A guidance for the enrichment of micropollutants from wastewater by solid-phase extraction before bioanalytical assessment

Abstract

Background

Wastewater can contain a complex mixture of organic micropollutants, with both chemical analysis and effect-based methods needed to identify relevant micropollutants and detect mixture effects. Solid-phase extraction (SPE) is commonly used to enrich micropollutants prior to analysis. While the recovery and stability of individual micropollutants by SPE has been well studied, few studies have optimized SPE for effect-based methods. The aim of the current study was to develop and evaluate two standard operating procedures (SOPs) for the enrichment of micropollutants in preparation for chemical analysis and bioanalysis, one covering a broad range of chemicals and the other selective for estrogenic chemicals.

Results

Pristine surface water spiked with > 600 micropollutants was used to develop a generic extraction method for micropollutants with a wide range of physiochemical properties, while water spiked with estrogenic chemicals was used to identify a selective extraction method. Three different SPE sorbents were tested, with recoveries of individual chemicals and effect in assays indicative of mutagenicity, estrogenic activity, and fish embryo toxicity assessed. The sorbent HRX at pH 7 was selected for the generic extraction method as it showed the best recovery of both individual chemicals and effect in the bioassays. The sorbent HLB at pH 3 showed optimal recovery of estrogenic chemicals and estrogenic activity. The two optimal SPE methods were applied to spiked and unspiked wastewater effluents, with the concentrations of detected chemicals and observed effects similar to those of previous studies. The long-term storage of both extracts and SPE cartridges for estrogens and estrogenic activity after extraction with the HRX and HLB methods were evaluated, with estrogenic effectiveness close to 100% after 112 days when HLB was used.

Conclusions

HRX is recommended for generic extraction, while HLB is optimal for the selective extraction of estrogenic micropollutants. However, if a laboratory only wants to use a single SPE sorbent, HLB can be used for both generic and selective extraction as it yielded similar chemical and effect recovery as HRX for a wide range of micropollutants. This paper is supplemented by the final SOP that includes a variant for generic extraction and one for the extraction of estrogenic chemicals.

Background

Human expectations of well-being, economic and societal growth, and lifestyle have fostered the development, production, and use of more than 350,000 chemicals [1]. A large amount of these chemicals and their unintended transformation products are released to surface waters by municipal and industrial wastewater, with potential adverse effects on the production of drinking water and on ecosystems [2,3,4,5,6,7]. Due to the diversity of chemicals present, monitoring a low number of prioritized micropollutants, e.g., according to the Water Framework Directive in surface waters, is insufficient to investigate and characterize wastewater pollution and mixture risks in its entirety [8]. Efficient monitoring methods should capture the total chemical burden in water samples, including the mixture effects of the many chemicals present. Effect-based methods, which include in vitro bioassays as well as plate-based in vivo assays, are recommended for monitoring chemical water quality as they can detect the mixture effects of all active chemicals in a sample [9]. Effect-based methods indicative of different stages of cellular toxicity pathways, including induction of xenobiotic metabolism, receptor-mediated effects, adaptive stress responses, and apical effects, have been applied to monitor wastewater, surface water, and drinking water quality [e.g., 10–12]. Thus, a combination of effect-based methods and wide-scope chemical screening methods is necessary to identify hazardous micropollutants and mixtures [8, 13,14,15].

Micropollutants are generally present in the nanogram per liter-to-microgram per liter concentration range in treated wastewater effluent, surface water, and drinking water [16,17,18]. While it is possible to inject samples directly in the liquid chromatography-mass spectrometry (LC-MS) system for analysis, detection of low micropollutant concentrations is not possible without enrichment. Further, while unenriched or native wastewater samples can be run in bioassays [19, 20], the effect of micropollutants cannot be differentiated from the effect of other components in water, such as metals and salts. Therefore, enrichment prior to both bioanalysis and chemical analysis is recommended, with solid-phase extraction (SPE) widely used to enrich micropollutants and exclude other water matrix components, such as metals, salts, and other inorganics. High extraction factors can be achieved with SPE, making it possible to detect micropollutants and test different water samples, including drinking water, recycled water, and surface water, in a range of bioassays [6, 10, 21]. Enrichment with SPE is established in chemical-analytical methods [22], but the regulatory use of SPE in effect assessment is not yet implemented in any directives. An important first step toward this is the standardization of the extraction method.

Assessing the applicability of SPE for bioanalytical and chemical analysis of micropollutants in the aquatic environment requires comprehensive knowledge on chemical recovery and their effects in bioassays [23]. As environmental samples contain thousands of known and unknown compounds, only effect-based methods can obtain information on the mixture toxicity of the bioactive substances in these samples [8]. Ideally, an enrichment method should therefore transfer the chemical mixture in the environment, minus the matrix components, to the extract unaltered to avoid under- or overestimating the adverse effect potential. Further, enrichment for bioanalytical purposes also needs to prevent any blank effects from, e.g., the SPE sorbent material or the solvent, as bioassays cannot differentiate between the effect from micropollutants in an extract and the effect from contaminants introduced during SPE [24]. A method producing high blank levels resulting in lower sensitivity or a method with poor compound recoveries is not suitable either.

A range of SPE sorbents have been used to extract micropollutants from water prior to bioanalysis, including HLB, HR-X, MCX, C18, and ENVI-Carb [e.g., 19, 25–28], with a recent review showing HLB as the most commonly applied SPE sorbent for water quality monitoring [29]. Several studies have evaluated the impact of factors, such as sample volume, SPE sorbent material, and SPE conditioning and elution solvents on bioactivity [30,31,32]. For example, differing activity in the same sample extracted using different SPE sorbents has been observed [19, 26]. Therefore, it is important to identify optimal extraction methods for a diverse set of micropollutants with different physiochemical properties, as well as toxicological relevant micropollutant classes, such as estrogenic chemicals. While chemical recovery by SPE can be evaluated using internal standards, effect recovery is more difficult to determine because internal standards cannot be added as their effect cannot be separated from micropollutants in the sample. Within the literature, different approaches have been used to evaluate effect recovery. Most studies spike a water sample with a chemical cocktail before SPE and compare the effect in the extract to the predicted effect based on the concentration of chemicals detected in the extract or the nominal concentration of spiked chemicals [e.g., 33, 34]. Another study attempted to compare the effect of unenriched water and SPE extracts for unspiked water samples, but the presence of other matrix components in the unenriched water samples made the comparison difficult [19]. Therefore, to properly evaluate effect recovery, it is necessary to determine the effect of a spiked sample before and after extraction [24].

Preserving the sample and ensuring the stability of the micropollutants from sample collection to analysis are of utmost importance. The sampling and preservation methods must be optimally tailored to the target analytes to avoid changes to the sample and the analytes. The stability of polar and non-polar micropollutants and their stabilization in water samples for chemical analysis are relatively well studied and standardized [17, 35, 36]. The on-site preservation procedures are always tailored to individual target substances. However, the complex mixtures in environmental samples do not allow for specific preservation. The addition of preservatives may cause unknown chemical reactions in the sample, potentially causing a change in toxicity. As an example, sample acidification is a common pre-treatment step to reduce microbial activity and increase the extraction of weakly acidic micropollutants [29]. Yu et al. [37] found that sample acidification altered androgenic activity and oxidative stress response in wastewater, while estrogenic activity was similar in both acidified and unacidified samples. Further, Šauer et al. [38] observed no difference in androgenic and anti-androgenic activities in wastewater effluent after acidification. Few systematic studies have considered the chemical stability of organic compounds on solid-phase materials [23, 39, 40], with the stability of estrogenicity and estrogenic micropollutants on SPE cartridges not comprehensively investigated.

The aim of this study was to develop and evaluate a standard operating procedure (SOP) for the enrichment of micropollutants contained in wastewater for chemical analysis and bioanalysis. Unlike chemical analysis, few studies have considered effect recovery or identified optimal SPE extraction methods for bioanalysis. The SOP should have two variants for unbiased enrichment of diverse micropollutants (approach A: generic extraction) and for selective extraction of estrogenic micropollutants (approach B: selective extraction). Estrogenic micropollutants were targeted due to their potential to induce effects on reproduction of fish and other aquatic species at low concentrations [41], with the selective method aiming to reduce matrix-related cytotoxicity and masking effects that may occur from extracting a wide range of chemicals. Initially, SPE methods were compared for generic and selective extraction of micropollutants from water by considering commonly used SPE sorbents and pristine surface water as the background for the recovery of > 600 micropollutants. The developed SOP was evaluated by applying it to different types of wastewater. Finally, the stability of estrogens and estrogenic activity in both SPE cartridges and solvent extracts were evaluated over time.

Materials and Methods

Spiked mixture

Two spiked mixtures, one with 620 organic micropollutants (referred to as “water mix,” short “WM,” Table S1) and another with 56 estrogen-active steroids (“steroid mix,” short “SM,” Table S1), were used in the current study to determine chemical and effect recovery. The two mixtures were combined for the generic extraction procedure (676 chemicals, Table S1), while the steroid mix was only spiked for the specific extraction procedure for estrogenic compounds. The final concentrations per chemical added to the water samples were 500 ng/L per chemical of the 620 WM chemicals and 10 ng/L per chemical of the 56 SM chemicals.

For spiking of the wastewater samples, the added mixture concentration was adjusted so as not to exceed an estradiol equivalent concentration of 10 ng/L to ensure that the chemical mixture was not overdosed with respect to the bioassays. The ratio of WS and SM was 1:9 due to high estrogenic potential of the xenoestrogens in the WM.

Sample collection

Both surface water and wastewater effluent samples were used for SPE method selection and validation. Pristine surface water from Wormsgraben in the Harz Mountains, Germany, was used as the water matrix for the SPE selection experiments. Using ultrapure water to develop new SPE methods is not recommended since possible interference with the natural sample matrix is not considered and the lack of relevant ionic strength and background natural organic matter may also lead to pH instability and precipitations. Wormsgraben water has previously been shown to have low micropollutant contamination [24], which is why it was selected in the current study. It is termed “pristine surface water” throughout the paper.

Wastewater effluent samples from industrial, municipal, and hospital wastewater treatment plants (WWTP) covering different levels of treatment were collected to assess the applicability of the selected SPE methods. Table 1 lists the different types of WWTPs, number of collected samples, and the applied sampling techniques. The samples were collected at the official sampling points of the WWTPs or as grab samples at the WWTP outlet. The grab samples were obtained with a stainless-steel bucket (10 L). All samples were stored in stainless-steel barrels (30 L) at 4 °C. The bucket and the barrels were cleaned with LC–MS grade water and methanol. The residual methanol was evaporated in a fume hood in the laboratory or before reuse of the bucket in the field. The bucket and the sampling containers were not rinsed with the samples. The two municipal WWTPs were sampled before and after the quaternary treatment step, which was ozonation or sand filtration. The sampling time offset before and after the treatment step was 15 min to consider the hydraulic residence time and to ensure that the same water parcel was sampled. The wastewater samples were filtered using glass fiber filters (Whatman GF/F, nominal pore size: 0.7 µm) within 48 h after sampling. As a result, any chemicals and effects associated with suspended particulate matter will not be retained by the extraction method. After filtration, the samples were aliquoted to subsamples of 1000 mL and 2000 mL and stored in brown borosilicate glass bottles (Schott) at 4 °C. Ultrapure water (Honeywell CHROMSOLV LC–MS) was processed similar to the WWTP samples to obtain blank samples. Additionally for the stability experiments, municipal wastewater (PN8) was sampled, with 60 L collected and stored at 4 °C.

Table 1 Types of sampled wastewater treatment plants (WWTP), sample IDs, and sampling types

Solid-phase extraction

Three types of SPE sorbents were selected for both the generic and selective extraction of spiked chemical in pristine water samples. The initial selection of SPE sorbents and extraction procedures was based on a literature study. For generic extraction, commercially available SPE cartridges Chromabond® HR-X (6 mL/200 mg, Macherey Nagel), Oasis® HLB (6 mL/200 mg, Waters), and Oasis® MCX (6 mL/150 mg, Waters) were used. For selective extraction, commercially available SPE cartridges Oasis® HLB (6 mL/200 mg, Waters), Supelclean™ ENVI-Carb (ECarb) (6 mL/500 mg, Merck), and TELOS® C18 (6 mL/500 mg, Kinesis) were used. The cartridges were conditioned using 1 × 5 mL ethyl acetate (LC–MS grade, Merck or Honeywell), 1 × 5 mL methanol (LC–MS grade, Merck or Honeywell), and 10 mL ultrapure water (Chromasolv™ Ultra LC–MS, Honeywell). The samples were extracted at pH 7 for the HLB7N, HLB7Y, and HRX7N methods and at pH 2 for the MCX2N method. The coding is related to the procedure, e.g., HLB7N encodes extraction with Oasis® HLB at pH 7 with no washing. The samples were extracted at pH 3 for the HLB3Y, ECarb3Y, and C183Y methods. An aliquot of 1000 mL of spiked and unspiked pristine surface water was extracted by SPE, with a final extract volume of 1 mL and consequently an extraction factor (EF) of 1000. Details of the washing step and elution conditions are provided in Table 2. The water:methanol mixture used for the wash steps was selected based on preliminary experiments outlined in Section S1 and Figure S1. The final extracts were dissolved in methanol.

Table 2 Overview of SPE material and procedures

After the development stage, HRX7N was used as the generic method and HLB3Y was used as the selective method and applied to wastewater samples to assess the recovery against a background of realistic matrix. An aliquot of 2000 mL was extracted by HRX7N, while 1000 mL was extracted by HLB3Y. The final EF was 1000 for both HRX7N and HLB3Y.

For stability experiments, 1000 mL of sample PN8 was extracted per cartridge (both HRX7N and HLB3Y), with a final EF of 1000. One batch was extracted and eluted on the same day as sampling (T0). The rest of the water was extracted on day 3 after sampling to simulate transport and laboratory processing delays. Some of the cartridges were eluted immediately after extraction (T3 cartridge), with the others stored at -20 °C and then eluted after 28 (T28 cartridge) or 112 days (T112 cartridge). The cartridges were supposed to be eluted after 56 days, rather than 112 days, but elution was delayed due to COVID-19 lockdowns. The sample extracts were analyzed within a week of elution. Further, some of the samples eluted immediately after extraction (T3) were stored at −20 °C for 28 (T28 extract) or 112 days (T112 extract) prior to analysis. Storage of extracts is standard, but storage of cartridges with later elution for preservation is suggested as a method for longer storage [39]. An overview of the study is provided in Fig. 1 with some details in Figure S2.

Fig. 1

Overview of the current study, which includes the development of non-selective and selective SPE methods, application of optimal SPE methods to wastewater samples, and assessment of the stability of SPE cartridges and extracts over time. Only unspiked wastewater was used for the stability experiments

Chemical analysis

The chemical analysis of all SPE extracts and reference standards was performed on an UltiMate 3000 LC-system (Thermo Scientific) coupled to a quadrupole-Orbitrap MS (QExactive Plus) with a heated ESI source (LC-HRMS) by a target screening approach. Some of the estrogen-active steroids were analyzed using an Agilent 1290 LC-system coupled to a Sciex QTrap 6500 with TurboV ESI-source in the Scheduled MRM™ mode. Further details, including instrumental settings and methods for data evaluation, can be found in Section S2, with information about method development in Finckh et al. [5]. The isotope-labeled internal standards were only spiked after SPE and only in the subsamples to determine the chemical recovery as they can contribute to the effect in the bioassays.

Bioanalysis

To represent different modes of action, the following set of bioassays was selected to evaluate effect recovery and stability. The chosen bioassays agree with the minimal requirements for bioassay batteries for effect monitoring [10, 13, 43,44,45]. All bioassays can be established in routine laboratories and applied on high-throughput platforms.

Non-specific activity:

• Fish egg test (48 h) with Danio rerio [46] and fish embryo test (96 h, FET) with Danio rerio [47] (development and application experiments, see Fig. 1);

• Combined algae test (CAT) with Raphidocelis subcapitata [48,49,50,51,52] (application experiment only)

Genotoxic activity:

Ames fluctuation test with strains Salmonella typhimurium TA98 und YG7108 [53,54,55,56] (development and application experiments).

Estrogenic activity:

• ERα-CALUX [57,58,59,60] (development and application experiments),

• ERα-GeneBLAzer [61] (development, application, and stability experiments).

The bioassays were carried out according to their published or standardized procedures. The specific details and differences to the standard test protocols, as well as bioassay data evaluation, are described in Section S3.

Dose metrics after SPE

In practice, the terms extraction, enrichment, or concentration factor are often used in different meanings, causing confusions about included enrichment and dilution steps. In consequence, bioassay (and chemical) data may not be intercomparable if the extraction factor were calculated differently. In this study, the effect data are reported in the relative extraction factor (REF) [29]. The REF includes all enrichment and dilution steps. As the REF is typically a net enrichment unless the water sample contains a very high load of micropollutants, it is sometimes also called relative enrichment factor.

The enrichment of micropollutants in the SPE is described by the extraction factor, often also called enrichment factor (Eq. 1),

$${\text{EF}}\text{ (extraction)\,=}\, \frac{{\text{V}}_{\text{water}}}{{\text{V}}_{\text{extract}}}$$

(1)

where EF (extraction) is the SPE extraction factor and V is a known volume of a water sample V_water in an adjusted final volume of the extract V_extract. A practical EF (extraction) is 1000, i.e., 1000 mL of water is enriched in 1 mL of extract. In certain cases, dosing in bioassays requires further enrichment, e.g., by evaporation of an aliquot of the extract and re-constitution in a lower volume of, e.g., medium. The additional enrichment (EF [enrichment]) is calculated by Eq. 2:

$${\text{EF}}\left({\text{enrichment}}\right){=}{\text{EF}}\text{(extraction)}\times\frac{{\text{V}}_{\text{extract}}}{{\text{V}}_{\text{enriched extract}}}$$

(2)

For example, an aliquot of 1000 µL of an extract with an EF (extraction) 1000 is evaporated and then reconstituted in 100 µL of the test medium or a solvent, which results in an EF (enrichment) of 10,000.

A defined aliquot of the extract is typically dosed in bioassays in dilution series to investigate the concentration-dependent effects. The extract is (i) dissolved in the test medium, or (ii) submitted in the bioassay together with, e.g., cells and test medium (Eq. 3) and consecutive evaporation of the remaining solvent. The dilution series is achieved in the first case by dilution in the test medium and in the second case by dosing of different volumes of the extract. Equation 3 covers both cases:

$${\text{DF}} \, \left({\text{bioassay}}\right)\,=\,\frac{{\text{V}}_{\text{extract}}}{{\text{V}}_{\text{bioassay}}}=\frac{{\text{V}}_{\text{extract}}}{{\text{V}}_{\text{medium}}\text{+}\left({\text{V}}_{\text{cells/organisms}}\right)\text{+}{\text{V}}_{\text{extract}}}$$

(3)

where DF is the dilution factor. The volume of the bioassay is the sum of the volumes of the different parts of the bioassays (i.e., medium, cells/organisms, and extract). In the case of cell-based assays, the volume of the V_cells ≈ 0 and thus negligible. Bioassay-specific dilutions steps must be considered in the calculation of DF.

The biological effect of environmental samples is dependent on the total concentration of all micropollutants in the mixture and their specific effect potentials [62]. Thus, the effect of the sample can be estimated by a dilution series and directly compared with the effect of reference measurements applying the REF [29]. The REF is a unit analog to concentration units (e.g., mol/L or ng/L), which is formally dimensionless according to the International System of Units (L/L) but is related to L_water/L_bioassay. The REF is calculated using the final EF (Eqs. 1 and 2) and the DF (Eq. 4):

$$\text{REF=EF}\left({\text{extraction}}\right)\times{\text{DF}}\left({\text{bioassay}}\right)$$

(4)

• REF = 1: Concentration of micropollutants in bioassay medium ≙ concentration in original sample.

• REF > 1 (e.g., REF 10): Micropollutants in the bioassay medium are enriched (e.g., tenfold enriched compared to the original sample).

• REF < 1 (e.g., REF 0.1): Micropollutants in the bioassay medium are diluted (e.g., tenfold diluted compared to the original sample).

The REF enables a direct comparison of the results of different bioassays and with reference measurements (single compound and mixture assessment). An EC₁₀ of REF 5 in the fish embryo assay can be compared with an EC₁₀ of REF 5 in the algae assay or an EC₁₀ of REF 5 in a cell-based bioassay. The simulated and real REF for dosing in bioassays is demonstrated in Figure S3.

Effect recovery

The results in the ERα-CALUX and ERα-GeneBLAzer as well as the algal toxicity assays are expressed as bioanalytical equivalent (BEQ) concentrations, which are calculated based on the effect of the sample in units of REF and the effect of the reference compound. The reference compound for the ER assays is 17β-estradiol and the BEQ are termed estradiol equivalent concentration (EEQ) and for the photosynthesis inhibition after 2 h and 24 h in the CAT, the reference compound was diuron and the effects were expressed as diuron equivalent concentration (DEQ).

The effect recovery of the SPE for BEQ was calculated using Eq. 5 [24]:

$${\text{E}}\text{ffect recovery}\text{ of} \, {\text{SPE}}{=}\frac{{\text{BEQ}}_{\text{bio, extract}}\left(\text{water+mix}\right)-{\text{BEQ}}_{\text{bio, extract}}\left({\text{water}}\right)}{{\text{BEQ}}_{\text{bio}}\left({\text{mix}}\right)}$$

(5)

where BEQ_bio,extract(water + mix) is the BEQ of the spiked extract, BEQ_bio,extract(water) is the BEQ of the unspiked extract, and the BEQ_bio(mix) is the BEQ of the spike mixture directly tested in the bioassay.

The effect data of the Ames fluctuation assay and the fish egg and fish embryo assay were transformed to toxic units (TU, Eq. 6):

$${\text{TU}}_{\text{bio}} = \frac{1}{{\text{EC}}_{50}}$$

(6)

where TU_bio is the biological toxic unit and EC₅₀ the effect concentration of the half maximum effect of the concentration–response curve. Using the TU, the effect recovery can be calculated for the Ames fluctuation and the fish egg and fish embryo assay with Eq. 7:

$${\text{E}}\text{ffect recovery}\text{ of} \, {\text{SPE}}{=}\frac{{\text{TU}}_{\text{bio, extract}}\left(\text{water+mix}\right)-{\text{TU}}_{\text{bio, extract}}\left({\text{water}}\right)}{{\text{TU}}_{\text{bio}}\left({\text{mix}}\right)}$$

(7)

Results and discussion

Selection of SPE method

Chemical recovery

Recovery of individual micropollutants by the different generic and selective SPE methods is shown in Tables S2 and S3, with an overview presented in Fig. 2 and Table S4. Not all the 676 spiked micropollutants were detected after extraction, with up to 464 of the spiked micropollutants detected after the different generic extraction techniques (Table S2). Chemicals that were not detectable due to issues with ionization, or high method detection limits, had to be omitted in the analysis of chemical recovery, but they might still contribute to the mixture effects and were therefore included in the tables. For the generic method, the best result was observed for HLB7N, with a median recovery of 102%. HLB7Y (median 129%) and HRX7N (median 130%) were significantly different from MCX2N (median 107%) and HLB7N (ANOVA with Dunnett’s post-test, α = 0.05, p < 0.05). The high recovery observed for some chemicals may be due to isobaric compounds or, given that 40 isotope-labeled compounds were used, a mismatch of the internal standard. The HLB7Y method was excluded from further investigation because the results were similar compared to HRX7N, but it had a larger scatter (Fig. 2). Chemical recovery in all three sorbents has previously been studied [e.g., 23, 63, 64]. For example, Schulze et al. [23] reported a median recovery of 96% by HRX, with 246 of the 251 spiked chemicals detected after extraction. Further, Osorio et al. [64] found average recoveries between 65 to 139% and 34% to 60% for charged and neutral micropollutants in HLB and MCX, respectively.

Fig. 2

Recovery of spiked micropollutants in pristine surface water using A the generic method for all micropollutants (n = 458) and B the selective method for estrogenic micropollutants (n = 25). Each box extends from the 25th to the 75th percentile, with the horizontal line indicating the median

For the selective methods, the recovery rates of estrogenic micropollutants were in the range of the desired recovery of 100% for C183Y and HLB3Y (Fig. 2, Table S4), with 25 of the 56 spiked micropollutants detected. Analysis of variance (ANOVA with Dunnett’s post-test, α = 0.05) showed that the median recovery of 54% of the estrogenic micropollutants extracted using the ECarb3Y method was significantly (p < 0.05) different from the other two methods C183Y (median: 98%) and HLB3Y (median: 103%). The method with the lowest variation in recovery was the C183Y method. Goeury et al. [65] also found good recovery of spiked hormones and xenoestrogens in tap water using HLB (93–108%), with lower recovery for C18 (< 5–104%).

Effect recovery

Effect recovery is presented in Table 3. The concentration–response curves are provided in Figures S4–S11, with the EC and EEQ values presented in Table S5. The effect recovery in the Ames fluctuation test with Salmonella typhimurium TA98 + S9 was only 14–15% in the HLB7N and HRX7N methods (no significant difference, Wilcoxon test, α = 0.05, p > 0.05, two tailed). No recovery could be calculated for MCX2N as the sample extract did not have an effect up to the maximum REF. The potentially mutagenic micropollutants in the water mix appear to strongly to the solid phases and could not be eluted with the chosen methods. Better effect recovery was observed for the fish embryo test with Danio rerio, with 60% in the HLB7N method and 85% in the HRX7N method (no significant difference, Wilcoxon test, α = 0.05, p > 0.05, two tailed). Similar to the Ames fluctuation test, no recovery could be calculated for MCX2N.

Table 3 Effect recoveries of spiked micropollutants in the applied biotests extracted with the extraction methods HLB7N, HRX7N, and MCX2N (spiked with the water mix and the steroid mix: [W&SM]) as well as HLB3Y, ECarb3Y, and C183Y (spiked with the water mix and the steroid mix [W&SM] or with the steroid mix only [SM])

The effect recoveries in ERα-CALUX ranged from 269% with the HRX7N method to 378% with the HLB7N method. The significantly higher effects in the spiked samples compared to the combined water and steroid mix (Table S5) and the resulting effect recovery > 200% are possibly due to xenoestrogenic micropollutants in the steroid mix and some apparently strongly estrogenic micropollutants in the water mix and the sample matrix (natural organic substances and macromolecules). The unspiked extract also had a variable response in ERα-CALUX (Figure S6), possibly contributing to the high effect recovery values. In the ERα-GeneBLAzer, the recoveries ranged from 146% with the MCX2N method to 217% with the HLB7N method. Although these values are also high, they show that the ERα-GeneBLAzer resulted in recoveries closer to the optimal 100% when used in the generic extraction process. The standard errors of the effects were up to twofold higher in the ERα-CALUX than those in the ERα-GeneBLAzer (Table S5). Overall, there was little significant difference in terms of extraction and test methods (Friedmann test with Dunn’s post-test, α = 0.05, p > 0.05). A significant difference in effect recovery was only determined between the MCX2N extracts tested in ERα-GeneBLAzer and the HLB7N extracts tested in ERα-CALUX (Friedmann test with Dunn’s post-test, α = 0.05, p = 0.016). The observed recovery for ERα-GeneBLAzer using HRX7N is higher than that in the study by Neale et al. [24], where 35% effect recovery was reported. While steroids were also included in the spiked mixture of 579 micropollutants in the study by Neale et al. [24], more steroids were spiked in the current study, potentially contributing to the improved recovery.

The effect recoveries of the selective extraction procedures for estrogenic substances were determined with samples that were spiked with both the water and steroid mix and the steroid mix only. No effect recovery was observed after extraction with ECarb3Y. The samples spiked with the water and steroid mix achieved a recovery of 280 ± 18% in HLB3Y and 22 ± 4% in C183Y in ERα-CALUX. The samples spiked with the steroid mix only had a recovery of 76 ± 33% in the HLB3Y method and 49 ± 40% in the C183Y method. Focusing on eight estrogenic compounds, Leusch et al. [34] found 98% recovery based on the predicted effect of chemicals detected in the extract in ERα-CALUX using HLB. The samples tested in the ERα-GeneBLAzer had a recovery of the water and steroid mix of 289 ± 12% in the HLB3Y and 90 ± 38% in the C183Y method. The effect recovery of the steroid mix was found to be 179 ± 71% in the HLB3Y method and 114 ± 43% in the C183Y method. Comparable to the generic methods, the standard error of the ERα-CALUX was up to three times higher than that of the ERα-GeneBLAzer (Table S5). The ERα-GeneBLAzer was therefore significantly more robust and reproducible than the ERα-CALUX.

Selecting appropriate generic and selective SPE methods

The prerequisite for selecting an appropriate SPE method is the best possible recovery of the amount and effect of spiked micropollutants, as well as a high level of precision. The different sorbent properties can affect the extraction and instrumental analysis of micropollutants. For example, chemicals can bind strongly with the solid-phase material, or they are only insufficiently retained and break through after reaching the solid-phase capacity [66, 67]. This leads to losses in chemical recovery. For instrumental analysis, matrix effects in mass spectrometry can cause ion suppression or an increase in the number of ions, resulting in lower or higher recoveries. The interval of acceptable recovery of commonly detected micropollutants is between approximately 60 and 120% [23, 24]. This means that the 25th and 75th percentiles (i.e., middle 50% of the recovery data) should be within this interval. The violation of the lower limit is to be given greater weight than that of the upper limit. A recovery that is too low causes an underestimation of the true substance concentration and consequently the adverse effect potential. For the generic method, only HRX7N and HLB7Y met the lower limit (Fig. 2A). As HRX7N also showed comparable or better effect recovery to HLB7N and MCX2N (Table 3), HRX7N was selected as the generic method. In principle, the HLB7N process can be used as an alternative as it also showed similar chemical and effect recovery. The MCX2N method has proven to be unsuitable due to the comparatively poor effect recovery results in the Ames fluctuation test, fish egg test, and fish embryo test. While a limited number of endpoints were considered in the current study, a previous study using large volume SPE with HRX sorbent found that effect recovery was within a factor of 2 of 100% recovery for assays indicative of activation of the aryl hydrocarbon receptor, binding to peroxisome proliferator-activated receptor gamma and hormone receptor-mediated effects [24]. Therefore, HRX7N should also be suitable for other endpoints.

The HLB3Y method was selected for the extraction of estrogenic compounds. HLB has previously been identified as the optimal sorbent for extraction of steroid hormones [68]. The chemical recoveries of the HLB3Y and C183Y methods were not significantly different (Fig. 2B), but effect recoveries in the HLB3Y method were higher than those from the C183Y method (Table 3). TELOS™ C18/ENV was previously identified as a suitable method for the enrichment of estrogenic substances [19], though it seems that C18 alone is not sufficient.

Applying SPE methods to wastewater samples

Chemical analysis

A total of 286 micropollutants were detected at least once in the unspiked extracts enriched with HRX7N with a maxiumum of 224 micropollutants per sample (Fig. 3A, Table S6). In the spiked extracts 328  micropollutants were detected at least once with a maximum of 288 micropollutants per sample (Fig. 3B, Table S6). The concentrations of individual micropollutants extracted with HRX7N are presented in Table S6. The number and concentration range detected in the unspiked samples are comparable to other wastewater effluent monitoring studies [69, 70]. While 458 micropollutants were detected by LC-HRMS, 172 micropollutants in the unspiked extracts and 130 micropollutants in the spiked extracts were not identified in any of the samples. This could be due to the concentrations in the extracts being below the detection limits. Sixty-three micropollutants were identified in the unspiked ultrapure water sample at low ng/L concentrations and are typical laboratory contaminants such as phosphoric acid esters (flame retardants and plasticizers), lauryl diethanolamide (cosmetics), or benzotriazole (corrosion inhibitors) or represent measurement artifacts due to carry over of substances from the previously measured calibration standards. Carryover was < 0.1% in all cases. The chemical concentrations in the unspiked and the spiked ultrapure water extracts were not significantly different (paired t test, p = 0.1426, with significant pairing [p < 0.05]), indicating that the spikes did not add substantially to the chemical burden. This was expected because the totally added chemical spike was adjusted to trigger not more than 10 ng/L EEQ, which is a very small chemical quantity given than many highly potent estrogens and xenoestrogens were contained in WM and SM. Consistent with this observation, the differences in the spiked and unspiked PN samples were small (paired t test with mostly no significant difference but significant pairing) and within the variability of the recoveries of two independent repeats. Hence, Fig. 3A and B is very similar and represents mainly the variability of replicate experiments, while it delivers more meaningful results for the bioassays as described in the following section.

Fig. 3

Comparison of the levels of detected micropollutants in A unspiked and B spiked samples extracted with the HRX7N generic method and extraction of estrogenic micropollutants in the unspiked samples with C the HRX7N generic method and D the HLB3Y selective method. Each box extends from the 25th to the 75th percentile, with the horizontal line indicating the median

Focusing on estrogenic micropollutants, only 7 of the 56 estrogenic micropollutants analyzed were detected in the unspiked samples (Fig. 3C and D, Table S7). The substances were bisphenol A, estrone, p-chlorocresol, 2-phenylphenol, butylparaben, methylparaben, and propylparaben. The remaining compounds were not detected in the wastewater samples, which might be attributable to high signal background and strong matrix effects.

Municipal, industrial, and hospital WWTP effluents were considered in the current study. Focusing on the unspiked samples extracted using HRX7N, the highest number of detected chemicals and median concentration were found in municipal wastewater (PN6), with 224 chemicals detected and a median concentration of 77 ng/L (Table S6). Sand filtration (PN7) did not significantly reduce the chemical load (Friedmann test, α = 0.05, p > 0.05), with 220 chemicals detected and a median concentration of 73 ng/L. In contrast, ozonation and sand filtration significantly reduced the median chemical concentration from 32 ng/L in PN2 to 19 ng/L in PN3 (Friedmann test with Dunn’s post-test, α = 0.05, p = 0.0044). A total of 177 chemicals were detected in hospital wastewater after membrane filtration (PN5), with a median concentration of 51 ng/L. As expected, a variety of pharmaceuticals were detected including not only sulfamethoxazole, candesartan, tramadol, lidocaine, and metoprolol but also hexa(methoxymethyl)melamine (HMMM) at a concentration of 5 μg/L. Higher concentrations of HMMM have been detected in surface water from China, stormwater runoff from the USA, and wastewater from Germany [reviewed in Ref. 71]. HMMM is used as a precursor for melamine resin surface coatings. It was also detected in industrial wastewater (PN1) in the current study, but at a concentration 30 times lower than that in PN5. A discharge of non-crosslinked HMMM due to intensive surface cleaning in hospitals may explain the high concentration.

Bioanalysis

The HRX7N extracts were tested in all assays, while the HLB3Y extracts were only analyzed in the two estrogenic assays, ERα-CALUX and ERα-GeneBLAzer. Neither the spiked nor unspiked extracts showed mutagenicity at concentrations up to cytotoxicity in the Ames fluctuation test with Salmonella typhimurium TA98 and YG7108 (Figures S12–S14). The lack of mutagenicity in the final effluent has been observed in other studies [72, 73]. While ozonation can form mutagenic transformation products, post-treatment processes, such as activated carbon, can typically remove mutagenic activity [74]. Further, a concentration–response curve was only observed in industrial effluent (PN 1) for the fish egg test and fish embryo test (Figure S15). The spiked PN2 and PN3 extracts also had a minor effect in the assay, with 10% effect observed at a REF of 20 and 75, respectively (Table S8). The low sensitivity of Danio rerio to the wastewater samples can be explained by the fact that the individual micropollutants in the mixtures are only slowly absorbed into the fish egg or fish embryo or not at all [75].

In contrast, all spiked and unspiked wastewater samples were active in the combined algae assay, with EC values provided in Table S9 and all concentration–response curves shown in Figures S16–S20. The results were expressed as DEQ (Table 4). Of the unspiked samples, industrial wastewater (PN1) was the most toxic, while hospital wastewater (PN5) and municipal wastewater after ozonation and sand filtration (PN3) were the least toxic. The DEQ values for municipal wastewater were similar to previously reported DEQ values [50, 76]. Growth inhibition after 24 h could only be evaluated in the ultrapure water sample and PN1 as cell growth with increasing concentration was observed for the other wastewater samples (Figures S18–S20). Cell growth reached a maximum in many of the unspiked and spiked samples examined, with a subsequent increase in inhibition due to increasing cytotoxicity.

Table 4 Bioanalytical equivalent concentrations (BEQ_bio) in unspiked (B) and spiked (S) samples extracted with HRX7N and HLB3Y

Removal efficiency after ozonation and sand filtration (PN2 and PN3) and sand filtration alone (PN6 and PN7) was evaluated using the approach outlined in Section S4. Removal efficiency after ozonation and sand filtration was 51% (photosynthesis inhibition after 2 h) and 96% (photosynthesis inhibition after 24 h), respectively, and led to a significant reduction in algae toxicity. The difference can be explained by herbicides (dominant in photosynthesis inhibition after 2 h), which were probably less degraded than micropollutants with non-specific effects. The removal efficiency by sand filtration was much lower, with adverse effects only reduced by 12% (photosynthesis inhibition after 2 h), 28% (photosynthesis inhibition after 24 h), and 13% (algal growth over 24 h). This observation agrees with previous studies, where secondary treatment was unable to efficiently remove photosynthesis-inhibiting herbicides, but ozonation is more suitable [28, 50, 77].

Estrogenic activity in the unspiked and spiked samples extracted with both the HRX7N and HLB3Y methods was analyzed using both ERα-CALUX and ERα-GeneBLAzer. The EC₁₀ values are reported in Table S10, with the EEQ values provided in Table 4. All concentration–response curves are shown in Figures S21–S48. A hierarchical cluster analysis (Figure S49) showed that both extraction methods and bioassays enabled reliable differentiation of unspiked samples with different potency levels. Three clusters were formed: (1) high toxicity (PN1-HLB, PN1-HRX, PN2-HLB, PN6-HLB, PN6-HRX), (2) medium toxicity (PN2-HRX, PN5-HRX, PN7-HLB, PN7-HRX), and (3) low/no toxicity (LC-HLB, LC-HRX, PN3-HLB, PN3-HRX, PN5-HLB, PN5-HRX). Municipal wastewater before ozonation (PN2) fell into both cluster 1 (PN2-HLB) and cluster 2 (PN2-HRX), which may be due to unmasking estrogenic potency by using the selective method HLB3Y. The classification of hospital wastewater after membrane filtration (PN5) into cluster 2 (PN5-HLB) and cluster 3 (PN5-HRX) may be based on an artifact of the analysis of sample PN5-HLB in ERα-CALUX. The measurement was associated with a high degree of uncertainty (Figure S38) and thus resulted in a relatively low toxicity value (Table 4). The comparison of the two assays shows that the ERα-GeneBLAzer detected up to three times higher EEQ_bio than that of the ERα-CALUX. In agreement with the literature [42], the difference of EEQ_bio can be explained by the difference in the potency of estrogens and xenoestrogens between the two assays [29]. As a result, only ERα-GeneBLAzer was used for the stability experiments.

Both extraction methods and assays showed good removal of estrogenic activity by ozonation and sand filtration (PN2 and PN3), with between 80% (ERα-GeneBLAzer-HLB) and 100% (ERα-CALUX-HLB and ERα-CALUX-HRX) removal efficiency observed. Between 18 and 60% removal was observed after sand filtration (PN6 and PN 7). Lower removal was observed with the generic extraction method, which can be explained by a higher proportion of matrix (e.g., organic macromolecules) and xenoestrogenic micropollutants in the HRX7N extracts. Conventional wastewater treatment processes can typically remove between 80 and > 99% of estrogenic activity [e.g., 10, 78, 79]. Ozonation was found to remove 99.7% of estrogenic activity [28], which fits with the current study. In contrast, Hamilton et al. [80] found limited removal of estrogenic activity after sand filtration using a suite of different assays indicative of estrogenic activity, supporting the findings here.

The results for samples extracted using the generic method were more heterogeneous than those of samples enriched using the selective method. The extracts from the HLB3Y process had a higher precision compared to the HRX7N process (Figure S50). The difference can be seen particularly clearly in the samples PN1 (industrial wastewater) and PN2 (municipal wastewater without ozonation). The precision of the HLB3Y method was very good, with little variation in measurements between replicates except for samples PN3 and PN6 (Figure S50). However, when selecting the extraction process, it must therefore be considered whether the additional effort of a specific process is scientifically and economically justifiable. In a laboratory that only analyzes estrogenic activity, the HLB3Y method may be the right choice. In laboratories that carry out a large number of different bioassays, the exclusive use of the HRX7N method may make sense, as the estrogenic activity is not necessarily underestimated.

Storage stability of estrogens and estrogenic activity

Chemical analysis

Only 27 estrogenic active compounds were detected in municipal wastewater effluent (PN8) after extraction (Table S11). The recovery of estrogenic micropollutants after storage was highly heterogeneous, with the results for the generic method more scattered than those for the selective extraction method. Percent recovery is shown in Table S12 and Figures S51 and S52. Recovery was calculated in relation to the reference sample at time T0 (HLB T0, HRX T0) and at time T3 (HLB T3, HRX T3). The detected concentrations in sample PN8 were in the lower range of the typical measurements, close to the detection limits and are therefore associated with an uncertainty that cannot be quantified. Overall, the compounds could be divided into three groups: (1) compounds with similar recovery at all three time points and were stable throughout the storage period (e.g., androsterone, clobetasol propionate, cortisone, and mometasone furoate for HLB3Y and 4-cumylphenol, canrenone and flunisolide for HRX7N); (2) compounds whose recoveries increased or decreased over time or were excessive (e.g., 4-phenylphenol, butylparaben, estrone, and triclosan for HLB3Y and 17α-ethinylestradiol and estriol for HRX7N); and (3) compounds with low recovery (17α-estradiol and 17α-ethinylestradiol for HLB3Y and 17α-estradiol, butylparaben, and ethylparaben for HRX7N). The chemical analysis data are summarized in Fig. 4 as a heatmap with a hierarchical cluster analysis. The result is two main clusters. In main cluster A (from left), all samples at time T112 are grouped into two subclusters (extracts and cartridges). The main cluster B includes three groups: the samples stored on cartridges from both methods at T28 with the extracts at T0, the extracts from both methods at T3, and the extracts from both methods at T28. This means that the samples stored on cartridges at T28 were statistically more similar to the extracts at T0 than the extracts at times T3 and T28. Overall, a general statement about the stability of estrogenic compounds cannot be made based on the highly heterogeneous measurement data.

Fig. 4

Heatmap and hierarchical cluster analysis of the recoveries of 27 estrogen-active micropollutants in PN8 (with T0 as reference), which were extracted using the HLB3Y and HRX7N methods and stored on cartridges or as extracts over 3, 28, and 112 days. The distances are Euclidian and clustered with complete linkage after scaling. Similar color values have a similar distance to their neighboring value (within the samples and between samples), while the red and blue values have a maximum distance from each other. The color values are independent of absolute concentrations

Several studies have considered the stability of individual micropollutants over time. Baronti et al. [81] found between 89 and 93% recovery of estriol, estradiol, ethinylestradiol, and estrone in river water samples extracted with graphitized carbon black SPE cartridges and stored for 60 days at -18 °C. Focusing on pharmaceuticals and drugs of abuse, González-Mariño et al. [82] found acceptable recovery of wastewater samples stored on HLB cartridges at -20 °C after 12 weeks, while Baker and Kasprzyk-Hordern [83] reported no degradation in wastewater samples stored on MCX cartridges for 6 weeks at −20 °C.

Bioanalysis

In contrast to the chemical analysis data, recommendations regarding sample storage can be made based on the bioanalytical assessment. Table 5 shows the recovery of estrogenic activity and cytotoxicity in ERα GeneBLAzer for sample PN8 after 3, 28, and 112 days. Recovery was calculated in relation to the reference sample at time T0 (HLB T0, HRX T0) and at time T3 (HLB T3, HRX T3). The EC₁₀ values are provided in Table S13. The recoveries of the estrogenic potency of the examined samples were all within two standard deviations away from the mean. The values in the HRX7N samples were systematically below the recoveries for the HLB3Y samples, which was likely due to the reduced matrix due to the washing step in the HLB3Y method. Moreover, the estrogenic activity of the samples was very stable over the entire test period of 112 days in the HLB3Y method. An exception is the sample PN8 HLB cartridge T112, which had a significant difference compared to T0 (ANOVA with Dunnett’s post-test based on the effect data, α = 0.05, p > 0.05). Previously, Murk et al. [84] found no change in estrogenic activity in water sample extracts stored in DMSO at -20 °C for 6 weeks. In contrast, Yu et al. [37] found a decrease in estrogenic activity in wastewater extracts stored in ethanol for 1 year regardless of whether the sample was acidified or not prior to SPE. The difference in stability may be due to the extract solvent, with Murk et al. [84] finding ethanol as not a suitable solvent for long-term extract storage as it can evaporate even at −20 °C.

Table 5 Estrogenic activity and cytotoxicity recoveries (%) relative to the activity at time T0 and T3 for wastewater effluent (PN8) in ERα-GeneBLAzer

Cytotoxicity was also very stable in the first 3 days after sampling (T3). The recovery of cytotoxicity fluctuated in both methods, but the fluctuations were generally within two standard deviations from the mean. Increased cytotoxicity was observed over time for HRX7N, particularly after 112 days (Table S13). The HRX7N extracts contained a higher total amount of micropollutants and matrix compared to the HLB3Y extracts. Therefore, transformation of coeluted micropollutants or matrix may have caused the increased cytotoxicity at T112.

Overall, the estrogenicity was maintained at almost 100% over the entire period of 112 days when the HLB3Y method was used. The result was independent of storage of the sample on the cartridge with elution at the final time point or as an extract. The recovery of estrogenic potency in the HRX7N method was slightly lower than that in the HLB3Y method, regardless of storage on the cartridge or as an extract.

Conclusions

Municipal and industrial wastewater is a major entry pathway for organic micropollutants into surface water. The aim of the project was to develop two SOPs applicable to both chemical analysis and bioanalysis, with one method for the enrichment of micropollutants with a wide range of physicochemical properties and another one for the selective enrichment of estrogenic micropollutants. At the end, we developed one SOP with only minor variants A and B (Supplementary Information B). Of the tested SPE sorbents, the HRX7N process achieved the best chemical and effect recovery and was selected for generic extraction, while HLB3Y method was optimal for the selective extraction of estrogenic compounds. These two procedures were applied to effluent samples from industrial, hospital, and municipal WWTPs, with the number and concentrations of detected chemicals and level of effect observed comparable to previous studies. Both variants of the SOP resulted in good reproducibility for chemical analysis and bioanalysis for real water samples.

Finally, long-term storage of samples as either extracts or cartridges was not found to decrease estrogenic activity, with extraction using HLB3Y resulting in very little change in estrogenic activity over 112 days.

While different SPE sorbents are recommended for generic and selective extraction, HLB showed similar chemical and effect recovery as HRX when developing the generic method. Hence, HLB could be used for generic extraction if a laboratory only wants to use a single SPE sorbent as both sorbents are similar co-polymer mixes with slight modifications. Although the SOP for generic extraction was only validated with three bioassays, it is likely that it will also be suitable for other bioassays. As we have also reported recoveries for individual chemicals, one can predict the recovery in a different bioassay by matching the chemical recovery with bioassay data of single chemicals in the bioassay of interest.

These SOPs represent an important step toward standardization and finally the possible implementation of effect-based methods and the assessment of whole mixtures in regulations.