Existing tests, databases, and literature data

Roche internal databases, the European Chemicals Agency (2018) database, and the scientific literature were searched for physicochemical, toxicological, and environmentally relevant data as well as measured environmental concentrations (MECs) for MPM and MPA. In addition, the IQVIA MIDAS Quantum subscription database (IQVIA 2018) was queried for sales amounts in kilograms of MPM and MPA in European countries for the period 2004 to 2017, with the aim of basing the environmental risk assessment on realistic use data.

The existing Syntex tests comprise the following internal, non–good–laboratory–practice (GLP) tests and external, GLP-compliant studies: vapor pressure of MPA; photodegradation of MPA in aqueous solutions; and biodegradability of ¹⁴C-labeled MPM (2 test substances with radiolabels in different moieties of the molecule) in river water and sediment. Descriptions of the test methods applied are given in the Supplemental Data. Additional tests following Organisation for Economic Co-operation and Development (OECD; 2018) test guidelines in compliance with GLP, ensuring the validity of the procedures and statistics (Caldwell et al. 2018), were commissioned by Roche. Because MPM is fully cleft to MPA and the excreted metabolites are 7-O-MPA glucuronide (MPAG) and MPA, all new tests were performed with the active moiety MPA. The following tests were performed: OECD test guideline 106 phase I batch equilibrium adsorption test (Organisation for Economic Co-operation and Development 2000); OECD test guideline 314B activated sludge biodegradation simulation test (Organisation for Economic Co-operation and Development 2008); OECD test guideline 201 algal growth inhibition test with Anabaena flos-aquae (cyanobacteria; Organisation for Economic Co-operation and Development 2011); OECD test guideline 211 chronic reproduction test with Daphnia magna (Organisation for Economic Co-operation and Development 2012); and a fish partial life cycle test consisting of an OECD test guideline 229 short-term reproduction test (Organisation for Economic Co-operation and Development 2012b) followed by an OECD test guideline 210 early life stage test with Danio rerio (Organisation for Economic Co-operation and Development 2013). Also, a ready biodegradability study according to OECD test guideline 301F was performed with MPM (Organisation for Economic Co-operation and Development 1992). Descriptions of the test methods applied are given in the Supplemental Data.

Predicted and measured environmental concentrations

For 24 documented European countries, the administered amount of MPA and the stoichiometric fraction of MPA from MPM administration in kg/yr were retrieved for the period 2004 to 2017 (IQVIA 2018), for the derivation of predicted environmental concentrations (PECs). For each combination of country and year, these amounts were summed and divided by the population in that country and year (European Commission 2018), and by 365 d, to derive year-specific daily per capita uses. In addition to single-country uses, we also derived year-specific European uses as the average daily per capita use over all countries.

An initial MPA sewage treatment plant (STP) PEC following the European Medicines Agency (EMA) guideline (Committee for Medicinal Products for Human Use 2015) was obtained by dividing the highest overall European MPA daily per capita use by a default 200 L of wastewater/inhabitant/day; no human metabolism or removal in STPs was included. When the EMA guideline default dilution factor was applied, the initial surface water PEC was 10 times lower (Committee for Medicinal Products for Human Use 2015). (Note that this default has been contested for single German rivers by Link et al. [2017]. On the other hand, data by Keller et al. [2014] suggest that the average and median dilution factors for European countries are >10, whereas Belgium, for example, has a lower dilution.) Refined MPA STP PECs per country were calculated by dividing the highest country-specific daily per capita MPA use by the country-specific daily water use (Keller et al. 2014). Refined MPA surface water PECs per country were then determined by deducting a predicted STP removal based on the OECD test guideline 314B (Organisation for Economic Co-operation and Development 2008) test results using SimpleTreat 4.0 model (Struijs 2015) for 80% of MPA mass (no wastewater treatment was assumed for the remaining 20%) and further division by the median country-specific dilution factor (Keller et al. 2014). For comparison with the initial PECs based on the EMA guideline, average European refined, population-, water-use-, and dilution-factor–weighted PECs for STPs and receiving waters were calculated as well.

The geographically based exposure to pharmaceuticals in the environment (ePiE; Oldenkamp et al. 2018) model was applied to derive distributions of PECs in European river catchments. The ePiE model was developed within the research project intelligence-led assessment of pharmaceuticals in the environment (iPiE 2019) and combines high-resolution georeferenced information on river flow and locations of STPs with information on number of inhabitants attached, API consumption, human metabolism, and fate of APIs during their passage through STPs and receiving surface waters. The model was run under annual mean, maximum, and minimum monthly flow conditions for the year 2015, and was applied to 4 climatically and hydrologically diverse river basins: the Rhine catchment, with a total population of approximately 58 million in 2009 (Uehlinger et al. 2009), receiving water and effluents from Switzerland, Liechtenstein, Austria, France, Germany, Luxembourg, Belgium, and The Netherlands; the Ouse Basin in northern England, UK; and the Turia and Guadalquivir catchments, in eastern and southern Spain. Average total MPA use data for these countries (except Liechtenstein, for which no separate use data are available) were entered into ePiE, as were physicochemical and environmental fate data, to derive catchment-wide PECs reflecting the geographical distributions of the modeled concentrations. We regarded the Rhine catchment as representative of climatically moderate northwestern Europe, with a significant share of >10% of the total European population (Uehlinger et al. 2009; European Commission 2018). The 2 Spanish catchments are representative of warmer and drier southern Europe; the Guadalquivir is a basin near the large cities of Cordoba and Seville, with a correspondingly high population/STP effluent contribution; the Turia catchment is smaller and more agriculturally influenced. Finally, the Ouse catchment is regarded as representative for more Atlantic-influenced western Europe.

The MECs for MPA were searched for in the scientific literature using the terms “mycophenol*”, “environment*”, and “concentration”, collated and percentage-ranked following Straub (2008).

Predicted no-effect concentrations

The predicted no-effect concentration (PNEC) for STPs (PNEC_STP) was derived by dividing the lowest activated sludge respiration inhibition no-observed-effect concentration (NOEC) by an assessment factor of 10. The surface (fresh) water PNEC_SW was derived in 2 ways, from the lowest chronic NOEC on the one hand and from the lowest chronic 10% effect concentration (EC10) on the other, from (sub)chronic tests with fish, daphnids, green algae, and cyanobacteria (for the latter 2 the NOErC/ErC10 for growth rate [r] was used), using an assessment factor of 10 (Committee for Medicinal Products for Human Use 2015).

Risk assessment

The potential risk for STPs and surface freshwater organisms was characterized by dividing the various PECs and MECs by the respective PNECs for MPA (Committee for Medicinal Products for Human Use 2015). Further risk assessments were performed for antibiotic resistance risk due to the described antibiotic activity of MPA, risk for environmental top predators, and also risk for humans through so-called secondary poisoning from uptake of water or fish.

Antibiotic resistance risk is assessed through comparing PECs or MECs with a PNEC for formation or maintenance of antibiotic resistance (PNEC_ABR). There is no universally accepted or regulatory algorithm for deriving a reliable PNEC_ABR for bacteria, so for the time being this is approximated by applying assessment factors to bacterial minimal inhibitory concentrations (MICs). Kümmerer and Henninger (2003) applied an assessment factor of 100 to the lowest bacterial MIC/antibiotic. In 2016, Bengtsson-Palme and Larsson used the lower 1st percentile (%ile) MIC/antibiotic and applied 2 assessment factors, a general factor of 10, and an additional factor that depends on the size of the MIC dataset and reflects the magnitude of uncertainty. To be fully consistent with Bengtsson-Palme and Larsson's (2016) approach to PNEC_ABR derivation, the authors were contacted for help; they (J. Bengtsson-Palme, Department of Infectious Diseases, Institute of Biomedicine, The Sahlgrenska Academy, University of Gothenberg, Gothenberg, Sweden, personal communication) state that “we only use species with 10 or more MIC observations, to avoid drawing conclusions from insufficient data [...]. The only species with >10 observations in the Noto et al. (1969) paper is S[taphylococcus] aureus. In addition, when the MIC1% are selected, only one MIC [J.O.Straub's italics] value is selected per species (the lowest). Finally, the uncertainty factor for low sampling numbers is calculated as: (observed lowest MIC) × (number of tested species) ÷ 41.” Both these provisional PNEC_ABR values were derived from the MPA MICs; the lower was selected and compared with the PECs and MECs.

The potential risk for (semi)aquatic top predators such as otters that consume fish and drink surface water was assessed by maximum tolerable daily intake (MTDI) calculated following Murray-Smith et al. (2012) and the European Union technical guidance for deriving environmental quality standards (European Commission 2011). In brief, acute and long-term mammalian toxicity results are divided by appropriate assessment factors, depending on the nature and duration of the tests, to derive an MTDI. For an otter, a default body mass of 10 kg, an intake of 1 kg fish, and 0.79 L of water/d is assumed; the concentration of the substance in fish results from the surface water PEC and the bioconcentration factor (BCF). The combined daily intake from fish and water is then compared with the MTDI (Murray-Smith et al. 2012).

Similarly, potential risk for humans from secondary poisoning is assessed by using the acceptable daily exposure (ADE) value for MPA (T. Pfister, unpublished data), which considers toxicological data as well as human experience and covers all routes of uptake, including sensitive subpopulations, over 365 d/annum, with the combined worst-case uptake through drinking water and fish. No removal of MPA in drinking water production is assumed, but a default intake of 2 L water and 115 g fish/person and day is assumed (European Commission 2011; Murray-Smith et al. 2012). This combined uptake is then compared with the ADE.

Physicochemical properties and environmental fate

Syntex derived many basic physicochemical properties for both MPM and MPA in the early 1990s. The results for MPA are mostly given in the present study. Syntex internal tests were performed without GLP compliance, but with full documentation (protocol, experimental data, analytics with validation, and calculation algorithms included in reports): solubility in buffered aqueous solutions, dissociation constant (pK_a) determined by titration, screening vapor pressure, n-octanol/water distribution coefficient (log D_OW) by the shake-flask method, and screening hydrolysis (Table 1; T.J. Lynch and N. Licato, Syntex, Palo Alto, CA, USA/F.Hoffmann-La Roche, Basle, Switzerland, unpublished data; V. Nicholson, Syntex, Palo Alto, CA, USA/F.Hoffmann-La Roche, Basle, Switzerland, unpublished data; V. Nicholson and N. Licato, Syntex, Palo Alto, CA, USA/F.Hoffmann-La Roche, Basle, Switzerland, unpublished data; V. Nicholson et al., Syntex, Palo Alto, CA, USA/F.Hoffmann-La Roche, Basle, Switzerland, unpublished data; A. Young and N. Licato, Syntex, Palo Alto, CA, USA/F.Hoffmann-La Roche, Basle, Switzerland, unpublished data).

The log D_OW values ≤2.28 at pH 5 to 9 (A. Young and N. Licato, unpublished data) suggest that MPA will not bioaccumulate significantly. Because no experimental bioaccumulation data have been located, quantitative structure–property (QSPR) models were used to estimate BCFs for MPA. The EPISuite Ver 4.11 software calculates a BCF (wet wt) for fish of 3.16, independent of ionization (US Environmental Protection Agency 2016). The SciFinder database models BCFs along most of the pH range, with the highest value of 486 at pH 1, which drops to 161 at pH 5, to 23.2 at pH 6, to 2.49 at pH 7, and to 1.0 at pH 8 to 10 (American Chemical Society 2018), echoing the log D_OW distribution (Table 1). The geometric average of the SciFinder values in the environmentally relevant range of pH 5 to 9 is a BCF of 6.22. These models support the log D_OW-based prediction of no significant bioaccumulation for MPA.

The low vapor pressure of <10^–6 hPa (V. Nicholson and N. Licato, unpublished data) suggests that MPA will not volatilize to a significant extent. The screening hydrolysis test resulted in 29.03 to 46.15% substance loss at pH 5 to 9 (A. Young and N. Licato, unpublished data). These losses suggest that under environmental conditions (default 12 °C water temperature in Europe) there will only be slow hydrolysis. In confirmation, Franquet-Griell et al. (2017a) recently reported a hydrolytic degradation constant for MPA of 0.0002 min^–1.

The photodegradation test shows k_photo values at pH 5 ranging from 0.0017 min^–1 in winter to 0.0059 min^–1 in summer, at pH 7 from 0.0049 min^–1 to 0.018 min^–1 and at pH 9 from 0.0087 min^–1 to 0.031 min^–1 (V. Nicholson, unpublished data). Marín-García (2015) studied ultraviolet (UV)-C photodegradation of MPA; at pH 7, 20 mg/L MPA showed negligible photodegradation but higher rates at pH 10 to 12 (Marín-Garcia 2015), suggesting that very-short-wave UV-C does not significantly degrade MPA at environmentally relevant pH levels. In contrast, Franquet-Griell et al. (2017a) reported MPA k_photo values of 0.0284 min^–1 for UV-C irradiation and 0.004 min^–1 in a Solar Box artificial sunlight reactor. The former rate suggests relatively rapid degradation under UV-C exposure, whereas the latter rate compares rather well with the data just given above developed under natural sunlight by V. Nicholson (unpublished data). Giebułtowicz and Nałęcz-Jawecki (2016) presented a UV-visual light spectrum of MPA in their Supporting Information with a λ_max at 215 nm and 2 sequentially lower peaks at 248 and 297 nm, confirming UV-C photosensitivity. Altogether, these data suggest a notable removal of MPA in surface waters through photodegradation, which, however, is strongly dependent on season, latitude, water depth, turbidity, and pH level.

The sediment/water fate test (Z. Yan, ABC Laboratories, Columbia MO, USA, on behalf of Syntex, Palo Alto, CA, USA/F.Hoffmann-La Roche, Basle, Switzerland, unpublished data) shows that the overall ¹⁴C mass balance for all systems and concentrations was satisfactory, with a minimum of 90.4 ± 0.9% and a maximum of 99.9 ± 5.6%. At the end of the test, in the carboxyl-¹⁴C-labeled MPM systems, 58.9 and 60.2% (for 1 and 5 mg/L dosing, respectively) of applied ¹⁴C activity was trapped as ¹⁴CO₂, whereas 4.8 and 6.7% remained in water and 17.6 and 20.3% was bound to sediment, plus some activity found in the samples for analyses. In the morpholine-¹⁴C–labeled MPM systems, 21.5 and 28.0% was trapped as ¹⁴CO₂, whereas 41.5 and 32.4% remained in the water and 12.0 and 12.4% was bound to sediment, plus the samples activity. The analysis of ¹⁴CO₂ and high-performance liquid chromatography (HPLC) data showed that full primary degradation took place in all systems and dosings within 7 d, resulting in several transformation products. The transformation products were further biodegraded, partially mineralized, or bound to sediment during the 64-d test period. HPLC cochromatography of day 64 samples with the respective ¹⁴C reference standards showed that both forms of ¹⁴C-labelled MPM had completely disappeared, leaving no significant peak at all in the carboxyl-¹⁴C-MPM systems and a small peak of an unidentified transformation product in the morpholine-¹⁴C-MPM systems. This particular transformation product was shown by HPLC to reach a peak of approximately 75 to 85% relative to initially applied by day 7 and then to decrease to 40 to 50% in the morpholine-¹⁴C-MPM systems, and to 5 to 10% in the carboxyl-¹⁴C-MPM systems. The report states, “Compared to the morpholine-¹⁴C radiolabeled metabolite, the carboxyl-¹⁴C radiolabeled metabolite (probably ¹⁴C-mycophenolic acid) is much more easily biodegraded and mineralized” (Z. Yan, unpublished data, p 23). This sediment/water fate test shows that the parent MPM is lost from the systems with a 50% disappearance time (DT50) of <2 d in all cases. The carboxyl-¹⁴C-MPM transformation product, tentatively identified by the authors as ¹⁴C-MPA, is nearly completely degraded by the end of the test, with a DT50 of approximately 14 d. The unidentified morpholine-¹⁴C-MPM transformation product is degraded much more slowly, but on day 64 this transformation product, the only peak remaining, had a much smaller area (no quantitative data available) on the chromatogram, compared with the reference standard. Altogether, this test suggests that MPM is transformed rapidly in natural river water/sediment systems, and may in part form nonextractable, bound sediment residues while the remainder in the aqueous phase is partially to nearly fully mineralized.

Franquet-Griell et al. (2017a) also investigated the biodegradation of MPA in an aerobic sequencing batch reactor with an initial concentration of 1 to 1.2 g/L activated sludge that was run for 5 cycles of 48 h each, after which the activated sludge was left to settle, the supernatant was decanted, and each cycle was supplied with fresh primary effluent spiked with the same MPA concentration. While in the first cycle, approximately 58% of MPA was still detected after 48 h; in the fifth cycle, no MPA at all was detected already after 24 h, showing the adaptation of the biomass to degrade MPA. Franquet-Griell et al. (2017a) calculated biodegradation rate constants k_biodeg for MPA in the first cycle of 0.0017 min^–1 and in the fifth cycle of 0.006 min^–1, specifically remarking about the “degradation capacity of aerobic activated sludge for [this] compound.”

The new adsorption test following OECD test guideline 106 (Organisation for Economic Co-operation and Development 2000) showed satisfactory mass balances between 95 and 107%; based on pretests, MPA was shown by analytics to be stable over the selected equilibration time of 10 h (recoveries from 0.01 M CaCl₂ after 24 h: 104% at 0.2 mg MPA/L and 97% at 20 mg MPA/L). The main test resulted in soil K_d values of 2.2, 2.8, and 5.0 L/kg and activated sludge K_d values of 9.3 and 13 L/kg; normalizing to organic carbon content, the highest K_OC for MPA was 37 L/kg in one of the activated sludges (V. Halász-Laky, Toxi-Coop ZRT, Balatonfüred, Hungary, on behalf of F.Hoffmann-La Roche, Basle, Switzerland, unpublished data). Hence, MPA does not sorb significantly to activated sludge in STPs, but is mobile and remains in the aqueous phase (Briggs 1973; McCall et al. 1980). Thus, MPA is not expected to be significantly transferred to soil with surplus sludge (Committee for Medicinal Products for Human Use 2015).

An OECD test guideline 301F (Organisation for Economic Co-operation and Development 1992) ready biodegradability test with MPM at a nominal concentration of 100 mg/L (A. Häner, unpublished data) showed primary degradation by HPLC without oxygen uptake, which was interpreted as hydrolysis of the mofetil ester without subsequent biodegradation of both moieties.

The new OECD test guideline 314B (Organisation for Economic Co-operation and Development 2008) activated sludge degradation test (T. Junker and M. Herrchen, ECT Oekotoxikologie, Flörsheim, Germany, on behalf of F.Hoffmann-La Roche, Basle, Switzerland, unpublished data) fulfilled all quality and validity criteria (see the Supplemental Data). The test item was rapidly mineralized, with biodegradation starting directly after application, and increasing rapidly at the beginning and more slowly afterward up to 82.2% of initially recovered radioactivity (iRR) at day 28 (Figure 2). Mineralization comprises evolved CO₂ (detected in the traps) as well as CO₂ dissolved in the sludge. The latter was highest at day 2 and decreased afterward, whereas the amount of evolved CO₂ increased continuously until the end of the study. At the start of the test (day 0), 82.5% iRR was detected in the liquor extracts; the liquor extract radioactivity decreased to 1.2% iRR at the end of the test. At no time point was a noteworthy amount of radioactivity (>2% iRR throughout the test period) detected in the liquor after extraction. Radioactivity in the solids extract was highest at day 0 (5.5% iRR) and decreased afterward to 0.3% iRR after 28 d. Nonextractable residues increased at the beginning up to 10.2% iRR on day 2, did not change substantially until day 14 (range: 11.1 to 13.1% iRR), and then fell to 7.3% iRR on day 28.

Using the curve for total CO₂ produced over time in this OECD test guideline 314B (Organisation for Economic Co-operation and Development 2008) test (T. Junker and M. Herrchen, unpublished data), a biodegradation rate constant k_biodeg of 0.0174 h^–1 or 0.00029 min^–1 can be calculated for MPA in STPs. Based on this k_biodeg and physicochemical properties for MPA, SimpleTreat 4.0 (Struijs 2015) calculated a removal of 12% in STPs, mostly through biodegradation (0.27% in the primary settler and 11.5% in the aeration tank), with a small fraction of 0.13% adsorbed to sludge, in agreement with the OECD test guideline106 (Organisation for Economic Co-operation and Development 2000) adsorption results mentioned in the previous paragraph (V. Halász-Laky, unpublished data). However, when the k_biodeg data developed by Franquet-Griell et al. (2017a) were entered in the sequencing batch reactor for k_biodeg of 0.0017 min^–1 in the first cycle, SimpleTreat calculated a removal of 43.6% in STPs (0.272% biodegradation primary settler, 43.24% aeration tank; 0.080% adsorbed); for k_biodeg in the fifth cycle of 0.006 min^–1, SimpleTreat calculated a removal of 73.1% in STPs (0.272% primary settler, 72.77% aeration tank; 0.0373% adsorbed; Struijs 2015). A high degradation rate is supported by data from Franquet-Griell et al. (2017b), who measured STP influents and effluents and compared solid phase extraction and macroporous ceramic passive samplers in Spain. The STP in question treats 65% of the wastewaters from Barcelona and surroundings (2 843 750 inhabitant equivalents) and receives urban waters, effluents from 3 large hospitals, and industrial waters; the STP performs biological treatment without nitrogen and phosphorus removal. Based on substance concentrations in effluent and influent, removal rates of >90% emerged for MPA (Franquet-Griell et al. 2017b).

As a further confirmation of biodegradability during wastewater treatment, weekly measurements of MPM and MPA concentrations in the influent and effluent of a small industrial activated sludge STP at a Roche production site were made during 8 wk of MPM production in 2008 (G. Cahill, F.Hoffmann-La Roche, Basle, Switzerland, internal memo dated 19 September 2008, unpublished Roche data). The removal was between 74 and 95% in 6 of these weeks, whereas in the 2 remaining wk there was no significant removal in one sample, but negative removal of –73% in the other. The memo does not give any information as to whether in the latter 2 wk, asynchronous sampling of influent and effluent may have led to no or negative removal, and the site has stopped production in the meantime. Still, including all 8 measurements, the average removal was 52% and the median removal was 83%, which clearly supports efficient biodegradation of MPM and MPA in activated sludge STPs. Therefore, the 3 SimpleTreat removal predictions in the previous paragraph were used for low, middle, and high removal scenarios and derivation of the corresponding refined PECs.

Use data

Use data for MPM and MPA for the years 2004 to 2017 were retrieved from the IQVIA (2018) MIDAS database for the following 24 European countries: Austria, Belgium, Bulgaria, Croatia, the Czech Republic, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Latvia, Luxembourg, The Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, Switzerland, and the United Kingdom. For Greece, Latvia, and Luxembourg, only retail data were available (i.e., no hospital use data); however, this is not regarded as a major drawback, because MPM and MPA have to be taken on a daily basis over an indefinite time and therefore most of the use amounts will be at home, not in hospitals. This assumption is also supported by Franquet-Griell et al. (2015, p 163). The 24 countries are estimated to comprise 504.7 million inhabitants in 2017, corresponding to approximately 93.4% of the total western and middle European population of 540.3 million (European Commission 2018). Therefore, the available data for MPA are regarded as representative for Europe.

PECs and MECs

Because the single sales amounts are the intellectual property of IQVIA, the original data may not be shared as such, but only in processed form. The initial, overall-use–based European MPA PECs and the refined average and single-country highest-use–based MPA PECs for STPs and receiving freshwaters were calculated as described in the Materials and Methods section. The refined PECs based on 3 different SimpleTreat removal predictions and ePiE PECs for the Rhine and Guadalquivir Rivers assuming only 12% STP removal are shown in Table 2. For the refined, country-specific, water-use- and dilution-factor–based PECs, note that the highest PECs for STPs and receiving waters do not necessarily denote the same country, because low average water use will result in a high STP PEC but in case of subsequent high dilution this may still lead to a low surface water PEC. For the initial– and refined-use–based PECs, the STP PECs ranged from 0.898 to 8.830 µg MPA/L, and the surface water PECs ranged from <0.001 to 0.532 µg MPA/L. The median refined European, population-, water-use-, and dilution-factor–weighted surface water PECs for 12, 43.6, and 73.1% STP removal were 0.058, 0.042, and 0.027 µg MPA/L, respectively. The ePiE distributions were calculated using all stretches in the respective ePiE catchments, that is, they included all those parts of rivers and tributaries upstream of the uppermost STP, where 0 concentrations are predicted. Under annual mean monthly flow conditions, the ePiE surface water PECs for the 4 catchments ranged from <0.001 to 5.495 µg MPA/L. This maximum corresponds to the 99th %ile of the Guadalquivir PECs, whereas the median Rhine and Guadalquivir ePiE PECs were 0.007 and <0.001 µg MPA/L. Under annual minimum monthly flow conditions, the worst-case ePiE PECs were higher, with median PECs of 0.021 and <0.001 µg/L for the Rhine and Guadalquivir, and 99th %ile PECs of 2.666 and 12.419 µg/L, respectively. The extreme low-flow ePiE PEC for the Guadalquivir is due to a high seasonal variability in river flow conditions. The ePiE mean-flow surface water PECs for the Turia and Ouse catchments (Table 2) ranged from <0.001 to 0.318 and <0.001 to 0.534 µg MPA/L, respectively, with the median in both cases being <0.001 µg/L.

Franquet-Griell et al. (2015) calculated an average STP effluent PEC of 2.008 µg MPA/L and a surface water PEC of 0.0774 µg MPA/L, based on the total MPA use in Catalunya over the last 3 yr before their publication, 41% removal in STPs, and the country-specific dilution factor for Spain from Keller et al. (2014). In view of the uncertainties included (e.g., modeled STP removal and the dilution factor, which encompasses extremely dry regions in central and southern Spain), both of these PECs correspond reasonably well with the ones derived in the present study.

We only located a few surface water MECs for MPA, all dating from recent years. In 2013, Rossi and Cheseaux reported the presence of MPA in 3/3 samplings from the Swiss Rhone River upstream of Lake Geneva, at 0.0014, 0.0015, and 0.0021 µg/L, with a limit of quantitation (LOQ) of 0.001 µg/L and high uncertainty due to the low number of samples. Still, the 3 MECs, with an average of 0.0017 µg/L, correspond reasonably well with the refined use–based PEC for the whole of Switzerland assuming 73.1% STP removal of approximately 0.005 µg/L (data not shown).

In 2016, Franquet-Griell et al. (2016) published 9/9 MPA surface water MECs for Spain, ranging from 0.0128 to 0.0562 µg/L (mean: 0.0217 µg/L), in the lower, heavily industrialized and urbanized River Llobregat in the south of Barcelona with a total of nearly 5 million people, and followed it through drinking water treatment to the finished water, where it was no longer detected (0/2; LOQ of 0.0001 µg/L; Franquet-Griell et al. 2016). In the River Besòs in Catalunya, Spain, which runs through “a heavily populated and industrialized area, receiving the authorized discharges of 27 [STPs], 219 industries and 12 hospitals” in northern Barcelona, Franquet-Griell et al. (2017c) measured cytostatic APIs. (Note that MPA is cytostatic only insofar as cell growth that depends on purine biosynthesis without salvaging pathways is inhibited, but not a cytostatic in the sense of classical cell division inhibitors, even though it is included in Franquet-Griell et al. 2017c.) In 2 sampling campaigns in 2014, MPA was the target compound detected at the highest levels and was present mainly in the lowest, most urbanized river stretches, in 7/19 samples at 0.0131 to 0.0895 µg/L (mean: 0.018 µg/L, with nondetects counted as half the LOQ) in May and in 4/19 samples at 0.0085 to 0.656 µg/L (mean: 0.0474 µg/L, with nondetects counted as half LOQ) in July, with the latter, higher concentrations attributed by the authors to low water flow in summer. The dilution factor of the River Besòs is given as 1.2, showing very low dilution and a high fraction of treated wastewater (Franquet-Griell et al. 2017c). The median combined MEC in Catalunya is just below 0.001 µg/L in view of many nondetects in the upper River Besòs, which is far lower than the refined use–based PEC for the whole of Spain assuming 73.1% STP removal of approximately 0.040 µg/L and lower again by another factor of 2 than the surface water PEC of 0.0774 µg MPA/L calculated by Franquet-Griell et al. (2015) themselves. The median MEC of just below 0.001 µg/L is close to the median mean-flow ePiE PECs for the Turia and Guadalquivir Rivers (both <0.001 µg/L), even though these river catchments are situated in different parts of Spain.

Giebułtowicz and Nałęcz-Jawecki (2016) measured MPA in 2 rivers in Poland, the smaller Utrata and the large Vistula, in both cases upstream and downstream of STPs, in the Vistula downstream of the capital Warsaw, over 1 yr. They detected MPA 38/60 times in river water, from below an LOQ of 0.0005 to 0.180 µg/L downstream of a large STP, and 0/9 times in finished drinking water from the city of Warsaw. The median of Giebułtowicz and Nałęcz-Jawecki's (2016) surface water MECs, including the nondetects, was close to 0.003 µg/L (mean: 0.0168 µg/L, with nondetects counted as half LOQ), which is more than a factor of 10 lower than the refined use–based PEC for Poland of just above 0.040 µg/L, assuming 73.1% STP removal.

A total of 110 published surface waters MECs for MPA have been located to date, with 61 above the LOQ. These values, including the nondetects, were compiled, back-distributed, and percentage-ranked according to Straub (2008), resulting in a compound median MEC of just below 0.002 µg/L (75th %ile ~0.017 µg/L, 90th %ile ~0.057 µg/L, 95th %ile ~0.123 µg/L, 99th %ile ~0.604 µg/L) and a maximum MEC of 0.656 µg MPA/L. The overall mean, with nondetects counted as half the LOQ, was 0.022 µg/L. In view of the low number of MECs and the restriction to Switzerland, Poland, and northeastern Spain, it is unknown whether these MECs and their distribution are representative for Europe. The single comparisons with refined PECs per country given in the previous paragraph consistently suggest an overestimation of the PECs by a factor of 3 for Switzerland, 13 for Poland, and 40 to 80 for Catalunya. The median MEC of <0.002 µg/L for the Rhone River is reasonably close, however, to the median mean-flow Rhine River ePiE PEC of 0.007 µg/L. The overall distribution of MECs in comparison with the refined and ePiE PECs just described seems to support the notion of a too conservative PEC derivation. However, the low total and possibly skewed distribution of MECs must be kept in mind before accepting such conclusions as definitive.

Singer et al. (2016) identified and measured MPA in the effluent of 6 Swiss STPs, with an LOQ of 0.030 µg/L. They detected MPA in all 6 effluents with MECs of 0.065 to 4.190 µg/L (median: 0.755 µg/L). Based on only 6 MECs, the range and median compare reasonably well with the refined use- and water-use–based STP PECs just described of 0.898 to 8.830 (median: 2.960) µg MPA/L.

Giebułtowicz and Nałęcz-Jawecki (2016) also detected MPA “in very low concentrations [close to the limit of detection of 0.0015 µg/L] in sediment samples collected close to STPs. However, [... they] concluded that the detected MPA is freely dissolved in sediment porewater, which constituted approximately 30% of sediment samples prior to freeze drying” (Giebułtowicz and Nałęcz-Jawecki 2016, p 144). This likely nonadsorption to sediment is in agreement with the new OECD test guideline 106 (Organisation for Economic Co-operation and Development 2000; V. Halász-Laky, Toxi-Coop ZRT, Balatonfüred, Hungary, on behalf of F.Hoffmann-La Roche, Basle, Switzerland, unpublished data).

Drinking water treatment removes MPA to untraceable concentrations (Franquet-Griell et al. 2016; Giebułtowicz and Nałęcz-Jawecki 2016). Giebułtowicz and Nałęcz-Jawecki (2016) did not detect MPA in tap water in 9/9 samples in Poland. Similarly, Franquet-Griell et al. (2016) showed that MPA from Llobregat River source water disappeared early in the drinking water treatment process and was not detected in 2/2 finished drinking water samples in Catalunya, Spain. Both groups had LOQs of 0.001 µg/L for MPA.

Ecotoxicity

Syntex had MPA tested for acute daphnid ecotoxicity in compliance with GLP according to US Food and Drug Administration guideline 4.08 (2006), a test that resulted in a 48-h median effect concentration (EC50) of 755 mg/L average measured concentration (AMC) and a 48-h NOEC of 440 mg/L AMC (J.W. Blasberg and J. Bucksath, ABC Laboratories, Columbia MO, USA, on behalf of Syntex, Palo Alto, CA, USA/F.Hoffmann-La Roche, Basle, Switzerland, unpublished data.). The public European Union European Chemicals Agency (2018) database gives a robust study summary for a static algal growth inhibition test according to OECD test guideline 201 (Organisation for Economic Co-operation and Development 2011) in compliance with GLP with MPA in green algae of the species Rhaphidocelis (Pseudokirchneriella) subcapitata. The algae were exposed to MPA concentrations first for 24 h in the dark, and then for 72 h under continuous illumination. Analytical determinations of MPA concentrations were performed at 0, 24, and 96 h. Due to decrease in MPA concentrations over time, the geometric mean measured concentrations (GMMCs) for the 24- and 96-h determinations were calculated. The ErC50 for MPA was 68 µg/L, the ErC10 was 12 µg/L, and the NOErC was 9 µg/L (all GMMC; European Chemicals Agency 2018). The Swedish FASS (2018) database gives the following ecotoxicity information (endpoints only) from Novartis for sodium-MPA: the same algal data as above in the Registration, Evaluation, Authorisation and Restriction of Chemicals regulation, acute daphnid EC50 >100 mg/L in D. magna (OECD test guideline 202; Organisation for Economic Co-operation and Development 1984) and acute fish 50% lethal concentration (LC50) in Cyprinus carpio of >100 mg/L (OECD test guideline 203; Organisation for Economic Co-operation and Development 1993); also, an activated sludge respiration inhibition test according to OECD test guideline 209 (Organisation for Economic Co-operation and Development 2010) resulted in a 3-h EC50 of 2213 mg/L and an EC10 of 69 mg/L (FASS 2018). Because the latter 3 tests are from the same contract laboratory as the algal test above, they were probably performed in compliance with GLP as well. In addition, there is one older non-GLP literature 16-h LC50 value for the hypersaline water brine shrimp Artemia salina of 98.4 mg MPA/L (Ďuračková et al. 1977).

For terrestrial plant ecotoxicity, Wright (1951) examined the phytotoxic effects of some antibiotics on germination and root growth of wheat (Triticum sp.), white mustard (Sinapis alba), and red clover (Trifolium pratense) on agar containing different concentrations of test substances. The germination of mustard and clover seeds, but not wheat seeds, was significantly inhibited by MPA in a concentration-dependent manner. However, MPA inhibited the root growth of all 3 plants with increasing concentration. Although no statistics are given in the publication, a concentration range of 1 to 5 ppm (1–5 mg/kg agar) emerges as the lowest-observed-effect concentration (LOEC) for both endpoints.

The new chronic ecotoxicity tests with cyanobacteria (D. Gilberg and G. Chambers, ECT Oekotoxikologie, Flörsheim, Germany, on behalf of F.Hoffmann-La Roche, Basle, Switzerland, unpublished data), daphnids (P. Egeler and J. Chambers, ECT Oekotoxikologie, Flörsheim, Germany, on behalf of F.Hoffmann-La Roche, Basle, Switzerland, unpublished data), and fish (D. Gilberg and G. Chambers, unpublished data) fulfilled the validity criteria of the respective OECD guidelines (see the Supplemental Data). In the new OECD test guideline 201 (Organisation for Economic Co-operation and Development 2011), exposure of the cyanobacterium A. flos-aquae resulted in a clear concentration–response relationship for both biological parameters growth rate and yield during the exposure period. The following endpoints were determined, all relating to GMMC: NOEC (both growth and yield) 83.9 µg MPA/L, LOEC 284 µg/L, ErC10 155 (95% confidence interval [CI] 137–171) µg/L, and ErC50 423 (95% CI 406–441) µg/L (D. Gilberg and G. Chambers, unpublished data).

In the new OECD test guideline 211 (Organisation for Economic Co-operation and Development 2012a) reproduction test with D. magna, a clear dose–response relationship was found for fecundity and reproduction, with a NOEC of 630 µg MPA/L, a LOEC of 1800 µg/L, and an EC10 of 929 µg/L (no 95% CI possible), all GMMC. No clear dose–response relationship was found for the endpoint length of parental daphnids (NOEC 630 µg/L, LOEC 1800 µg/L, both GMMC), nor for the 2 endpoints mortality/immobility of parental daphnids and intrinsic rate of population increase, which both showed a NOEC at the highest tested concentration of 7530 µg MPA/L GMMC and a LOEC of >7530 µg/L (D. Gilberg and G. Chambers, unpublished data). Thus, the overall NOEC for the daphnids is 630 µg MPA/L and the overall EC10 is 929 µg/L, both GMMC.

In the fish partial life cycle test test with D. rerio, at the end of the OECD test guideline 229 (Organisation for Economic Co-operation and Development (2012b) test phase at 23 d, no concentration–response relationship was observed for the parameters survival of the parental fish, fecundity, and fertilization success for the F0 generation. The NOEC of all endpoints was the highest tested nominal concentration of 100 µg MPA/L, corresponding to 73 µg/L GMMC. The EC10 for survival of the parental fish was 18.9 µg/L GMMC, whereas the EC10 for fecundity and fertilization success was calculated to be >73 µg MPA/L GMMC (D. Gilberg and G. Chambers, unpublished data). At the end of the OECD test guideline 210 (Organisation for Economic Co-operation and Development 2013) test phase at 34 d (total duration of the test 57 d), no concentration–response relationship was observed for the parameter hatching success (NOEC 73 µg/L, EC10 >73 µg GMMC MPA/L), but a clear concentration–response relationship was observed for posthatch success (survival) and number of healthy fish, with a NOEC of 10.9 µg MPA/L and an EC10 of 5.8 µg MPA/L (GMMC; no CI), as well as length of the surviving F1 fish, with a NOEC of 1.32 µg/L and an EC10 of 31.9 (CI = 1.3–78) µg MPA/L, and finally wet weight of the surviving F1 fish with a NOEC of 1.32 µg/L and an EC10 of 10.6 (CI = 8–14.8) µg MPA/L (D. Gilberg and G. Chambers, unpublished data). At the highest tested concentration of 0.1 mg MPA/L nominal concentration (73 µg/L GMMC), only 2 of 75 hatched fish survived until the end of the test, showing that the test concentration range was well chosen. The most sensitive endpoints in this partial life cycle test were survival and growth in terms of length and mass of the young F1 fish; the effects were comparatively small but statistically significant at p = 0.05. In contrast, the F0 generation was impacted to a lesser extent. Thus, the overall NOEC for all endpoints in both F0 and F1 generations was 1.32 µg MPA/L, and the lowest EC10 was 5.8 µg MPA/L, both GMMC.

These relatively low (sub)chronic endpoints in fish are supported by Wu et al. (2006), who described a significant inhibition of angiogenesis of the intersegmental blood vessels of embryonic D. rerio exposed to MPA concentrations of ≥0.9 µmol/L (288 µg/L) in the short time window of 32 to 48 h post fertilization (hpf). In view of the rapid embryogenesis of D. rerio, Gao et al. (2014) exposed dechorionated zebrafish eggs from 2 to 72 hpf (when morphogenesis is basically complete) to different concentrations of MPA. As endpoints, heart rate and rhythm were assessed at 52 hpf, and pericardial edema, circulation, hemorrhage, and thrombosis were observed at 72 hpf. The MPA exposure resulted in pericardial edema at 6.92 µmol/L (2217 µg/L) and in abnormal body shape, axis shortening, enlarged yolk sac, and decreased motility at the lowest tested concentration of 1.38 µmol/L (442 µg/L); with the concentrations tested in that study, no NOEC was found. Thus, both Wu et al. (2006) and Gao et al. (2014) support developmental toxicity of MPA at concentrations <300 µg/L in short-term embryotoxicity studies, in agreement with the partial life cycle test reported in the present study. In addition, Gao et al. (2014) also exposed D. rerio embryos at 72 hpf in an acute toxicity assay over 24 h; this test resulted in an LC50 of 55.4 µmol MPA/L (17.75 mg/L), which is clearly lower than the results reported for juvenile carp over 96 h in an OECD test guideline 203 (Organisation for Economic Co-operation and Development 1993) test of >100 mg/L (European Chemicals Agency 2018) and may evidence increased sensitivity of embryonic stages.

On a broader scale of comparison, developmental toxicity as observed in D. rerio was also observed in Wistar rat embryos, murine embryonic stem cells, and mouse fibroblasts exposed to MPA (Eckardt and Stahlmann 2010), and in chicken embryos (Veselý and Veselá 1991, publication in Czech language, seen only as the abstract). Thus MPA has been seen to cause proliferation and cellular and developmental toxicity to mammalian and avian embryos, stem cells, or fibroblasts at concentrations of ≥31 µg/L in mammalian systems, in agreement with the effects seen in fish.

Sewage treatment and aquatic PNECs

The microorganism PNEC for STPs was derived from the activated sludge respiration inhibition test according to OECD test guideline 209 (Organisation for Economic Co-operation and Development 2010), which resulted in a 3-h EC50 of 2213 mg/L and an EC10 of 69 mg/L (European Chemicals Agency 2018), by applying an assessment factor of 10 to the EC10, resulting in a PNEC_STP of 6900 µg MPA/L. This finding is indirectly supported by the only tested concentration of 20 µg MPA/L in the OECD test guideline 314B (Organisation for Economic Co-operation and Development 2008) biodegradation study, in which biodegradation began without any delay or lag phase (T. Junker and M. Herrchen, unpublished data), and also by a ready biodegradability study with MPM, which showed no inhibition of activated sludge activity at a nominal concentration of 100 mg MPM/L (A. Häner, unpublished data).

The NOEC-based aquatic PNEC (PNEC_NOEC) was derived from the lowest chronic NOEC from the tests with green algae, cyanobacteria, daphnids, and fish with an assessment factor of 10, resulting in a PNEC_NOEC of 0.132 µg MPA/L based on the fish NOEC. The EC10-based PNEC_EC10 was derived from the same tests, resulting in 0.58 µg MPA/L, also based on fish, which were the most sensitive species among the 4 groups tested.

Aquatic environmental risk assessment

For STPs, PECs or MECs were compared with the PNEC STP by calculating the risk quotient PEC/PNEC. The highest refined PEC_STP was 8.83 µg/L (Table 2), and the highest STP effluent MEC was 4.19 µg/L (Singer et al. 2016); the PNEC_STP based on OECD test guideline 209 (Organisation for Economic Co-operation and Development 2010; FASS 2018) was 6900 µg/L. Thus, the highest PEC-based risk quotient for STPs was 0.00128, and the highest MEC-based risk quotient was 0.0006. These low risk quotients suggest no significant risk for STPs from the use of MPM and MPA.

Similarly, for receiving freshwaters, the initial EMA PEC was 0.299 µg/L, and the average and median refined PECs, depending on STP removal, were quite close together, with 0.082 and 0.058 µg/L for 12% removal, 0.058 and 0.042 µg/L for 43.6% removal, and 0.037 and 0.027 µg/L for 73.1% removal; the highest refined PECs per country were 0.532, 0.383, and 0.245 µg/L for the 3 removal rates, respectively. The aquatic PNEC_NOEC was 0.132 µg/L, and the PNEC_EC10 was 0.580 µg/L. The corresponding risk quotients for the PNEC_EC10 are shown in Table 3 (showing the PNEC_NOEC risk quotient as well would inflate the table even more). Some cases with a risk quotient >1 suggest there may be risk to surface waters. The situation becomes even more complex with a distribution of all MECs, of the population-weighted refined use-based PECs for the 24 countries and of the ePiE PECs, the latter including both %ile distributions and mean-, low-, and high-flow conditions. Therefore, these risks are shown graphically in Figure 3.

All PECs or MECs with a risk quotient >1 (i.e., >0.132 µg MPA/L for PNEC_NOEC or >0.580 µg MPA/L for PNEC_EC10) signify potential risk. The initial use-based Europe-wide PEC (black filled circle in Figure 3A) shows a risk quotient of 2.27 for the PNEC_NOEC but no significant risk (risk quotient 0.516) for the PNEC_EC10. The percent-ranked refined, total-MPA-use-, water-use-, and dilution-factor–based PECs per country assuming 12% (black triangles pointing down) or 73.1% (black triangles pointing up) removal in STPs showed no significant risk based on the PNEC_NOEC for 21 of 24 countries, or >94% of instances referring to total population, with 3 countries (<6% of total population) potentially at risk. All the latter 3 countries are characterized by a combination of above average total MPA use and below average water use times receiving water dilution. The average and median refined PECs were 0.082 and 0.088 µg MPA/L, respectively, for low (12%) STP removal, and the average and median refined PECs were 0.037 and 0.027 µg MPA/L, respectively, for high (73.1%) STP removal. Hence, for both averages and medians, the PNEC_NOEC indicates no significant risk. Based on the PNEC_EC10, none of the 24 countries in both low- and high-removal scenarios show potential risk.

The mean-flow ePiE PEC distribution for the Guadalquivir, Ouse, Rhine, and Turia catchments are shown in Figure 3B, as thick green, brown, orange, and mauve lines, with the low-flow PECs in the same colors to the right and the high-flow PECs to the left. All ePiE PEC lines have in common that the graphed values only start the PECs at ≥0.0005 µg/L, which corresponds to half the LOQ for the MECs; lower values are not shown. The mean-flow ePiE PEC distribution for the Rhine catchment (orange lines) suggests no significant risk for approximately 89% based on the PNEC_NOEC (left vertical red line, Figure 3B) and, vice versa, potential risk for approximately 11% of the whole Rhine catchment; based on the PNEC_EC10 (right vertical red line), no significant risk is indicated up to approximately 97.5% and a potential risk for approximately 2.5% of the whole catchment. Under low flow conditions, the percentages at no risk decreased to approximately 72 and approximately 92% based on the PNEC_NOEC and the PNEC_EC10, respectively. The mean-flow ePiE PEC distribution for the Guadalquivir catchment (green lines) suggests no significant risk for approximately 78% based on the PNEC_NOEC and >89% based on the PNEC_EC10; for the worst-case low-flow Guadalquivir PECs, no significant risk appeared for approximately 68% based on the PNEC_NOEC and for approximately 80% based on the PNEC_EC10. The lower inclination of the Guadalquivir PECs in comparison with the Rhine in Figure 3 is likely the result of a higher spatial variation in concentrations; this can be due to geographical variation in flow and/or population density (consider the presence of Cordoba and Seville in this midsize river basin), plus all the factors that influence concentrations in a warm Mediterranean climate zone. For the Turia catchment (mauve lines) mean flow, no risk emerged for 96.5% for PNEC_NOEC and >99% for PNEC_EC10; for Turia low flow, there was no risk for 90 and 96%, respectively. Lastly, for the Ouse catchment (brown lines) mean flow, there was no significant risk for 91% for PNEC_NOEC and >99% for PNEC_EC10; for Ouse low flow, there was no risk for 80 and 93%, respectively.

All available MECs were percentage ranked and graphed and are shown in Figure 3C. Of the total of 110 MECs (bright blue dots), 51 (46.4%) were below the LOQ. The remaining MECs increased up to a maximum of 0.656 µg MPA/L. Five of the 110 MECs (4.6%) were above the PNEC_NOEC of 0.132 µg/L, whereas only 1 (0.91%) was higher than the PNEC_EC10 of 0.580 µg/L.

Figure 3 shows first that the distribution of the MECs groups nicely with the ePiE PECs: quantifiable (>0.001 µg/L) MECs start at approximately 50% of the total number of samples, which is not too far from that of the ePiE PECs. Also, the general inclination and highest MEC fit quite well into the ePiE PECs. However, this fit might be an artifact, seeing that the MECs are from Switzerland (specifically, a part that does not drain into the Rhine catchment), Poland (which is not yet incorporated into the ePiE), and northeastern Spain (which is not part of the Turia or Guadalquivir catchments and climatically not the same as southern Spain). Hence, the fit may represent a genuine property of use, STP removal, and environmental fate of MPA, but it might be coincidence, or somewhere in between. Also, the comparison of the refined PECs per country, in particular with high (73.1%) STP removal, may be tricky, because compound MECs from only 3 countries, with widely varying numbers of measurements, are being compared with a distribution of 24 single-country average PECs. Still, reasonable commonalities are the upper (99th %ile) ends of all distributions, which range between 0.1 and nearly 10 µg/L and which are certainly dependent on the total amount of MPA introduced into all models and samples. Also, the European Union-wide PECs without removal and with low and high removal in STPs agree quite well overall with the ePiE PECs and the MECs, but also with the single-country PECs with low and high removal. Although it is recognized that the comparisons are not straightforward and may be equivocal, it is proposed that there is sufficient overlap to accept some commonalities, specifically in the upper regions of PECs and MECs, so that a comparison of the PEC and MEC distributions seems reasonably well founded.

Antibiotic resistance risk assessment

Noto et al. (1969) tested the activity of MPA against various microorganisms, both fungi and some strains of bacteria. For 12 strains of S. aureus they determined MICs between 31.25 and 125 mg MPA/L; for Staphylococcus epidermidis and 2 strains each of Shigella flexneri, Proteus vulgaris, Escherichia coli, Pseudomonas aeruginosa, and Salmonella enteritidis, they found MICs of 125 to >500 mg MPA/L. These high MIC values characterize MPA as a rather weak antibiotic. Higher inhibition of MPA was shown against pathogenic fungi of the genera Candida, Willia, Cryptococcus, Microsporum, Aspergillus, and Trichophyton, with MICs ranging from 3.9 to 500 mg/L (Noto et al. 1969). Based on the 23 bacterial MICs given by Noto et al. (1969), a provisional PNEC_ABR for MPA can be derived following Kümmerer and Henninger (2003) by dividing the lowest MIC by 100, resulting in a first PNEC_ABR of 312.5 µg/L. The procedure described by Bengtsson-Palme and Larsson (2016) results in a second provisional PNEC_ABR for MPA, calculated by J. Bengtsson-Palme, of “76 µg/L (which would be rounded down to 64 µg/L”; J. Bengtsson-Palme, personal communication to J.O. Straub, 25 July 2018). Selecting the lower of the 2 values results in a provisional PNEC_ABR of 64 µg MPA/L, based on a small MIC dataset.

This provisional PNEC_ABR of 64 µg MPA/L can now be compared with the PECs for STPs and surface water and with the highest MECs available. The highest use-based refined PEC_STP was 8.83 µg MPA/L (Table 2), whereas the highest STP effluent MEC was 4.19 µg/L (Singer et al. 2016); the corresponding risk quotients were 0.138 and 0.065, respectively, suggesting no significant antibiotic resistance risk for STPs. For surface waters, all PECs including the highest, the Guadalquivir River low-flow 99th %ile ePiE PEC, were below the PNEC_ABR, and thus all risk quotients were ≤0.194, all refined PEC-per-country risk quotients were ≤0.0083, and all mean- and low-flow median ePiE PEC risk quotients were ≤0.00033; finally, the highest MEC risk quotient was 0.0103. Both models and measurements imply that throughout Europe, even in the worst-case scenario of a southern Mediterranean catchment with extremely low river flow in summer, no significant antibiotic resistance risk emerges for MPA.

Secondary poisoning for (semi)aquatic top predators and humans

Risk assessments for secondary poisoning of both (semi)aquatic top predators and human consumers of water and fish depend on mammalian toxicity data. Co-author T. Pfister collated and assessed such data for the derivation of an ADE value for MPA (T. Pfister, unpublished data). According to these same data, based on chronic and subchronic studies with mice, rats, dogs, and monkeys, adverse effects on the hematopoietic and lymphoid system were identified as the leading changes; the rat was the most sensitive species, and the no-observed-adverse-effect level (NOAEL) after chronic (6 or 12 mo) oral treatment was established at 2 mg/kg body weight/d. The anticipated therapeutic effect, immunosuppression, was achieved at or below no-effect dose levels for toxicity in (sub)chronic studies in rat and monkey, as assessed in vitro by the effect of serum from dosed animals. In terms of genotoxicity, MPM did not induce point mutations or primary DNA damage in the presence or absence of metabolic activation. More recent studies with extended exposure conditions have shown that MPM possesses a clastogenic potential, which becomes expressed only at highly cytotoxic dose levels, apparently as a consequence of purine synthesis inhibition. It was not carcinogenic in mice dosed orally for 104 wk at 25, 75, or 180 mg/kg body weight/d, or in rats dosed for at least 104 wk at 3, 7, or 15 mg/kg body weight/d. On the contrary, in secondary pharmacology studies (T. Pfister, unpublished data), MPM demonstrated in vitro antitumour effects in lymphocyte and erythrocyte systems, and MPM prolonged survival times in mice with large cell lymphoma. In a female fertility and reproduction study in rats dosed orally, the highest dose of 4.5 mg/kg body weight/d caused malformations (mainly of the head and eyes) in the pups in the absence of maternal toxicity, whereas there was no effect on fertility; the NOAEL was 1.5 mg/kg body weight/d. In teratology studies, rats and rabbits were dosed orally daily, resulting in deformities including head and ventral wall abnormalities in the rat and cardiovascular, kidney, and lung effects in the rabbit; the no-observed-effect level (NOEL) for teratogenic changes was 2 mg/kg body weight/d for rats and 30 mg/kg body weight/d for rabbits (T. Pfister, unpublished data).

The ADE was derived based on the available nonclinical and clinical data, with the critical effects identified as 1) immunosuppression and secondary effects thereof and 2) teratogenicity; in addition, the relevance of positive results in genotoxicity tests was assessed (T. Pfister, unpublished data). Based on the NOAEL of 2 mg/kg body weight/d in the 12-mo oral toxicity study in the rat, an ADE was derived by multiplying this NOAEL by a default body weight of 60 kg for humans and dividing by adjustment factors, namely, 6.2 for extrapolation from rats to humans, 10 for variability between individuals, 1 for study duration of 1 yr for rodents, 5 for severity of systemic toxicity (hematopoiesis, immunotoxicity), and 3 for use of an established NOAEL instead of an NOEL; this derivation results in a first ADE of 0.1 mg MPM/60-kg person/d. Based on the NOEL of 2 mg/kg body weight/d in the teratogenicity study in the rat, an ADE was derived by multiplying this NOEL by a default body weight of 60 kg for humans and dividing by adjustment factors, namely, 6.2 for extrapolation from rats to humans, 10 for variability between individuals, 1 for reproductive studies in which the whole period of organogenesis was covered, 10 for a teratogenic effect without maternal toxicity, and 1 for use of an established NOEL; this derivation resulted in a second ADE of 0.19 mg MPM/60-kg person/d (T. Pfister, unpublished data). The weak clastogenic potential observed at higher doses was considered to be irrelevant at these exposure levels. Because the first ADE was lower than, and therefore also protective of, the second, 0.1 mg MPM/60-kg person/d was selected as the ADE. Stoichiometrically, this ADE for MPM corresponds to 0.0739 mg MPA/d, which was rounded to 0.075 mg or 75 µg MPA/d. The MTDI for an otter was derived by normalizing the ADE for a 60-kg human to the otter with a default body mass of 10 kg, resulting in an MTDI of 0.0125 mg or 12.5 µg MPA/10-kg otter/d.

For the top predator risk assessment through secondary poisoning, an otter is assumed to consume 1 kg of fish plus 0.79 L water every day (European Commission 2011; Murray-Smith et al. 2012). The fish will take up MPA from the water with the highest modeled BCF, the geometric average of the SciFinder values in the pH range of 5 to 9, of 6.22 (see the previous section, Physicochemical properties and environmental fate). Using the Rhine River mean-flow 99th %ile ePiE PEC of 0.671 µg/L, which is close to the highest MEC available of 0.656 µg/L, and the BCF of 6.22, results in 4.17 µg MPA/kg fish. In addition, the otter drinks 0.79 L of water at 0.671 µg MPA/L or 0.53 µg MPA/d. Thus, the total MPA uptake for an otter is 4.7 µg MPA/d. This is below the MTDI of 12.5 µg MPA/d, with a risk quotient of 0.376. Using the Rhine River low-flow 99th %ile ePiE PEC, in contrast, results in a risk quotient of 43.2, suggesting potential risk. As worst-case scenarios, the highest Guadalquivir River mean- and low-flow 99th %ile ePiE PECs result in risk quotients of 88.9 and 201, respectively. These risk quotients suggest potential risk in extreme exposure scenarios to top predators like otters, which are reported to occur in the whole of Europe including the Iberian Peninsula and the Guadalquivir catchment (International Union for Conservation of Nature 2019). As for the aquatic risk assessment, the high ends of the ePiE PEC distributions signal potential risk, whereas the median refined and ePiE PECs suggest no significant risk.

For the human secondary poisoning risk assessment, a default 60-kg person is expected to consume 2 L of drinking water/d (European Commission 2011; Murray-Smith et al. 2012). Assuming as a worst case that no elimination of MPA takes place during drinking water treatment (despite the evidence to the contrary cited previously), this corresponds to 1.312 µg MPA/d. Using the data calculated just above for the otter, an amount of 115 g fish meat is predicted to contain 0.480 µg MPA. Therefore, a human is expected to consume a total of 1.792 µg MPA/d through secondary poisoning, which is well below the ADE of 75 µg/d, with a risk quotient of 0.024. For human secondary poisoning, even the extreme Guadalquivir River low-flow 99th %ile ePiE PEC of 12.419 µg MPA/L does not result in a significant risk, with a risk quotient of 0.442. Hence, there is no significant risk for human secondary poisoning.

Persistence, bioaccumulation, and toxicity assessment

The water/sediment test (Z. Yan, unpublished data), the OECD test guideline 314B (Organisation for Economic Co-operation and Development 2008) biodegradation test (T. Junker and M. Herrchen, unpublished data), and the k_biodeg results from Franquet-Griell et al. (2017a) suggest that MPA is not persistent. Furthermore, measured log D_OW values of 2.28 at pH 5, 0.48 at pH 7, and ≤–1.54 at pH 9 (V. Nicholson et al., unpublished data) suggest that MPA is not bioaccumulative, which is supported by modeled BCFs (US Environmental Protection Agency 2016; American Chemical Society 2018). Lastly, MPA is toxic based on mammalian carcinogenic, mutagenic, or reprotoxic properties (T. Pfister, unpublished data) and on chronic ecotoxicity in green algae and fish with NOECs <10 µg/L (European Chemicals Agency 2018; D. Gilberg and G. Chambers, unpublished data). Overall, even though it is toxic, MPA is not classified as persistent, bioaccumulative, or of specifically high ecotoxicity.