Tandem field and laboratory approaches to quantify attenuation mechanisms of pharmaceutical and pharmaceutical transformation products in a wastewater effluent-dominated stream

Abstract

Evolving complex mixtures of pharmaceuticals and transformation products in effluent-dominated streams pose potential impacts to aquatic species; thus, understanding the attenuation dynamics in the field and characterizing the prominent attenuation mechanisms of pharmaceuticals and their transformation products (TPs) is critical for hazard assessments. Herein, we determined the attenuation dynamics and the associated prominent mechanisms of pharmaceuticals and their corresponding TPs via a combined long-term field study and controlled laboratory experiments. For the field study, we quantified spatiotemporal exposure concentrations of five pharmaceuticals and six associated TPs in a small, temperate-region effluent-dominated stream during baseflow conditions where the wastewater plant is the main source of pharmaceuticals. We selected four sites (upstream, at, and progressively downstream from effluent discharge) and collected water samples at 16 time points (64 samples in total, approximately twice monthly, depending on flows) for 1 year. Concurrently, we conducted photolysis, sorption, and biodegradation batch tests under controlled conditions to determine the major attenuation mechanisms. We observed 10-fold greater attenuation rates in the field compared to batch tests, demonstrating that connecting laboratory batch tests with field measurements to enhance predictive power is a critical need. Batch systems alone, often used for assessment, are useful for determining fate processes but poorly approximate in-stream attenuation kinetics. Sorption was the dominant attenuation process (t_1/2<7.7 d) for 5 of 11 compounds in the batch tests, while the other compounds (n=6) persisted in the batch tests and along the 5.1 km stream reach. In-stream parent-to-product transformation was minimal. Differential attenuation contributed to the evolving pharmaceutical mixture and created changing exposure conditions with concomitant implications for aquatic and terrestrial biota. Tandem field and laboratory characterization can better inform modeling efforts for transport and risk assessments.

Graphical Abstract

1. INTRODUCTION

Wastewater effluent-dominated streams are becoming increasingly common globally, including in temperate regions, due to climate change and demands on water resources (Rice et al., 2013; Rice and Westerhoff, 2015). Pharmaceuticals commonly present in effluent-dominated streams generate unique, elevated-concentration, complex exposure conditions for aquatic and terrestrial biota with potential ecological implications (Arnnok et al., 2017; Bartelt-Hunt et al., 2009; Bradley et al., 2014; Schultz et al., 2011; Zhi et al., 2020). Most pharmaceutical studies have focused on active ingredients parent compounds, with comparatively little attention to corresponding pharmaceutical transformation products (TPs) (Godoy et al., 2015; Grabicova et al., 2017; Santos et al., 2013). Risk assessments typically assume transformation results in decreased toxicological risks (Chen et al., 2014). Unfortunately, growing evidence indicates that TPs of some pharmaceuticals (e.g., diclofenac, metformin) can exhibit similar or greater toxicity than parent compounds (Fu et al., 2020). TPs can be formed in the human body, within the wastewater treatment plants, or in-stream (e.g., photolysis, biotransformation) along surface/subsurface flow paths (Houtman et al., 2013; Rúa-Gómez and Püttmann, 2013; Scheurer et al., 2012; Tisler et al., 2019). Parent pharmaceuticals and TPs contribute to complex mixtures that can vary substantially through space and time (Writer et al., 2013b). The evolving spatiotemporal nature of complex pharmaceutical mixtures in receiving waters makes assessment of impact challenging (Carpenter and Helbling, 2018; Fairbairn et al., 2016; Zhi et al., 2020).

Although many pharmaceutical compounds in wastewater effluent dominated streams are considered pseudo-persistent in the aquatic environment because the attenuation rate is balanced by continuous influx (Daughton, 2003), it is critical to understand the relevant contributing mechanisms and corresponding attenuation kinetics for pharmaceuticals and TPs. Relatively few studies have evaluated the occurrence and spatiotemporal distributions of different pharmaceutical mixtures in surface water including lakes, rivers, and streams worldwide (Cantwell et al., 2018; López-Serna et al., 2012; Writer et al., 2013b). Differential attenuation of individual pharmaceutical compounds/TPs contributes to the complex mixtures that create spatially and temporally unique exposure conditions for aquatic and terrestrial biota. Pharmaceuticals and TPs may undergo attenuation processes including physical/chemical processes (e.g., dilution, sorption, photolysis), and biodegradation (Andreozzi et al., 2003; Boix et al., 2016; Schaper et al., 2019). The primary mechanism of pharmaceutical removal is thought to be sorption to sediments and stream biofilms (Writer et al., 2013a). Some pharmaceuticals (e.g., citalopram, venlafaxine) are differentially persistent compared to their respective TPs (Writer et al., 2013a), contributing to the evolving complex mixture along the stream. To date, attenuation mechanisms of parent compounds have received the greatest focus; thus, studies on attenuation of TPs in aquatic systems are limited but critical when considering evolving complex mixtures.

Combining laboratory and field measurements is critical to characterize and understand the attenuation of evolving complex pharmaceutical mixtures. The majority of research on potential attenuation processes of pharmaceutical compounds has been conducted under controlled laboratory studies (Li et al., 2013; Rúa-Gómez and Püttmann, 2013). Although these studies help explain the observed persistence and widespread detection in surface waters and are valuable to understanding mechanisms, laboratory-based experiments yield different transformation kinetics compared to real-world conditions, making it difficult to translate results into attenuation rates in the stream (Fu et al., 2020; Li and Mclachlan, 2019). For example, spiking standardized bottle tests with pharmaceuticals at a concentration markedly higher than their environmental concentrations can strongly influence biodegradation and yield different degradation rates than unspiked tests, lessening the environmental relevance of test results (Li and Mclachlan, 2019). Although in-stream assessment of chemical fate is informative, spiking of additional chemicals into a natural system is often impractical; furthermore, biodegradation kinetics can be influenced by the amount of chemical added and spiking can also change the microbial population (Fahlman et al., 2018; Li and Mclachlan, 2019). For field studies, long-term frequency measurements can better capture the variation of chemical concentration and compositions in the effluent and along the stream reach compared to one-time snapshot sampling (Zhi et al., 2020). To our knowledge, no prior studies have evaluated in-stream attenuation of pharmaceutical compounds using controlled laboratory tests under environmentally relevant (i.e., unspiked) conditions combined with a long-term effluent-dominated field study from an existing ecosystem with a single continuous point-source (i.e., wastewater discharge).

In recent prior work, we demonstrated that spatiotemporal changes of complex pharmaceutical mixtures in a temperate-region wastewater effluent-dominated stream are primarily driven by input variation from the wastewater treatment plant (WWTP) and differential attenuation level of individual pharmaceuticals (Zhi et al., 2020). Yet, the mechanisms that contribute to differential compound attenuation or formation of corresponding TPs remain a critical knowledge gap. Therefore, the objective of this work was to combine laboratory and field experimental approaches to determine the prominent attenuation mechanisms of pharmaceuticals and corresponding TPs that contribute to evolving complex pharmaceutical mixtures and concomitant unique exposure conditions to aquatic organisms. Here, we quantified pharmaceutical/TP attenuation dynamics in the field at baseflow conditions over a one-year time period and concurrently quantified laboratory batch attenuation/transformation (i.e., sorption, biotransformation, photolysis) of the field-collected, un-spiked water. We hypothesized that batch tests can predict the major transformation mechanisms, but in-stream attenuation rates would be faster due to the complex conditions in the natural environment; thus, both approaches in tandem are critical to developing a comprehensive study and will yield synergistic understanding. This research will enable better future predictions of the fate and transport of complex pharmaceutical parent and TP mixtures in wastewater effluent-dominated streams and can inform modeling efforts.

2. Materials and Methods

2.1. Chemicals

All pharmaceutical standards (described fully in SI) were purchased from Sigma Aldrich and used as received. The solvent, methanol, acetonitrile, water, and formic acid were all optima LC-MS grade (Fisher Chemical). Pharmaceutical compounds including antidepressants, antidiabetics, anticonvulsants, as well as their major TPs, were selected based on U.S. Geological Survey preliminary results (Figure S.1; Meppelink et al., 2020), and were the most frequently detected compounds with high concentrations but with unclear in-stream behaviors. Selected TPs (that may be formed in the stream, WWTP, or human body) for this study were distinct from our previous work that mainly focused on pharmaceutical parent compounds, rather than TPs (Zhi et al., 2020).

2.2. Environmental Sampling

Muddy Creek, Iowa, USA (Latitude 41°42’00”, Longitude 91°33’46”) has a drainage area of 22.5 km² comprised of both agricultural and urban land use. Site details were fully described previously (Zhi et al., 2020). Muddy Creek is a temperate region, gaged stream reach where streamflow under the majority of conditions is dominated by wastewater effluent from a single point-source and minimal expected non-point sources of pharmaceuticals, representing an ideal field site to study the behavior of pharmaceutical compounds. Under baseflow conditions (as were sampled, Figure S.5), effluent substantially contributed to Muddy Creek streamflow (e.g., with a median effluent/DS1 flow ratio of 91% during the 12-month collection periods in our prior study)(Zhi et al., 2020). The effluent is discharged from the WWTP in the city of North Liberty, Iowa, USA. North Liberty has an estimated population (World Population Review, 2018) of 19,240 and is the second fastest-growing city in Iowa. The current wastewater discharge averages 9,274 m³/day (Figure S.4). During September 2018–August 2019, stream water samples and effluent samples were collected at least once a month during baseflow conditions at four sampling sites: 1) approximately 0.1 km upstream of WWTP outfall (US1); 2) wastewater effluent outfall (Effluent); 3) 0.1 km downstream of WWTP outfall (DS1); and 4) 5.1 km downstream of WWTP outfall (DS2) (Figures S.2, S.3). Stream samples were collected using the single vertical at centroid-of-flow (VCF) method, established and described in ‘Collection of Water Samples’ (Section 4.1.3A) of the USGS National Field Manual for the Collection of Water-Quality Data (U.S. Geological Survey, 2006). Effluent samples were straight from the point of discharge as effluent exited the outfall pipe (Figure S.3b). All stream sampling sites were selected following the USGS guidelines (U.S. Geological Survey, 2006) for optimally locating flowing-water sampling sites (Section 4.1.2.A). Preliminary in-field lateral and vertical measurements (dissolved oxygen, pH, water temperature, and conductivity) performed at multiple cross sections upstream and downstream from each stream sampling site demonstrated well-mixed conditions and that the VCF-method is a valid approach to achieve sufficiently representative samples (Meppelink et al., 2020; U.S. Geological Survey, 2006). The target sampling regime was biweekly during baseflow conditions; however, as local precipitation caused elevated streamflow, only sixteen sets of baseflow samples were collected during the study period (Figure S.5). Dissolved oxygen, pH, conductivity, and water temperature were measured in situ during each sample collection using a HACH multimeter (see SI).

2.3. Sample processing

For each sampling event, 250 mL water was collected in an acid-washed amber glass bottle at each of the four sampling sites in the field, filtered through 0.7 μm glass fiber filters (GF/F, Whatman Inc., Piscataway, New Jersey), spiked with an isotopically-labeled surrogate mix (n=5 compounds, see SI), and concentrated via Strata X-CW solid phase extraction (SPE) cartridges (500mg, 6mL, Phenomenex, Torrance, CA, USA) in the laboratory. The surrogate mix was added after filtration of the water sample but prior to SPE to generate an accurate representation of pharmaceutical concentrations in the aqueous phase rather than particles; surrogate recovery was applied to reported concentrations in the manner of our prior work (Zhi et al., 2020). Glass filters were chosen to remove particles and minimize possible chemical interference. Details for SPE processes were described previously (Zhi et al., 2020) and are fully described in SI. The final solution was transferred into a glass screw cap HPLC vial whereupon 10 μL caffeine-¹³C₃ (10 mg/L) was added as an internal standard to evaluate matrix effects during instrumental analysis. Samples were stored at −20°C until analysis, typically within 14 days.

2.4. Laboratory Batch tests

Batch tests were conducted to quantify relevant pharmaceutical attenuation mechanisms including photolysis, sorption, and biotransformation. Parent compounds and TPs were quantified during a 24-h period for photolysis tests, and a 14-day period for sorption / biodegradation tests. All experimental groups were conducted in replicate (n=3 or n=4) concurrently, and sampled at the same times to permit matched-pairs statistical design.

Batch tests consisted of six experimental groups: (1) photolysis, (2) sorption, (3) biodegradation, (4) sorption combined with biodegradation, (5) positive control (to ensure microbial survivability), and (6) an abiotic negative control (Figure S.7). Photolysis was quantified in parallel experiments with 10 mL of raw creek water and 10 mL filter-sterilized creek water (0.2 μm PTFE), both from site DS1. 10 mL water was placed in a 50 mL glass beaker in a SunTest CPS+ solar simulator with irradiance of 750W m⁻² (Atlas Material Testing Technology, Mount Prospect, IL) where water temperature was maintained ~15°C (Kral et al., 2019) (Figure S.6). 1 mL samples were collected at seven pre-determined intervals over the course of 24 h. The remaining water at the end of each test was collected and measured to determine loss via evaporation. Each condition was conducted in duplicate.

For the sorption treatment, the ratio of water:sediment was adjusted to 10:3 by weight based on design of previously published research (Martínez-Hernández et al., 2014; Scheytt et al., 2005) to better simulate natural conditions in the bed sediment (i.e., not intended to emulate suspended solids in the water column) of the effluent-dominated stream during baseflow conditions. Sediment was collected from DS1 and autoclaved at 120 °C for 30 minutes to sterilize (Bank et al., 2008). 10 mL sterile creek water filtered with 0.2 μm pore diameter PTFE filter and containing 3 g autoclaved sediment were added into a 10 mL amber bottle and sealed. For biodegradation experiments, 10 mL raw (unfiltered) creek water was added into a 10 mL amber bottle and sealed. Headspace was retained in the bottle for sufficient oxygen supply based on the five-day carbonaceous biochemical oxygen demand (CBOD5, 0.035 mL oxygen). For the sorption combined with biodegradation experiment, 3 g autoclave-sterilized sediment and 10 mL raw creek water were added into a 10 mL amber bottle. For the microbial growth positive control to ensure that bacteria were able to thrive under the experimental conditions, 10 mL raw creek water was added to an autoclaved amber glass vial and to measure the optical density at a wavelength of 600nm (OD600) to observe microbial growth (Table S.3). For the abiotic control, 10 mL filter-sterilized creek water was added into a 10 mL amber bottle and sampled to observe any abiotic degradation. We also conducted a desorption experiment with 3g sediment and 10 mL pH 7 DI water, where no pharmaceuticals above the MRLs were detected.

All bottles were placed on a shaker table at 100 rpm under room temperature conditions (20 °C). Sampling occurred at 6 pre-determined intervals over the course of 14 days where 1 mL samples were collected using sterile needle and syringes (Fisher) and filtered with 0.2 μm pore diameter PTFE filters, whereupon samples were transferred into a glass HPLC vial and 10 μL caffeine-¹³C₃ (10 mg/L) was added as an internal standard. Samples were stored at −20°C until analysis. Sediment analysis of organic matter content (0.61%) and total carbon content (0.30%) was conducted in the Sediment Lab at the Iowa Geological Survey (Iowa City, Iowa, USA) to measure physicochemical properties.

2.5. Analytical Methods

Sample analysis was performed by LC-MS/MS using a 1290 Agilent Infinity HPLC and Triple Quadrupole with an Electrospray Ionization (ESI) source and Mass Hunter software. An Agilent Eclipse Plus C18 column (4.6×150mm; 5μm particle size) coupled with a guard column of the same material, at a flow rate of 0.6 mL/min was used for separation at 50°C. The total method was 18 min and details are fully described in the SI. Samples (composition of 50% acetonitrile and 50% H₂O) were kept chilled in the autosampler tray at 10 °C and the injection volume was 20 μL. The ESI source was operated under positive mode conditions (Table S.4). For quantification and confirmation, two multiple reaction monitoring (MRM) transitions were monitored for each analyte in dynamic MRM mode (Table S.5). A five-point standard calibration curve was performed between 50–500,000 ng/L. Isotopically labeled surrogates were used for recovery and the internal standard was used to check for the matrix effects during instrumental analysis. Method report limits (MRLs) were quantified via a previous published method (Klarich et al., 2017) and ranged 26–72 ng/L (Table S.6).

2.6. Quality Assurance / Quality Control

Fresh, disposable nitrile gloves and freshly acid-washed amber glass bottles were used every time at each sampling location. All environmental samples were spiked with 5 isotopically-labeled surrogates (that correspond to target compounds in the study and represent a range of chemical properties) to not only quantify target compounds and compensate for matrix effects, but also to account for potential loss or errors during sample preparation. Recovery of water samples spiked with the suite of pharmaceuticals determined via our previously published method (Zhi et al., 2020) was typically >85%. Field blanks using organic free water (Thomas Scientific) were processed during the sampling campaign. One field blank was performed at each sampling site on different sampling dates over the course of the study. At least one instrumental blank sample was analyzed with every 16 environmental injections. No pharmaceuticals above MRLs were detected in the field or instrumental blanks. The pH conditions for lab experiments were neutral (~pH 7) and were similar to field conditions (pH 7.5–8.3). Our sampling approach and analytical method have been previously published (Zhi et al., 2020), and was cross-checked with an established analytical method from the U.S Geological Survey (Furlong et al., 2014).

2.7. Data Analysis

GraphPad Prism 8 (La Jolla, California) was used for all statistical analysis. Normality tests (Shapiro-Wilk test) were conducted to analyze data distributions (α=0.05). Matched-paired t-tests or nonparametric tests (Wilcoxon matched-pairs signed rank tests) assessed statistical differences (α=0.05) between parent and TPs, seasonality and sampling locations. A pseudo first-order curve was assumed and fitted for field attenuation kinetics and batch tests to determine half-lives (t_1/2 = ln2/k) and attenuation rates based on previous literature (Wick et al., 2009). Linear regression coefficients of determination (R²) were ≥0.99 for the calibration curves for all detected analytes.

3. Results and Discussion

3.1. Occurrence of pharmaceuticals and associated transformation products

Consistently high concentrations of both pharmaceutical parent compounds and their respective transformation products (TPs) in the WWTP effluent and the corresponding stream reach below the WWTP outfall suggests that complex mixture evolution is driven by variation in the effluent and differential in-stream attenuation of individual pharmaceuticals and TPs—rather than in-stream transformation from parent to TPs (Figure 1). In the effluent, venlafaxine was the parent pharmaceutical present at the highest concentration, followed by metformin, citalopram, bupropion, and carbamazepine. The same rank order occurred at DS1; thus, with only 100 m of transport, in-stream attenuation processes did not have a large impact on pharmaceutical relative composition during baseflow conditions. At DS2 (5.1 km downstream from the effluent discharge), in contrast, metformin was the parent pharmaceutical present at the highest concentration, followed by venlafaxine, bupropion, carbamazepine, and citalopram thus illustrating differential in-stream attenuation of compounds due to all processes occurred with longer stream transport distances. The concentrations of citalopram and bupropion were higher than their associated TPs in 100% of samples (p<0.02 and p<0.0001, respectively, for all three sites [effluent, DS1, DS2]; Table S.13). Most literature reports hydroxybupropion at higher concentrations than bupropion in wastewater effluent and surface water (Bai et al., 2018; Metcalfe et al., 2010; Writer et al., 2013b), but at least one prior study was consistent with our results (Deo, 2014). The concentrations of metformin and guanylurea were not different at the effluent or either downstream sites, suggesting no differential attenuation occurred (p>0.5 for all three sites; Table S.13); no substantive inputs of pharmaceuticals from non-point sources are expected based on land use along the reach (Table S.2).

Figure 1:

(a)-(e) Pharmaceutical compounds in Muddy Creek, Iowa, USA September 2018-August 2019 (n=16 sampling events at baseflow conditions). Adjacent three bars for each date represent site Effluent, DS1 and DS2 from left to right. Blue represents the parent compound, orange represents the Phase I product, and yellow represents the Phase II product. (f) Median concentrations (ng/L) of investigated compounds in the Effluent and at downstream sites DS1 and DS2. US site was excluded due to low detection frequencies and trace level concentrations (Table S.12). One data point was missing during December sampling event because the water sample at DS1 was broken. The uneven sampling dates causes the distribution of the x-axis labels.

Desvenlafaxine concentrations were always higher than parent compound venlafaxine in wastewater effluent and accounted for >50% (ranged 22–86%, median 70%) of the total concentration of venlafaxine, desvenlafaxine, and N,O-didesmethylvenlafaxine (hereafter, “DDV”; an inactive Phase II product of venlafaxine (Colvard, 2014)). This is likely because desvenlafaxine is both a major active Phase I product from venlafaxine and also a prescribed pharmaceutical (Gasser et al., 2012). Desvenlafaxine concentrations were consistently and significantly higher than venlafaxine in the effluent (p=0.0084), but concentrations of the two compounds were not significantly different at DS1 (p=0.15) or DS2 (p=0.08), suggesting differential attenuation between venlafaxine and desvenlafaxine. The concentration of DDV was consistently lower than venlafaxine or desvenlafaxine along the stream (p<0.009; Table S.13). Transformation from desvenlafaxine to DDV may be reversed by biodegradation in the treatment systems (Gasser et al., 2012), but information is lacking on the presence of DDV in aquatic environments. Carbamazepine and dihydroxycarbazepine concentrations were not significantly different in the effluent and DS1 (p>0.09), but dihydroxycarbazepine concentration was significantly higher than carbamazepine at DS2 (p=0.02; Table S.13), suggesting differential attenuation.

Overall, only trace levels of pharmaceuticals and TPs were sporadically detected at US1 above the WWTP outfall (median ≤13 ng/L; detection frequency=19%), indicating the wastewater effluent was a consistent pharmaceutical point-source during baseflow conditions year-round in this system as we documented previously (Zhi et al., 2020). Because the land uses between the effluent input and DS2 do not substantially differ from the upstream, non-point contributions of pharmaceuticals along this reach are likely minor compared to the WWTP effluent. Consistent detection frequencies of metformin (50%) and guanylurea (25%) but low concentrations (median ≤3 ng/L) at US1 indicated unknown upstream sources, such as septic systems, stormwater runoff, or leaking pipes (Kolok et al., 2014; Masoner et al., 2019). Indeed, metformin was detected in 73% of 49 urban stormwater samples across the United States (Masoner et al., 2019).

3.2. Spatiotemporal patterns of Parent-to-Product ratios (PtPs)

Quantifying spatiotemporal patterns of parent-to-product ratios (PtPs; i.e., the measured concentration of parent compound divided by the measured concentration of individual TP) is critical to differentiate between in-stream parent-to-product transformation and differential attenuation of individual parent compounds and TPs (Mahler et al., 2021; Writer et al., 2013a). For example, increasing PtP ratios along the stream can indicate either greater product than parent loss, or back conversion from product to parent. Conversely, decreasing PtP ratios along the reach could indicate greater parent than product loss, or transformation of parent to product in the stream. Unchanging PtP ratios suggests similar attenuation, or a lack of attenuation. At our study site, the parent pharmaceutical compounds and related TPs attenuated to different extents in the stream reach between the effluent discharge and downstream sites. For example, both citalopram and desmethylcitalopram were significantly attenuated at DS1 and DS2 (p<0.0001), likely due to rapid sorption and dilution (Beretsou et al., 2016; Klement et al., 2018). Citalopram PtPs at DS2 were significantly lower than PtPs in the effluent or DS1 (p<0.0001; Figure 2), indicating differential attenuation between citalopram and desmethylcitalopram. Indeed, the median citalopram and desmethylcitalopram attenuation was 41% and 36% at DS1, and 92% and 83% at DS2, respectively, indicating greater citalopram loss than TP loss. Similar relationships between citalopram and desmethylcitalopram in wastewater effluent and surface water have been reported previously (Writer et al., 2013b). For venlafaxine and associated TPs, we quantified the venlafaxine to DDV ratios (venlafaxine PtPs) and desvenlafaxine to DDV ratios (desvenlafaxine PtPs; Figures 2 and S.8). Both venlafaxine PtPs and desvenlafaxine PtPs significantly decreased from effluent to DS2 (venlafaxine p=0.02; desvenlafaxine p=0.008), indicating faster attenuation of venlafaxine and desvenlafaxine compared to DDV or possible transformation from parent compounds to products in the stream.

Figure 2:

Spatiotemporal patterns of pharmaceutical parent-to-product ratios (PtPs) along the stream during the one-year period (16 sampling events in total due to weather / flow conditions). Citalopram PtPs at DS2 were consistently lower than PtPs at other sites, indicating significant attenuation compared to other pharmaceutical PtPs. One data point was missing during December sampling event due to a broken water sample at DS1. The distribution of x-axis labels is due to the sampling dates distribution.

We also observed seasonal patterns in the concentrations of metformin and its TP guanylurea (Figures 2 and S.9), which is consistent with our previous work (Zhi et al., 2020) and studies in Greek wastewaters where the highest concentrations of guanylurea were detected in winter (Kosma et al., 2015), although the reason for this season trend is unclear. Moreover, metformin and guanylurea are largely resistant against further degradation in surface water (Markiewicz et al., 2017a, 2017b), but a recent study demonstrated that the resident microbial community in the sediments can degrade both compounds, forming biuret as a product (Posselt et al., 2020). Citalopram, carbamazepine, and venlafaxine PtPs also exhibited temporal patterns under different water temperature conditions (i.e., when ambient water temperatures measured at site US1 [native water unimpacted by the anthropogenic effluent temperature effects] >10 °C during our sampling events were considered “warm conditions,” whereas ≤10 °C at US1 were considered “cool conditions”; Table S.11; Figures S.9–S.11; Zhi et al., 2020). Both venlafaxine PtPs (p=0.04) and carbamazepine PtPs (p=0.0074) were higher under warm conditions compared to cool conditions at all three sites (matched-pairs analysis), whereas citalopram PtPs were higher under cool conditions compared to warm conditions at all sites (p=0.04). Overall, the spatiotemporal patterns of PtPs along the stream reach during baseflow conditions indicated the changing composition of complex pharmaceutical mixtures (Figures 1, 2, S.8), creating unique exposure conditions to aquatic and terrestrial biota.

3.3. Tandem laboratory batch tests and field measurements increase the understanding of attenuation processes

Our batch tests demonstrated that sorption was the major attenuation process for 7 of the 11 pharmaceuticals/TPs measured, including citalopram (t_1/2=0.33±0.01d, p=0.0042), desmethylcitalopram (t_1/2=0.30±0.07d, p=0.0051), venlafaxine (t_1/2=0.94±0.14d, p=0.0052), and desvenlafaxine (t_1/2=1.86±0.30 d, p=0.0048). The other 4 compounds, however, remained >50% in the aqueous phase during the 14-day laboratory experiment (Figure 3). This is consistent with our field measurements in terms of the compound rank order. Biodegradation as a specific attenuation process only impacted bupropion (t_1/2=11.5±1.92 d, p=0.0567) and desmethylcitalopram (t_1/2=9.86±1.41 d, p=0.0306), indicating the generally poor biodegradability of most compounds in the aqueous phase. The predicted biodegradation half-lives for the studied pharmaceuticals/TPs ranged 3.36–6.54 d (U.S. EPA; Table S.1), which is much longer than the estimated travel time in our stream reach, but still underestimated the persistence of these compounds in the aqueous phase measured via batch tests. Metformin and guanylurea are both polar and mobile compounds in the aquatic environment and not readily biodegradable under typical conditions (Kosma et al., 2015; Markiewicz et al., 2017a). In our batch tests, however, ~50% of guanylurea was removed during the 14-day period in the sorption group. Contrasting sorption behavior between metformin and guanylurea has been previously reported (Briones and Sarmah, 2018); guanylurea exhibits a higher affinity to organic matter than metformin, while metformin partitions to negatively-charged sorption sites (e.g., clay minerals). Metformin, carbamazepine, and dihydroxycarbazepine were stable in our batch systems, consistent with our field data. Although there was no significant difference between the sorption group and sorption/biodegradation combined group (p>0.05) for the majority of compounds (8 out of 11 compounds), the remaining 3 compounds (desvenlafaxine, guanylurea, and DDV) exhibited enhanced attenuation in the sorption+biodegradation combined experimental group, suggesting that sediments can enhance the elimination of pharmaceuticals in the aqueous phase. This phenomenon may be because microbes in the water column can become enriched in sediments and increase biodegradation in sediments as biofilms, thereby enhancing the elimination of chemicals in the water column (Desiante et al., 2021; Kunkel et al., 2008; Yamamoto et al., 2009).

Figure 3:

Attenuation kinetics of pharmaceutical compounds in laboratory experimental batch systems. Half-lives where kinetics were first-order are included in the figure. Each figure includes four different processes: sorption between water and sediments, biodegradation in the water compartment, combined biodegradation and sorption between water and sediments, and an abiotic control group. Each group contained four replicates. Quantitative parameters are in Table S.17. Metformin, carbamazepine and dihydroxycarbazepine were not removed in the aqueous phase of the batch tests. Photolysis results are included separately in SI for clarity.

With the exception of desvenlafaxine (t_1/2= 0.54 d) and DDV (t_1/2= 1.51 d; Figure S.16), photolysis did not significantly impact any of the pharmaceuticals/TPs under raw, unfiltered water conditions. Interestingly, ~40% of citalopram (p=0.0042) and ~60% of desmethylcitalopram (p=0.0192) dissipated via photolysis under the filtered conditions whereas both remained present in the aqueous phase under the raw-water laboratory conditions. These results suggest that particle screening may be a potentially important mechanism that furthers the persistence of these compounds and warrants further investigation. A previous study demonstrated that citalopram degraded less than 0.5% at pH 7 in aqueous phase during a 30-d exposure period under simulated sunlight (Kwon and Armbrust, 2005); however, citalopram may migrate from the aqueous phase to sediments due to high sorption potential in real environments (Klement et al., 2018). Negligible photolysis of bupropion, venlafaxine, and relatively slow photolysis of desvenlafaxine (estimated t_1/2 =0.75 d) in surface water has been previously reported; no photolysis is known to occur for metformin or guanylurea (Santoke et al., 2012; Trautwein and Kümmerer, 2011), but photolysis can be significant to some effluent-derived pharmaceuticals including diclofenac/products (Jaeger et al., 2019). Hydrolysis is not believed to be an important attenuation mechanism of pharmaceutical compounds (Lam et al., 2004), and losses appeared negligible in the negative abiotic control groups (p>0.05).

Field results exhibited faster in-stream attenuation rates when compared to laboratory batch tests. For example, citalopram, desmethylcitalopram, and venlafaxine had half-lives of 0.33±0.06 d, 0.23±0.001 d, and 0.91±0.13 d in batch tests, respectively, whereas faster attenuation occurred in the field (0.06±0.04 d, 0.19±0.13 d, 0.15±0.17 d, respectively; Table S.16). Similarly, biodegradation was the main loss process for bupropion in the batch tests (t_1/2=11.5±1.92 d); however, we observed faster attenuation in the field (t_1/2=0.18±0.18 d; Table S.16) This phenomenon is likely due to greater opportunities for physical and biological attenuation processes in the field such as dilution, dispersion, hyporheic exchange, possible plant/phytoplankton uptake (Bradley et al., 2016; Keefe et al., 2019; LeFevre et al., 2015; Romeijn et al., 2021; Schmadel et al., 2016) that make complex field conditions more conducive to in-situ biochemical attenuation processes (e.g., sorption/biodegradation processes become more efficacious when combined with the enhanced physical attenuation like hyporheic exchange). Additionally, at the pH conditions at field study site DS2 (pH=7.6–8.3; Table S.11), bupropion (pk_a=8.2) may undergo electrostatic interactions with bed sediments and suspended sediments. We did not observe strong/significant correlations, however, between bupropion concentrations and water pH at DS2 (Pearson r=0.22, p=0.60; Figure S.12). The pH conditions for lab experiments were neutral (~pH 7) and were similar to field conditions (pH 7.5–8.3); most of investigated compounds have pKa values greater than 8.2 and thus pH conditions were thus unlikely important to dissipation rates.

We demonstrated through this work that although carefully-controlled laboratory tests can determine major attenuation mechanisms, batch tests and field conditions can yield substantially different attenuation half-lives; consequently, a single approach may be insufficient to adequately predict behaviors of pharmaceuticals and TPs in the environment and we recommend the tandem approached embodied herein. Conducting research studies at well-established field sites (i.e., in this study, (Zhi et al., 2020)) using tandem lab and field approach can help us better understand and contextualize results of other studies that either examine field or lab, with the limitations of each approach in proper context. For example, several recent studies have examined the limitations of OECD guidelines for standard laboratory batch tests, as the guideline spiking conditions can generate different attenuation kinetics (e.g., first-order kinetics vs. zero-order kinetics with a 15–28 d lag phase)(Coll et al., 2020; Li and Mclachlan, 2019). Although standardized bottle tests can be useful for assessing spill scenarios, these spiked-batch tests are less relevant for assessing biodegradation under natural conditions where chemicals already exist continuously. Likewise, conducting field studies alone can make elucidating attenuation mechanisms difficult (Schultz et al., 2010). Thus, connecting field studies to laboratory batch tests can improve the context for a regulatory framework for natural water systems, especially in surface waters with high ambient levels of chemicals (i.e., effluent-dominated systems). Moreover, tandem field work and laboratory-controlled tests can inform/contextualize existing work that only employed a single lab/field approach, and also improve the framework for risk assessment of pharmaceuticals in effluent-dominated systems. High spatiotemporal-resolution sampling synergized with comprehensive analytical analysis and the application of appropriate simulation models would provide the most comprehensive assessment of the potential risks of pharmaceuticals in effluent-dominated systems.

Based on our prior research in this study reach, pharmaceuticals in the water and sediment of our field site can be assumed at pseudo-equilibrium under baseflow conditions due to consistent influx of treated wastewater (Zhi et al., 2020). Consequently, the sediment in the creek bed is potentially both a major sink and source of chemicals, via sorption and desorption (Kolok et al., 2014). We therefore estimated the partitioning coefficients (K_d) of four investigated parent compounds (described in the SI, Table S.19); these results further support sorption as a major attenuation process for compounds (e.g., citalopram, venlafaxine). Compounds with the lowest K_d values (i.e., metformin) are expected to exhibit minimal sorption. The mid-range K_d values of carbamazepine and venlafaxine, and the low K_d value for citalopram suggests that sorption to sediment is a progressively more influential attenuation process for these compounds, consistent with our batch tests—except for carbamazepine. Carbamazepine is known to exhibit a low sorption affinity (K_d= 0.21 L/kg) to sandy sediments (Scheytt et al., 2005), indicating theoretical partitioning coefficient may sometimes overestimate the sorption. Although we did not quantify sediment pharmaceutical concentrations, an analogous USGS study (Boulder Creek, Colorado, USA), reported high concentrations of citalopram (up to 14.95 ng/g) and venlafaxine (up to 26.07 ng/g) in the sediments, whereas bupropion concentration was one order of magnitude lower in the sediments (up to 2.12 ng/g; Schultz et al., 2010). Thus, determining the theoretical partitioning of pseudo-persistent pharmaceuticals is important for contextualizing laboratory batch and field results while improving predictive power for sites beyond the initial study location.

Tandem batch tests and field measurements demonstrated that, although the temporal variability of the complex mixture evolution was mainly due to effluent contributions, the spatial mixture evolution was driven by differential attenuation via sorption rather than in-stream product formation (i.e., biodegradation, photolysis, hydrolysis). For example, citalopram has a high reported sorption capacity to sediment (K_d= 1.9×10⁴ L/kg) under similar organic matter content and pH conditions (TOC=0.93%, pH 7.8, respectively) as our system (Kwon and Armbrust, 2008). Biodegradation in general is considered a less-important processes in the water column due to the relatively-low abundance of microbes in surface water (Fatta-Kassinos et al., 2011), although microorganisms can degrade pharmaceuticals to different extents in batch experiments (Casas et al., 2015; Lin et al., 2010; Xu et al., 2011). In contrast, the sediment compartment is known to contain a comparatively high density of microorganisms, representing a potential major site of pharmaceutical loss via biodegradation and sorption, both in the lab and field studies (Martínez-Hernández et al., 2014; Radke and Maier, 2014). The sediment, as a sink, may also release pharmaceutical compounds via desorption even when the compound is not present in the WWTP effluent at a particular time, which may occur at our study site and further contribute to the pseudo-persistence of exposure conditions (Kolok et al., 2014). We do, however, recognize that this finding is applicable to baseflow conditions but may not represent elevated-flow conditions where stormflow can enhance the mobilization of pharmaceuticals, thus affecting sorption processes. The focus of this study, and indeed the vast majority of hydrologic conditions experienced at this site (Fig. S.5), is baseflow, wherein the aforementioned sorption description would be applicable under pseudo-persistent conditions (Kolok et al., 2014). Physical attenuation, such as dilution and groundwater exchange, also contribute to concentration attenuation in the water column. A few studies have focused on the attenuation processes occurring in hyporheic zones, where pharmaceuticals have been frequently detected (Lewandowski et al., 2011; Posselt et al., 2020; Schaper et al., 2019). Biodegradation of pharmaceuticals in hyporheic zones is considered more efficient than in WWTPs and surface water due to longer residence times and higher microbial diversity, which also contributes to the evolving complex pharmaceutical mixtures in the aquatic environment (Lewandowski et al., 2011; Peter et al., 2019). In summary, multiple mechanisms can contribute to the evolving parent-product mixture in the stream, generating complex exposure conditions for aquatic and terrestrial biota with potential implications. Combining laboratory and field approaches improves characterization of processes and kinetics, and can inform future modeling efforts or context for established guidelines on chemical fate (e.g., OECD batch degradation tests).

4. CONCLUSIONS

The tandem field and laboratory research at the Muddy Creek research site have documented important conclusions regarding pharmaceutical attenuation.

• Pharmaceuticals and TPs were attenuated faster in the small effluent-dominated stream field site compared to laboratory batch tests, indicating that, although batch tests alone can illuminate mechanisms, tandem laboratory batch tests and field measurements are complementary for comprehensive assessment of attenuation behaviors in real-world systems and help inform a modeling framework for effluent-dominated streams.

• Parent pharmaceuticals and their TPs were mainly derived from point-source release (i.e., WWTP discharge) rather than in-stream product formation.

• Sorption to sediment was the predominant attenuation process (rather than biodegradation or photolysis) during baseflow conditions. Trends in sorption were consistent between batch, field, and theoretical results.

• TPs were present in high concentrations along with parent compounds in WWTP discharge and surface water and contribute to evolving complex pharmaceutical mixtures. Thus, TPs are an important component to understanding the fate and effects of pharmaceuticals.

Supplementary Material

SUPPLEMENTARY MATERIAL. Additional method details, statistical analysis, quality assurance / control, additional detailed data / results / analysis in figures and tables.

Highlights:

• Pharmaceuticals and transformation products persisted in effluent-dominated stream.

• Pharmaceuticals and transformation product mixtures changed over time.

• Sorption was the dominant in-stream attenuation mechanism during baseflow conditions.

• Field study attenuation rates were faster than laboratory rates.

• Tandem lab and field studies are critical to understanding attenuation mechanisms.