Mixture toxicity of the antiviral drug Tamiflu® (oseltamivir ethylester) and its active metabolite oseltamivir acid

Aquatic Toxicology 96 (2010) 194–202

Contents lists available at ScienceDirect

Aquatic Toxicology journal homepage: www.elsevier.com/locate/aquatox

Mixture toxicity of the antiviral drug Tamiflu® (oseltamivir ethylester) and its active metabolite oseltamivir acid Beate I. Escher a,b,∗ , Nadine Bramaz b , Judit Lienert b , Judith Neuwoehner b , Jürg Oliver Straub c a b c

The University of Queensland, National Research Centre for Environmental Toxicology (Entox), 39 Kessels Rd, Brisbane, Qld 4108, Australia Eawag, Swiss Federal Institute of Aquatic Science and Technology, 8600 Dübendorf, Switzerland F.Hoffmann-La Roche Ltd, Corporate Safety, Health and Environmental Protection, 4070 Basel, Switzerland

a r t i c l e

i n f o

Article history: Received 29 August 2009 Received in revised form 20 October 2009 Accepted 23 October 2009 Keywords: Environmental risk assessment Pharmaceuticals Algae Metabolite Mixture Tamiflu®

a b s t r a c t Tamiflu® (oseltamivir ethylester) is an antiviral agent for the treatment of influenza A and B. The prodrug Tamiflu® is converted in the human body to the pharmacologically active metabolite, oseltamivir acid, with a yield of 75%. Oseltamivir acid is indirectly photodegradable and slowly biodegradable in sewage works and sediment/water systems. A previous environmental risk assessment has concluded that there is no bioaccumulation potential of either of the compounds. However, little was known about the ecotoxicity of the metabolite. Ester hydrolysis typically reduces the hydrophobicity and thus the toxicity of a compound. In this case, a zwitterionic, but overall neutral species is formed from the charged parent compound. If the speciation and predicted partitioning into biological membranes is considered, the metabolite may have a relevant contribution to the overall toxicity. These theoretical considerations triggered a study to investigate the toxicity of oseltamivir acid (OA), alone and in binary mixtures with its parent compound oseltamivir ethylester (OE). OE and OA were found to be baseline toxicants in the bioluminescence inhibition test with Vibrio fischeri. Their mixture effect lay between predictions for concentration addition and independent action for the mixture ratio excreted in urine and nine additional mixture ratios of OE and OA. In contrast, OE was an order of magnitude more toxic than OA towards algae, with a more pronounced effect when the direct inhibition of photosystem II was used as toxicity endpoint opposed to the 24 h growth rate endpoint. The binary mixtures in this assay yielded experimental mixture effects that agreed with predictions for independent action. This is consistent with the finding that OE exhibits slightly enhanced toxicity, while OA acts as baseline toxicant. Therefore, with respect to mixture classification, the two compounds can be considered as acting according to different modes of toxic action, although there are indications that the difference is a toxicokinetic effect, not a true difference of mechanism of toxicity. The general mixture results illustrate the need to consider the role of metabolites in the risk assessment of pharmaceuticals. However, in the concentration ratio of parent to metabolite excreted by humans, the experimental results confirm that the active metabolite does not significantly contribute to the risk quotient of the mixture. © 2009 Elsevier B.V. All rights reserved.

1. Introduction Oseltamivir ethylester phosphate is used under the trade name of Tamiflu® as an antiviral agent for the treatment and prophylaxis of influenza A and B. Its mechanism of action is related to the inhibition of the influenza virus neuraminidase (Roche, 2006). With the appearance of the bird flu in humans in 2007, and the H1N1 pandemic in 2009, the sales and use of this compound have increased tremendously. Most countries also keep a stock of Tamiflu® to treat

∗ Corresponding author at: The University of Queensland, National Research Centre for Environmental Toxicology (Entox), 39 Kessels Rd, Brisbane, Qld 4108, Australia. Tel.: +61 7 3274 9180; fax: +61 7 3274 9003. E-mail address: [email protected] (B.I. Escher). 0166-445X/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.aquatox.2009.10.020

up to 50% of their population in case of an emergency (Singer et al., 2008). In June 2006, a new “Guideline on the Environmental Risk Assessment of Medicinal Products for Human Use” (EMEA, 2006) was released in the European Union, which requires an environmental risk assessment for all marketing authorization applications. It is also of interest to investigate whether compounds that are already on the market pose a hazard to the environment, particularly those with growing market shares like Tamiflu® . A large number of different pharmaceuticals have been detected in surface waters (Ternes, 1998; Kolpin et al., 2002). The parent compound of Tamiflu® has not been detected in surface waters yet. However, its active metabolite oseltamivir acid (also called oseltamivir carboxylate) was detected in low ng/L concentrations during the flu season in Japan (Ghosh et al., 2009; Söederströem

B.I. Escher et al. / Aquatic Toxicology 96 (2010) 194–202

et al., 2009). Oseltamivir acid is relatively persistent, showing only indirect photodegradation (Bartels and von Tumpling, 2008), slow degradation in surface waters, but increased (microbial) degradation in sediment/water systems (Accinelli et al., 2007; Sacca et al., 2009). Tamiflu® manufacturer Roche performed a prospective environmental risk assessment according to the EMEA guideline with a few modifications to account for pandemic use conditions (Straub, 2009). This study concluded that Tamiflu® does not pose an environmental risk. The risk quotient remains below one even under influenza pandemic conditions, despite very high predicted environmental concentrations in surface waters, due to its relatively low aquatic ecotoxicity (Straub, 2009). Hutchinson et al. (2009) complemented this study by assessing the risk and PBT characteristics of oseltamivir acid for the marine environment according to the Technical Guidance document of the EU (European Commission, 2003). They found no environmental risk for the use of Tamiflu® in a pandemic situation. According to the EMEA guideline, risk assessment needs to be performed not only on the parent compound but also on the human metabolites, provided they exceed 10% of the parent or the parent is a pro-drug (EMEA, 2006). Oseltamivir ethylester (OE) is a pro-drug that lacks antiviral activity (Goodman and Gilman, 2006). Approximately 80% of OE is bioavailable after oral administration (He et al., 1999). OE is hydrolyzed by esterases in the liver to oseltamivir acid (OA) under the release of ethanol with a yield of 75% and is the active antiviral agent (DrugBank, 2006; Goodman and Gilman, 2006; Roche, 2006; Straub, 2009). No further metabolism occurs and elimination is primarily via urine with 60–70% of an oral dose appearing in the urine as the active metabolite (He et al., 1999). The environmental risk assessment of Tamiflu® accounted for the metabolite formation by performing toxicity tests both with OE and with a mixture of OE and OA in the ratio of 1:4, which corresponds to the ratio excreted in urine (Straub, 2009). The chronic toxicity assays with algae, daphnia and fish performed with this mixture resulted in “No Observed Effect Concentrations” (NOEC) of 10, >1, and >1 mg/L, respectively (Straub, 2009). The PNEC (predicted no effect concentration) of 0.1 mg/L was derived from the fish early life stage test with Danio rerio using an assessment factor of 10. Even for worst-case exposure under pandemic conditions, the risk analysis indicated no significant risk to surface water or sewage works (Straub, 2009). A more thorough investigation of the mixture toxicity of OE and OA would be helpful to support the conclusions of this environmental risk assessment because the mixture expected to be excreted to wastewater was used in the risk assessment without resolving the constituents of the mixture. A predictive model of the mixture toxicity of pharmaceuticals and their transformation products had indicated that OA does not significantly influence the overall toxic potential (Lienert et al., 2007). However, the newly published ecotoxicity data (Straub, 2009) and a more thorough analysis as well as an improved prediction of the physicochemical parameters of OE and OA for the present study gave rise to the assumption that OA may contribute substantially to the overall mixture toxicity. Therefore it was the aim of this study to perform binary mixture experiments with OE and OA with a focus on learning more about their mixture effect and how a parent compound interacts in mixtures with its metabolite but also to substantiate the risk assessment of Tamiflu® . It is relatively laborious to perform systematic mixture toxicity experiments with chronic tests and given that the NOECs were relatively high, no clear answers would be expected. Mixture toxicity experiments were performed using an acute algal toxicity assay because algae proved to be a more sensitive aquatic species than daphnids or fish (Straub, 2009). The algal toxicity test was complemented by a non-specific standard bacterial toxicity screening

195

assay, the bioluminescence inhibition test with the marine bacterium Vibrio fischeri (Escher et al., 2008a), which yields information about baseline toxicity and the inhibition of energy transduction. A sound mixture toxicity analysis requires some information on the modes of toxic action of the mixture constituents. Compounds that have the same target site and act according to the same mode of action are likely to follow the mixture toxicity concept of concentration addition, while compounds that have different target sites act according to independent action (Altenburger et al., 2003). Therefore, prior to studying the mixtures, a mode of action analysis was performed with the single compounds to set up the appropriate hypotheses for the mixture toxicity experiments. The paper is concluded with considerations on the inclusion of metabolites into the risk assessment of pharmaceuticals highlighting the example of Tamiflu® and the results of the mixture study are related to the recently published environmental risk assessment for Tamiflu® (Straub, 2009). 2. Materials and methods 2.1. Chemicals Oseltamivir ethylester phosphate (CAS 204255–11–8 for phosphate salt, CAS 196618–13–0 for free base) and d-tartrate salt of oseltamivir carboxylate (CAS 187227–45–8 for the OA zwitterion) were kindly provided by F.Hoffmann-La Roche Ltd, Basel, Switzerland. To avoid any ambiguity related to the molecular weight of the salt and the speciation in solution, all data are reported in molar units. For comparison with literature data, the molecular weights are 410.4 g/mol for OE phosphate salt and 357 g/mol for OA tartrate salt. For structures and physicochemical properties refer to Table 1. 2.2. Chemical analysis The concentrations of OE in the bioassays were quantified with HPLC (Summit HPLC System; Dionex, Olten, Switzerland) and UV detection at 220 nm (UVD 340-U, Dionex, Olten, Switzerland). A reversed-phase C18 column (125/4 Nucleodur Gravity 5m. Macherey-Nagel, Oensingen, Switzerland) was used for separation. The eluent was composed of buffer (10 mM ortho-phosphoric acid at pH 7) and acetonitrile (70:30). 2.3. Bioluminescence inhibition in V. fischeri The 30-min bioluminescence inhibition test with the marine bacterium V. fischeri was used for assessing baseline toxicity and specific interference with the energy metabolism. It was performed according to ISO guideline 11348–3 (International Organization for Standardization, 1998) with modifications as described in (Escher et al., 2005b) using freeze-dried bacteria (Dr. Lange, Düsseldorf, Germany). The mixture experiments were performed in 96-well microtiter plate format after it was confirmed that measured concentrations were equal to nominal concentrations (Escher et al., 2008a). The luminescence output of the bacteria was measured prior to addition of the sample and after 30-min incubation (MicroLumatPlus, Berthold Technologies, Regensdorf, Switzerland). The inhibition of bioluminescence was calculated as described in ISO guideline 11348–3. The effect concentrations EC50 were derived from log-logistic concentration response curves (Escher et al., 2005a, 2008b). Full concentration–effect curves were determined for all binary mixture ratios (exact ratios given in Table 3). Mixture experiments were performed in a minimum of triplicates and alongside single compounds. The reported data relate to the best fit of a single concentration–effect curve over all accumulated data (3–10 replicates).

196

B.I. Escher et al. / Aquatic Toxicology 96 (2010) 194–202

Table 1 Structures and physicochemical properties of oseltamivir ethylester and its human metabolite oseltamivir acid. log Kow a

log Klipw

OE (Oseltamivir ethylester)

1.21

OA (Oseltamivir acid)

0.006

a b c d e

b

d

pKa (base)c

pKa (acid)c

fneutral

log Dlipw (pH 7)e

1.61

7.60

–

0.201

1.06

0.52

7.81

3.78

0.865

0.46

Octanol–water partition coefficient, average of various QSARs as cited in Straub (2009). Liposome–water partition coefficient calculated with Eq. (1). Acidity constant, estimated using SPARC (Hilal et al., 2005). Fraction of neutral species at pH 7 (Schwarzenbach et al., 2003). Liposome–water distribution ratio at pH 7 calculated with Eq. (2).

2.4. Algal toxicity Direct and indirect effects on photosynthesis were evaluated with the 24 h chlorophyll fluorescence test with the green algae Desmodesmus subspicatus using the chlorophyll fluorometer ToxYPAM according to Escher et al. (2005a). Mixture experiments were performed with Pseudokirchneriella subcapitata in 96-well microtiter plate format (Escher et al., 2008a) after confirmation that measured concentrations equaled nominal concentrations. Two different algal species were used to ensure consistency with the baseline toxicity quantitative structure activity relationship (QSAR). Growth conditions were identical for each of the two strains and both were used in the exponential growth phase. A Maxi-Imaging-PAM (IPAM) (Walz GmbH, Effeltrich, Germany) was used to determine the yield of photosynthesis Y after 2 and 24 h, and optical density (Spectramax® Plus 384, Molecular Devices Corporation, Sunnyvale, USA) was measured to derive the growth rate during 24 h. The EC50 values for the inhibition of the photosynthetic yield after 2 h (EC502h IPAM ) or after 24 h (EC5024h IPAM ) and the growth rate inhibition (EC5024h growth ) were derived from log-logistic concentration–effect curves as described previously (Escher et al., 2008a). The mixture experiments were performed analogously to those described above, however different concentration ratios were used (see Table 5). 2.5. Speciation and hydrophobicity indicators The liposome–water distribution ratio at pH 7, Dlipw (pH 7), is a better descriptor for bioaccumulation and hydrophobicity of ionizable compounds than the commonly used octanol–water partition coefficient Kow (Escher and Hermens, 2002) and was therefore used for the QSARs. Speciation has to be accounted for when estimating the liposome–water distribution (Schwarzenbach et al., 2003). The liposome–water partition coefficient of the neutral species Klipw,neutral was calculated from the estimated octanol–water partition coefficient log Kow with Eq. (1), which is valid for polar compounds (Vaes et al., 1997; Escher et al., 2006). log Klipw,neutral = 0.904 log Kow + 0.515

(1)

for charged species The corresponding Klipw,charged (charged = anionic or cationic) was assumed to be approximately one order of magnitude lower than that of the corresponding neutral species, i.e. log Klipw,charged = log Klipw,neutral − 1 (Escher and Sigg, 2004). The Dlipw (pH 7) was computed with Eq. (2), where fneutral refers to the fraction of neutral species. log Dlipw (pH7)=fneutral log Klipw,neutral +



1−fneutral log Klipw,charged (2)

2.6. Baseline toxicity versus specific mode of toxic action The toxic ratio TR (Eq. (3)) is a measure of the specificity of a mode of toxic action and is defined as the quotient of the predicted baseline effect concentration EC50baseline toxicity of a given compound to its experimentally determined EC50experimental (Verhaar et al., 1992). TR < 10 corresponds to baseline toxicity, and TR ≥ 10 to a specific mode of toxic action (Verhaar et al., 1992). The higher the TR, the higher the intrinsic potency of a chemical, i.e. the more pronounced is a specific mode of toxic action. TR =

EC50baseline toxicity EC50experimental

(3)

The EC50baseline toxicity values were predicted with QSARs taken from the literature (European Commission, 2003; Escher et al., 2005a). Published QSARs are typically based on the Kow as hydrophobicity descriptor. As discussed above, the Kow is not a suitable hydrophobicity descriptor for ionized species. However, we have earlier demonstrated (Escher et al., 2002) that QSARs based on the Dlipw (pH 7) are valid for the whole spectrum of neutral, positively and negatively charged molecules, provided that the mode of toxic action is baseline toxicity. Therefore Kow -based literature QSARs can be rescaled to the liposome–water distribution ratio of all species at pH 7, log Dlipw (pH 7), as the hydrophobicity descriptor by inserting equation 1 in each QSAR equation (Escher and Schwarzenbach, 2002). The resulting QSARs based on log Dlipw (pH 7) as hydrophobicity descriptor are listed in Table 2. The QSAR for the bioluminescence inhibition test in microtiter plate format was experimentally determined for a set of baseline

B.I. Escher et al. / Aquatic Toxicology 96 (2010) 194–202

197

Table 2 Measured and modeled EC50 values of the parent compound OE for various aquatic endpoints. Organism Bacteria Vibro fischeri

Algae Selenastrum capricornutum (exp.) or Chlorella vulgaris (QSAR) Algae Desmodesmus subspicatus Waterflea Daphnia magna Fish Cyprinus carpio (exp.) or Pimephales promelas (QSAR) a b c d e

Endpoint

EC50experimental (mM)

QSAR equation

EC50, parent, baseline QSAR (mM)

a

log (1/EC50baseline toxicity (M)) = 0.7 log Dlipw (pH 7) + 1.54c

4.5

1.1b

log (1/EC50baseline toxicity (M)) = 0.91 log Dlipw (pH 7) + 0.63d

27.7

log (1/EC50baseline toxicity (M)) = 0.91 log Dlipw (pH 7) + 1.10c log (1/EC50baseline toxicity (M)) = 0.77 log Dlipw (pH 7) + 1.89d log (1/LC50baseline toxicity (M)) = 0.83 log Dlipw (pH 7) + 1.46d

9.4

30-min bioluminescence inhibition 72 h growth inhibition

10

PSII inhibition after 24 h 48 h immobilization 96 h lethality

15.5a b

0.08

0.24b

TRe 0.4

25

0.6

2.1

26

4.9

20

Data from this study, performed in vials with experimental verification of exposure concentration by HPLC. Data from Straub (2009). QSAR from Escher et al. (2005a). Baseline toxicity QSAR rescaled from Kow -based QSARs from the Technical Guidance Document (TGD) of the EU (European Commission, 2003). Toxic ratio (TR) calculated with Eq. (3).

toxicants in ref. (Escher et al., 2008a) (Eq. (4)). log(1/EC50baseline toxicity,Vibrio fischeri (M))=0.84 log Dlipw (pH 7)+1.69 (4) The EC50baseline toxicity for the combined algae test is given in equation 5 for the endpoint 24 h growth rate and in equation 6 for the endpoint 24 h IPAM (Escher et al., 2008a). log(1/EC50baseline toxicity,24h growth (M))=0.95 log Dlipw (pH 7) + 1.16

(95% confidence interval: 7.9–11.2 mM) for the test performed according to ISO 11348–3 (Table 2) and 10.3 mM (95% confidence interval: 9.9–10.7 mM) using the microtiter plate format (Table 3). These values correspond to an EC50 value of 4.3 (4.0–4.6) g/L of OE. According to ISO 11348–3, all tested concentrations were confirmed by HPLC and refer to aqueous concentrations in the bioassay. The good agreement between the two test setups and between measured and nominal concentrations allowed us to perform all further metabolite and mixture experiments in the microtiter plate format reporting nominal concentrations.

(5) log(1/EC50baseline toxicity,24h IPAM (M)) = 0.84 log Dlipw (pH 7) + 1.07 (6) 2.7. Mixture toxicity evaluation The mixture experiments were evaluated with the isobologram method described by Altenburger and Boedeker (1990). Here the experimental EC50 value that refers to total molar concentration of both components of the binary mixture for a given mixture ratio was multiplied with the fraction of each mixture component. The two axes of the isobologram were constructed from EC50 times fraction of OA and EC50 times fraction of OE, respectively. The line connecting the two axis intercepts in this isobologram is then equivalent to the model of concentration addition (Altenburger and Boedeker, 1990). Although it is not common practice, it is also possible to plot the prediction for independent action into an isobologram, in order to differentiate between independent action and true antagonism. In this case, for each mixture ratio, a prediction of EC50 was performed independently with the equations described in Backhaus et al. (2000). The fractions of OE and OA were multiplied with this predicted value and the result was plotted in the isobologram along with the experimental results. 3. Results and discussion 3.1. Toxicity of the parent compound OE in the bioluminescence inhibition test in V. fischeri The 30-min EC50 values in the 30-min bioluminescence inhibition test with the marine bacterium V. fischeri of the parent compound oseltamivir ethylester phosphate (OE) were 10.5 mM

3.2. Toxicity of the parent compound OE in the algal toxicity test In the chlorophyll fluorescence test with the green algae D. subspicatus, measuring photosynthesis inhibition after 24 h using the ToxY-PAM, the EC50 was 15.5 mM (95% confidence interval 14.9–16.2 mM), which corresponds to 6.4 (6.1–6.7) g/L (Table 2). P. subcapitata was more sensitive with an EC5024h growth of 0.51 mM (95% confidence interval 0.48–0.55 mM) corresponding to 210 mg/L phosphate salt and an EC5024h IPAM of 0.78 mM (95% confidence interval 0.76–0.81 mM) corresponding to 319 mg/L phosphate salt (Table 4). These EC50 values are very high, but consistent with the low hydrophobicity of the compound. The values are in the same range as the EC50growth of 463 mg/L (1.1 mM) in the 72 h algal growth inhibition test with P. subcapitata (Straub, 2009). 3.3. Toxicity of the metabolite OA The metabolite oseltamivir acid (OA) was almost equipotent to the parent oseltamivir ethylester phosphate (OE) with an EC50 of 6.64 mM (95% confidence interval: 6.55–6.70 mM) for V. fischeri, which corresponds to 2.4 g/L for the used tartrate salt of OA (Table 3). In contrast, in the algal toxicity assay, the EC5024h growth of 8.3 (7.7–8.8) mM and the EC5024h IPAM of 19.3 (18.2–20.4) mM of OA are 16 and 25 times less toxic, respectively, than the corresponding EC50 values for the parent OE (Table 4). The slope of the concentration–effect curve of OA was much steeper than that of OE. The onset of toxicity for OE was relatively fast with first signs of photosynthesis inhibition after 2 h. In contrast, the metabolite OA showed <20% of PSII inhibition after 2 h exposure at concentrations that lead to 100% inhibition of photosynthesis and growth rate after 24 h (data not shown). This is an observation that could not be rationalized with any mechanistic explanation but will be relevant for the mixture toxicity experiments discussed below.

198

B.I. Escher et al. / Aquatic Toxicology 96 (2010) 194–202

Table 3 Descriptors of the log-logistic concentration–effect curves (Escher et al., 2005a) and toxic ratio (TR) analysis for the parent OE and the metabolite OA and their binary mixtures in the bioluminescence inhibition test with Vibrio fischeri using the 96-well plate format. Mixture OE:OA

OE fraction (%)

log (1/EC50experimental (M))

OE Mixture 1:0.2 Mixture 1:0.4 Mixture 1:0.6 Mixture 1:0.8 Mixture 1:1 Mixture 0.8:1 Mixture 0.6:1 Mixture 0.4:1 Mixture 0.3:1 Mixture 0.2:1 OA

100% 83% 71% 63% 56% 50% 44% 38% 29% 25% 17% 0%

1.99 1.94 1.91 1.90 1.93 1.94 2.00 2.02 2.04 2.06 2.10 2.18

a

± ± ± ± ± ± ± ± ± ± ± ±

0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01

Slope of log-logistic fit 2.2 2.5 3.7 6.3 8.3 6.9 7.4 9.1 4.4 36.8 11.9 25.5

± ± ± ± ± ± ± ± ± ± ± ±

0.1 0.2 0.3 0.7 1.0 0.5 0.9 1.3 0.6 225 2.5 3.9

EC50experimental (mM) (95% confidence intervals)

log (1/EC50baseline ) (M)a

TR

10.3 (10.1–10.5) 11.6 (11.2–12.0) 12.3 (11.9–12.7) 12.5 (12.2–12.8) 11.9 (11.7–12.1) 11.5 (11.4–11.6) 10.1 (9.9–10.3) 9.5 (9.3–9.7) 9.0 (8.7–9.4) 8.6 (6.9–10.7) 7.9 (7.7–8.0) 6.6 (6.5–6.7)

2.57

0.26

2.08

1.27

QSAR Eq. (4).

Table 4 Concentration–effect curves and toxic ratio (TR) analysis for the parent OE and the metabolite OA in the algal toxicity test with Pseudokirchneriella subcapitata using the 96-well plate format. Slope of log-logistic fit

EC50experimental (g/L)b

EC50baseline (mM)a

TR

Endpoint 24 h growth rate OE 0.51 (0.48–0.55) OA 8.27 (7.75–8.82)

2.7 2.5

0.21 (0.20–0.26) 2.95 (2.76–3.15)

7.0 25.4

13.7 3.1

Endpoint 24 h IPAM OE 0.78 (0.76–0.81) OA 19.28 (18.22–20.39)

3.1 1.8

0.32 (0.31–0.34) 6.88 (6.51–7.27)

10.8 34.4

14.0 1.8

EC50experimental (mM) (95% confidence intervals)

a b

QSAR Eqs. (5) and (6). Based on phosphate salt for OE and tartrate salt for OA.

3.4. Speciation and hydrophobicity indicators The liposome–water partition coefficient of the neutral species Klipw,neutral was calculated from the estimated octanol–water partition coefficient log Kow of 1.21 for the parent OE and 0.006 for the metabolite OA (Straub, 2009) using Eq. (1). Note, however, that the limit of applicability of such an equation might be reached considering that OE and OA both have a Kow that is outside the test set domain of the QSAR equation. The predicted log Klipw,neutral is 1.61 for OE and 0.46 for OA (Table 1). OE is a weak ammonium base and is therefore positively charged at ambient pH. With an estimated acidity constant of 7.6 (OE) and 7.81 (OA) for the basic amino group (estimated with SPARC (Hilal et al., 2005)), the fraction of neutral

species at pH 7 is 20% for OE. 16% of OE is in its cationic form, resulting in a log Dlipw (pH 7) of 1.06 (Eq. (2)). OA with its acidity constant of the carboxylic acid of 3.78 and its retained aliphatic amine function is a zwitterion and overall neutral at pH 7. The log Dlipw (pH 7) of OA can therefore be assumed to be equal to log Klipw,neutral . 3.5. Toxic ratio analysis Analysis of the literature data for the parent compound OE revealed that the toxic ratio (TR) varies from 0.4 to 26 for the different test species (Table 2). In principle, TR > 10 would be the cut-off value for specific toxicity. However, for the three species with TR 20–26, there are differences between the tested species and

Fig. 1. Toxic ratio (TR) analysis for (A) the bioluminescence inhibition test with Vibrio fischeri and (B) the 24 h growth rate endpoint in the algal toxicity assay with Pseudokirchneriella subcapitata. The experimental data of the parent OE and the metabolite OA are depicted with black diamonds. The line corresponds to the baseline toxicity QSAR for the given endpoint, derived from the experimental data of the baseline toxicants depicted with empty circles (Escher et al., 2008a). (Drawn line: baseline toxicity, TR = 1; broken line: specific mode of toxic action TR ≥ 10).

B.I. Escher et al. / Aquatic Toxicology 96 (2010) 194–202

199

Fig. 2. (A) Isobologram for the binary mixtures of the parent OE and the metabolite OA in the bioluminescence inhibition test with Vibrio fischeri. The drawn line corresponds to the prediction for concentration addition, the broken line to the prediction for independent action. (B) Deviation of the experimental data from the predictions for concentration addition (deviation = (EC50experimental − EC50CA prediction )/EC50experimental , empty circles) and independent action (deviation = (EC50experimental − EC50IA prediction )/EC50experimental , filled circles).

the baseline QSARs that are rescaled from Kow -based QSARs, all of which adds to the uncertainty of the resulting TR value. Therefore no clear conclusions can be drawn. In contrast, the TR remained below 10 (Fig. 1) in all endpoints from literature where the baseline QSAR was established with the same experimental setup, and was defined with experimental Dlipw (pH 7) values as hydrophobicity descriptors. The calculated toxic ratios (TR) of OE and OA in the 30-min bioluminescence inhibition test with the marine bacterium V. fischeri were 0.27 and 1.27, respectively, indicating that both are baseline toxicants towards V. fischeri (Table 3, Fig. 1A). Note that OE and OA are at the lower end of hydrophobicity in the baseline toxicity QSAR indicating an overall low toxicity. While the TR values of OA for both 24 h endpoints in the algal toxicity assay were still in the range to be classified as baseline toxicant, the TR of OE were about 10 times higher than that of OA in both 24 h endpoints, thus marginally but significantly exceeding the threshold value of TR = 10 (Table 4, Fig. 1B). This indicates clear differences in the mode of action between OE and OA or a significant error in the estimation of Dlipw (pH 7) of OE. The latter is rather unlikely because if Dlipw (pH 7) of OE was underestimated the TR in V. fischeri would be <0.1, which is an unrealistic value. The difference in the time to effect discussed above supports the conclusion that there is a difference in mode of toxic action.

symmetric with a maximum of 20–30% deviation from the model (Fig. 2B). The TR analysis had clearly indicated that both, parent OE and metabolite OA, act as baseline toxicants. Therefore, the expected mixture toxicity model is concentration addition (Escher et al., 2002), which was not congruent with the experimental observations. A major difference between OE and OA is the gradient of their concentration–effect curves (Fig. 3). While OE has a slope typical for a baseline toxicant, the slope of OA is unusually steep (Table 3). The mixtures vary in slope depending on the composition. Mixtures with a higher OA content have a correspondingly higher slope (Table 3). For typical binary mixtures, the predictions for concentration addition and independent action are overlapping. Exceptions are binary mixtures of components with largely different slopes as it is observed here. Nevertheless, CA overpredicts the toxicity of the mixture by no more than 25%. Cedergreen et al. (2008) recently compared CA and IA as models for mixtures of chemicals with different molecular target sites and found that neither of the models was more accurate (Table 5). In contrast, in the algal toxicity assay the isobologram for the endpoint 24 h growth rate clearly shows subadditivity with the experimental data perfectly overlapping with the prediction of independent action (Fig. 4A). For the endpoint 24 h IPAM in the

3.6. Mixture toxicity Ten different mixture experiments with binary mixtures and varying ratios of OE/OA were performed with the bioluminescence inhibition test of V. fischeri. The resulting EC50 values are listed in Table 3. When plotting the data in form of an isobologram (Fig. 2A) it is evident that the mixture experiments yield higher EC50 than the prediction for concentration addition (CA), which corresponds to the straight line connecting the EC50 of the two mixture components (Altenburger and Boedeker, 1990). This combination effect is termed subadditivity (Altenburger and Boedeker, 1990). However, the experimental mixture toxicities are still higher than predictions for independent action (IA). The predictions for IA were performed independently from the isobologram analysis according to Backhaus et al. (2000) and a mixture EC50 derived from the prediction for IA was also plotted in the isobologram (as indicated by the broken line in Fig. 2A). The deviation of the experimental data from the predictions of concentration addition (filled circles) and independent action (empty circles) is approximately

Fig. 3. Concentration–effect curves for the parent OE, the metabolite OA, and the equimolar mixture in the bioluminescence inhibition test with Vibrio fischeri.

200

B.I. Escher et al. / Aquatic Toxicology 96 (2010) 194–202

Table 5 Concentration–effect data for the parent OE and the metabolite OA and their binary mixtures in the algal toxicity test with Pseudokirchneriella subcapitata. Mixture OE:OA

OE Mixture 1:1 Mixture 0.3:1 Mixture 0.1:1 Mixture 0.05:1 Mixture 0.02:1 Mixture 0.01:1 OA

OE fraction (%)

100% 50% 25% 9.1% 4.8% 2.0% 1.0% 0%

Endpoint 24 h growth rate

Endpoint 24 h IPAM

log (1/EC5024h growth (M))

Slope of log-logistic fit

log (1/EC5024h IPAM (M))

Slope of log-logistic fit

3.10 2.77 2.19 1.81 1.70 1.69 1.65 1.72

3.10 3.00 1.52 1.54 1.85 1.59 1.53 1.75

3.29 3.00 2.64 2.28 2.18 2.08 2.07 2.08

2.66 1.76 0.87 0.95 1.39 2.20 1.94 2.45

same bioassay, the deviation from concentration addition is even more pronounced, even the model of IA overestimates the mixture toxicity (Fig. 4B). The mixture toxicity analysis thus confirms that OE and OA act according to a different mode of toxic action in algae, with the parent OE being slightly more toxic than baseline toxicity. OE is an aliphatic amine and fully protonated to its cationic form at ambient pH. Similar observations of a slightly enhanced toxicity were reported for various ß-blockers (Escher et al., 2006) as well as fluoxetine and its metabolite norfluoxetine (Neuwoehner et al., 2009). All these compounds have in common that they are aliphatic amines with acidity constants as high as 9–10. A time and effect pattern similar to baseline toxicants has also been reported (Neuwoehner et al., 2009). While this phenomenon is apparently typical for aliphatic amines, it remains unexplained. At present, the hypothesis that this is a toxicokinetic effect rather than a toxicodynamic effect, i.e. an effect that is related to the mode of toxic action, is investigated (Escher and Hermens, 2002). NMR experiments with Scenedesmus vacuolatus indicated that the internal pH in algae is slightly higher than pH 7, the external pH of the experiments (Küsel et al., 1990). Therefore, we hypothesize that internally more of the neutral species is formed and thus there is a higher total internal concentration of OE than we expect from using Dlipw (pH 7) as an indicator of bioaccumulation into algae. Such a model is similar to the ion-trapping model that can explain the pH dependence of toxicity of substituted phenols (Escher and Hermens, 2004). 3.7. Role of the metabolite for the risk assessment of Tamiflu® According to the guideline on the environmental risk assessment of medicinal products for human use, the environmental risk assessment needs to include the metabolites if a metabolite is formed in more than 10% (EMEA, 2006). For this reason Straub

(2009) used a mixture of OE:OA of 1:4, which corresponds to the ratio excreted by humans, for ecotoxicity testing and derived the risk quotient from these experimental data. Such an approach is valid as is demonstrated by the mixture toxicity experiments that were performed in the present study. An alternative approach for implementing the metabolite in environmental risk assessment can be used if no experimental data for the excreted mixture are available. To assess the role of the metabolite for the risk assessment of the parent compound, the first step is to assess the relative potency (RP) of the metabolite compared to the parent compound (Eq. (7)). RP =

EC50parent EC50metabolite

(7)

The RP of OA was 1.6 (i.e. OA being slightly more potent than OE) in the bacterial bioluminescence inhibition test, but only 0.04–0.06 in the algal toxicity assay (i.e. the metabolite OA being 20 times less potent than OE). For mixtures of compounds that act concentration-additive, the risk quotient of the mixture, RQmixture , is the equivalent to the risk quotient of the parent compound, RQparent , multiplied by the total toxic potential of the mixture TPmixture (Lienert et al., 2007) (Eq. (8)). Strictly spoken, this equation holds only for mixtures of concentration-additive acting compounds. In the mode of action analysis described above the deviation of the experimental toxicity from the predictions of CA has been of high relevance. Since this deviation is at maximum a factor of three, it is negligible for the derivation of the RQ given all the other uncertainties pertaining to the derivation of RQ. RQmixture = TPmixture · RQparent

(8)

Fig. 4. Isobologram for the binary mixtures of the parent OE and the metabolite OA in the algal toxicity test with Pseudokirchneriella subcapitata; (A) 24 h growth rate, (B) 24 h IPAM. The drawn line corresponds to the prediction for concentration addition (CA), the broken line to the prediction for independent action (IA).

B.I. Escher et al. / Aquatic Toxicology 96 (2010) 194–202

The total toxic potential of the mixture TPmixture sums up the relative potencies of all metabolites normalized to the fractions excreted fexcreted,i (Eq. (9)). TPmixture = fexcreted,parent +



fexcreted,i · RPi

(9)

i

This equation is simplified to i = OA in case of Tamiflu® . TPmixture was calculated to be 1.4 for the bioluminescence inhibition test and 0.25 for the algal toxicity assay for an excreted ratio of OE:OA of 1:4. This demonstrates that the metabolite, although it is present in three times higher concentrations than the parent compound, does not influence the RQ dramatically in case of Tamiflu® (given that the risk assessment is based on algae as most sensitive species, not on bacteria). 4. Conclusion The findings of the present study support the conclusions of the previous environmental risk assessment of Tamiflu® (Straub, 2009) but it additionally gives insight into the modes of toxic action to non-target organisms and mixture interactions. In the bacterium V. fischeri, both OE and OA, were baseline toxicants acting almost concentration-additive in binary mixtures (maximum deviation from the expected mixture model is 30%). In contrast, in green algae, OE was specifically active and OA was a baseline toxicant. Consequently, the model of independent action described the mixture effect better than that of concentration addition, which is consistent with mixture theory. While it is interesting to understand the intricacies of mixture toxicity and modes of toxic action, the overall message is that the mixture effects were close to concentration addition. This is a prerequisite for summing up risk quotients of mixtures from the risk quotients of the single components. In case of Tamiflu® it is clearly important to account for the metabolite in risk assessment since the metabolite is formed in a large fraction (75%). However, while the metabolite is equipotent to the parent compound in the bacterial assay, it is around 20 times less toxic than the parent compound in the algal toxicity assay. Since the risk assessment is based on the most sensitive species, here algae, the contribution of the metabolite to the mixture risk quotient is negligible. But this result cannot be generalized and the relationship between fraction formed/excreted and toxic potential of the metabolite must explored prior to making conclusions. We propose that mixture toxicity of parent compounds with their metabolites should be included in the future systematically in the risk assessment of pharmaceuticals. The pharmacokinetics are generally well researched and identities and fractions of metabolites excreted into wastewater readily available. If no experimental data are available, the data gaps in toxicity can be reasonably well closed by QSAR predictions and then simple mixture toxicity models can be applied. Acknowledgements This study was partially funded by the Swiss Federal Office for the Environment (FOEN) and by the European Union under the 6th Framework Program in the STREP ERAPharm (SSPI-CT2003-511135). We thank Manuela Richter and Simon Galenda for experimental assistance. We thank F.Hoffmann-La Roche Ltd for the donation of the oseltamivir phosphate and oseltamivir acid tartrate. We thank Eva Holt for reviewing the manuscript. References Accinelli, C., Caracciolo, A.B., Grenni, P., 2007. Degradation of the antiviral drug oseltamivir carboxylate in surface water samples. Int. J. Environ. Anal. Chem. 87, 579–587.

201

Altenburger, R., Boedeker, W., 1990. Evaluation of the isobologram method for the assessment of mixtures of chemicals. Combination effect studies with pesticides in algal biotests. Ecotoxicol. Environ. Saf. 20, 98–114. Altenburger, R., Nendza, M., Schuurmann, G., 2003. Mixture toxicity and its modeling by quantitative structure–activity relationships. Environ. Toxicol. Chem. 22, 1900–1915. Backhaus, T., Altenburger, R., Boedeker, W., Faust, M., Scholze, M., Grimme, L.H., 2000. Predictability of the toxicity of multiple mixtures of dissimilarly acting chemicals to Vibrio fischeri. Environ. Toxicol. Chem. 19, 2348– 2356. Bartels, P., von Tumpling, W., 2008. The environmental fate of the antiviral drug oseltamivir carboxylate in different waters. Sci. Total Environ. 405, 215– 225. Cedergreen, N., Christensen, A.M., Kamper, A., Kudsk, P., Mathiassen, S.K., Streibig, J.C., Sorensen, H., 2008. A review of independent action compared to concentration addition as reference models for mixtures of compounds with different molecular target sites. Environ. Toxicol. Chem. 27, 1621–1632. DrugBank, “Oseltamivir phosphate (APRD01148)”, http://www.drugbank.ca/cgibin/getCard.cgi?CARD=APRD01148.txt. In: D.S. Wishart, Knox, C., Guo, A. C., Shrivastava, S., Hassanali, M., Stothard, P., Chang, Z., Woolsey, J., (Eds.), DrugBank: A Comprehensive Resource for in Silico Drug Discovery and Exploration. Nucleic Acids Res., 34, Suppl. 1, D668–D672, 2006. EMEA, 2006. Guideline on the Environmental Risk Assessment of Medicinal Products for Human Use CHMP/SWP/4447/00. The European Agency for the Evaluation of Medicinal Products, London. Escher, B., Schwarzenbach, R.P., 2002. Mechanistic studies on baseline toxicity and uncoupling as a basis for modeling internal lethal concentrations in aquatic organisms. Aquat. Sci. 64, 20–35. Escher, B.I., Bramaz, N., Eggen, R.I.L., Richter, M., 2005a. In-vitro assessment of modes of toxic action of pharmaceuticals in aquatic life. Environ. Sci. Technol. 39, 3090–3100. Escher, B.I., Bramaz, N., Maurer, M., Richter, M., Sutter, D., von Känel, C., Zschokke, M., 2005b. Screening test battery for pharmaceuticals in urine and wastewater. Environ. Toxicol. Chem. 24, 750–758. Escher, B.I., Bramaz, N., Mueller, J.F., Quayle, P., Rutishauser, S., Vermeirssen, E., 2008a. Toxic equivalent concentrations (TEQs) for baseline toxicity and specific modes of action as a tool to improve evaluation of ecotoxicity tests on environmental samples. J. Environ. Monitor. 10, 612–621. Escher, B.I., Bramaz, N., Quayle, P., Rutishauser, S., Vermeirssen, E., 2008b. Monitoring of the ecotoxicological hazard potential by polar organic micropollutants in sewage treatment plants and surface waters using a mode-of-action based test battery. J. Environ. Monitor. 10, 622–631. Escher, B.I., Bramaz, N., Richter, M., Lienert, J., 2006. Comparative ecotoxicological hazard assessment of beta-blockers and their human metabolites using a modeof-action based test battery and a QSAR approach. Environ. Sci. Technol. 40, 7402–7408. Escher, B.I., Eggen, R.I.L., Schreiber, U., Schreiber, Z., Vye, E., Wisner, B., Schwarzenbach, R.P., 2002. Baseline toxicity (narcosis) of organic chemicals determined by membrane potential measurements in energy-transducing membranes. Environ. Sci. Technol. 36, 1971–1979. Escher, B.I., Hermens, J.L.M., 2002. Modes of action in ecotoxicology: their role in body burdens, species sensitivity, QSARs, and mixture effects. Environ. Sci. Technol. 36, 4201–4217. Escher, B.I., Hermens, J.L.M., 2004. Internal exposure – linking bioavailability to effects. Environ. Sci. Technol. 38, 455A–462A. Escher, B.I., Sigg, L., 2004. Chemical speciation of organics and of metals at biological interfaces. In: Van Leeuwen, H.P., Köster, W. (Eds.), Physicochemical Kinetics and Transport at Biointerfaces. John Wiley & Sons, Chichester, pp. 205–271. European Commission, Technical Guidance Document in Support of Commission Directive 93/67/EEC on Risk Assessment for New Notified Substances, Commission Regulation (EC) No 1488/94 on Risk Assessment for Existing Substances, and Directive 98/8/EC of the European Parliament and of the Council Concerning the Placing of Biocidal Products on the Market. Office for Official Publications of the European Communities, Luxembourg, 2003. Ghosh, G.C., Nakada, N., Yamashita, N., Tanaka, H., 2009. Oseltamivir carboxylate – the active metabolite of oseltamivir phosphate (Tamiflu®), detected in sewage discharge and river water in Japan. Environ. Health Perspect., doi:10.1289/ehp.0900930 (Online 28 September 2009). Goodman, L.S., Gilman, A. (Eds.), 2006. Goodman and Gilman’s The Pharmacological Basis of Therapeutics. McGraw-Hill. He, G., Massarella, J., Ward, P., 1999. Clinical pharmacokinetics of the prodrug oseltamivir and its active metabolite Ro 64-0802. Clin. Pharmacokin. 37, 471–484. Hilal, S.H., Karickhoff, S.W., Carreira, L.A., SPARC – Performs automated reasoning in chemistry. Accessible at http://ibmlc2.chem.uga.edu/sparc/. University of Georgia, Athens, GA, 2005. Hutchinson, T.H., Beesley, A., Frickers, P.E., Readman, J.W., Shaw, J.P., Straub, J.O., 2009. Expanding the environmental risk assessment for oseltamivir (Tamiflu® ) under pandemic use conditions to the coastal marine compartment. Environ. Int. 35, 931–936. International Organization for Standardization, 1998. Water Quality—determination of the inhibitory effect of water samples on the light emission of Vibrio fischeri (luminescent bacteria test), EN ISO 11348-3, Geneva, Switzerland. Kolpin, D.W., Furlong, E.T., Meyer, M.T., Thurman, E.M., Zaugg, S.D., 2002. Pharmaceuticals, hormones and other organic contaminants in U.S. streams,

202

B.I. Escher et al. / Aquatic Toxicology 96 (2010) 194–202

1999–2000: a national reconnaissance. Environ. Sci. Technol 36, 1201– 1211. Küsel, A., Sianoudis, J., Leibfritz, D., Grimme, L.H., 1990. The dependence of the cytoplasmic pH in aerobic and anaerobic cells of the green algae Chlorella fusca and Chlorella vulgaris on the pH of the medium as determined by 31 P in vivo NMR spectroscopy. Arch. Microbiol. 153, 254–258. Lienert, J., Güdel, K., Escher, B.I., 2007. Screening method for ecotoxicological hazard assessment of 42 pharmaceuticals considering human metabolism and excretory routes. Environ. Sci. Technol. 41, 4471–4478. Neuwoehner, J., Fenner, K., Escher, B.I., 2009. Physiological modes of action of fluoxetine and its human metabolites in algae. Environ. Sci. Technol. 43, 6830–6837. Roche, Tamiflu (oseltamivir phosphate) Capsules and for Oral Suspension. Complete Production Information. F.Hoffmann-La Roche Ltd, Basel, Switzerland, 2006. Sacca, M.L., Accinelli, C., Fick, J., Lindberg, R., Olsen, B., 2009. Environmental fate of the antiviral drug Tamiflu in two aquatic ecosystems. Chemosphere 75, 28–33. Schwarzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2003. Environmental Organic Chemistry, second edition. Wiley, New York, NY, USA. Singer, A.C., Howard, B.M., Johnson, A.C., Knowles, C.J., Jackman, S., Accinelli, C., Caracciolo, A.B., Bernard, I., Bird, S., Boucard, T., Boxall, A., Brian, J.V., Cartmell, E., Chubb, C., Churchley, J., Costigan, S., Crane, M., Dempsey, M.J., Dorrington, B.,

Ellor, B., Fick, J., Holmes, J., Hutchinson, T., Karcher, F., Kelleher, S.L., Marsden, P., Noone, G., Nunn, M.A., Oxford, J., Rachwal, T., Roberts, N., Roberts, M., Sacca, M.L., Sanders, M., Straub, J.O., Terry, A., Thomas, D., Toovey, S., Townsend, R., Vouivoulis, N., Watts, C., 2008. Meeting report: risk assessment of tamiflu use under pandemic conditions. Environ. Health Perspect. 116, 1563–1567. Söederströem, H., Järhult, J.D., Olsen, B., Lindberg, R.H., Hiroaki, T., Fick, J., 2009. Detection of the Antiviral Drug Oseltamivir in Aquatic Environments. PLoS ONE 4, e6064, doi:10.1371/journal.pone.0006064. Straub, J.O., 2009. An environmental risk assessment for oseltamivir (Tamiflu® ) for sewage works and surface waters under seasonal influenza and pandemic use conditions. Ecotoxicol. Environ. Saf. 72, 1625–1634. Ternes, T.A., 1998. Occurence of drugs in German sewage treatment plants and rivers. Chemosphere 32, 3245–3260. Vaes, W.H.J., Urrestarazu-Ramos, E., Hamwick, C., van Holstein, I., Blaauboer, B.J., Seinen, W., Verhaar, H.J.M., Hermens, J.L.M., 1997. Solid phase microextraction as a tool to determine membrane/water partition coefficients and bioavailable concentrations in in-vitro systems. Chem. Res. Toxicol. 10, 1067–1072. Verhaar, H.J.M., van Leeuwen, C.J., Hermens, J.L.M., 1992. Classifying environmental pollutants. 1. Structure–activity relationships for prediction of aquatic toxicity. Chemosphere 25, 471–491.