From in vitro to in vivo: Integration of the virtual cell based assay with physiologically based kinetic modelling

Accepted Manuscript From in vitro to in vivo: Integration of the virtual cell based assay with physiologically based kinetic modelling

Alicia Paini, Jose Vicente Sala Benito, Jos Bessems, Andrew Worth PII: DOI: Reference:

S0887-2333(17)30164-9 doi: 10.1016/j.tiv.2017.06.015 TIV 4034

To appear in:

Toxicology in Vitro

Received date: Revised date: Accepted date:

29 October 2015 26 February 2017 16 June 2017

Please cite this article as: Alicia Paini, Jose Vicente Sala Benito, Jos Bessems, Andrew Worth , From in vitro to in vivo: Integration of the virtual cell based assay with physiologically based kinetic modelling, Toxicology in Vitro (2017), doi: 10.1016/ j.tiv.2017.06.015

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Title: From in vitro to in vivo: Integration of the Virtual Cell Based Assay with Physiologically Based Kinetic modelling

Authors: Alicia Paini a*, Jose Vicente Sala Benito a, Jos Bessems a, Andrew Worth a

Affiliation:

Chemical Safety and Alternative Methods Unit, EURL ECVAM, Directorate F – Health, Consumers and

PT

a

NU

* Corresponding author: [email protected]

SC

RI

Reference Materials, Joint Research Centre, European Commission, Ispra, Italy

Joint Research Centre

MA

Directorate F – Health, Consumers and Reference Materials Chemical Safety and Alternative Methods Unit

D

Via E. Fermi 2749, TP 126

tel.+39-0332-78 3986

AC

CE

fax +39-0332-78 9963

PT E

I-21027 Ispra (VA), Italy

1

ACCEPTED MANUSCRIPT

AC

CE

PT E

D

MA

NU

SC

RI

PT

Graphical Abstract

2

ACCEPTED MANUSCRIPT Abstract [Max 200 words]

Physiologically Based Kinetic (PBK) models and the Virtual Cell Based Assay can be linked to form so called Physiologically Based Dynamic (PBD) models. This study illustrates the development and application of a PBK model for prediction of estragole-induced DNA adduct formation and hepatotoxicity in humans. To address the hepatotoxicity, HepaRG cells were used as a surrogate for

PT

liver cells, with cell viability being used as the in vitro toxicological endpoint. Information on DNA

RI

adduct formation was taken from the literature. Since estragole induced cell damage is not directly caused by the parent compound, but by a reactive metabolite, information on the metabolic pathway

SC

was incorporated into the model. In addition, a user-friendly tool was developed by implementing the PBK/D model into a KNIME workflow. This workflow can be used to perform in vitro to in vivo

MA

NU

extrapolation and forward as backward dosimetry in support of chemical risk assessment.

Highlights

D

1. VCBA is coupled with PBK model linking intracellular concentration to external dose.

PT E

2. The integrated model is applied to perform IVIVE of estragole-induced cell damage.

CE

3. The IVIVE process is automated by means of a KNIME workflow.

AC

Keywords [max 6 Kw]

VCBA, PBK, PBD, Estragole, DNA adduct, Risk Assessment, KNIME workflow

3

ACCEPTED MANUSCRIPT Introduction The use of alternatives to animal testing is becoming increasingly important in the safety assessment of chemicals. Various cellular models have been developed, such as human cell lines (e.g. HepG2, HepaRG), freshly isolated hepatocytes, co-cultures, 3D models, and tissue/organ on chip models (Soldatow et al., 2013; Ramanan et al., 2014). In order to use

PT

human in vitro concentration-response data for human risk assessment, a stronger focus on internal exposure is necessary. In such an approach, in vitro concentrations (points of

RI

departure) would be compared to simulated or measured (e.g. by biomonitoring) human in

SC

vivo systemic exposure concentrations. The resulting margin of internal exposure (MOIE) would then be used to characterise the risk (Bessems et al., accepted; Bessems et al.,

NU

2015).

MA

Computational approaches that aid the risk characterization are: i) Quantitative structure– activity relationship (QSAR) models, which relate the biological activity of a compound to its molecular structure; ii) Physiologically based kinetic and dynamic (PBK/D) modelling, which

D

can be used to simulate the biokinetics and dynamics of chemicals in model organisms or

PT E

humans; and iii) other biokinetic mathematical models, such as the virtual cell based assay (VCBA), to simulate fate of chemical taking into account the setup of in vitro experiments,

CE

leading to intracellular simulation of the chemical under study. Despite considerable progress in the development of these in silico approaches, there is still

AC

a need to develop practical tools for risk assessors, who increasingly need to rely on in vitro to in vivo extrapolations of chemical toxicity in order to carry out animal-free safety assessments. Recent work has illustrated several approaches to perform Quantitative In vitro to In vivo Extrapolation (QIVIVE) (Broeders, 2014). Uncertainties in the extrapolation from in vitro concentrations to in vivo human exposure conditions, including the influence of differences between species and exposure routes, can be taken into account by an integrated approach. Such an approach is to link the two models - the VCBA with the PBK model. Briefly, the VCBA model describes and predicts what is happening in an in vitro 4

ACCEPTED MANUSCRIPT system, especially the fate of a chemical within the well, taking into account partitioning with protein, lipids, and plastic binding (Zaldivar et al., 2010, 2011, 2012, 2015). The model also features a growth model, which takes into account the cell growth phases (G1, S, G2, M phases). Other features taken into account are the partitioning of a compound within the cell and the toxicity to the cell. The latter part of the model is based on the prediction of two

PT

parameters, the no-effect concentration (NEC) and the killing rate (kr), which are linked to experimental cell viability (Zaldivar et al., 2010, 2011, 2012, 2015). The main property

RI

simulated is the real, intracellular, concentration of a specific chemical affecting the cell.

SC

Examples of simulation for different chemicals and specific cell lines can be found in (Kramer et al., 2010; Broeders et al., 2015; Armitage et al. 2014).

NU

On the other end we have the PBK/D model, which combines kinetic and dynamic

MA

interactions of a specific compound after oral, inhalation, or dermal exposure by applying a set of differential equations. To be more specific, PBK models describe the body as a set of interconnected compartments, which represent the human organs and blood circulation,

D

describing the absorption, distribution, metabolism, and excretion (ADME) (Krishnan &

PT E

Andersen, 2001). PBD models additionally include a description of the interaction of the chemical or its reactive metabolite with the receptor mediating the adverse effect.

CE

In order to illustrate the integration of the VCBA outputs with a PBK model we chose the alkenylbenzene class of chemicals, including methyleugenol, safrole, and estragole, for

AC

which several PBK/D models have been built for both rodents and humans (Punt et al, 2008; Punt et al., 2009; Paini et al., 2010, Martati et al., 2011; Al-Subeihi, et al., 2011, Punt et al., 2016). Estragole, which was selected as the gold compound for this exercise, is found in herbs and spices, to which humans are exposed at low doses via the diet. Once absorbed estragole can undergo detoxification and bioactivation through phase I and II enzymatic pathways resulting in the compound being either excreted or converted to a reactive carbocation which

5

ACCEPTED MANUSCRIPT binds covalently to DNA. Estragole is known to produce tumours in rodents exposed to high dose levels (Miller et al., 1983) and it has been shown to be genotoxic and carcinogenic when tested in isolated form at high doses in animal experiments. The main metabolic pathway for sstragole is represented in figure 1. The main pathway for bioactivation of estragole proceeds by initial hydroxylation on the allyl side chain by cytochrome P450

PT

enzymes resulting in the formation of 1'-hydroxyestragole, the proximate carcinogenic metabolite. 1'-Hydroxyestragole can be detoxified by glucuronidation or via oxidation

RI

(Phillips et al., 1981, Lyer et al., 2003), the latter followed by glutathione conjugation.

SC

Alternatively, sulfonation of 1'-hydroxy estragole by sulfotransferases gives rise to the unstable metabolite 1'-sulfooxyestragole which decomposes to the reactive carbocation that

NU

covalently binds to DNA, RNA and protein (Phillips et al., 1981, Gardner et al., 1996). Several adducts are formed during the reaction of 1'-sulfooxyestragole with the guanine

MA

base in DNA including N 2-(trans-isoestragol-3'-yl)-2'-deoxyguanosine, N 2-(estragole-1'-yl)2'-deoxyguanosine, 7-(trans-isoestragol-3'-yl)-2'-guanine and 8-(trans-isoestragol-3'-yl)-2'-

D

deoxyguanosine (Phillips et al., 1981, Punt et al., 2007). The major guanine adduct formed is

PT E

N 2-(trans-isoestragol-3'-yl)-2'-deoxyguanosine (E-3'-N 2-dGuo) which is considered to play a role in the genotoxic and carcinogenic effects induced by estragole (Smith et al., 2002,

AC

CE

Phillips et al., 1981).

6

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

PT E

D

Figure 1 Estragole metabolic pathway in rodents and humans.

In the present work, we illustrate how by linking the VCBA to the PBK model developed for

CE

estragole (Punt et al., 2009; 2016) we can achieve a new refined PBK/D model that can be used for in vitro to in vivo extrapolation of endpoints, cell viability and DNA adducts; a

AC

schematic representation of the approach is given in figure 2. The PBK/D model was implemented in KNIME to achieve an open source automated ready to use tool for extrapolation.

7

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

MA

Figure 2. Schematic overview of the link between the VCBA and a PBK model and the possibility to extrapolate from an in vitro concentration to an in vivo external dose, and from an external dose to

D

internal response. Light grey arrow represents forward extrapolation (external dose to internal effect),

AC

CE

PT E

black arrow reverse extrapolation (internal effect to external dose).

8

ACCEPTED MANUSCRIPT Materials and methods Estragole (Cas. 140-67-0) and DMSO were purchased from Sigma. TMRE, Toto 3, and Hoechst 33342 for cellomic analysis were purchased from Invitrogen. Cryo preserved Human Cell Line HepaRG was obtained from INSERM’s laboratory U522 and a cell culture bank had been established in-house at the Joint Research Centre. The concentration range

PT

of estragole was 0-75 mM starting from a 100mM stock solution in cell exposure medium. Viability and mitochondrial membrane potential in HepaRG

RI

For the purpose of the current study, an in-house experiment using HepaRG exposed to

SC

different concentrations of the selected chemicals for 24hrs was carried out. Briefly, HepaRG cells were seeded at a density of 2.6 × 104 cells/cm2 in a growth medium composed of

NU

Williams E medium supplemented with 10% FCS, 100 units/mL penicillin, 100 μg/mL

MA

streptomycin, 5 μg/mL insulin, 2 mM glutamine and 5 × 10−5 M hydrocortisone hemisuccinate. Further culturing was carried out for 2 more weeks with the same medium supplemented with 2% DMSO in 75 cm2 culture flask. The medium was renewed every 2 to

D

3 days. After differentiation, HepaRG cells were detached by gentle trypsinization, and then

PT E

seeded at a density of 4-5 × 104/well in 96 well microtiter plates (BD Biosciences) to allow the selection of hepatocyte-like populations. The cells were used for testing within one week

CE

after plate seeding. Compounds were diluted in culture medium with 5% HyClone Fetalclone III serum to obtain a final concentration of DMSO of 0.1%, 100 µL of each concentration for

AC

each compound were used. After 24h of exposure, treated HepaRG cells were stained with TMRE, Toto 3, and Hoechst 33342 for 30 minutes. This assay was performed three times in triplicate (biological & technical replicates). Viability and mitochondrial membrane potential (mmp) were assessed with a high-content analysis (HCA) approach using Cellomics™ ArrayScan vTi (Thermo Scientific, Pittsburgh, PA, USA). A magnification of 10 image field per well for two fluorescence channels with the XF93 filter set were collected. Cell count analysis was performed using the Target Activation

9

ACCEPTED MANUSCRIPT Bioapplication v.4 from Cellomics™ Scan Software (Mennecozzi et al., 2011) on 10 fields taking into account around 0.8-1 × 104cells. PBK model extension

To study the bioactivation and detoxification of estragole a human PBK model was built by Punt et al., (2009). We used this published human PBK model as the basis of our work. The

PT

PBK model calculates the internal concentration of estragole at organ level (CL) in the human body following a specified exposure scenario. We included the Virtual Cell (VC)

RI

model, which is the cellular part of the VCBA, into the PBK model of Punt et al., (2009) to

SC

simulate the intracellular concentration.

NU

The Virtual Cell Model

In the VC, the cell is modelled as four compartments: lipid, protein, water and mitochondria

MA

(Worth et al., in present issue) where the uptake of the chemical across the cell membrane occurs by passive diffusion into the aqueous compartment followed by distribution into the

D

three other compartments of the cell by instantaneous partitioning. When the chemical

PT E

enters into the cell a toxicodynamic process occurs which is governed by two parameters: No Effect Concentration (NEC), killing rate (kr). These parameters are calculated prior to the PBK-VCBA simulation by optimization of predicted cell viability by the VCBA with the

CE

experimental (HTS) in vitro data. In this way the model can predict the cell viability in the liver after a given dose. The parameters of the heptatocyte (HepaRG) VC are given in Table

AC

1. A cell growth model was also derived, however it was set to initial G1 phase, since the HepaRG cell line does not proliferate. Furthermore, the in vitro fate part of the VCBA needs some input parameters (LogP, air and water degradation, Molar Volume, MW, Henry constant, and atomic diffusion) which were computed by EPI Suite (version 4.1), however experimental values were preferred wherever available.

10

ACCEPTED MANUSCRIPT Table 1. Parameters of the HepaRG virtual cell model included in the PBK model. Parameter type

Abbreviation used in the

Value

Units

Ref.

model

Uptake rate

faq,

0.72

fL,

0.012

fP

0.268

rup

% weight

35.208

Elimination rate

rel

35.208

Wet weight

W

1.79 10

Cell cycle* duration

G1 M

Phase I & II metabolism:

(from Punt et al., 2009)

1.4

4’ Allylephenol estragole-2′,3′-oxide

1’-hydroxyestragole

1’-oxoestragole

CE

glucuronidation

PT E

_________________

1’-sulfooxyestragole

(Zaldívar et al., 2015)

3

H -1

h

21

0.7

290

1.6

83

2.6

350

0.6

708

9.4

354

0.06

727

D

3’-hydroxyanethole

-6

-1

Km

MA

1’-Hydroxyestragole

m

1.19 10

NU

Mortality

Vmax

Gr

-15

849 .

-2

-9

1.67 10

-1

Lm h

RI

V

.

SC

Volume of the cell

.

-2

Lm h

PT

Mass fraction of compartment fx (aqaqueous, L-lipids, Pproteins)

AC

*HepaRG cell do not proliferate the models keeps only one cell cycle step G1. **Scaled Vmax (app) expressed as μmol h−1 (g tissue)−1, calculated from the in vitro Vmax based on a microsomal protein yield of 32 mg/g tissue for liver and an S9 protein yield of 143 mg/g liver (Punt et al., 2009). ***Km in µM.

The concentration of estragole reaching the liver will enter the cell. By simulation the virtual cell model takes into account the total number of moles of a compound (ntot) in the cell as the sum in the different compartments, water, protein, lipid, and mitochondria. The model representing the cell was added to the PBK model, and is described by the following equations:

11

ACCEPTED MANUSCRIPT ntot  naq  nPC  nLC  nmit  Vaq .Caq  VPSC .C PC  VLC .C LC  Vmit Cmit 

1

where the Vi's refer to the compartment volumes (l) and the Ci's refer to the concentration in the -1

compartments (mol.l ), water, protein, lipid and mitochondria. Also the total number of moles of a chemical can be expressed as:

ntot  W .Cb MW

PT

2 -1

where W is the cell weight (g), MW is the molecular weight of the chemical (g.mol ) and Cb is the -1

RI

chemical concentration in the cell (g.gw.w ).

K PC 

C PC /PC  C LC /LC  ; K LC  ; C aq Caq

NU

partition coefficients:

SC

The chemical is assumed to be in equilibrium between the different compartments with fixed values

MA

The time evolution of this substance in the cell can be calculated by a simple mass balance, -2

-1

assuming that the uptake and elimination rates rup and rel (l.cm .s ) are proportional to the surface

PT E

dntot  V 2 3 rup .C dis  rel .C aq  dt

D

area of the cell (passive diffusion) and the transfer occurs through the aqueous compartment only as:

3

-1

where Cdis and Caq refer to the chemical concentration (mol.l ) outside of the cell and in the aqueous -1

dW   dCb  Cb W  dt dt  

4

AC

dntot 1  dt MW

CE

compartment of the cell (mol.l ), respectively. Appling the chain rule of differentiation to eq 2 we have:

and rearranging terms we obtain:

dCb MW .V 2 3 rup .Cdis  rel .Caq   Cb dW  dt W W dt

5

the last term represents the dilution due to growth of the cell that in the case of non-proliferating (Fe=0) cells this can be neglected.

12

ACCEPTED MANUSCRIPT Since the concentration in the aqueous fraction C aq is not a value that is measured, then we have to convert in terms of Cb using the partitioning approach. The wet weight, W can also be expressed as a function of the volumes of the different compartments:

W  .V   Vaq  VPC  VLC  Vmit 

6

On the other hand:

PT

VPC  WPC  PC

RI

VLC  WLC  LC

10

NU

Vaq  Waq  aq

8 9

SC

Vmit  Wmit  mit

7

where W PC, W LC, W mit, and W aq are the masses of proteins, lipids and aqueous compartments in the

MA

cells and ρP, ρL, ρmit and ρaq their densities.

To find the relation between Caq and Cb we have to combine ntot in eqs 1-2, the partition coefficients

Cb

 f aq f f f MW .  LC K LC LC   PC K PC PC   mit   PC  mit  aq  LC

PT E

C aq 

D

and eqs 7-10, then we have:

11

   

CE

where fi refer to the mass fraction of each compartment (aqueous, lipid, protein) in the cell. Replacing

AC

this equation into eq.5 and rearranging we obtain:

dC b  MW .V 2 3   rup C dis dt  W 

   V23    f aq f f f  LC K LC LC   PC K PC PC   mit  W .  PC  mit    aq  LC

    1 dW  rup C b   C b    W dt        12

However in this case uptake and elimination rates are not constant, but depend on the status of the cell and take into account the differences in growth. The variation of the wet weight, W, as a function of time can be obtained, assuming constant composition and hence density, as:

13

ACCEPTED MANUSCRIPT dW dV  dt dt

13

4

If we consider spherical shape 𝐴 = 3 𝜋𝑟 3 and von Bertalanffy’s growth curve (Zaldivar et al., 2011), 𝑟(𝑡) = 𝑟∞ − (𝑟∞ − 𝑟0 )𝑒 −𝛼𝐺 𝑡

13

PT

where 𝑟0 and 𝑟∞ refer to the initial and final cell radius and 𝛼𝐺 is the von Bertalanffy’s growth

RI

rate. Then we have:

14

SC

dV 2  4   r  (r  r0 ) exp(  G  t ) (r  r0 )   G  exp(  G  t ) dt

NU

However, the introduction of this term considers only a single cell developing during the simulation. To consider the whole population of the cells, instead of a single cell, we take

MA

average values for the weight, its derivative, and the surface depending on the four stages: G1, S, G2, M. The model in Eq. 12 has several parameters that need to be evaluated. The

D

uptake and elimination rates, rup and rel, and the partition coefficients, KL and KP, depend on

PT E

the compound; whereas the remaining parameters depend on the type of cell. The metabolism occurs inside the cell and, therefore, this process is included in the chemical

Fmet  k met  Cb

CE

mass balance inside the cell. In any case, we can assume the same principle and write: 15

AC

where Cb refers to the cell internal concentration (g gww-1) and kmet is the metabolism rate constant (s-1). However, a better expression is provided using the Michaelis-Menten equation:

Fmet 

Vmax  Cb K M  Cb

16

where Vmax (s-1) represents the maximum rate achieved by the system and KM is the substrate concentration at which the reaction rate is half of Vmax. Equation 16 was used to

14

ACCEPTED MANUSCRIPT describe phase I metabolism of estragole, the values of Vmax and Km for each path of estragole metabolism were taken from Punt et al. (2009), and values are reported in table 1. The complete R script, linking PBK with the VCBA can be found in the Appendix.

PT

Optimization error

The optimization error was calculated using the following equation by the VCBA:

RI

𝑛 𝑒𝑥𝑝 𝑛 𝑒𝑥𝑝 𝑒𝑟𝑟𝑜𝑟 = ∑𝑖=1 (𝑣𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑒𝑥𝑝 − 𝑣𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑠𝑖𝑚 )2 = ∑𝑖=1 (𝑣𝑖𝑎𝑏𝑖𝑙𝑖𝑡𝑦𝑒𝑥𝑝 − (𝑛𝑐𝑡𝑜𝑡,𝑡 /𝑛𝑐𝑡𝑜𝑡,𝑡=0 )𝑠𝑖𝑚 )2

SC

17

With viability defined as the % of cells surviving at the end of the experiment. This step is

NU

performed first to obtain the NEC and Kr to be used in the joint PBK/D –VCBA. For more

PBD model – DNA adduct formation

MA

information on the optimization step, see Zaldivar et al. (present issue).

D

To calculate the DNA adducts formed from the estragole intracellular concentration, a linear

PT E

equation passing through zero was used: Y = A*X

18

CE

Y is the number of DNA adducts (#Adducts) formed and X is the concentration of 1’-

AC

sulfoxyestragole (nmol/g liver), A is a factor described in Paini et al (2012b) which allows to extrapolate from the levels of the sulfoxy derivative to the number of DNA adducts (Paini et al., 2012a), and AMHES is the concentration of 1’-sulfoxyestragole (nmol/g liver). So we can introduce the following equation for DNA adduct formation: #Adducts = 32.6*(AMHES) (Punt et al., 2016) KNIME workflow Development of a KNIME workflow automates the modelling thereby facilitating the application in the risk assessment of chemicals. KNIME is a user-friendly graphical

15

ACCEPTED MANUSCRIPT workbench for the analysis process such as data access, data transformation, investigation, visualization and reporting (http://www.knime.org/). R is a language and environment for statistical computing and graphics (http://www.r-project.org/). The PBK model from Punt et al (2009) was converted from Berkeley Madonna (www.berkeleymadonna.com) to R language in order to be compatible with the KNIME R snippet node within the workflow. The PBD

PT

model was implemented as an open source platform using KNIME and R programs (which are all freely available). The workflow is divided into an input zone (where the end user will

RI

input the values for each descriptor/parameter), a core zone (where the PBD model is

SC

stored) and output zone, where graphical representation and table to extrapolate from the simulate curve can be found (figure 3). This workflow is available from the COSMOS Space

AC

CE

PT E

D

MA

NU

website (http://cosmosspace.cosmostox.eu/).

16

CE

PT E

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC

Figure 3. KNIME workflow to perform IVIVE, for human exposure, based on the estragole PBD model.

17

ACCEPTED MANUSCRIPT Results To run the VCBA, to obtain the NEC, and kr, the EPI suite software (version 4.1) was used to achieve the parameters needed, (Table 2). Furthermore, in vitro testing to optimize the model was performed. Table 2. Physicochemical parameters (EPI Suite predictions) for estragole, used as input to the

Molar Volume

Atomic Diffusion

140-67-0

3.47

136.9925

93.56

MW (g/mol)

Water Degradation

Air Degradation

Henry constant

148.2

2.14E-07

4.91E-05

4.68E+01

NEC

Kr

0.00533

0.0469

SC

RI

Log Kow

MA

Estragole

Cas #

NU

Chem. Name

PT

VCBA.

Figure 4A and 4B show the estragole concentration response for cell viability and mmp

D

tested in HepaRG. The results are the average of five biological replicates, for each

PT E

biological replicate only one measurement was possible (so 1 technical replicate). Cell viability and mitochondrial membrane potential (mmp) were measured by means of Cellomics™. After exposure of 24 hrs to estragole a rapid decrease was registered for both

AC

CE

readouts. Cell viability starts to drop at 12.5 mM and mmp at 25 mM.

18

ACCEPTED MANUSCRIPT 140 120

% Viability

100 80 60 40 20 0 0

0.10 0.20 0.39 0.78 1.56 3.13 6.25 12.5

25

50

75

PT

-20

SC

RI

Estragole mM

120

NU

100

MA

% mmp

80 60 40

PT E

D

20 0 0

0.10 0.20 0.39 0.78 1.56 3.13 6.25 12.5

25

50

75

CE

Estragole mM

Figure 4. Viability (4A) and mmp (4B) in percentage results after 24 hr exposure to different

AC

concentrations of estragole (mM) in HepaRG, (Y-axis I normalized to control at 0 mM, for Cell viability is represented by % of cell death and for mmp by ΔΨ).

The Cell viability experimental results were used to optimize the NEC and kr resulting in the following values 0.00533, and 0.0469, respectively (see also Table 2) The KNIME workflow allows the user to process the simulation in a fast and automated way. Thus, the parameters (listed in table 2, plus NEC and kr) obtained were inputted in the

19

ACCEPTED MANUSCRIPT KNIME workflow developed for estragole, that can be divided into three parts (left, central, right). On the left side the nodes, are accessible by the user to input the dose range (e.g. 01000 mg/kg bw) that they wish to apply and the simulation time (e.g. 0-24hrs). The central part (the core) represents the R script, the mathematical representation of the model. Finally, the right side of the workflow is built with nodes that elaborates the model outputs into a

PT

graphical representation (figure 5A and B). Within the PBK/D model KNIME workflow (figure 3) an extension was made to be able to translate the in vitro concentration response curve to

RI

an external dose response curve and by extrapolation the actual quantified effect (in vitro) to

SC

an external dose (in vivo). This is called in vitro to in vivo extrapolation (IVIVE), or reverse dosimetry. The model can be also applied in forward dosimetry going from the selected

NU

exposure dose to the internal effect. For instance for the average dietary human daily intake for estragole (0.07 mg/kg bw, EU-SCF 2001) a viability value of 1 was found, meaning that

MA

this dose would not harm the liver cells, assuming HepaRG cells are representative of the human liver, and adducts formed are 0.00084/108 nt (nucleotide, nt). However if we increase

D

the dose to 100 mg/kg bw, a reduction in fraction of viable cells to 0.7 is observed, indicating

PT E

that estragole at this dose is inducing liver cell damage, with the number of adducts increasing to 0.133/108nt.

CE

Due to the fact that estragole is known not to induce cell damage directly (only at high concentration and/or doses) we studied the possibility to extend the human PBK/D model to

AC

simulate the metabolites of estragole, by taking the PBD model from Paini et al., 2012b and revised in Punt et al., (2016) and by correcting the simulation using the intracellular concentration. This was done as follows: 1. Estragole enters the liver and reaches the hepatocytes (concentration in the liver, CL). 2. Estragole enters the cells by passive diffusion and partitions between the cellular compartments. The model simulates the intracellular concentration, within the cell model we included the equation describing metabolism for phase I and II.

20

ACCEPTED MANUSCRIPT 3. Then the starting concentration for metabolism of estragole becomes the intracellular concentration obtained with the VCBA, since metabolism is occurring in the cells. We assumed that HepaRG is a good in vitro surrogate of hepatocytes. 4. By introducing in the PBK/D model a simple linear formula, the number of DNA adducts was estimated based on the relationship between sulfonation in humans (Punt et al., 2016)

PT

and DNA adduct formation measured in vivo in rodents (Paini et al., 2012a). The DNA adduct simulation was previously done in rodents and the PBD model was validated in vivo

RI

(Paini et al., 2010, 2012b). The results showed that after exposure to estragole for 24 hrs a

SC

rapid decrease was simulated for cell viability (Figure 5A). The model was also able to

AC

CE

PT E

D

MA

NU

simulate DNA adduct formation, showing an increase with exposure dose (Figure 5B).

A

21

B

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

PT E

Figure 5. Estragole dose response predictions using the automated human estragole PBK/D model to perform IVIVE (KNIME workflow), readouts for A. Cell viability (viability is normalized to 1, with no

AC

nucleotide).

CE

concentration of estragole 100% of the cell as viable) and B. DNA adducts (Y-axis in adducts /10^8

22

ACCEPTED MANUSCRIPT Discussion Estragole is an alkenylbenzene that can be found in spices and herbs, and due to its aniseed it was once used as a flavouring agent in confectionary. Although it is prohibited to add estragole to food products as such, due to its genotoxic properties in rodent studies (Miller et al., 1983), it is still found in the daily human diet due to its presence as a natural

PT

constituent of foods (e.g.in pesto sauce, since it is the main constituent of basil leaves). However, it can also be added to cosmetics within safe limits. Thus there is a need to

RI

develop tools to address the safety of this chemical (group). To carry out in vitro to in vivo

SC

extrapolation (IVIVE), a physiologically based kinetic/dynamic (PBK/D) model previously developed for estragole (Punt et al., 2009; 2016; Paini et al., 2010, 2012a), was selected as

NU

a case stud for integration with the Virtual cell based assay (VCBA). The joint PBK/D-VCBA model describes the relationship between the tissue dose, early chemical-tissue interactions,

MA

and resulting toxic effect(s); thus it can be used to predict the toxicologically effective target organ dose.

D

The aim of the present study was to refine the PBK/D model to be able to i) include the

PT E

mathematical description of the virtual cell in the PBK model to represent the link between in vitro and in vivo exposures; ii) link an adverse outcome (such as cytotoxicity) and; iii)

cell.

CE

simulate the number of DNA adducts based on the simulated concentration reaching the

AC

This combination of models was previously used by Gajewska et al. (2015). In the latter study, a multi-scale computational model was built linking a PBK model for caffeine with the Virtual Cell-Based Assay to relate an external oral and dermal dose with the measured in vitro HepaRG cell viability. The model simulation showed that cell viability remained almost unchanged for external human doses of caffeine (5–400  mg) by both oral and dermal absorption routes, in single exposure mode only. In the present work estragole was chosen as a well-studied compound, since in vitro data for cell viability was obtained in house using the HepaRG cell line, while DNA adducts in 23

ACCEPTED MANUSCRIPT humans were taken from the literature (Paini et al., 2012b; Punt et al., 2016). We have used the HepaRG cell line, since it was known to be a good candidate for these studies to mimic real hepatocytes (Zanelli et al., 2012). HepaRG cells are terminally differentiated hepatic cells derived from a human hepatic progenitor cell line that retain many characteristics of primary human hepatocytes, including metabolic competence. The novel part was to extend

PT

the human PBK/D model from Paini et al., 2012b (revised by Punt et al., 2016) and correct it for the intracellular concentration. The results showed that after exposure of 24 hrs to

RI

estragole a rapid decrease was simulated for cell viability. The concentration at which this

SC

occurred (12.5 mM) can be regarded as the point of departure for extrapolation. At the same time the concentration of estragole that entered the cell was lower, thus, the DNA adduct

NU

formation was also lower compared to what was reported by Punt et al., (2016); as an example, for an exposure dose of 0.01 mg/kg the simulation resulted in 0.001

MA

adducts/108nts versus 1.2 adducts/108nts reported in Punt et al.,, showing a 2 fold lower prediction as compared to what was previously reported this could be due to the correction

D

of the intracellular concentration.

PT E

For the specific case of estragole, we could investigate further the liver zonation in human as done previously by Paini et al (2012a) in rodents, which is especially interesting for the DNA

CE

adduct formation. Another important piece of information given by the simulation was the amount of estragole reaching the mitochondria. So far we were able only to link the

AC

experimental output to the simulated concentration (data not shown), and not yet simulate the mitochondrial membrane potential, although this will be addressed as a further refinement in the future. The current work did not apply Monte Carlo modelling to simulate interindividual human variation in DNA adduct formation in the population since it was recently published in Punt et al., (2016). Another interesting study would be the integration of repair mechanisms into the PBK/D model. Finally, we have developed a user-friendly tool implementing the newly developed PBK/D model

into

a

KNIME

workflow,

published

in

COSMOS

KNIME

webportal 24

ACCEPTED MANUSCRIPT (http://cosmosspace.cosmostox.eu/). This tool (workflow) is made freely available for endusers to automate the extrapolation from an in vitro effective concentration to the corresponding human exposure dose. The workflow could in principle be adapted to include route to route extrapolation or interspecies and intraspecies extrapolation, as well as low to high dose extrapolation.

PT

We believe that the methodology described supports efforts to shift away from the traditional approach to risk assessment based on animal experimentation to a more scientifically robust

AC

CE

PT E

D

MA

NU

SC

RI

approach based on automated in vitro testing and mathematical modelling.

25

ACCEPTED MANUSCRIPT Acknowledgements This research was partially funded by the Seventh Framework Programme (FP7/2007–2013) COSMOS (Integrated In Silico Models for the Prediction of Human Repeated Dose Toxicity of Cosmetics to Optimize Safety) Project (http://www.cosmostox.eu/) and by Cosmetics Europe. The

AC

CE

PT E

D

MA

NU

SC

RI

PT

authors thank Dr Ans Punt (Wageningen University) for constructive comments on the manuscript.

26

ACCEPTED MANUSCRIPT References

Alhusainy WAAM, Paini A, Punt A, Louisse J, Spenkelink A, Vervoort J, Delatour T, Scholz G, Schilter, B, Adams T, van Bladeren PJ, and Rietjens IMCM. (2010) Identification of nevadensin as an important herb-based constituent inhibiting estragole bioactivation and physiology-based biokinetic modeling of its possible in vivo effect. Toxicology and Applied Pharmacology 245, 179-190.

PT

Alhusainy, W, van den Berg, SJPL, Paini, A, Campana, A, Asselman, M, Spenkelink, A, Punt, A, (...), Rietjens, IMCM, (2012) Matrix modulation of the bioactivation of estragole by constituents of different

RI

alkenylbenzene-containing herbs and spices and physiologically based biokinetic modeling of

SC

possible in vivo Effects Toxicological Sciences, 129 (1), pp. 174-187.

Alhusainy, W, Paini, A, van den Berg, JHJ, Punt, A, Scholz, G, Schilter, B, van Bladeren, PJ, (...),

NU

Rietjens, IMCM, (2013) In vivo validation and physiologically based biokinetic modeling of the inhibition of SULT-mediated estragole DNA adduct formation in the liver of male sprague-dawley rats

MA

by the basil flavonoid nevadensin Molecular Nutrition and Food Research, 57 (11), pp. 1969-1978. Al-Subeihi AAA, Spenkelink B, Rachmawati N, Boersma MG, Punt A, Vervoort J, van Bladeren PJ, and Rietjens IMCM. (2011) Physiologically based biokinetic model of bioactivation and detoxification

D

of the alkenylbenzene methyleugenol in rat. Toxicology in Vitro, 25(1), 267-285.

PT E

Armitage JM, Wania F, Arnot JA (2014) Application of mass balance models and the chemical activity concept to facilitate the use of in vitro toxicity data for risk assessment. Environ. Sci. Technol., 48 (16)

CE

(2014), pp. 9770–9779

Bessems J, Paini A, Gajewska M, Worth A. (Manuscript submitted to Toxicology) The margin of

AC

internal exposure (MOIE) concept for route-to-route extrapolation of cosmetics. Bessems J, Coecke, S, Gouliarmou, V, Whelan, M, Worth, A (2015) EURL ECVAM Strategy for Achieving 3Rs Impact in the Assessment of Toxicokinetics and Systemic Toxicity. http://publications.jrc.ec.europa.eu/repository/bitstream/JRC96418/eurl%20ecvam%20toxicokinetics% 20strategy.pdf Broeders JJ, Parmentier C, Truisi GL, Jossé R, Alexandre E, Savary CC, Hewitt PG, Mueller SO, Guillouzo A, Richert L, van Eijkeren JC, Hermens JL, Blaauboer BJ. (2014) Biokinetics of chlorpromazine in primary rat and human hepatocytes and human HepaRG cells after repeated exposure.Toxicol In Vitro: S0887-2333(14)00202-1.

27

ACCEPTED MANUSCRIPT Broeders, JJ,

Hermens JL, Blaauboer BJ, Zurich MG (2015) The in vitro biokinetics of

chlorpromazine and diazepam in aggregating rat brain cell cultures after repeated exposure. Toxicol. in Vitro, 30 (1) (2015), pp. 185–191. EU-SCF (Scientific Committee on Food). (2001) Opinion of the scientific committee on food on estragole (1-allyl-4-methoxybenzene),. Gajewska M, Paini A, Sala Benito J, Burton J, Worth A, Urani C, Briesen H & Schramm KW (2015). In

PT

vitro to in vivo correlation of the skin penetration, liver clearance and hepatotoxicity of caffeine. Food and Chemical Toxicology 75, 39-49.

RI

Gardner I, Bergin P, Stening P, Kenna JG, Caldwell J, (1996) Immunochemical detection of covalently

SC

modified protein adducts in livers of rats treated with methyleugenol. Crit. Rev. Toxicol., 9 713 721. Iyer LV, Ho MN, Shinn WM, Bradford WW, Tanga MJ, Nath SS, Green CE, (2003) Glucuronidation of

NU

1’-hydroxyestragole (1’-HE) by human UDP-glucuronosyltransferases UGT2B7 and UGT1A9. Toxicol. Sci., 73, 36 43.

MA

Jeurissen SMF, Punt A, Boersma MG, Bogaards JJ, Fiamegos YC, Schilter B, van Bladeren PJ, Cnubben NHP, Rietjens IMCM, (2007) Human cytochrome P450 enzyme specificity for the bioactivation of estragole and related alkenylbenzenes. Chem. Res. Toxicol., 20 798 806.

D

Kramer NI (2010) Measuring, modelling and increasing the free concentration of test chemicals in cell

PT E

assays (PhD Thesis) Utrecht University (2010) Krishnan K and Andersen ME (2001) Physiologically Based Pharmacokinetic Modeling in Toxicology. th

Edition, A. Wallace Hayes (Ed), Taylor & Francis,

CE

In Principles and Methods in Toxicology, 4 Philadelphia. pp 193- 241.

AC

Martati E, Boersma MG, Spenkelink A, Khadka DB, Punt A, Vervoort J, van Bladeren PJ, and Rietjens IMCM. (2011) A physiologically based kinetic (PBK) model for safrole bioactivation and detoxification in rat and comparison to those for estragole and methyleugenol. Mennecozzi M, Landesmann B, Harris GA, Liska R, Whelan M (2011) Hepatotoxicity screening taking a mode-of-action approach using HepaRG cells and HCA. Altex Proc, 1/12 (2011) WC8 Miller EC, Swanson AB, Phillips DH, Fletcher TL, Liem A, Miller JA (1983). Structureactivity studies of the carcinogenicities in the mouse and rat of some naturally occurring and synthetic alkenylbenzene derivatives related to safrole and estragole. Cancer Res. 43(3):1124-34.

28

ACCEPTED MANUSCRIPT Paini, A, Punt, A, Viton, F, Scholz, G, Delatour, T, Marin-Kuan, M, Schilter, B, (...), Rietjens, IMCM. (2010) A physiologically based biodynamic (PBBD) model for estragole DNA binding in rat liver based on in vitro kinetic data and estragole DNA adduct formation in primary hepatocytes Toxicology and Applied Pharmacology, 245 (1), pp. 57-66 Paini, A, Punt, A, Scholz, G, Gremaud, E, Spenkelink, B, Alink, G, Schilter, B, (...), Rietjens, IMCM.

PT

(2012a) In vivo validation of DNA adduct formation by estragole in rats predicted by physiologically based biodynamic modelling Mutagenesis, 27 (6), pp. 653-663.

RI

Paini, A, (2012b) Generation of in vitro data to model dose dependent in vivo DNA binding of

SC

genotoxic carcinogens and its consequences: the case of estragole. WUR library PhD Thesis pp. 7085. http://library.wur.nl/WebQuery/clc/1984285

NU

Phillips DH, Miller JA, Miller EC, Adams B. (1981) Structures of the DNA adducts formed in mouse liver after administration of the proximate hepatocarcinogen 1'-hydroxyestragole Cancer Res., 41, 176

MA

186.

Punt A, Delatour T, Scholz G, Schilter B, Van Bladeren PJ, Rietjens IMCM. (2007) Tandem mass

D

spectrometry analysis of N2-(trans-isoestragol-3’-yl)-2’-deoxyguanosine as a strategy to study species

Res. Toxicol., 20 991 998.

PT E

differences in sulfotransferase conversion of the proximate carcinogen 1’-hydroxyestragole. Chem.

Punt A, Freidig AP, Delatour T, Scholz G, Boersma MG, Schilter B, van Bladeren PJ, Rietjens IMCM.

CE

(2008). A physiologically based biokinetic (PBBK) model for estragole bioactivation and detoxification in rat. Toxicol. Appl. Pharmacol. 231:248-259.

AC

Punt A, Paini A, Boersma MG, Freidig AP, Delatour T, Scholz G, Schilter B, van Bladeren PJ, Rietjens IMCM (2009). Use of physiologically based biokinetic (PBBK) modeling to study estragole bioactivation and detoxification in humans as compared with male rats. Toxicol. Sci. 110:255-269. Punt A, Jeurissen SM, Boersma MG, Delatour T, Scholz G, Schilter B, van Bladeren PJ, Rietjens IMCM. (2010) Evaluation of human interindividual variation in bioactivation of estragole using physiologically based biokinetic modeling Toxicol. Sci. 113, 337– 348. Punt A, Paini A, Spenkelink A, Scholz G, Schilter B, van Bladeren PJ, Rietjens IM, (2016) Evaluation of interindividual human variation in bioactivation and DNA adduct formation of estragole in liver predicted

29

ACCEPTED MANUSCRIPT by physiologically based kinetic/dynamic (PBK/D) and Monte Carlo modelling. Chemical Research in Toxicology, 29, 659-668. Ramanan, V, Scull, MA, Sheahan, TP, Rice, CM, & Bhatia, SN. (2014). New Methods in Tissue Engineering:

Improved

Models

for

Viral

Infection.

Annual

Review

of

Virology,

1,

475–499.

http://doi.org/10.1146/annurev-virology-031413-085437

Smith RL, Adams TB, Doull J. (2002) Safety assessment of allylalkoxybenzene derivatives used as

PT

flavouring substances—methyl eugenol and estragole. Food Chem. Toxicol., 40 851 870.

RI

Soldatow, VY, LeCluyse, EL, Griffith, LG, & Rusyn, I. (2013). In vitro models for liver toxicity testing.

Ullrich

D,

Fischer

G,

Katz

N,

Bock

KW.

SC

Toxicology Research, 2(1), 23–39. http://doi.org/10.1039/C2TX20051A (1984)

Intralobular

distribution

of

UDP-

NU

glucuronosyltransferase in livers from untreated, 3-methylcholanthrene- and phenobarbital-treated rats. Chem. Biol. Interact., 48,181 190.

MA

Worth et al., (present issue)

Zaldívar Comenges, JM, Mennecozzi M, Marcelino Rodrigues R, Bouhifd M. 2010. A biology-based

D

dynamic approach for the modelling of toxicity in cell-based assays. Part I: Fate modelling. EUR

PT E

24374 EN.

Zaldívar Comenges, JM, Baraibar, J., 2011, A biology-based dynamic approach for the reconciliation of acute and chronic toxicity tests: Application to Daphnia magna. Chemosphere 82, 1547-1555.

CE

Zaldívar Comenges, JM, Wambaugh, J., Judson, R, (2012) Modeling in vitro cell based assays

AC

experiments: cell population dynamics. Chapter in book Zaldívar Comenges, JM, Joossens E, Sala Benito JV, Worth A. Paini A, (2015) Theoretical and mathematical foundation of the Virtual Cell Based Assay. Toxicology in vitro. Spceial issue VCBA

30

ACCEPTED MANUSCRIPT Figure Legend Figure 1 Estragole metabolic pathway in rodents and humans. Figure 2. Schematic overview of the link between the VCBA and a PBK model and the possibility to extrapolate from an in vitro concentration to an in vivo external dose, and from an external dose to internal response. Light grey arrow represents forward extrapolation (external dose to internal effect), black arrow reverse extrapolation (internal effect to external dose). Figure 3. KNIME workflow to perform IVIVE, for human exposure, based on the estragole PBD model.

PT

Figure 4. Viability (4A) and mmp (4B) in percentage results after 24 hr exposure to different concentrations of estragole (mM) in HepaRG, (Y-axis I normalized to control at 0 mM, for Cell viability is represented by % of cell death and for mmp by ΔΨ).

AC

CE

PT E

D

MA

NU

SC

RI

Figure 5. Estragole dose response predictions using the automated human estragole PBK/D model to perform IVIVE (KNIME workflow), readouts for A. Cell viability (viability is normalized to 1, with no concentration of estragole 100% of the cell as viable) and B. DNA adducts (Y-axis in adducts /10^8 nucleotide).

31