TD models

Toxicology and Applied Pharmacology 305 (2016) 118–126

Contents lists available at ScienceDirect

Toxicology and Applied Pharmacology journal homepage: www.elsevier.com/locate/ytaap

Analysis of real-time mixture cytotoxicity data following repeated exposure using BK/TD models S. Teng a, C. Tebby a, S. Barcellini-Couget b, G. De Sousa c, C. Brochot a, R. Rahmani c, A.R.R. Pery d,e,⁎ a

Models for Toxicology and Ecotoxicology Unit, INERIS, Parc Technologique Alata, BP 2, 60550 Verneuil-en-Halatte, France ODESIA Neosciences, Sophia Antipolis, 400 route des chappes, 06903 Sophia Antipolis, France INRA, ToxAlim, 400 route des Chappes, BP, 167 06903 Sophia Antipolis, Cedex, France d AgroParisTech, UMR 1402 INRA-AgroParisTech Ecosys, 78850 Thiverval Grignon, France e INRA, UMR 1402 INRA-AgroParisTech Ecosys, 78850 Thiverval Grignon, France b c

a r t i c l e

i n f o

Article history: Received 10 March 2016 Revised 6 June 2016 Accepted 13 June 2016 Available online 15 June 2016 Keywords: TK/TD models Hepatotoxicity Repeated dose toxicity Cell viability Impedance Mixture

a b s t r a c t Cosmetic products generally consist of multiple ingredients. Thus, cosmetic risk assessment has to deal with mixture toxicity on a long-term scale which means it has to be assessed in the context of repeated exposure. Given that animal testing has been banned for cosmetics risk assessment, in vitro assays allowing long-term repeated exposure and adapted for in vitro – in vivo extrapolation need to be developed. However, most in vitro tests only assess short-term effects and consider static endpoints which hinder extrapolation to realistic human exposure scenarios where concentration in target organs is varies over time. Thanks to impedance metrics, real-time cell viability monitoring for repeated exposure has become possible. We recently constructed biokinetic/toxicodynamic models (BK/TD) to analyze such data (Teng et al., 2015) for three hepatotoxic cosmetic ingredients: coumarin, isoeugenol and benzophenone-2. In the present study, we aim to apply these models to analyze the dynamics of mixture impedance data using the concepts of concentration addition and independent action. Metabolic interactions between the mixture components were investigated, characterized and implemented in the models, as they impacted the actual cellular exposure. Indeed, cellular metabolism following mixture exposure induced a quick disappearance of the compounds from the exposure system. We showed that isoeugenol substantially decreased the metabolism of benzophenone-2, reducing the disappearance of this compound and enhancing its in vitro toxicity. Apart from this metabolic interaction, no mixtures showed any interaction, and all binary mixtures were successfully modeled by at least one model based on exposure to the individual compounds. © 2016 Elsevier Inc. All rights reserved.

1. Introduction With the ban of marketing cosmetic products tested on animals in March 2013 by the seventh amendment of the European cosmetic directive, the cosmetic industry has to put additional efforts into developing alternatives to animal testing in order to assess the safety of its products and especially to address the toxicity of mixtures. Cosmetic risk assessment has indeed to deal with mixture toxicity and these products are used repeatedly, on a long-term scale, which means that

Abbreviations: BK, biokinetics; TD, toxicodynamics; IA, independent action; CA, concentration addition; NEC, no effect concentration; BP2, benzophenone-2; NCI, normalized cell index; CHPH, cryopreserved human primary hepatocytes; NOEC, No observed effect concentration. ⁎ Corresponding author at: AgroParisTech, UMR 1402 INRA-AgroParisTech Ecosys, 78850 Thiverval Grignon, France. E-mail address: [email protected] (A.R.R. Pery).

http://dx.doi.org/10.1016/j.taap.2016.06.018 0041-008X/© 2016 Elsevier Inc. All rights reserved.

their toxicity has to be assessed in the context of repeated exposure. There is therefore a need for in vitro assays allowing long-term repeated exposure. With regard to acute toxicity, a general concordance between cytotoxicity assays and acute LD50 in animals has been observed, and it has been validated by ICCVAM (ICCVAM, 2006) and OECD (OECD, 2010) for estimating the starting dose for acute oral systemic toxicity tests. However, reliable chronic in vitro assays that are predictive of in vivo chronic toxicity are still missing. Numerous in vitro models have been generated over the past years for assessing toxicity (Adler et al., 2011) and, more particularly, hepatotoxicity, since the liver is the organ that is most commonly affected by repeated oral exposure to xenobiotics. These in vitro systems, which mechanistically mimic targeted organs to a certain extent, show some drawbacks, nevertheless, such as limited lifespan or instable phenotypes. Moreover, in vitro – in vivo extrapolation would be more relevant if the analysis of the effects could account for varying exposure concentration resulting in

S. Teng et al. / Toxicology and Applied Pharmacology 305 (2016) 118–126

dynamic changes as cosmetics concentration at target organs varied over time as a consequence of discontinuous exposure to cosmetics. The recent development of impedance metrics enables these requirements to be met. Impedance metrics provides a real-time electronic sensing system (RT-CES™) using impedance to monitor cell proliferation, cell spreading, real-time monitoring of cell dynamics and, more importantly, cell viability. We have recently proposed BK/ TD models for analyzing repeated-dose cytotoxicity of HepaRG cells measured by impedance metrics (Teng et al., 2015), shifting from static to dynamic analysis of cytotoxicity tests. Mathematical models were used to describe the kinetics, and the dynamics of three cosmeticsrelated compounds in the biological system and their threshold concentration (No Effect Concentration or NEC) below which no toxic effect occurs have been estimated. In the present study, we propose application of these models to analyze effects produced by mixtures using impedance metrics. The two usual toxicological concepts for analyzing mixture data have been incorporated into our models: i) concentration additivity (Loewe and Muischnek, 1926) and ii) independent action (Bliss, 1939). These are based on the mechanism of action of each component of the mixture. If the components act in a similar way, then the concentration addition (CA) mechanism is suitable for predicting the overall effect of the mixture and the contribution of each component in proportion to its relative toxic potency. If they act independently, i.e., if one component of the mixture does not affect the response of the other, then the independent action (IA) model, also called the response addition model, is suitable for predicting the overall mixture effect (European Commission, 2009; US EPA, 2000). Different approaches have been developed for analyzing mixture response at a given point in time. For instance, graphical methods have been derived to evaluate whether the mixture response is in accordance with concentration additivity (Kortenkamp and Altenburger, 1998; Suhnel, 1992). Another broadly accepted approach for characterizing mixture responses is the response surface model, which has been developed for CA and IA (Goldoni and Johansson, 2007; Vettori et al., 2006). These models are poorly suited to the analysis of real-time impedance data obtained with cells exposed to mixtures, because these data must be analyzed while taking the entire dynamic response into account instead of using multiple analyses at each time point. Here, we propose a new methodology that accounts for the overall dynamics of the cell population in order to analyze the response of cells exposed to binary and ternary mixtures using BK/TD models under both CA and IA hypotheses. We applied the models to the binary and ternary mixtures of the three compounds studied individually in our previous study (Teng et al., 2015): coumarin, isoeugenol and benzpohenone-2 (BP2). To properly analyze the data, interactions were not only considered for effects but also for metabolism. Indeed, we have shown that isoeugenol and benzophenone-2 are likely to disappear as a result of exposure to the medium through cell metabolism. Any interaction pertaining to metabolism may impact the exposure concentration in the in vitro system. These metabolic interactions were implemented in our mixture BK/TD models. Therefore, we are proposing a new methodology that represents a shift from a static classical dose response mixture analysis to a dynamic one that accounts for kinetic interactions when necessary. 2. Materials and methods

119

measurement), the plate was coated with type I collagen. The cells were thawed as preconized by Biopredic in a William's E medium, (Life technologies) containing the Biopredic's thawing supplement. Cells were maintained at 37 °C and 5% CO2 in an incubator. After 24 h, the cells were shifted to the differentiation medium: William's E medium containing 1% DMSO (Sigma®, St Quentin Fallavier, France) and supplemented with 10 MIU/mL of streptomycin, 10 mg/mL of penicillin 10% fetal bovine serum (Life technologies™), 100 IU/mL insulin (Eli Lilly, Indianapolis, USA), and 50 μM hydrocortisone (Sigma®, St Quentin Fallavier, France). The medium was renewed every 2 to 3 days. When the impedance measurements stabilized, after checking for cell adhesion and differentiation, which occurred on around day 7, the HepaRG cells were exposed to the compounds of interest in the same medium with a final concentration of 1% of DMSO. 2.2. Chemicals Coumarin (CAS no.: 91-64-5), isoeugenol (CAS no.: 97-54-1) and benzophenone-2 (2.2′,4,4′-tetrahydroxybenzophenone (BP2), CAS no.: 131-55-5) were purchased from Sigma® (St Quentin Fallavier, France). Chemicals were prepared in DMSO as a 200× stock solution. The compounds were diluted in culture medium containing 0.5% DMSO giving a final concentration 1% DMSO. In addition, controls (with no and high toxicity expected, respectively) were monitored with differentiation medium alone and differentiation medium with 2% sodium dodecyl sulfate (SDS) added. 2.3. Cell exposure HepaRG cells were exposed for 4 weeks. The medium was renewed every two to three days. Based on preliminary experiments, ratios applied for the mixture concentrations were determined based on the individual compounds' NOEC ratios. The concentrations were as follows in the mixture of coumarin + isoeugenol: 4 + 4; 4 + 2; 2 + 1; 1 + 0.5; 0.5 + 0.25; 0.25 + 0.125 mM; coumarin + benzophenone-2: 4 + 2; 2 + 1; 1 + 0.5; 0.5 + 0.25; 0.25 + 0.125; 0.125 + 0.0625 mM; isoeugenol + benzophenone-2: 2 + 2; 1 + 1; 0.5 + 0.5; 0.25 + 0.25; 0.125 + 0.125; 0.0625 + 0.625 mM and coumarin + benzophenone2 + isoeugenol: 2.67 + 1.33 + 1.33; 1.33 + 0.67 + 0.67; 0.67 + 0.33 + 0.33; 0.33 + 0.17 + 0.17; 0.17 + 0.083 + 0.083 mM. Each nominal concentration exposure was performed in duplicate. Four replicates were respectively allocated to the positive and negative controls. 2.4. Impedance measurements Cell responses were evaluated by HepaRG cells impedance which was measured by using the xCELLigence™ system (ACEA Biosciences, Roche® Diagnostics). The system measures electrical impedance across inter-digitated micro-electrodes placed on the bottom of the 96-well cell culture E-plates® RTCA. The impedance measurements were displayed as Cell Index (CI) (Eq. (1)), which provided quantitative biological information about the cell population, cyto-morphological changes and viability over time (Atienza et al., 2005; Ceriotti et al., 2007; Ke et al., 2011; Solly et al., 2004). The CI was calculated using the following equation:
 Rcell ð f i Þ −1 R0 ð f i Þ i¼1; :::;N

CI ¼ max

ð1Þ

2.1. Cell culture Differentiated HepaRG® cells were harvested as described by Teng et al. (2015). In summary, the cells were purchased from Biopredic International (Rennes, France) and prepared from differentiated cryopreserved vial stock (HPR116 – lot No 070). Before seeding with 0.5 × 105 cells per well on an E-plate™ (a 96-well plate designed for impedance

where N is the number of frequency points at which impedance is measured, and Rcell(fi) and R0(fi), respectively, are the frequency-dependent electrode resistance with and without cell at time point t (wells with 50 to 100 μL of PBS) (Solly et al., 2004). The CIs were normalized as described by Nawaz et al. (2014). To summarize, the CI value at time point t was divided by its value at the

120

S. Teng et al. / Toxicology and Applied Pharmacology 305 (2016) 118–126

reference time point, i.e., the last time before exposure of the compound to cells. In this way, the normalized cell index (NCI) value was set to 1 at the beginning of exposure. An NCI of 0 means that no cells are attached to the plates. An increase in the NCI is indicative of proliferation or spreading of the cells, whereas a decrease of the NCI is indicative of the detachment or death of the cells (Atienza et al., 2005; Solly et al., 2004; Xing et al., 2005; Xing et al., 2006). 2.5. Data analysis 2.5.1. BK/TD models for single compounds. We have shown (Teng et al., 2015) that the disappearance of the compounds from the culture medium must be taken into account. For coumarin, the concentration decrease was described by considering physico-chemical unspecific reaction (such as evaporation of the compounds or unspecific binding to well wall or unspecific protein in the medium) written as a firstorder decrease (Eq. (2)), whereas for the two other compounds (isoeugenol and benzophenone-2), the decrease was explained by metabolism by the exposed cells (Eq. (3)).

2.5.2. BK/TD models for mixtures based on the CA concept. The analysis of the mixture data was based on the mechanistic BK/TD models for the single compounds as described in the previous section. According to the CA concept, the mechanisms involved in the overall response are supposed to be similar for each component of the mixture. In our approach, the effect of the mixture is considered to be equivalent to the effect of one reference component at a concentration equal to the sum of all concentrations of the components with a multiplicative factor per component. This factor accounts for the relative toxicity of the component to the toxicity of the reference one. For instance, in the case of a binary mixture constituted by a compound 1 and a compound 2 at concentration C2, the effects of compound 2 in this mixture are equivalent to the effects of compound 1 at concentration x C2, with x the equivalent toxicity factor between compound 2 and compound 1. More generally, the effect of a mixture composed of n components can be expressed as:

E mix

dC coum ¼ −k  C coum dt

¼ E C1 þ

ð3Þ

where k is the unspecific decrease constant, Vmax,i (mmol/min/ 0.65 × 105 cells) the maximum metabolism rate of HepaRG cell of the I compound, KM,i (mM) the Michaelis Menten constant, N the number of cells at the time point, N0 the number of cells at reference time, Ccoum (mM) the concentration of coumarin, Ci (mM) the concentration of the I compound. Eq. (4) has been used to describe the survival of the cells in the wells. The rate of the cell population decrease was assumed to be proportional to the difference between the exposure concentration and the no effect concentration (NEC) and is characterized by b, the killing rate. dN ¼ −b  ðC i −NECÞ  N if C i NNEC dt

ð4Þ

For benzophenone-2, the NCI had not only to account for the cell survival (Eq. (4)) but also for the transient spreading phenomenon (Eq. (5)), which occurred at subtoxic concentration (Solly et al., 2004; Xing et al., 2006) dL ¼ γ  ðL0 þ lim  ðC BP2 −NEC ÞÞ−L dt

ð5Þ

where L is the surface of the spreading cells, L0 the initial surface of the cells, γ the spreading rate, lim the spreading coefficient, and CBP2 the concentration of benzophenone-2. The values of the parameters estimated in Teng et al. (2015) and described above are summarized in Table 1.

ð6Þ

i¼2

ð2Þ

V max;i  C i dC i N ¼−  N0 dt K M;i þ C i

! n X ðxi  C i Þ

where Ci denotes the concentration of the substance i in the mixture constitutive of n compounds and xi the equivalent toxicity factor of the ith substance compared to the reference substance, substance 1. E accounts for the concentration-effect relationship of substance 1. For a given binary mixture, the value of the parameter x was estimated through the simultaneous fitting of the single compound data obtained in Teng et al. (2015) for the two compounds of the binary mixture. All the other parameters were re-estimated at the same time. Note that the estimated parameters NEC and b were related between the two compounds through x as the proportionality factor (see Table 2). As we have shown in our previous work, the disappearance of isoeugenol and BP2 is likely to be due to metabolism by the exposed cells. The metabolisms of these compounds have been characterized by de Sousa et al. (2016) in cryopreserved human primary hepatocytes and this metabolism data was therefore the basis for the study of our mixture's metabolism, as this cell type has all of the cellular organelles, intra-cellular protein and membrane transporters, which makes them suitable for metabolism studies (Hewitt et al., 2007). The metabolic interaction studies showed that the metabolism of benzophenone-2 was inhibited by isoeugenol, but that the metabolism of isoeugenol was, conversely, not affected by the presence of benzophenone-2 (DeSousa et al., accepted in food and chemical toxicology). In short, cryopreserved human primary hepatocytes (CHPH) from three donors were thawed as described by De Sousa et al. (1997) and exposed to 10 μM of benzophenone-2 and to isoeugenol at the following concentrations: 2.5; 5; 10; 30; 60 and 120 μM for 20, 40 and 60 min. The experiments were performed twice with duplicates for each. We accounted for this interaction in the equations for disappearance of the compounds by introducing an apparent constant of inhibition, Ki,app (Eq. (7)), which was estimated based on these metabolism data (see Supplementary Data).

Table 1 Parameters estimated from the chronic BK/TD models of the single compounds studied by Teng et al. (2015).

Coumarin Isoeugenol BP2

b (L/mmol/h)

k (h−1)

kmet (mmol/L/h/0.65 · 105 cells)

NEC (mM)

lim

gamma (h−1)

6.90 · 10−3 [6.50 · 10−3 − 7.4 · 10−3] 4.08 · 10−2 [1.59 · 10−2 − 5.02 · 10−2] 0.162 [0.158–0.167]

5.15 · 10−2 [4.94 · 10−2 − 5.21 · 10−2] –

–

0.147 [0.104–0.150] 0.216 [0–0.292] 0.272 [0.256–0.291]

–

–

1.6*

–

–

7.9*

2.5*

0.1*

0.091 [0.061–0.103]

–

−2

2.51 · 10 [2.16 · 10−2 − 2.82 · 10−2]

95% confidence intervals are in brackets and the fixed parameters are specified with an asterisk.

Km (μM)

34.19*

S. Teng et al. / Toxicology and Applied Pharmacology 305 (2016) 118–126

121

Table 2 BK/TD model equations for all mixtures studied under the CA hypothesis. BK/TD models

Coumarin + isoeugenol

Coumarin + BP2

BK

dC coum ¼ −k  C coum dt V C iso dC iso ¼ − K;max;iso  NN0 dt M;iso þC iso

dC coum ¼ −k  C coum dt V C BP2 dC BP2 ¼ − K;max;BP2 dt M;BP2 þC BP2

TD CA

dN dt

¼ −b  ðC coum þ xCI  C iso −NECÞ  N

TD IA BK/TD models BK

dN dt

¼ ½−bcoum  ðC coum −NEC coum Þ−biso  ðC iso −NEC iso Þ  N

Isoeugenol + BP2 dC iso dt dC BP2 dt

TD CA TD IA

dN dt

¼− ¼

V ;max;iso C iso K M;iso þC iso V

i;app

dC coum ¼ −k  C coum dt V C iso dC iso ¼ − K;max;iso  NN0 dt M;iso þC iso

 NN0

ÞðK M;BP2 þC BP2 Þ



N N0

dC BP2 dt

¼ −b  ðC iso þ xIB  C BP2 −NECÞ  N dL ¼ γ  ðL0 þ ; lim  ðC BP2 − NEC xIB ÞÞ− dt

L dN dt

N

Coumarin + isoeugenol + BP2

;max;BP2 C BP2

C iso

ð1þK

 NN0

dN ¼ −b  ðC coum þ xCB  C BP2 −NECÞ  N dt dL ¼ γ  ðL0 þ ; lim  ðC BP2 − NEC xCB ÞÞ−L dt dN ¼ ½−bcoum  ðC coum −NEC coum Þ−bBP2  ðC BP2 −NEC BP2 Þ dt

¼ ½−biso  ðC iso −NEC iso Þ−bBP2  ðC BP2 −NEC BP2 Þ  N

¼

V ;max;BP2 C BP2 C iso ÞðK M;BP2 þC BP2 Þ i;app

ð1þK

 NN0

dN ¼ −b  ðC coum þ xCIB1  C iso þ xCIB2  C BP2 −NECÞ  N dt dL ¼ γ  ðL0 þ ; lim  ðC BP2 − NEC xCIB ÞÞ−L dt dN ¼ ½−bcoum  ðC coum −NEC coum Þ−biso  ðC iso −NEC iso Þ−bBP2 dt

 ðC BP2 −NEC BP2 Þ  N

Vmax, iso, Vmax,BP2 are the maximum metabolism rate of isoeugenol and BP2; Km,iso, Km,BP2 are the Michaelis' constant for isoeugenol and BP2, b the killing rate of the overall mixture; NEC the no effect concentration, i.e. the threshold concentration of the overall mixture concentration below which there is no significant decrease of cell viability; xCI, xCB, xIB are respectively the x factor of the concentration of isoeugenol relative to the concentration of coumarin, the concentration of benzophenone-2 relative to the concentration of coumarin, the concentration of benopenone-2 relative to the concentration of isoeugenol. For the ternary mixture, xCIB1, xCIB2 are the x factor of the concentration of isoeugenol relative to the concentration of coumarin and the concentration of benzophenone-2 relative to the concentration of coumarin. For the IA TD models, each compound had its own dynamic parameter, i.e., bcoum, biso, bBP2 are the killing rate, and NECcoum, NECiso, NECBP2 are the respective threshold concentrations of coumarin, isoeugenol and benzophenone-2.

While isoeugenol's kinetic equation remained the same, BP2's kinetic equation thus became: V max;BP2  C BP2 dC BP2 N  ¼
   N0 C iso dt 1þ  K M;BP2 þ C BP2 K i;app

ð7Þ

of NCI between two successive time points. We used the BroydenFletcher-Goldfarb-Shanno (BFGS) method implemented in the “Rvmmin” package (Nash, 2014) of the R software. 95% confidence intervals were calculated through bootstrapping with 1000 resampled datasets. 3. Results

where Vmax,BP2 denotes the maximum metabolism rate of BP2 in HepaRG cells, KM,BP2 the Michaelis-Menten constant of BP2, Ki,app the apparent inhibition constant of isoeugenol on BP2 metabolism in human primary hepatocytes. Ki,app which represents the apparent capacity of isoeugenol to inhibit BP2 metabolism is assumed to be similar in CHPH and in HepaRG cells, as both cell types expressed the phase I and phase II enzymes as well as some transporters (Anthérieu et al., 2010, 2012; Hart et al., 2010; Hewitt et al., 2007; Jossé et al., 2008; Kotani et al., 2012; Szabo et al., 2013). 2.6. BK/TD models for mixtures based on the IA concept According to IA definition, each compound exerts its own effect, and each compound's effect contributes independently to the overall response. The dynamics of the number of cells can be calculated by: " # n X dN ¼− bi  ðC i −NEC i Þ  N dt i¼1

with C i −NEC i ¼ 0 if C i bNEC i

ð8Þ

where bi is the killing rate of the ith compound, NECi the No Effect Concentration of the ith compounds and N the number of cells. If BP2 was involved in the mixture, the models also accounted for the cell spreading phenomenon with Eq. (5). Moreover, as with CA, inhibition of benzophenone-2 metabolism by isoeugenol was accounted for in the equations for disappearance. 2.7. Parameter estimations As presented by Teng et al. (2015), the data were homogenized by spacing out the measures every 4 h and the data after the medium renewal were removed. The parameters' values were estimated together using the least squares method applied to fit the variations

We failed to fit simultaneously, with a proportionality ratio x, the single compound impedance data for the mixtures coumarin + isoeugenol and coumarin + BP2. This is consistent with the single compound parameters values, as the ratio between the killing rates b differed substantially from the ratio of the no effect threshold concentrations NEC (for instance, biso/bcoum was more than 8 times higher than NECcoum/NECiso), whereas we would expect to find a common ratio under the CA concept. These mixtures cell viability were predicted well by the IA BK/TD models (Figs. 1 and 2) using the single compound parameters, although regarding the mixture of coumarin and isoeugenol at a concentration of 1 + 0.5 mM for coumarin and isoeugenol, respectively, the model overestimated the toxicity after the fourth renewal (Fig. 1). This lack-of-fit was deemed acceptable, as it has already been observed in HepaRG cells exposed repeatedly to a compound. In the single compound analyses, we have previously observed the same trend toward overprediction after four exposures of isoeugenol, i.e., the model predicted the viability for the first four exposures, then it over-predicted the cytotoxicity. This phenomenon has been attributed to the cellular capacity to compensate for the toxicity induced by an exogenous cellular stressor (Teng et al., 2015). For the coumarin and BP2 mixture, as shown in Fig. 2, the IA model was able to capture the cell viability over time, although a drastic drop in NCI decrease is observed at the concentration of 1 mM and 0.5 mM for coumarin and BP2, respectively. This type of sudden cell viability decrease has already been reported when HepaRG was exposed for the fourth time to 0.5 mM of BP2 alone. When present within the mixture of 1 + 0.5 mM, this phenomenon only happened after the second exposure (Fig. 2). Contrary to the other binary mixture, for the isoeugenol + BP2 mixture, their metabolic interactions and impact on the compound's kinetics have to be taken into consideration. The CA BK/TD mixture model

122

S. Teng et al. / Toxicology and Applied Pharmacology 305 (2016) 118–126

Fig. 1. Predictions by IA BK/TD models of the coumarin + isoeugenol mixture. Lines represent the predictions, and points account for measures.

was able to provide parameters that fit the two single compound datasets simultaneously. They were implemented in the BK/TD model and were not able to predict the overall mixture effect. They overestimated the toxicity both at the concentrations of 0.25 + 0.25 (see Fig. S1 of Supplementary Data) of 0.5 + 0.5 mM (see Fig. S2 of Supplementary Data). In contrast to the CA model, predictions by the IA model considering the inhibition of BP2 metabolism by isoeugenol (detailed analyses of these data being shown in Supplementary Data) were acceptable and are presented in Fig. 3. As regards the ternary mixture impedance data, the drop described for the mixture of coumarin and BP2 at 1 + 0.5 mM (Fig 2) as well as for BP2 alone at 0.5 mM (Teng et al., 2015) also occurred in the ternary mixture at the concentration of 0.67 + 0.33 + 0.33 mM for coumarin, isoeugenol and BP2, respectively. This drop was under-predicted by

the IA model, even considering the inhibition of BP2 metabolism by isoeugenol (Fig. 4). 4. Discussion The dose response of mixture analyses is described by one of the two concepts, CA or IA, and relies on the mechanism of mixture toxicity. Consequently, there is a need to define the criteria that define similar action. As is often the case in human toxicology, “dissimilar action” is considered to be the simple negation of “similar action,” but IA is usually regarded as the default assessment without further evidence that the underlying mechanisms do indeed satisfy any explicit dissimilarity criteria (European Commission, 2009). However, according to US EPA (2000), a case-by-case study selection should be applied based on the knowledge of the mode of action of the mixture constituents.

Fig. 2. Predictions by IA BK/TD models of the mixture coumarin + BP2. Lines represent the predictions, and points account for measures.

S. Teng et al. / Toxicology and Applied Pharmacology 305 (2016) 118–126

123

Fig. 3. Predictions by IA BK/TD models of the mixture isoeugenol + BP2, considering the inhibition of BP2 metabolism by isoeugenol. Lines represent the predictions, and points account for measures.

Regarding our mixture analyses, the detailed biological and mechanistic mechanisms of the observed toxicity have not yet, to our knowledge, been fully described. The assumption of CA or IA could not therefore be assessed in a straightforward manner without modeling the predictions under these hypotheses. Thus, we used BK/TD models under either CA or IA concept to analyze real-time in vitro mixture cytotoxicity data. In the case of CA, either the common fitting of the BK/TD models to the data of the individual compound was not possible (this was the case for all the mixtures except for isoeugenol + BP2) or the model with the estimated parameters was not able to predict the effects of the mixture (as was the case with isoeugenol + BP2). Only the IA BK/TD models were capable of predicting the cell viability of HepaRG cells exposed to the studied mixtures. As for the individual compounds (Teng et al., 2015), metabolism data were necessary either to estimate the required parameters or to

account for metabolic interactions between the mixture components resulting in higher exposure concentrations than in single-compound experiments. The mixture data (isoeugenol + BP2) did not deviate consistently from the CA BK/TD model, we therefore concluded that the components within the mixture acted independently within the mixtures studied. The inhibition data on CHPH were analyzed according to the IC50 shift method (Grimm et al., 2009). We concluded that BP2 metabolism was inhibited reversibly (Supplementary Data) by isoeugenol. This inhibition has been taken into account by employing a Michaelis-Mententype equation in the isoeugenol + BP2 IA model, which was able to describe the data (Fig. 3), whereas if the inhibition had not been considered, the cytotoxicity would have been under-predicted (Fig. 5). With regard to the ternary mixture impedance data, however, accounting for this inhibition was not sufficient for predicting the

Fig. 4. Predictions by IA BK/TD models considering the reversible inhibition of BP2 metabolism by isoeugenol within the ternary mixture. Lines represent the predictions, and points account for measures.

124

S. Teng et al. / Toxicology and Applied Pharmacology 305 (2016) 118–126

Fig. 5. Predictions of NCI of HepaRG cells over time following 4 weeks of exposure to the isoeugenol + BP2 mixture using the IA model without considering inhibitions between the two substances. Lines represent the predictions, and points account for measures.

experimental data. This could be partly due to the difference in the cellular material used in the metabolism experiments (CHPH) and in the impedance measurements (HepaRG cells), which suggests that the parameters values we used from data on primary hepatocytes should be updated for HepaRG cells. The inhibition data available for our mixtures and substances were obtained on CHPH, although inhibition experiments are usually conducted in hepatic microsomes to determine in vitro IC50 and Ki values (Pelkonen et al., 2005). The use of CHPH seemed to be relevant in our study because isoeugenol and BP2 are metabolized by phase II enzymes which, except for UGT, are not present in human liver microsomes (Li, 2001). However, metabolism and transport activity between CHPH and HepaRG cells have been reported to be different (Kotani et al., 2012; Le Vee et al., 2006, 2013; Szabo et al.,

2013), and these two cell types are structurally different. Indeed, HepaRG cells, seeded at high density (this being the case in our experiments), will differentiate into hepatocyte-like cells at a rate of about 55% and into biliary-like cells (Cerec et al., 2007), which make them closer to the hepatic physiological entity. Due to these differences, additional experiments for characterizing the inhibition of BP2 metabolism by isoeugenol in HepaRG cells may be required to improve the analysis of mixture data using HepaRG cells for real-time cytotoxicity assessment. As an illustration, we simulated predictions with IA models for the ternary mixture within which the BP2 concentration was assumed to be constant over time (which would be the consequence of an extremely low Ki value). This latter model provided a good prediction of the real-time cell viability at the tested concentrations of the ternary

Fig. 6. Predictions by IA BK/TD models assuming the concentration of BP2 did not vary over time within the ternary mixture (which means that metabolism would be fully inhibited). Lines represent the predictions, and points account for measures.

S. Teng et al. / Toxicology and Applied Pharmacology 305 (2016) 118–126

mixture and especially at the concentration of 0.67 + 0.33 + 0.33 mM (Fig. 6). Thus, kinetics may explain part of the lack of predictive capacity of ternary data by our model. Toxicodynamics may also be involved. The simultaneous presence of coumarin, isoeugenol and BP2 might enhance their toxicity, since the decrease of cell viability at 0.67 + 0.33 + 0.33 mM of coumarin, isoeugenol and BP2, respectively, was similar to the cytotoxicity observed at 0.5 + 0.5 mM of isoeugenol and BP2. Preliminary analyses of impedance data, showed a drastic NCI drop when BP2 was applied singly (Teng et al., 2015) or within mixture containing BP2 (i.e. coumarin + BP2 (Fig. 2); isoeugenol + BP2 (Fig. 3); coumarin + isoeugenol + BP2 (Fig. 4)) at the intermediate tested concentrations. It could be explained by the sudden decrease of the cellular capacity to detoxify after a few exposures to a mixture containing BP2 (see the Results section) or to this substance alone (Teng et al., 2015). In terms of modeling, this phenomenon could be translated within the BK/TD model by considering that, once the cellular detoxifying capacity is overwhelmed, then the NEC would decrease. Nonetheless, we could not apply this approach because the concentrations we tested did not cover the intermediate concentrations broadly enough in order to estimate the parameters characterizing such a decrease in NEC. Regarding the methodology used in the mixture dose-response analysis, the experimental design is important. Traditionally, the analysis of mixtures of multiple chemicals involves response surface methodology (Greco et al., 1995) with a ray experimental design, because it not only requires less experimental data than a full factorial design but also explores various concentrations, resulting in varying biological responses, including low biological response levels, which enable a threshold response to be estimated. A methodology based on a single fixed-ratio ray design using even fewer experimental data points than a ray design, called the Single Chemical Required (SCR) methodology developed by Casey et al. (2004) and Gennings et al. (2002), showed that a single fixed-ratio mixture experimental design enables the estimation of an additivity or non-additivity response provided the ratio has been chosen in a proportion that is relevant to the biological question. Therefore, in the present study, we used a single ray design, with all of the ratios in the mixtures being fixed based on the NOEC ratio of the single compound constituting the mixture in order to avoid a disproportionate contribution of one of the components. With this design and our models, we were able to assess whether IA or CA were the most likely modes of interaction. In this study, without knowing the specific mode of action for the studied compounds' toxicities, we were able to investigate both CA and IA concepts using BK/TD models to analyze real-time in vitro mixture cytotoxicity data. However, additional knowledge on the compounds' toxicity mechanisms would have made it possible to confirm our results. The BK/TD models proposed in this paper permits to account for the kinetics of the studied compounds and relate the latter to the biological response. One of our kinetic models considered the unspecific phenomena related to physico-chemical properties as a first order equation, and this was relevant for coumarin. For other compounds it may be necessary to describe these reactions in further details, for instance to account for saturable processes related to binding to the plastic or protein, as it has been done by Hamon et al. (2014) to model the effect of cyclosporine A in human renal cells. With such approaches, kinetics data would be necessary to be able to estimate the parameters of the model, which would be difficult only based on effects data. These BK/TD mixture models could also be integrated in in vivo chronic toxicity assessment by performing an extrapolation of the in vitro cytotoxicity to actual in vivo hepatotoxicity. Péry et al. (2013) already showed that in vitro - in vivo extrapolation can be performed to estimate a threshold dose of acetaminophen, which is known to be hepatotoxic, by coupling in vitro BK/TD models to predict cytotoxicity with a PBPK model. These latter models employ a physiologically relevant

125

compartmental approach to describe in vivo absorption distribution metabolism and excretion processes. They can predict the in vivo concentration of the parent as well as the metabolites in the compartment of interest (e.g., the liver). To assess the effects of mixtures of isoeugenol and BP2, the metabolic interaction between these two compounds in the liver should be taken into account. Indeed, BP2 present in a mixture with isoeugenol would be found at higher concentrations and for a longer period of time than if it were present alone. Using PBPK models, we would then be able to integrate the effects of the mixture at the target level but also on the fate of each component in the body. This would increase the relevance of the mixture risk assessment. Conflict of interest statement All authors declare that they have no conflicts of interests with the contents in this article with respect to financial support or financial relationship. Transparency document The Transparency document associated with this article can be found, in online version. Acknowledgments The research leading to the results presented here has received funding from the European Community's Seventh Framework Program (FP7/2007–2013) and from Cosmetics Europe through the COSMOS project under grant agreement n° 266835. Authors are also very grateful to two anonymous reviewers who greatly helped to improve the manuscript. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.taap.2016.06.018. References Adler, S., Basketter, D., Creton, S., Pelkonen, O., van Benthem, J., Zuang, V., Andersen, K.E., Angers-Loustau, A., Aptula, A., Bal-Price, A., et al., 2011. Alternative (non-animal) methods for cosmetics testing: current status and future prospects-2010. Arch. Toxicol. 85, 367–485. Anthérieu, S., Chesné, C., Li, R., Camus, S., Lahoz, A., Picazo, L., Turpeinen, M., Tolonen, A., Uusitalo, J., Guguen-Guillouzo, C., et al., 2010. Stable expression, activity, and inducibility of cytochromes P450 in differentiated HepaRG cells. Drug Metab. Dispos. 38, 516–525. Anthérieu, S., Chesné, C., Li, R., Guguen-Guillouzo, C., Guillouzo, A., 2012. Optimization of the HepaRG cell model for drug metabolism and toxicity studies. Toxicol In Vitro 26, 1278–1285. Atienza, J.M., Zhu, J., Wang, X., Xu, X., Abassi, Y., 2005. Dynamic monitoring of cell adhesion and spreading on microelectronic sensor arrays. J. Biomol. Screen. 10, 795–805. Bliss, C.I., 1939. The toxicity of poisons applied jointly. Ann. Appl. Biol. 26, 585–615. Casey, M., Gennings, C., Carter, W.H., Moser, V.C., Simmons, J.E., 2004. Detecting interaction(s) and assessing the impact of component subsets in a chemical mixture using fixed-ratio mixture ray designs. J. Agric. Biol. Environ. Stat. 9, 339–361. Cerec, V., Glaise, D., Garnier, D., Morosan, S., Turlin, B., Drenou, B., Gripon, P., Kremsdorf, D., Guguen-Guillouzo, C., Corlu, A., 2007. Transdifferentiation of hepatocyte-like cells from the human hepatoma HepaRG cell line through bipotent progenitor. Hepatology 45, 957–967. Ceriotti, L., Ponti, J., Colpo, P., Sabbioni, E., Rossi, F., 2007. Assessment of cytotoxicity by impedance spectroscopy. Biosens. Bioelectron. 22, 3057–3063. De Sousa, G., Fontaine, F., Pralavorio, M., Botta-Fridlund, D., Letreut, Y., Rahmani, R., 1997. Insecticide cytotoxicity and CYP1A1/2 induction in primary human and rat hepatocyte cultures. Toxicol In Vitro 11, 451–457. De Sousa, G., Teng, S., Salle-Siri, R., Pery, A., Rahmani, R., 2016. Prediction of the metabolic clearance of benzophenone-2, and its interaction with isoeugenol and coumarin using cryopreserved human hepatocytes in primary culture. Food Chem. Toxicol. 90, 55–63. European Commission, 2009. State of the Art Report on Mixture Toxicity. Gennings, C., Carter, W.H., Campain, J.A., Bae, D., Yang, R.S.H., 2002. Statistical analysis of interactive cytotoxicity in human epidermal keratinocytes following exposure to a mixture of four metals. J. Agric. Biol. Environ. Stat. 7, 58–73.

126

S. Teng et al. / Toxicology and Applied Pharmacology 305 (2016) 118–126

Goldoni, M., Johansson, C., 2007. A mathematical approach to study combined effects of toxicants in vitro: evaluation of the bliss independence criterion and the Loewe additivity model. Toxicol In Vitro 21, 759–769. Greco, W.R., Bravo, G., Parsons, J.C., 1995. The search for synergy: a critical review from a response surface perspective. Pharmacol. Rev. 47, 331–385. Grimm, S.W., Einolf, H.J., Hall, S.D., He, K., Lim, H.-K., Ling, K.-H.J., Lu, C., Nomeir, A.A., Seibert, E., Skordos, K.W., et al., 2009. The conduct of in vitro studies to address time-dependent inhibition of drug-metabolizing enzymes: a perspective of the pharmaceutical research and manufacturers of America. Drug Metab. Dispos. 37, 1355–1370. Hamon, J., Jennings, P., Bois, F.Y., 2014. Systems biology modeling of OMICs data: effect of cyclosporine a on the Nrf2 pathway in human renal cells. BMC Syst. Biol. 8, 76. Hart, S.N., Li, Y., Nakamoto, K., Subileau, E., Steen, D., Zhong, X., 2010. A comparison of whole genome Gene expression profiles of HepaRG cells and HepG2 cells to primary human hepatocytes and human liver tissues. Drug Metab. Dispos. 38, 988–994. Hewitt, N.J., Lechón, M.J.G., Houston, J.B., Hallifax, D., Brown, H.S., Maurel, P., Kenna, J.G., Gustavsson, L., Lohmann, C., Skonberg, C., et al., 2007. Primary hepatocytes: current understanding of the regulation of metabolic enzymes and transporter proteins, and pharmaceutical practice for the use of hepatocytes in metabolism, enzyme induction, transporter, clearance, and hepatotoxicity studies. Drug Metab. Rev. 39, 159–234. ICCVAM, 2006. In Vitro Cytotoxicity Test Methods for Estimating Starting Doses for Acute Oral Systemic Toxicity Tests (NIH Publ. No 07-4519). Jossé, R., Aninat, C., Glaise, D., Dumont, J., Fessard, V., Morel, F., Poul, J.-M., GuguenGuillouzo, C., Guillouzo, A., 2008. Long-term functional stability of human HepaRG hepatocytes and use for chronic toxicity and genotoxicity studies. Drug Metab. Dispos. 36, 1111–1118. Ke, N., Wang, X., Xu, X., Abassi, Y.A., 2011. The xCELLigence system for real-time and labelfree monitoring of cell viability. Methods Mol. Biol. Clifton NJ 740, 33–43. Kortenkamp, A., Altenburger, R., 1998. Synergisms with mixtures of xenoestrogens: a reevaluation using the method of isoboles. Sci. Total Environ. 221, 59–73. Kotani, N., Maeda, K., Debori, Y., Camus, S., Li, R., Chesne, C., Sugiyama, Y., 2012. Expression and transport function of drug uptake transporters in differentiated HepaRG cells. Mol. Pharm. 9, 3434–3441. Le Vee, M., Jigorel, E., Glaise, D., Gripon, P., Guguen-Guillouzo, C., Fardel, O., 2006. Functional expression of sinusoidal and canalicular hepatic drug transporters in the differentiated human hepatoma HepaRG cell line. Eur. J. Pharm. Sci. 28, 109–117. Le Vee, M., Noel, G., Jouan, E., Stieger, B., Fardel, O., 2013. Polarized expression of drug transporters in differentiated human hepatoma HepaRG cells. Toxicol In Vitro 27, 1979–1986.

Li, A.P., 2001. Screening for human ADME/Tox drug properties in drug discovery. Drug Discov. Today 6, 357–366. Loewe, S., Muischnek, H., 1926. Über Kombinationswirkungen. Naunyn-Schmiedebergs Arch. Für Exp. Pathol. Pharmakol. 114, 313–326. Nash, J.C., 2014. Rvmmin: Variable Metric Nonlinear Function Minimization. Nawaz, A., Razpotnik, A., Rouimi, P., de Sousa, G., Cravedi, J.P., Rahmani, R., 2014. Cellular impact of combinations of endosulfan, atrazine, and chlorpyrifos on human primary hepatocytes and HepaRG cells after short and chronic exposures. Cell Biol. Toxicol. 30, 17–29. OECD, 2010. Guidance Document no 129 on Using Cytotoxicity Tests to Estimate Starting Doses for Acute Oral Systemic Toxicity Tests. Pelkonen, O., Turpeinen, M., Uusitalo, J., Rautio, A., Raunio, H., 2005. Prediction of drug metabolism and interactions on the basis of in vitro investigations. Basic Clin. Pharmacol. Toxicol. 96, 167–175. Péry, A., Brochot, C., Zeman, F.A., Mombelli, E., Desmots, S., Pavan, M., Fioravanzo, E., Zaldívar, J.-M., 2013. Prediction of dose-hepatotoxic response in humans based on toxicokinetic/toxicodynamic modeling with or without in vivo data: a case study with acetaminophen. Toxicol. Lett. 220, 26–34. Solly, K., Wang, X.B., Xu, X., Strulovici, B., Zheng, W., 2004. Application of real-time cell electronic sensing (RT-CES) technology to cell-based assays. Assay Drug Dev. Technol. 2, 363–372. Suhnel, J., 1992. Zero interaction response surfaces, interaction functions and difference response surfaces for combinations of biologically-active agents. Arzneim. Forsch. 42–2, 1251–1258. Szabo, M., Veres, Z., Baranyai, Z., Jakab, F., Jemnitz, K., 2013. Comparison of human hepatoma HepaRG cells with human and rat hepatocytes in uptake transport assays in order to predict a risk of drug induced hepatotoxicity. PLoS One 8, e59432. Teng, S., Barcellini-Couget, S., Beaudouin, R., Brochot, C., Desousa, G., Rahmani, R., Pery, A.R.R., 2015. BK/TD models for analyzing in vitro impedance data on cytotoxicity. Toxicol. Lett. 235, 96–106. US EPA, 2000. Supplementary Guidance for Conducting Health Risk Assessment of Chemical Mixtures. Vettori, M.V., Goldoni, M., Caglieri, A., Poli, D., Folesani, G., Ceccatelli, S., Mutti, A., 2006. Antagonistic effects of methyl-mercury and PCB153 on PC12 cells after a combined and simultaneous exposure. Food Chem. Toxicol. 44, 1505–1512. Xing, J.Z., Zhu, L., Jackson, J.A., Gabos, S., Sun, X.-J., Wang, X.-B., Xu, X., 2005. Dynamic monitoring of cytotoxicity on microelectronic sensors. Chem. Res. Toxicol. 18, 154–161. Xing, J.Z., Zhu, L., Gabos, S., Xie, L., 2006. Microelectronic cell sensor assay for detection of cytotoxicity and prediction of acute toxicity. Toxicol. In Vitro 20, 995–1004.