatment plant receiving the same wastewater. The removal of estrogenic activity (98%) and androgenic activity (87%) by an MBR has been previously reported, which was similar to the performance of a CAS plant treating the same influent water .
The reuse of biosolids from wastewater treatment plants in agriculture is another possible source of environmental contamination. Studies on trace chemical contaminant concentrations in the biomass phase of MBRs, however, are limited. Research is needed to determine whether significant sorption occurs on the biomass or whether these chemicals are removed by biodegradation. Hu et al.  studied three pilot-scale and two lab-scale MBRs for the removal of estrogens in MBR biomass and aqueous phases and reported that the elimination of overall estrogenicity is attributed mainly to bacterial metabolism and physical binding to biomass in MBRs. No information is available regarding the removal of androgens or androgenicicity by MBRs. Urase et al.  investigated the adsorption of ibuprofen on CAS biosolids at various pH levels and found that removal was low at neutral pH but high at acidic pH. Ibuprofen is an acidic chemical with an acid dissociation constant (pKa) of 4.91, and thus, in neutral solutions, ibuprofen is ionized, and in acidic solutions, it has increased hydrophobicity, resulting in adsorption to hydrophobic biomass particulates. Accordingly, it is important to clarify the factors affecting the removal of EDCs and PPCPs in wastewater and whether sorption or biodegradation is the primary mode of removal.
The aim of the present study was to gain a better understanding of the fate, behavior, and removal of EDCs and PPCPs in a decentralized MBR by investigating the biomass and aqueous components of the MBR. A sample pretreatment method was developed and optimized for the extraction of the chemicals prior to analysis. A combination of chemical and biological analyses was used to obtain detailed data on the levels and nature of the contaminants at different stages of the treatment process: influent, MBR biomass, MBR aqueous phase, and final effluent. A yeast screen bioassay was used to determine the overall estrogenic and androgenic activities of samples, and gas chromatography/mass spectrometry (GC-MS) was used to analyze for selected target compounds present in the wastewater. The targeted analytes included the pharmaceuticals ibuprofen and salicylic acid, the personal care product triclosan, the EDC 4-tert-octylphenol, and a range of natural and synthetic estrogens and androgens.
MATERIALS AND METHODS
All chemicals were purchased from Sigma-Aldrich unless otherwise stated and were ultrapure or analytical grade. Chlorophenol red-β-D galactopyranoside was supplied by Boehringer Mannheim. The isotopically labeled standards D2-androsterone, D4-salicylic acid, D3-ibuprofen, D4-estrone, D4–17β-estradiol, and D4–17α-ethynylestradiol were purchased from CDN Isotopes. Acetonitrile (anhydrous spectroscopy grade), acetone (spectroscopy grade), and methanol (analytical grade) were purchased from Ajax Chemicals or Asia Pacific Specialty Chemicals.
Samples (7.5 L) of raw sewage (MBR influent), mixed liquor from the MBR, and MBR effluent from a sewage treatment plant in New South Wales, Australia, were collected in 2.5-L glass bottles. The MBR is a pilot-scale (25 equivalent population [EP], 4 kl/d) single circular tank with a 0.2-micron submerged membrane. The feed water was the diverted influent from a larger conventional sewage treatment plant (3,500 EP, 0.8 ml/d) located on the same site, after screening and grit removal. Electrochlorination followed the MBR process. The hydraulic retention time of the MBR was 1 to 2 d, and the SRT was 20 d. No wet-weather events occurred prior to sampling. The samples were all taken at 10:00 AM on March 4, 2008 (Australian summertime).
Because of spiking of deuterated surrogate standards (which include known estrogenic and androgenic chemicals) for the chemical analysis, different samples had to be used for the biological and chemical analysis. Triplicate samples of 1 L (aqueous phase) or 0.5 g (solid phase) of the influent raw sewage, MBR biomass, MBR aqueous, and MBR effluent were subjected to biological and chemical analysis. Deuterated surrogate standards (300 ng) were added to each sample used for chemical analysis. For quality control, triplicates of spiked deionized water and spiked artificial laboratory MBR biomass were extracted parallel to the treatment of the real samples and analyzed by the respective method. Triplicates of unspiked deionized water and unspiked artificial laboratory MBR biomass were used as blanks. The relative recoveries were between approximately 90 and 110% for salicylic acid, ibuprofen, triclosan, estrone, 17α-estradiol, and 17β-estradiol for both the aqueous phase and the solid phase and 146 ± 36% (solid phase) and 165 ± 14% (aqueous phase) for 4-tert-octylphenol.
Fenoprop (m = 4.3 μg) was added to each sample, blank, and calibration standard as an internal standard prior to derivatization and chemical analysis. The solid and liquid phases of each sample were separated by filtration. The aqueous component of each sample and the aqueous and biosolid component of the MBR mixed liquor were investigated. The aqueous phases and the biomass component of the MBR mixed liquor sample, which was retained during the first filtration step, were prepared as described here. For filtration of the samples, Whatman No. 1 filter papers were used followed by Whatman GF/A glass-fiber filter papers (pore size 1.6 μm) and finally Whatman GF/F filter papers (pore size 0.7 μm). The GF/F filtrates were then acidified with sulfuric acid to a final pH of 2 to 3 and spiked with deuterated standards.
The biomass extraction method was modified from a method previously published by Ternes et al. , who investigated the concentrations of iodinated contrast media, musk fragrances, and pharmaceuticals present in sewage sludge in different German and Swiss sewage treatment plants. The procedure involved sample pretreatment by freeze-drying, ultrasonic solvent extraction, filtering, and diluting the sample in water, cleanup using solid phase extraction (SPE), concentrating by evaporation, reconstituting in solvent, and finally biological (yeast screen bioassay) and chemical (derivatization: GC-MS) analysis. Samples were frozen for 24 h in a freezer at −20°C and then dehydrated for 3 d in a freeze dryer. The samples were then crushed to a fine powder using a mortar and pestle and spiked with deuterated steroids. Chemicals were extracted using ultrasonic solvent extraction. This was done by adding 4 ml of methanol to the biomass sample, ultrasonicating for 5 min in an ultrasonic bath (Ultrasonics Australia Pty Ltd), followed by centrifuging (Beckman Coulter Allegra X-12R) at 3,000 rpm (1,207 g) for 5 min. The supernatant was removed and kept in a separate centrifuge tube. This procedure was repeated three times using 2 ml of methanol and twice with 2 ml of acetone. The supernatants were then combined and dried down to approximately 200 μl using a Thermo Savant SDP121P Speedvac Concentrator (Thermo Scientific) at an operating temperature of 35°C and a pressure of 0.02 Torr. The 200-μl samples were diluted in 150 ml of deionized water and adjusted to pH 2 by adding diluted sulfuric acid. In order to prevent clogging of the SPE cartridges, the samples were then filtered using a Millipore filtration kit with a Whatman GF/F glass-fiber filter.
Cleanup using solid-phase extraction
Solid-phase extraction was used for extracting and concentrating all the prepared aqueous and biomass samples. Oasis HLB 6-cc solid-phase extraction cartridges (absorbent material: divinylbenzene-N-vinylpyrrolidone copolymer) were used. The cartridges were conditioned with 5 ml of acetone, 10 ml of methanol, and 10 ml of deionized water adjusted to pH 2 to 3 prior to extraction. All filtered samples were then passed through the SPE cartridges at a flow rate of 5 ml/min. The cartridges were dried under nitrogen and then eluted with 4 ml of acetone/methanol mixture 1:1 v/v, and the extracts were dried to complete dryness under a gentle stream of nitrogen. The sample extracts were redissolved in 1 ml of acetonitrile for bioassay analysis and 450 μl of acetonitrile for GC-MS analysis.
Biological analysis: Estrogenic and androgenic activities. A yeast screen bioassay was used to test for estrogenic and androgenic activity in samples following the protocol described by Routledge et al. . The yeast assays were carried out in a type II laminar airflow cabinet to minimize aerosol formation. Aliquots of each sample (4 × 10 μl) were transferred to a 96-well optically flat-bottom plate (Nunc, USA). Each plate used for analysis contained 4 × 10 μl sample, at least one row of blanks (200-μl assay medium only), and one row of acetonitrile (10 μl acetonitrile and 200 μl assay medium) as well as two rows of calibration standards for 17β-estradiol (concentration range 10 μg/L-9.53 ng/L) or testosterone (50 μg/L-195.3 ng/L) for the estrogenic and androgenic assays, respectively. The seeded assay medium (200 μl medium containing recombinant yeast and the chromogenic substrate chlorophenyl red-β-D-galactopyranoside) were then dispensed to each sample well using a multichannel pipettor. The plates were sealed with autoclave tape and shaken vigorously for 2 min on a titer plate shaker and read at 570 nm (for color) and 620 nm (for turbidity) prior to incubation at 32°C in a naturally ventilated heating cabinet. For the next 2 d, the plates were shaken vigorously on the plate shaker for 2 min to mix and disperse the growing cells, read at 570 and 620 nm, and returned to the 32°C incubator. On the fourth day, after incubating for 3 d, the plates were shaken for 2 min and left for approximately 1 h to allow the yeast to settle. The plates were then read at absorbances of 570 and 620 nm using the plate reader. The plates were left at room temperature and read again later to verify the results. Photographs were taken of each plate with a Nikon D50 SLR camera every day after each plate reading. For the androgenic assay, the assay plates were incubated at 32°C for 48 h and then incubated at room temperature and read when the data with the greatest contrast were obtained. The actual estrogenic and androgenic activities were then calculated based on the dilution factor. Blanks were run on each plate for all samples in each assay and were always less than the limit of detection, <0.02 ng/L in the estrogenic assay and <0.1 ng/ L in the androgenic assay.
Chemical analysis: Analysis of EDCs and PPCPs. Chemical analysis was carried out by GC-MS using an Agilent HP5973 mass spectrometer with negative electron impact ionization combined with an Agilent HP 6890 gas chromatograph. An HP-5MS capillary column was used (5% phenyl methyl siloxane; nominal length 30 m, nominal internal diameter 250 μm, nominal film thickness 0.25 μm) with helium as the carrier gas at a constant flow rate of 1.2 ml/min and pressure of 12.43 psi. The temperature program was 100°C, held for 3 min, ramped to 150°C at 30°C/min, held for 1 min, ramped to 205°C at 3°C/min, ramped to 260°C at 10°C/min, and finally held for 23 min. An injection volume of 1 μl was used in splitless mode with an inlet temperature of 210°C using an Agilent 7683B autosampler. All quantitative results were calculated by integration of the peak areas obtained by monitoring the respective ion fragments using selective ion monitoring mode. Because of their high polarity and low volatility, the analytes required chemical derivatization prior to GC-MS analysis. Bis(trimethylsilyl)trifluoroacetamide-trimethylchlorosilane (BSTFA-TMCS) was used as the derivatizing agent to form the trimethylsilyl derivatives. The samples (450 μl of standard solutions or redissolved dried solid-phase extraction extracts in 450 μl acetonitrile) and 80 μl of BSTFATMCS (99:1) were heated at 70°C for 100 min prior to analysis by GC-MS.
RESULTS AND DISCUSSION
Biological analysis: Estrogenic and androgenic activities
After incubation, the control wells appear light orange in color because of the background expression of β-galactosidase (β-gal) and turbid because of the growth of the yeast. Positive wells are indicated by a deep red color accompanied by turbid yeast growth. Clear cells (containing no growth) indicate lysis of the cells, and the color may vary. To correct for turbidity, the following equation was applied to the data obtained for each well: Corrected Value = Chemical Absorbance at 570 nm – (Chemical Absorbance at 620 nm – Blank Absorbance [i.e., acetonitrile] at 620 nm). An average value was then taken for the four results obtained for each sample and for the two results obtained for each calibration curve. Calibration curves developed for 17β-estradiol and testosterone for each plate were plotted, giving concentration versus absorbance at 570 nm. The concentration of estrogenic or androgenic activity in each well was then calculated as 17β-estradiol equivalents (EEq) or testosterone equivalents (TEq) from the standard curves in μg/L. These values were then divided by 1,000 (since the samples were preconcentrated 1,000 times), giving final concentrations in ng/L. A mean value was calculated for each sample in four wells. The mean and standard deviation were then calculated for each set of triplicate samples. The standard deviation is represented as error bars in the figures. Levels of estrogenic and androgenic activities in influent, MBR biomass, and MBR aqueous and effluent samples are shown in Figures 1 and 2.
The raw wastewater showed an estrogenic activity of 4.82 ± 0.38 ng/L EEq (Fig. 1). Coleman et al.  investigated the same MBR previously and detected slightly lower influent concentrations of 2.41 ng/L EEq. Results in a similar range were reported from Singapore with raw wastewater activity of 2.5 to 7.1 ng/L EEq . Studies in England by Kirk et al.  reported activities in mainly domestic wastewater that were generally an order of magnitude higher, ranging from 15 up to 80 ng/L EEq for different sewage treatment plant influents. Kirk et al.  also revealed differences in estrogenic activities for samples taken at different times, ranging from 15 up to 40 ng/L EEq for one particular plant. Leusch et al.  conducted a study of STPs in Queensland, Australia, and found levels of <4to 185 ng/L EEq in raw sewage samples. Factors such as differences in population densities, treatment technologies, socioeconomic factors, and climatic differences should be considered when comparing different studies.
The treatment of the wastewater by MBR technology led to a decrease in estrogenic activity of 78%. Percentage removals were calculated from the influent and effluent concentrations. The estrogenic activity of the biomass was 1.63 ± 0.24 ng/L EEq (see Fig. 1), indicating that some sorption to the MBR biomass occurred. However, when comparing the influent concentrations and the concentrations absorbed to the biomass, it can be seen that the elimination of the overall estrogenic activity in the MBR was attributable mainly to biodegradation. This conclusion was also drawn by Holbrook et al.  during CAS treatment with waste-activated sludge concentrations of 3.3 ng/L EEq.
Figure 2 shows the androgenic activity in raw effluent, MBR aqueous, MBR biomass, and effluent samples. Androgenic activity levels in the raw influent were found to be 1177 ng/L TEq.
Coleman et al.  investigated the same MBR in 2006 and reported a value somewhat lower than this (478 ng/L TEq). The recent increase in androgenic activity may be due to an increase in population in this area over time as well as seasonal variations. Kirk et al.  reported lower levels of estrogenic and androgenic activity in influent and effluent after rainfall events. The androgenic values for the raw samples are significantly higher than those of the estrogenic component, being approximately 240 times greater than estrogenic activity. Coleman et al.  reported values of 74 to 240 times higher for androgenic activity compared to estrogenic activity in raw samples from four sewage treatment plants operated by the same water utility. Leusch et al.  also reported that androgenic activity in raw and treated sewage in Queensland, Australia, was much higher than estrogenic activity, being on average 50- to 100-fold higher. Kirk et al.  have suggested that most of the androgenic activity in municipal sewage with a predominantly domestic input is most likely caused by androgens excreted by humans. Androgen levels in humans are generally higher than estrogen levels . Concentrations of androgens in sewage would therefore be expected to be much higher than those of estrogens. A removal efficiency of 98% was seen for androgenic activity. The activity of 588.3 ng/g TEq dry biomass detected suggests that adsorption is the main mechanism of removal for compounds with androgenic activity. A significant decrease in androgenic activity in the effluent sample compared to the MBR aqueous sample has been observed. This is due to the fact that there is adsorption of the chemicals to the biomass coupled by the electrochlorinaton treatment process post MBR.
Wastewater treatment with MBR technology led to a significant decrease in estrogenic activity (78%) and in androgenic activity (98%). These results are comparable to a previous study by Coleman et al.  on the same MBR where removal rates of 88% for estrogenic activity and 98% for androgenic activity were observed. The measured removal of 78% is also comparable to those MBR removal efficiencies measured in similar studies reporting removal rates of 68%  and of 69% . The higher efficiency might be caused by the hot temperatures during the Australian summer and the longer hydraulic retention times (24 h compared to the MBR in the study by Holbrook et al.  of 8.5 h).
Coleman et al.  also measured and compared estrogenic and androgenic removal rates with CAS treatment of the same source water. The results showed that the CAS process and the MBR process are comparable for removal of estrogenic and androgenic activities from raw sewage. General literature comparison and studies investigating both MBR and CAS treatment technology [4,14,20] do not reveal significant differences between MBR and CAS for the capability of removing estrogenic activity. Drewes et al.  showed that the total estrogenic activity was removed by an average of 96% during secondary treatment and that two pilot-scale MBRs achieved the lowest concentrations of individual EDCs in secondary treated effluents, showing a better performance than the full-scale CAS facility operated in the same service area.
Values of 21 and 1.04 ng/L EEq were observed in the final effluent, which may be biologically significant since some studies have shown reproductive abnormalities in fish at ng/L levels (0.1 ng/L estrogenic activity and >1 ng/L androgenic activity) [6,21]. Furthermore, as the present study showed, chemicals with estrogenic and androgenic activity were sorbed to the MBR biomass, and therefore it could be assumed that the application of processed sludge with absorbed estrogenic and androgenic chemicals in agriculture may risk contamination of soil and groundwater. It would therefore be prudent to identify the specific chemicals responsible for the observed biological activity, as this may provide more information on toxicity and the source of the pollution.
Chemical analysis: EDCs and PPCPs
Statistical analysis was performed by taking the average value of each set of triplicate samples and the standard deviation. Standard deviations are represented by the error bars in each figure. Figure 3 shows the levels of EDCs and PPCPs detected in influent, MBR aqueous, MBR biomass, and effluent samples.
All the investigated PPCPs were detected in all samples. This is not surprising, as these PPCPs are in widespread use. All the detected contaminants were removed efficiently from the wastewater during their passage through the MBR with removal efficiencies ranging from 93% (triclosan) up to 99% (salicylic acid, ibuprofen) for the PPCPs (Fig. 4). Removal efficiencies were determined from the influent and effluent concentrations.
The high concentrations detected in raw wastewater for ibuprofen (6,474 ± 862 ng/L or 6.47 ± 0.82 μg/L) and for salicylic acid (32,152 ± 400 ng/L or 32.15 ± 0.40 μg/L) (Fig. 3) are at the higher limits but still in the same range as reported in other studies investigating raw wastewater where concentrations of ibuprofen ranging from 1.2 μg/L  to 6.77 μg/L  and salicylic acid from 2.72 μg/L  up to 38.5 μg/L  have been reported. The high values are not surprising, as ibuprofen and aspirin are highly consumed pharmaceuticals in Australia with annual dispensed masses of 20 × 103 and 14 × 103 kg (data from 2004), respectively . Despite the fact that salicylic acid is a relatively polar and acidic compound (log KOW = 2.26, pKa = 3), an accumulation on biomass was observed with a concentration of 260 ± 22 ng/g dry biomass, while the concentration in the MBR aqueous component was 46 ± 5 ng/L. A possible explanation for the sorption to biomass is that salicylic acid may form chelate complexes that adsorb to the biomass. The low concentration of ibuprofen in the MBR biomass (11 ± 1 ng/g dry biomass) indicates that biodegradation is the main mechanism for ibuprofen removal from the aqueous phase and that the role of sorption is less important. Other studies revealed similar results. Joss et al.  concluded that sorption onto sludge is not significant for compounds with sorption coefficients (kd values) <300L/kg, and therefore the main mode of removal for ibuprofen with a kd value of 7 L/kg is biodegradation.
The antibacterial agent triclosan, which is widely applied in soaps and detergents, was found at a concentration of 1.29 ± 0.16 μg/L in the raw wastewater (Fig. 3). This concentration is in the same magnitude as those reported in Canada  and in the United States . Studies from Japan report concentrations lower than those detected in most of the Western countries, which has been attributed to the lower usage of triclosan in Japan . A high concentration of triclosan was observed in the MBR biomass (475 ± 30 ng/g dry biomass), indicating that removal by sorption onto sludge is the principal mode of removal. Similar results were observed by Stasinakis et al. , who conducted a comprehensive study on the fate and toxicity of triclosan in activated sludge systems to investigate the role of biodegradation and sorption on its removal. Batch experiments by Stasinakis et al.  showed that a high fraction of triclosan was removed quickly from the dissolved phase by sorption to biomass and that its concentration in the dissolved phase remained constant, whereas the fraction adsorbed on biomass was further but not completely degraded.
Octylphenol is a degradation product of octylphenol polyethoxylate, which is a nonionic surfactant widely used in both domestic and industrial applications, such as the manufacture of plastics and elastomers. The concentration of 4-tert-octylphenol in the investigated MBR influent was 839 ± 41 ng/L (Fig. 3), which was slightly lower than reported in some previously published studies [23,30,31]. The levels of 4-tert-octylphenol previously reported in wastewater have ranged from several hundred ng/L up to several μg/L, depending on the wastewater source. Reports indicate that the higher the industrial fraction of the sewage, the higher the concentration of 4-tert-octylphenol. The fact that the source of the wastewater in the present study is almost entirely from domestic sources would explain these results. Clara et al.  measured concentrations of 118 to 680 ng/L of 4-tert-octylphenol in raw sewage of a rural area, whereas Lee et al.  reported concentrations of up to 3.08 μg/L for wastewater with a high industrial fraction. The measured concentration of octylphenol in the biomass was 145 ± 7 ng/g dry biomass, indicating that adsorption to biomass was the main pathway of removal. The adsorption of 4-tert-octylphenol to biomass was also observed by others [31,32] with concentrations of up to several μg/L reported . This can be explained by the low polarity of 4-tert-octylphenol (log KOW = 5.5) and its high partition coefficient, kd > 1,000 L/kg . As a general rule, the more hydrophobic a chemical, the greater the proportion that will adsorb to the biomass . Lee et al.  reported a removal efficiency of 84.7% and attributed it mainly to the result of accumulation to biomass rather than biodegradation, whereas Clara et al.  calculated that biodegradation was responsible for >90% of 4-tert-octylphenol removal.
All the detected contaminants were removed efficiently from the wastewater during treatment using an MBR with removal efficiencies ranging from 93% (triclosan) up to 99% (salicylic acid, ibuprofen), as shown in Figure 4. Surprisingly, twice as much triclosan was measured in the final effluent compared to the MBR aqueous sample. This could be because the samples were not perfectly representative of each other, as they were all taken at the same time, or may be due to the fact that some biodegradation may have occurred for the mixed liquor MBR samples during transportation compared to the cleaner effluent samples.
The MBR removal efficiencies in the cited literature [2,11,22,27,34] are similar to those in the present study and slightly higher than those for CAS treatment, indicating that MBR treatment is a promising technology for the removal of the investigated trace organics. A summarized comparison of the removal efficiencies for the present study, for other MBR studies, and with conventional activated sludge (CAS) studies is shown in Figure 5.
Ibuprofen was removed almost completely from the wastewater during treatment with a removal efficiency of 99%. Other studies reported similar findings during MBR treatment, whereas removal with CAS treatment was reported to be in a lower range of 60 to 95% [2,22,34]. The excellent removal efficiency of ibuprofen for MBR treatment and the higher values compared to CAS treatment can be explained by the longer SRT that can be achieved by MBR systems compared to CAS. Clara et al.  observed a strong influence of the SRT for the biodegradation of ibuprofen and determined a critical SRT of 5 d for significant biodegradation of ibuprofen (at 10°C). The studied MBR has a SRT of 30 d that would explain the high removal rates. An almost complete removal of 99% was achieved for salicylic acid. Similarly high removal efficiencies were reported for CAS treatment [23,24,30], indicating that salicylic acid is easily biodegraded. The investigated MBR showed a triclosan removal rate of 93%. A study investigating the fate of triclosan during MBR treatment carried out by Snyder et al.  reports efficient removals of >99%. This is higher than removal rates reported during CAS treatment ranging from 58%  up to 90% , with most reported removals being between 80 and 90% . The decrease from the influent concentration of 839 ± 41 ng/L down to the effluent concentration of 16± 4 ng/L for 4-tert-octylphenol during MBR treatment gave a removal rate of 98%. Clara et al.  investigated the removal of 4-tert-octylphenol by MBR treatment, reporting 45 to 98% removal. Despite the high removal rates for the investigated PPCPs, environmentally relevant concentrations for ibuprofen, salicylic acid, triclosan, and 4-tert-octylphenol of 11 ± 2, 15 ± 5, 89 ± 11, and 16± 4 ng/L, respectively, were found in the effluent samples. Several studies reported the occurrence of these compounds in surface water, with major sources being sewage treatment plant effluents [31,32]. In general, PPCP concentrations measured in surface waters are well below concentrations that are known to cause acute toxicity to aquatic organisms. However, chronic exposure to pharmaceutically active substances and/or endocrine-disrupting chemicals such as octylphenols has the potential for more subtle effects, such as metabolic or reproductive changes on nontarget organisms . In addition to discharged effluents, another major environmental source of the trace organics could be the use of the processed sludge in agriculture, especially for chemicals with high tendency to adsorb to biosolids and a high stability against biodegradation such as alkylphenols .
Chemical analysis of estrogens and androgens
The only estrogen detected in any of the samples was estrone. All other estrogens (estradiolacetate, 17α-estradiol, 17β-estradiol, 17α-ethynylestradiol, and estriol) were below the limit of detection (1.7 ng/L) in all samples. Estrogens are naturally excreted by humans, either unconjugated or primarily as inactive glucuronide or sulfate conjugates. However, these conjugates can be rapidly cleaved and metabolized into their steroidal active parent compounds by enzymes. The absence of estradiolacetate (at detectable levels) in the wastewater samples can be explained by the fact that its use as a pharmaceutical is not common (not within the top 50 of dispensed pharmaceuticals in Australia ). 17α-Estradiol was also not detectable in all the samples. No reports exist of these compounds in Australian wastewaters. The occurrence of estriol in wastewater has been reported in several studies up to 318 ng/L , whereas it was undetected in other studies . The main active compound of the contraceptive pill 17α-ethynylestradiol had not been detected in previous studies [27,39], and, if detected, the concentrations were in the very low ng/L range. 17β-Estradiol was also undetected in previous studies [23,39] or was found in the low ng/L range  with a maximum concentration of 16 ng/L . Batch experiments by Ternes et al.  showed that 17β-estradiol is degraded quickly and easily to estrone. A concentration of 39 ± 5 ng/L was measured in the present study for estrone in the raw wastewater. Estrone is, in general, apart from the less potent estriol, the most abundant steroidal estrogen in treated wastewaters, with concentrations reported in the range of 15 to 54 ng/L [39,41]. According to Johnson and Sumpter , the most likely pathways for estrone to occur in sewage are the degradation of the glucuronide and sulfonide conjugates and of 17β-estradiol in the sewer system.
The investigated MBR removed 87% of the received influent concentration of estrone (see Fig. 4), which is a promising result. Reported removal rates for CAS plants have generally ranged from 64  to 85%  as depicted in Figure 5. A similar high removal efficiency of 91% was reported by Hu et al. . However, a lower removal efficiency of 64% was observed by Cespedes et al. . In general, no significant trend can be observed between CAS and MBR treatment when comparing values reported in the literature. Negative removal efficiencies for estrone have also been reported  and are believed to be caused by further degradation of 17β-estradiol and cleavage of glucuronide and sulfonide conjugates of estrone during wastewater treatment. The log Kd values measured by Carballa et al.  suggest that sorption to biomass may play a role for estrone removal from the aqueous phase. However, no estrone was detected in the present study in the biomass phase. This indicates that the main mechanism of removal for estrone was biodegradation. Only a few studies could be found investigating the adsorption of estrone to biomass in sewage treatment plants. Reported concentrations of adsorbed estrone range from less than 1 ng/L  up to 11.8 ng/g dry biomass .
Estrone was found at a concentration of 5 ± 2 ng/L in the final MBR effluent. This is similar to many other studies, where estrone was detected in concentrations between 3 ng/L  and 17 ng/L , while estriol and 17β-estradiol were either not detected or in the very low ng/L range [2,27,44]. The low concentrations reported for 17β-estradiol in the effluent is thought to be due mainly to the fast degradation of 17β-estradiol to estrone. Estrone was also the main estrogen detected in a study of German rivers . Hohenblum et al.  investigated surface water samples in Austria that confirmed the presence of estrone with mean concentrations of 0.58 ng/L. As seen in the present study and others and summarized by Johnson and Sumpter , estrone has been identified as the most environmentally important estrogen despite its lower potency compared to 17β-estradiol, as it is detected more frequently and in higher concentrations than 17β-estradiol. However, 17β-estradiol and 17α-ethynylestradiol can play a minor but still important role for the estrogenic activity. Estriol seems to give rise to less of a concern for the environment, as it is, in general, released in low concentrations in the environment and shows a relatively low potency compared to other steroidal hormones .
Testosterone was detected in the influent at a concentration of 98 ng/L and in the MBR biomass at a concentration of 248 ng/g. It was not detected in any other samples. As suggested by Kirk et al. , the main pathway for androgens entering the sewage system with domestic input is most likely through human excretion. Concentrations of androgens in sewage are expected to be much higher than for estrogens because of the higher excretion rates in humans. However, except for testosterone, no other androgens were detected in any of the samples; this could be due to the limited sensitivity of the GC-MS method (limit of detection = 1.7–17.5 ng/L). The measured concentrations indicate that sorption may play an important role in the removal process, which can be explained by the potential of testosterone for sorption to biomass with its relatively high log KOW value of 3.3 . Very little literature is available on the target androgens in wastewater, presumably because there is much less concern about androgenic activities compared to estrogenic activities caused by trace contaminants in the environment. However, the present study and others [2,27] suggest that androgens are removed effectively during secondary treatment and that the levels in secondary effluent of testosterone, etiocholanone, and androsterone are not higher than a few ng/L. No data on the most potent androgen dihydrotestosterone could be found.
Comparison of results: Chemical and biological analysis
A comparison of the results from the chemical and biological analysis was made. Chemicals detected by chemical analysis (GC-MS) known to show endocrine-disrupting potential are 4-tert-octylphenol and estrone. In order to assess which chemicals contributed to the estrogenic activity, EEqs were calculated with the respective potency (potency of 17β-estradiol = 1) of 0.1 for estrone  and 0.0001  for 4-tert-octylphenol. Results are shown in Figure 6.
Using the calculated EEq values of the chemicals analyzed by GC-MS, the results compare well with overall estrogenic activity measured by the yeast screen assay. Estrone is the major contributor for the influent, MBR aqueous, and effluent sample, which was similar to results reported by Salste et al. . Therefore, the removal efficiencies of estrone (87%) and the decrease of the overall estrogenic activity (78%) were in the same range. The activity of 4-tert-octylphenol is expected to be a minor contribution because of its much lower potency of 0.0001 times that of 17β-estradiol. Studies from Japan  reported 17β-estradiol as the main contributor to estrogenic activity in wastewater influent, while for the effluent, estrone was the major contributor. These differences may partially be explained by the fact that the temperatures were higher in Australia during summer, which may have caused faster transformation of 17β-estradiol to estrone in the sewer system. Temperature can significantly affect the rate of degradation of hormones during activated sludge treatment . The estrogenic activity for the biomass component could not be explained by chemical analysis. It is very likely that nonylphenols, which are reported to adsorb to biomass up to several μg/g dry biomass with a relative potency of 0.00014 , contributed to this activity. Also, it could not be ruled out that concentrations less than the analytical detection limit (<1.7 ng/g dry biomass) for the potent estrogens 17β-estradiol and 17α-ethynylestradiol were sorbed to the biomass. The high measured overall androgenic activity could not be explained by the results obtained by chemical analysis for the androgens. This could be partially due to the presence of the target androgens below the analytical detection limit or due to other unidentified chemicals with androgenic activity. Since some studies have detected the androgens androstendione and testosterone at concentrations less than 200 ng/L and other studies have observed high androgenic activity (> 1,000 ng/L) [14,18], it is likely that other chemicals could be present in sewage showing androgenic activity, either at high concentrations or with strong androgenic potency.
Trace contaminants belonging to the classes of PPCPs and EDCs were extracted successfully from aqueous and solid phases of wastewaters undergoing treatment by MBR. The extraction method involved SPE for aqueous samples and ultrasonic solvent extraction and cleanup by SPE for the biomass component. Gas chromatography coupled with mass spectrometry was used for chemical analysis and was suitable for detecting most of the target analytes in the low ng/L and ng/g range in aqueous and solid samples. Estrogenic and androgenic activity was measured using the estrogen and androgen yeast screen bioassays. The PPCPs ibuprofen, triclosan, and salicylic acid and the EDC 4-tert-octylphenol were detected in raw wastewater in the μg/L range. Estrone and testosterone were the only steroidal estrogen and androgen detected. The MBR showed high removal rates for ibuprofen, salicylic acid, triclosan, and 4-tert-octylphenol of 99, 99, 93, and 98%, respectively, and 87% of estrone was removed. The estrogenic activity was reduced from 4.82 ng/L EEq for the raw influent to 1.04 ng/L EEq, giving a removal efficiency of 78%. Androgenic activity was several orders of magnitude higher (1,177 ng/L TEq) in the raw wastewater and decreased by 98% to 21 ng/L TEq. These high removal rates were attributed mainly to the high SRT of the MBR of 30 d. All removal rates were comparable or better than those from CAS processes reported from different studies. Despite the high removal efficiency, measurable amounts of the PPCPs and of estrone were released into the environment, with the effluent still showing estrogenic and androgenic activity. For the influent, the MBR aqueous, and the effluent, estrone was identified as the major compound responsible for the estrogenic activity. The biomass component showed significant concentrations of salicylic acid, triclosan, and 4-tert-octylphenol. Estrogenic and androgenic activity was also measured in the biomass extracts, indicating that processed sludge applied to land could be a possible source of environmental contamination. When comparing influent, MBR biomass, MBR aqueous, and effluent concentrations, it appeared that biodegradation was the main removal pathway, but sorption to biomass should also be considered, particularly for triclosan and 4-tert-octylphenol. The high removal rates are encouraging from the point of view of implementing MBR systems as package plant units for the decentralized treatment of effluent from single households or clusters of homes and in the potential application of the treatment of effluent for alternative water management practices such as water reuse.
The authors wish to thank MidCoast Water, New South Wales, Australia, for their help and cooperation in supplying samples.