Pregnancy-specific physiologically-based toxicokinetic models for bisphenol A and bisphenol S

Abstract

Predictions from physiologically based toxicokinetic (PBTK) models can help inform human health risk assessment for potentially toxic chemicals in the environment. Bisphenol S (BPS) is the second most abundant bisphenol detected in humans in the United States, after bisphenol A (BPA). We have recently demonstrated that BPS, much like BPA, can cross the placental barrier and disrupt placental function. Differences in physicochemical properties, toxicokinetics, and exposure outcomes between BPA and other bisphenols prevent direct extrapolation of existing BPA PBTK models to BPS. The current study aimed to develop pregnancy-specific PBTK (p-PBTK) models for BPA and BPS, using a common p-PBTK model structure. Novel paired maternal and fetal pregnancy data sets for total, unconjugated, and conjugated BPA and BPS plasma concentrations from three independent studies in pregnant sheep were used for model calibration. The nine-compartment (maternal blood, liver, kidney, fat, placenta and rest of body, and fetal liver, blood and rest of body) models simulated maternal and fetal experimental data for both BPA and BPS within one standard deviation for the majority of the experimental data points, highlighting the robustness of both models. Simulations were run to examine fetal exposure following daily maternal exposure to BPA or BPS at their tolerable daily intake dose over a two-week period. These predictive simulations show fetal accumulation of both bisphenols over time. Interestingly, the steady-state approximation following this dosing strategy achieved a fetal concentration of unconjugated BPA to levels observed in cord blood from human biomonitoring studies. These models advance our understanding of bisphenolic compound toxicokinetics during pregnancy and may be used as a quantitative comparison tool in future p-PBTK models for related chemicals.

Graphical Abstract

Introduction

Bisphenols are a large class of chemicals structurally identified as having two hydroxyphenyl rings. Many bisphenols are considered endocrine disrupting chemicals (EDCs) (Gore et al. 2014). They are widely used in the manufacturing of polycarbonate plastics, epoxy resins, dental sealants, and plastic and paper consumer products (Liao and Kannan 2013; Liao et al. 2012a), and are pervasively present in dust and soil (Kwak et al. 2018; Liao et al. 2012b). Due to consumer concerns and heightened regulations regarding the use of bisphenol A (BPA) in some countries (Jalal et al. 2018), industrial and consumer products producers have resorted to using less studied bisphenol alternatives in their products (EPA 2015). Such BPA-alternatives include bisphenol S (BPS), which is structurally similar to BPA, and is becoming just as environmentally prevalent (Rochester and Bolden 2015). As a consequence, BPS is the second leading bisphenol found in humans following BPA (Liao et al. 2012a; Philips et al. 2018; Ye et al. 2015). Bisphenols can be detected in urine, blood, breast milk, amniotic fluid and cord blood, highlighting the ubiquitous exposure humans have to these chemicals (Asimakopoulos et al. 2016; Lehmler et al. 2018; Liao et al. 2012b; Philips et al. 2018; Rocha et al. 2018; Xue et al. 2015; Ye et al. 2015). Several studies have shown that even at low concentrations, exposure to BPA during gestation can result in negative effects on the development of the fetus (Gingrich et al. 2020; Veiga-Lopez et al. 2018). The detection of BPS in human fetal cord blood (Kolatorova et al. 2018), the positive association between BPS exposure and prolonged gestational length (Wan et al. 2018), and the fact that in mammals, fetal exposure to BPS can alter reproductive (Gingrich et al. 2018; Kolla et al. 2018), metabolic (Pu et al. 2017b), and behavioral outcomes (Catanese and Vandenberg 2017), warrant research into the precise toxicokinetic mechanisms of these emerging bisphenol chemicals during pregnancy.

Physiologically based toxicokinetic (PBTK) mathematical models integrate toxicokinetic processes such as chemical absorption, distribution, metabolism, and excretion (ADME). The main advantage of PBTK models over the classical compartmental approaches to understanding chemical toxicokinetics is the ability of PBTK models to extrapolate outside of the conditions or population that was evaluated experimentally (Tsamandouras et al. 2015). The quantitative predictive and extrapolative capabilities of PBTK models can inform health risk assessments for chemical and pharmaceutical exposure (Ke et al. 2018; Zhuang and Lu 2016). Chemical toxicokinetics during pregnancy are more complex with the inclusion of the maternal, placental, and fetal compartments (Ke et al. 2018). Moreover, ethical constraints do not allow for any toxicokinetic studies other than biomonitoring to be conducted in pregnant women. The use of refined fetal surgery techniques in a sheep animal model represents unique opportunities to monitor the maternal, amniotic, and fetal compartments; key elements of pregnancy-specific PBTK (p-PBTK) models (Ke et al. 2018). Importantly, sheep are excellent models to study placental function (Fowden et al. 2015; Mourier et al. 2017) and have been used for the study of feto-maternal transfer of drugs (Krishna et al. 2002; Ngamprasertwong et al. 2016) and EDCs (Corbel et al. 2013; Corbel et al. 2015), as they allow for the simultaneous and longitudinal characterization of the pregnancy multi-compartment model in real time.

The toxicokinetics of BPA have been extensively studied and modeled in both animals and humans (Fisher et al. 2011; Karrer et al. 2018; Kawamoto et al. 2007; Poet and Hays 2018; Sharma et al. 2018; Vom Saal et al. 2014). Primary metabolism (conjugation) for BPA occurs in the liver and the intestine (Domoradzki et al. 2003). In rodents, BPA undergoes substantial enterohepatic recirculation. However, in monkeys and humans, the rapid metabolism and extensive renal excretion of BPA metabolites means that a negligible amount of conjugated BPA is able to undergo enterohepatic recirculation (Doerge et al. 2010; Volkel et al. 2002). In pregnancy, both conjugation and deconjugation reactions also occur in fetal tissues, primarily the fetal liver, but these processes occur at varying rates during different developmental windows. In the early developmental stages deconjugation dominates with conjugation barely occurring (Lucier et al. 1977). However, in the case of BPA, conjugation has been shown to increase from 512-fold lower to 13-fold lower when compared with maternal conjugation rates from early to late pregnancy (Corbel et al. 2015).

Despite the breadth of work on BPA, only a limited number of studies have investigated the toxicokinetics of BPS during pregnancy (Gingrich et al. 2019; Grandin et al. 2018). Of the two available BPS toxicokinetic models (Karrer et al. 2018; Oh et al. 2018) only one is physiologically-based (Karrer et al. 2018) and it is based on a non-pregnant sheep dataset. This non-pregnant BPS model was derived by a substitution of parameters from a previously calibrated BPA PBTK model with parameter values derived from quantitative structure-activity relationships (QSARs) for BPS, but was neither formally calibrated, nor validated. Recently, BPS was reported to reach higher systemic concentrations than BPA in humans (Khmiri et al. 2020), representing a need to better distinguish toxicokinetic characteristics between bisphenols, for which PBTK models are uniquely suited. Therefore, the objective of our current study was to improve the understanding of pregnancy toxicokinetics for bisphenols through the development of physiologically relevant multi-compartment p-PBTK models for BPA and BPS. Both p-PBTK models were developed using three independent pair-matched maternal and fetal sheep exposure cohort datasets (Corbel et al. 2013; Gingrich et al. 2019; Grandin et al. 2018).

Materials and Methods

Datasets

Experimental datasets used in this work were obtained from previously published bisphenol toxicokinetic studies in pregnant sheep (Corbel et al. 2013; Gingrich et al. 2019; Grandin et al. 2018). For model calibration, three independent datasets were used (two for each bisphenol). Dataset #1 from Gingrich et al. (Gingrich et al. 2019), reported total (conjugated plus unconjugated) bisphenol concentrations for BPA and BPS in the maternal and fetal plasma and was used for calibrating both bisphenol models. In brief, toxicokinetic data was obtained from pregnant Polypay x Dorset sheep (singleton pregnancies only) that underwent fetal catheterization surgery at gestational day (GD) 115. Females (n = 3) were injected with a single subcutaneous dose of BPS (0.5 mg/kg) or a combination of BPA and BPS (n = 3; 0.5 mg/kg for each chemical) and data were collected over a 72-h period. No differences in toxicokinetic parameters (maximum concentration reached, time of maximum concentration, half-life, area under the curve, area under the first moment curve, mean residence time, and total body clearance) between single-chemical exposure and mixture dosing were reported, so all BPS values (n = 6) were combined.

Additionally, two other toxicokinetic studies in pregnant sheep which presented data for conjugated and unconjugated bisphenols (Corbel et al. 2013; Grandin et al. 2018) were used during model calibration. For BPA, dataset #2 was obtained from Corbel et al. (Corbel et al. 2013), who used pregnant Lacaune sheep (unreported fetal number) that underwent fetal catheterization surgery between GD 108 and 117. In separate experiments, females (n = 8) and fetuses (n = 8, unreported sex) were dosed with an intravenous (IV) infusion over 24 hours of unconjugated BPA or BPA-glucuronide (conjugated, BPA-G) at a dose of 2.0 and 3.54 mg/kg/day respectively in the mother, and 5.0 and 3.54 mg/kg/day respectively in the fetus, assuming a 2.5 kg fetus. Plasma concentrations were collected over a 46-h period and the steady state plasma concentration over the final 3-h of infusion was reported. For BPS, dataset #3 was obtained from Grandin et al. (Grandin et al. 2018) which included pregnant Lacaune sheep (unreported fetal number) that underwent fetal catheterization surgery between GD 109 and 113. A dual dosing strategy was used, where pregnant females (n = 8) and their fetuses received simultaneous IV doses. First the mother received a dose of 2.7 mg/kg BPS-glucuronide (conjugated, BPS-G) and the fetus was administered a dose of 5 mg deuterated BPS (BPS-d8). This procedure was followed by a simultaneous administration of 5 mg/kg BPS to the mother and 17.5 mg BPS-G-d8 to the fetus. Plasma concentrations were reported over a 72-h period. Dataset #3 was collected at somewhat regular intervals, though not always at the exact same time point (Grandin et al. 2018). As such, these data could not be directly aggregated to yield mean and standard deviation values. Instead, the plasma data for each animal was interpolated using a cubic spline. The most representative time points were selected, and all the interpolated time-concentration curves were sampled at the selected time points and aggregated together. The time points used were either those containing the most data points across animals for all time points except the first and the last ones, or time points lying within each sheep’s interpolation region, for the first and last timepoint, to prevent extrapolation.

Model development

To establish informative and useful p-PBTK models for BPA and BPS, we developed a minimal generic p-PBTK model for an unconjugated bisphenol and its conjugate metabolite that includes 6 compartments for the mother (liver, fat, kidney, placenta, blood and rest of the body) and 3 compartments for the fetus (liver, blood and the rest of the body) (Figure 1). All relevant biological processes were included, namely conjugation (metabolism) in the maternal and fetal livers, maternal urinary and biliary excretion, and deconjugation in the fetal liver (Nishikawa et al. 2010). The two coupled sub-models of identical structure for the unconjugated and conjugated bisphenols were connected through liver metabolism in the mother and the fetus, as well as deconjugation in the fetal liver, with one sub-model used for the parent compound (BPA or BPS) and another for the conjugate (BPA_conj or BPS_conj). In the case of BPS, a duplicate model was developed for deuterated BPS and BPS_conj to account for fetal administration. Subsequently, we determined the physiological parameters for an average pregnant sheep, with a single fetus, at the gestational age where the experimental data were generated. This was done for the generic model, as well as for separately parametrized and calibrated individual instances of the generic model for both unconjugated BPA and BPS, and their respective conjugated metabolites. We based our model structure and types of compartments and processes to be inclusive of the only two, to our knowledge, published BPA p-PBTK models (Kawamoto et al. 2007; Sharma et al. 2018). All compartments were considered perfusion limited for both unconjugated and conjugated bisphenol sub-models. The most common bisphenol conjugate is glucuronide, although others, such as sulfate, exist (Ho et al. 2017). Due to a lack of available data on non-glucuronide conjugates, all conjugates for each parent compound were combined into a single conjugate parameter (BPA_conj/BPS_conj) which was calibrated against glucuronide-conjugate data.

Figure 1. PBTK model scheme.

Nine-compartment PBTK pregnancy model scheme for BPA and BPS. Compartments included are maternal fat (F), liver (L), kidney (K), blood (A), and rest of body (R), placenta (PL), and fetal blood, liver, and rest of body. Model includes subcutaneous first order absorption constant (K_a), placental to fetal (kt₁) and fetal to placental (kt₂) blood diffusion rates, chemical concentration in arterial blood (C_A), maximum rate of enzymatic reaction (V_max) and Michaelis-Menten constant (K_m), renal excretion (KEL_R and KEL_{R_conj}), biliary excretion (KEL_L) – resulting in fecal elimination, and for each of the nine-compartments, the respective blood flow rate (Q-tissue), partition coefficients (P-tissue), tissue concentrations (C-tissue), and venous concentration leaving the tissue (CV-tissue), for unconjugated (left panel) and conjugated (right panel) bisphenols. The addition of _f denotes fetal-specific parameters. BPS molecular structures below represent the difference in unconjugated and conjugated molecular size.

Model equations

The equations listed in this section describe both the BPA and the BPS p-PBTK models and are the same between the unconjugated and the conjugated forms of the compounds. The model equations for the maternal unconjugated bisphenol models are described below. All transport equations were perfusion limited. Equations for conjugation represented saturable metabolism in both maternal and fetal livers and the equation for deconjugation in the fetal liver, as well as maternal urinary and biliary excretion equations were modeled as first order processes. All physiological and biochemical parameter units can be found in Tables 1, 2, and 4.

Table 1.

Physiological parameters in pregnant sheep

Parameter	Abbreviation	Value	Units
Body weight	BW	76.251	kg
Total cardiac output2	QCC	6.9	L/h/kg BW
Fractional blood flow to fat3	QFC	8.5	%
Fractional blood flow to kidney3	QKC	17	%
Fractional blood flow to liver2	QLC	18.3	%
Fractional blood flow to placenta4	QPLC	8	%
Fractional volume of fat3	VFC	0.168	L/kg BW
Fractional volume of kidney3	VKC	0.0046	L/kg BW
Fractional volume of liver3	VLC	0.016	L/kg BW
Fractional volume of blood3	VBC	0.057	L/kg BW
Fractional volume of feto-placental unit1	VPLEFC	0.078	L/kg BW
Fractional volume of fetus4	VEFC	0.0525	L/kg BW

Values listed obtained from references:

¹Gingrich et al., 2019

²Craigmill, 2003

³Upton et al., 2008

⁴Makowski et al., 1968.

Table 2.

BPA and BPS physicochemical parameters

Parameter	Value	Units	Reference
Dose	0.5*10−3	g/kg BW	Gingrich et al. 2019
Ka-BPS	0.183	L/h	Fit to model
MW-BPS	250.3	g/mol	NCBI PubChem
MW-BPS-G	426	g/mol	Karrer et al. 2018
Ka-BPA	0.204	L/h	Fit to model
MW-BPA	228.3	g/mol	Karrer et al. 2018
MW-BPA-G	404	g/mol	Karrer et al. 2018

BW: body weight, G: glucuronide, Ka: absorption rate constant, MW: molecular weight

Table 4.

Rate constants

Parameter	Abbreviation	BPS	BPSconj	BPA	BPAconj
Bioavailability (%)	FSC	43	-----	12.9	-----
Biliary excretion (h−1)	KELL	-----	0.061	-----	2.052
Renal excretion (L/h)	KELR	0.023	4.093	0.035	0.375
Maternal Michaelis-Menten constant (mg/L)	Km	4.79	-----	3.46	-----
Maternal maximum rate of metabolism (mg/h/kg0.75)	Vmax	8185.66	-----	3,458.40	-----
Placental to fetal transfer	kt1	0.075	-----	6.733	-----
Fetal to placental transfer	kt2	0.113	-----	5.412	-----
Fetal deconjugation (L/h)	Kd_f	-----	2.80	-----	1.41
Fetal Michaelis-Menten constant (mg/L)	Km_f	2.74	-----	6.85	-----
Fetal maximum rate of metabolism (mg/h/kg0.75)	Vmax_f	1,000.26	-----	4,312.04	-----

$$V_K\frac{d{C}_K}{dt}=Q_K\cdot \left(C_A-\frac{C_K}{P_K}\right)\text{Maternal Kidney (K)}$$

$$V_L\frac{d{C}_L}{dt}=Q_L\cdot \left(C_A-\frac{C_L}{P_L}\right)-\frac{V_{max}\cdot C_L∕P_L}{(K_m+C_L∕P_L)}\text{Maternal Liver (L)}$$

$$V_F\frac{d{C}_F}{dt}=Q_F\cdot \left(C_A-\frac{C_F}{P_F}\right)\text{Maternal Fat (F)}$$

$$V_R\frac{d{C}_R}{dt}=Q_R\cdot \left(C_A-\frac{C_R}{P_R}\right)\text{Maternal Rest of body (R)}$$

$$V_B\frac{d{C}_A}{dt}=\sum Q_{T\_all}\cdot \frac{C_{T\_all}}{P_{T\_all}}-Q_C\cdot C_A-KE{L}_R\cdot C_A+F_{SC}\cdot K_a\cdot A_{SC}\text{Maternal Blood (B)}$$

$$V_{PL}\frac{d{C}_{PL}}{dt}=Q_{PL}\cdot (C_A-C_{VPL})-k{t}_1{C}_{VPL}+k{t}_2{C}_{A\_f}\text{Feto-placental (PL)}\text{transfer}$$

$$C_{VT}=\frac{C_T}{P_T}\text{Concentration in venous}\text{blood exiting tissue (T)}$$

V_T is the volume of tissue T, Q_T is the blood perfusion, C_T is the chemical concentration, P_T is the blood:tissue partition coefficient, and C_VT is the concentration in the venous blood exiting the tissue. C_A is the chemical concentration in the arterial blood. kt₁ and kt₂ are the diffusion rates from maternal placental blood to fetal blood and fetal blood to maternal placental blood, respectively. Subscript _f denotes fetal tissues. T_all is used in the maternal blood compartment to describes the sum of all tissue compartments. F_SC and A_SC represent the bioavailability of subcutaneous administration and remaining unabsorbed subcutaneous dose, respectively, and K_a is the first order rate constant for subcutaneous absorption. KEL_R is the rate of renal excretion. V_max is the maximum reaction rate, and K_m the Michaelis-Menten constant.

The model equations for the maternal conjugated bisphenol models are described as follows. Equations that are the same as in the unconjugated bisphenol models have been omitted.

$$V_L\frac{d{C}_L^{(c)}}{dt}=Q_L\left(C_A^{(c)}-\frac{C_L^{(c)}}{P_L^{(c)}}\right)+\frac{V_{max}\cdot \frac{C_L}{P_L}}{\left(K_m+\frac{C_L}{P_L}\right)}-KE{L}_L^{(c)}{C}_{VL}^{(c)}{V}_L\text{Maternal Liver}$$

$$V_B\frac{d{C}_A^{(c)}}{dt}=\sum Q_{T\_all}\cdot \frac{C_{T\_all}^{(c)}}{P_{T\_all}^{(c)}}-Q_C\cdot C_A^{(c)}-KE{L}_R^{(c)}\cdot C_A^{(c)}\text{Maternal Blood}$$

Here, (c) in superscript denotes the conjugated compound, KEL_L is the rate of biliary excretion. All the other symbols have the same meaning as in the unconjugated bisphenol models.

The model equations for the fetal unconjugated bisphenol models are described below.

$$V_{L\_f}\frac{d{C}_{L\_f}}{dt}=Q_{L\_f}\cdot \left(C_{A\_f}-\frac{C_{L\_f}}{P_L}\right)-\frac{V_{\max \_f}\cdot C_{L\_f}∕P_L}{(K_{m\_f}+C_{L\_f}∕P_L)}+K_{d\_f}{C}_{L\_f}^{(c)}∕P_L^{(c)}\text{Fetal Liver}$$

$$V_{R\_f}\frac{d{C}_{R\_f}}{dt}=Q_{R\_f}\cdot \left(C_{A\_f}-\frac{C_{R\_f}}{P_{R\_f}}\right)\text{Fetal Rest of body}$$

$$V_{B\_f}\frac{d{C}_{A\_f}}{dt}=\sum Q_{T\_f\_all}\cdot \frac{C_{T\_f\_all}}{P_{T\_f\_all}}-Q_{C\_f}\cdot C_{A\_f}+k{t}_1{C}_{VPL}-k{t}_2{C}_{A\_f}\text{Fetal Blood}$$

Maternal liver partition coefficient (P_L) was used in the fetus, as well.

The equations for the conjugated bisphenol models in the fetus are described below.

$$V_{L\_f}\frac{d{C}_{L\_f}^{(c)}}{dt}=Q_{L\_f}\left(C_{A\_f}^{(c)}-\frac{C_{L\_f}^{(c)}}{P_L^{(c)}}\right)+\frac{V_{\max \_f}\cdot C_{L\_f}∕P_L}{(K_{m\_f}+C_{L\_f}∕P_L)}-K_{d\_f}{C}_{L\_f}^{(c)}∕P_L^{(c)}\text{Fetal Liver}$$

$$V_{B\_f}\frac{d{C}_{A\_f}^{(c)}}{dt}=\sum Q_{T\_f\_all}\cdot \frac{C_{T\_f\_all}^{(c)}}{P_{T\_f\_all}^{(c)}}-Q_{C\_f}\cdot C_{A\_f}^{(c)}\text{Fetal Blood}$$

Parametrization

The generic bisphenol model was first partially parametrized with the pregnant sheep physiological parameters obtained from the literature (Table 1), inclusive of fractional blood flows and organ volumes. Following this procedure, two separate model instances were created for BPA and BPS using their respective physiochemical parameters (Table 2) and the tissue:blood partition coefficients (Table 3), which were calibrated within ranges of one order of magnitude around values either obtained from the literature (Craigmill 2003; Gingrich et al. 2019; Karrer et al. 2018; Makowski et al. 1968; NCBI ; Upton 2008), or estimated from the available log octanol:water partition parameters for compounds with similar partitioning (Choi and Lee 2017; Chow et al. 2016; Lyons et al. 2013) and calibration was performed within those ranges. Partition coefficients for the rest of the body for both BPA and BPS models were calibrated within the minimum and maximum values for all other tissues. All physiological parameters were assumed to be time-invariant due to the nature of the experimental data, which was collected over a short period of time during mid-late pregnancy.

Table 3.

Passive biochemical parameters (tissue/blood partition coefficients)

Compartment	Abbreviation	BPS	BPSconj	BPA	BPAconj
Adipose	PF	0.031	0.0027	1.160	0.220
Kidney	PK	0.017	0.0049	0.858	3.180
Liver	PL	2.300	2.4700	4.350	6.760
Rest of Body (maternal)	PR	0.013	0.0019	0.044	0.154
Placenta	PPL	0.106	0.0020	0.880	0.680
Fetal Rest of Body (fetal)	PR_f	0.005	0.1680	0.006	0.500

Calibration

The calibration for both the BPA and the BPS models was carried out in four steps: 1) fetal conjugated bisphenol calibration, 2) maternal conjugated bisphenol calibration, 3) maternal complete calibration, and 4) feto-placental transfer and fetal complete calibration. Maternal body weight used was dependent on which of the three experimental datasets the model was being calibrated against. During the fetal conjugated bisphenol calibration, the appropriate fetal conjugated bisphenol IV administration experiment was used to partially calibrate the fetal model, namely the conjugated bisphenol partition parameters for the fetal liver and the rest of the body. The maternal conjugated bisphenol calibration relied on the maternal conjugated bisphenol IV administration data and was used to partially calibrate the maternal model, namely the remaining conjugated bisphenol partition coefficients (maternal kidney, fat, and rest of body), as well as urinary and biliary excretion rates. The complete maternal calibration relied on the maternal unconjugated bisphenol IV and total bisphenol subcutaneous administration data from all three datasets. These were used to fully calibrate the maternal model, namely the unconjugated bisphenol partition coefficients, and metabolism and urinary excretion rates (Table 4). During this step of the calibration, the feto-placental transfer of the unconjugated bisphenol was not accounted for to minimize the number of calibrated parameters. Feto-placental transfer and total fetal calibration relied on the maternal unconjugated bisphenol IV and subcutaneous administration to fully calibrate the feto-placental diffusion rates and the remainder of the fetal parameters. These parameters were mainly unconjugated bisphenol partition coefficients for the rest of the body and metabolism and deconjugation rates in the fetal liver. Except for the rest of body partition coefficient, all other partition coefficients corresponding to the same tissue between the mother and the fetus were assumed equal.

Our p-PBTK models required the use of blood-to-plasma partition coefficients as parameters, since the experimentally derived calibration datasets reported plasma concentrations. The blood-to-serum partition coefficient for BPA in rats has been experimentally determined as 1.10 (Shin et al. 2004), and blood-to-plasma partition coefficient for BPA and BPA-glucuronide in humans has been computationally estimated to be 1.05 and 0.83, respectively (Edginton and Ritter 2009). Since calibrating the blood:plasma partition coefficient in the BPA model for both conjugated and unconjugated BPA within the range of 0.80 to 1.20 did not affect the model results in a significant way, blood:plasma partition coefficients for both conjugated and unconjugated BPA were fixed to 1, simplifying the modeling procedure. We observed similar results for BPS, and have thus fixed the blood:plasma partition coefficients of both conjugated and unconjugated BPS to 1.

Calibration of unknown parameter values was performed using sequential least square quadratic programming with random restart (Bonnans et al. 2006). Sequential least squares quadratic programming is a formal optimization technique known to perform well for systems requiring constrained nonlinear optimization, which was the case for our developed models. Here, each calibration procedure was repeated 500 times, each time starting from a randomly selected point within the allowable ranges of the calibrated parameters. The calibration most closely matching the datasets, using the lowest mean absolute percentage error score as the selection criteria, was chosen as the final calibration.

Extrapolation of maternal and fetal body burdens

Dosing regimens simulating daily repeated maternal and fetal exposures to both BPA and BPS were run with the calibrated ovine models using the reference dose for BPA set by the U.S. Environmental Protection Agency (50 μg/kg/day) (EPA 2012). Simulations were run over a two-week period.

Computing software

The current model was coded in, and all simulations run using the Python programming language and the Python package Tellurium version 2.1.5 developed for reproducible dynamical modeling of biological networks (Medley et al. 2018). The full model code is available at https://github.com/BhattacharyaLab/BisphenolPBTK

Sensitivity analysis

Global sensitivity analyses of the fetal plasma compartment kinetics for both unconjugated and conjugated BPA and BPS were performed to identify the most influential parameters determining fetal bisphenol kinetics. Sensitivity analysis was performed using the variance-based Sobol method (Saltelli et al. 2010), as implemented within the SALib python library (Herman and Usher 2017). Parameters determining fetal kinetics were examined between 50% and 150% of the nominal values listed in Table 3 and Table 4, and were sampled using the Saltelli sampling scheme with N = 1,000 generated samples (Saltelli 2002). To examine simulated fetal kinetics with both a loading (absorption) and an elimination phase, subcutaneous dosing from Dataset #1 was selected, as described in Datasets. The sensitivity analysis was repeated every half-hour for 48 h of simulation time (excluding 0 h). The parameters included in the sensitivity analysis were the fetal hepatic metabolism parameters (V_{max_f} and K_m_f), deconjugation rate constant (K_{d_f}) and rest of body partition coefficients (PR_{_f} and PR_{_f}^(c)). Additionally, we repeated the sensitivity analysis by also adding the feto-placental transfer parameters (kt₁ and kt₂).

Results

Calibration

Simulations of the fully calibrated BPA model for both maternal and fetal compartments were compared to experimental dataset #1 following a single subcutaneous administration of BPA to the mother (Figure 2A maternal compartment, and Figure 2D fetal compartment). A full simulation was also performed for dataset #2 following either a 24-h IV infusion of BPA and BPA-G to the mother (Figure 2B and Figure 2C, respectively), or 24-h IV infusion of BPA and BPA-G to the fetus (Figure 2E and Figure 2F, respectively). All simulations matched the experimental data ± one standard deviation from the individual data points for total, unconjugated and conjugated BPA (Corbel et al. 2013; Gingrich et al. 2019).

Figure 2. Simulated toxicokinetic plots of BPA for maternal and fetal circulation.

Experimentally obtained total BPA measurements in sheep maternal plasma over time (gestational age: 114.8 ± 0.8 days; diamonds in panel A) following a single subcutaneous injection of BPA (0.5 mg/kg; dataset #1 (Gingrich et al. 2019)) were used to calibrate simulated total BPA in maternal plasma over time (A; solid blue line). The BPA model was also calibrated against dataset #2 (circles) of unconjugated (B; orange line) and conjugated (C; dotted blue line) BPA from sheep plasma following steady state conditions (continuous intravenous infusion, 2 mg/kg BW/day BPA, gestational age: 108 - 117 days, (Corbel et al. 2013)). This process was repeated for the fetal compartment using fetal plasma concentrations of total (D), unconjugated (E) and conjugated (F) BPA, obtained from the dataset #1 and dataset #2 used for the maternal compartment calibration. All independent dataset values fall within the simulated ranges.

Similar to BPA, simulations of the fully calibrated BPS model were compared to experimental dataset #1 following a single subcutaneous injection of BPS to the mother (Figure 3A - maternal compartment, and Figure 3D - fetal compartment), or dataset #3 following a single IV bolus of BPA and BPA-G to the mother (Figure 3B and Figure 3C, respectively), or a single IV bolus of BPS-d8 and BPS-G-d8 to the fetus (Figure 3E and Figure 3F, respectively). Except for fetal IV boluses of BPS-d8 and BPS-G-d8, all data points were consistent with the experimental datasets ± one standard deviation from the individual data points for total, unconjugated and conjugated BPS (Gingrich et al. 2019; Grandin et al. 2018). However, adjustments to our model code could be made to account for a smaller fetus, for which were able to achieve a better simulated fit to these data (Supplemental Figure 3).

Figure 3. Simulated toxicokinetic plots of BPS for maternal and fetal circulation.

The model was calibrated through experimentally obtained total BPS measurements in sheep maternal plasma over time (gestational age: 114.8 ± 0.8 days; diamonds in panel A) following a single subcutaneous injection of BPS (0.5 mg/kg; dataset #1 (Gingrich et al. 2019)). Simulated total BPS in maternal plasma over time (A; solid blue line) was estimated from this model. The BPS model was also calibrated against dataset #3 (circles) of unconjugated (B; orange line) and conjugated (C; dotted blue line) BPS from sheep plasma following a single bolus intravenous injection (5 mg/kg BPS, gestational age: 109-113, (Grandin et al. 2018)). Insets have been included for these parameters to better demonstrate dataset fit to the simulation. The calibration process was repeated for the fetal compartment using fetal plasma concentrations of total (D), unconjugated (E) and conjugated (F) BPS, obtained from the dataset #1 and dataset #3 used for the maternal compartment calibration (Gingrich et al. 2019; Grandin et al. 2018). All dataset values except conjugated, and to a lesser extent unconjugated fetal BPS, fall within the simulated values.

Due to its robustness, full simulations of dataset #1 (Gingrich et al. 2019), separated into total, conjugated, and unconjugated forms of bisphenols, were run for both BPA and BPS (Supplemental Figure 1 and Supplemental Figure 2, respectively). This was necessary to estimate the breakdown of unconjugated and conjugated bisphenols, which was not available from the original dataset.

Extrapolation of maternal and fetal body burdens in an ovine model

Simulations showing repeated daily subcutaneous exposure to BPA and BPS are shown in Figure 4 and Figure 5, respectively. Maternal exposure was consistent with known toxicokinetic parameters for BPA (Gingrich et al. 2019), where unconjugated BPA was cleared from circulation within a 24-h period (Figure 4A, right panel). In the fetal compartment, we observed a gradual accumulation of total, unconjugated and conjugated BPA (Figure 4B), plateauing around a mean of 0.28 ng/ml unconjugated BPA at 14 days of daily exposure (Figure 4B, right panel, solid black line). Like BPA, total, unconjugated and conjugated BPS also rapidly clears from maternal blood (Figure 5A) and accumulate in the fetal compartment, but total fetal BPS accumulation does not plateau within the 14-day exposure window (Figure 5B, left panel). The BPS model simulates fetal blood concentrations at a mean of 0.45 ng/ml unconjugated BPS by 14 days of exposure (Figure 5B, right panel).

Figure 4. Simulation of BPA body burden following two weeks of daily dosing in an ovine model.

Using the BPA-model code, a simulation was run to predict maternal (A) and fetal (B) circulating BPA concentrations over 2 weeks following daily oral BPA administration (arrows) at the EPA reference dose for BPA (50 μg/kg body weight/day, (EPA 2012)). Simulated concentrations include total BPA (solid blue line), conjugated BPA (dotted blue line), and unconjugated BPA (solid orange line). Simulated total and conjugated BPA concentrations appear to nearly overlap and are indistinguishable from each other. To the right of each graph (shaded boxes), a magnified view of unconjugated BPA concentrations is provided. Simulated unconjugated BPA following a single dose is represented by a dotted orange line and exposure average of unconjugated BPA following daily dosing for 2 weeks is represented by the solid black line.

Figure 5. Simulation of BPS body burden following two weeks of daily dosing in an ovine model.

Using the BPS-model code, a simulation was run to predict maternal (A) and fetal (B) circulating concentrations of BPS following 2 weeks of daily oral BPS administration (arrows) at the EPA reference dose for BPA (50 μg/kg body weight/day, (EPA 2012)). Simulated concentrations include total BPS (solid blue line), conjugated BPS (dotted blue line), and unconjugated BPS (solid orange line). In maternal circulation, simulated total and conjugated BPS concentrations appear to nearly overlap and are indistinguishable from each other. To the right of each graph (shaded boxes), a magnified view of unconjugated BPS concentrations is provided. Simulated unconjugated BPS following a single dose is represented by a dotted orange line and exposure average of unconjugated BPS following daily dosing for 2 weeks is represented by the solid black line.

Sensitivity analysis

A global sensitivity analysis was run to investigate the main effect of all relevant fetal parameters over time and results are shown in Figure 6 (for BPA) and Figure 7 (for BPS). The dynamics of the main and interaction effects over time are shown in Supplemental Figure 4. For both bisphenols, the main effect (%) of the placental to fetal transfer parameter kt₁ were the highest among the parameters evaluated for both unconjugated and conjugated BPA and BPS. For BPA, the main effect of the fetal to placental transfer parameter kt₂ increased over time while other parameters like fetal hepatic deconjugation (K_{d_f}) and the rate of enzymatic reaction (V_{max_f}) remained constant. For BPS, the main effect of kt₂ was lower than for BPA, but also increased over time. The contribution of other parameters that determine fetal plasma kinetics, such as metabolic (V_{max_f}, K_{m_f}) and deconjugation (K_{d_f}) parameters tended to increase over time. Metabolic parameters (V_{max_f}, K_{m_f}) were more important for fetal plasma kinetics of unconjugated BPA, until ~15 h where they begin to plateau. For unconjugated BPS, the main effect of both V_{max_f} and K_{m_f} was higher than K_{d_f} throughout the 48-h period. The rest of body partition coefficient for unconjugated bisphenols (PR_{_f}) had a minor contribution to output variance in determining both BPA and BPS fetal plasma kinetics, however the rest of body partition coefficient for conjugated bisphenols (PR_{_f}^(c)) was especially important for determining conjugated BPA and BPS plasma kinetics.

Figure 6. Global sensitivity analysis of the fetal compartment for BPA model.

Global sensitivity analysis of fetal compartment parameters for the BPA model run over a 48-h period. Parameters were assessed for unconjugated BPA with (A) and without placental transfer (B) and conjugated BPA with (C) and without placental transfer (D). Sensitivity analysis was depicted as the main effect or percent of output variance for each parameter and was computed every half-hour for 48 h. Parameters evaluated were: K_{d_f} (rate of fetal hepatic deconjugation, solid black line), V_{max_f} (maximum rate of enzymatic reaction, solid red line), K_m (Michaelis-Menten constant, solid blue line), fetal rest-of-body partition coefficient for both unconjugated (PR_{_f}, hashed black line), and conjugated BPA (PR_{_f}^(c), hashed red line), and placental transfer to (kt₁, dotted black line) and from (kt₂, dotted red line) the placenta. The shaded area around the parameter lines depicts a 95% confidence interval.

Figure 7. Global sensitivity analysis of the fetal compartment for BPS model.

Global sensitivity analysis of fetal compartment parameters for the BPS model run over a 48-h period. Parameters were assessed for unconjugated BPS with (A) and without placental transfer (B) and conjugated BPS with (C) and without placental transfer (D). Sensitivity analysis was depicted as the main effect or percent of output variance, for each parameter and was computed every half-hour for 48 h. Parameters evaluated were: K_{d_f} (rate of fetal hepatic deconjugation, solid black line), V_{max_f} (maximum rate of enzymatic reaction, solid red line), K_{m_f} (Michaelis-Menten constant, solid blue line), fetal rest-of-body partition coefficient for both unconjugated (PR_{_f}, hashed black line) and conjugated BPS (PR_{_f} ^(c), hashed red line), and placental transfer to (kt₁, dotted black line) and from (kt₂, dotted red line) the placenta. The shaded area around the parameter lines depicts a 95 % confidence interval.

Discussion

Here, we have reported successful development of p-PBTK models for BPA and BPS, the two most common circulating bisphenols in humans. Both BPA and BPS models were calibrated against and accurately simulated the toxicokinetic plots of total, unconjugated, and conjugated bisphenols from three published toxicokinetic datasets for bisphenols in pregnant, mid-late term, sheep (Corbel et al. 2013; Gingrich et al. 2019; Grandin et al. 2018). Additionally, by extrapolating the dosing regimen to a repeated dosing model, we were able to simulate the potential effects of daily intake on the steady state of bisphenol concentrations in both the mother and the fetus.

All partition coefficients used for the BPA and BPS models fall within the ranges of those published for bisphenols (Karrer et al. 2018; Kawamoto et al. 2007), or were fit to these parameters within one order of magnitude of the original parameter. However, given the lack of available tissue BPA or BPS concentration data as part of the toxicokinetic datasets used, other than maternal and fetal plasma, our models did not strictly satisfy the property of identifiability, without introducing stringent constraints on most of the parameters (Slob et al. 1997). The partition coefficient for the rest of body for both BPA and BPS models was derived by calibration to the experimental data within the ranges of values available for all other partition coefficients. Additionally, this procedure accounts for any uncertainty in the values of other partition coefficients. Variability and uncertainty of partition coefficients are not usually considered in the development of PBTK models. This omission could represent a significant pitfall for most PBTK models, where partition coefficients are chain-cited across publications (Csanady et al. 2002; Mielke et al. 2011; Yang et al. 2015) without noted consideration for the level of uncertainty associated with the experimental or computational derivation of those parameters. For this reason, we have chosen to calibrate partition coefficients within ranges of up to one order of magnitude around reported experimentally or computationally derived partition coefficients (Karrer et al. 2018; Kawamoto et al. 2007). Despite this, the nominal values used for most parameters in the BPA model were equivalent to those found in the literature (Kawamoto et al. 2007).

Both BPA and BPS models were developed by modifying a previously published p-PBTK model for BPA in mice as an initial scaffold (Kawamoto et al. 2007). However, during our model parameterization, the single-compartment fetal model used by Kawamoto et al. (Kawamoto et al. 2007) was unable to produce a meaningful fit. Therefore, we expanded the complexity of the fetal sub-model to include liver and rest of body compartments, as well as fetal liver metabolism and deconjugation. Additionally, maternal brain, uterus, mammary tissue, and well- and poorly-perfused tissue compartments were combined, as separating these tissues did not contribute to increasing the goodness of the fit. Merging of tissues allowed for higher emphasis on important toxicokinetic parameters unique to pregnancy, such as placental chemical transport and fetal metabolism. Dosing strategy was changed from oral gavage, as in Kawamoto et al. (Kawamoto et al. 2007), to subcutaneous and intravenous in accordance with the available calibration datasets. Despite the use of high quality, robust datasets for calibration (Corbel et al. 2013; Gingrich et al. 2019; Grandin et al. 2018), the data in the current model did not have sufficient time resolution to provide information about the process of enterohepatic recirculation, which was therefore excluded from the models. Although there is evidence for enterohepatic recirculation of BPA in rodents (Kawamoto et al. 2007; Kurebayashi et al. 2003; Tominaga et al. 2006), the evidence of this process occurring in large mammals, like sheep, is lacking (Corbel et al. 2013; Gingrich et al. 2019; Grandin et al. 2018). Importantly, there are reports of little to no enterohepatic recirculation of BPA in humans (Teeguarden et al. 2005; Thayer et al. 2015) and non-human primates (Tominaga et al. 2006).

A recent human p-PBTK model for BPA (Sharma et al. 2018), which lacks in vivo data calibration and only utilizes end-point human biomonitoring data, includes an amnion compartment validated against early and late pregnancy amniotic fluid biomonitoring BPA concentrations in humans (Ikezuki et al. 2002). Compared to the fetus, both BPA and BPS are found at a lower concentration in amniotic fluid (Gingrich et al. 2019), and are present only as minor fractions of the initial doses in experimental studies (Doerge et al. 2011; Gingrich et al. 2019; VandeVoort et al. 2016; Vom Saal et al. 2014). This observation, coupled with the limited data available from the amniotic compartment, led us to exclude it from the current model to reduce complexity without losing valuable information. Sharma et al. (Sharma et al. 2018) additionally included a placental deconjugation process. Placental conjugation and deconjugation of BPA have been experimentally reported to occur in balance (Corbel et al. 2015), likely making this a negligible process for p-PBTK modeling of bisphenols.

All in vivo generated toxicokinetic data used for calibration of the BPA model fit within one standard deviation of each of the simulated data points individually for both maternal and fetal compartments. This was true also of the BPS model, except for fetal unconjugated and conjugated BPS plasma toxicokinetic plots. These mismatches likely occurred because of limitations in dataset #3 (Grandin et al. 2018). Grandin et al. (Grandin et al. 2018) administered a single, high-dose IV bolus of BPS-d8 to all fetuses without scaling to fetal body weight which is inherently variable, and did not report twin pregnancies, which are common in sheep and could lead to more variable toxicokinetic data. Additionally, the plasma concentrations were sampled at irregularly spaced intervals, and the first data point is recorded after a large portion of the dose had either been metabolized and distributed into fetal tissues or transferred through the placenta back into maternal circulation. By adjusting our model code to account for a smaller fetus, which could also indirectly account for smaller loading dose if initial metabolism or distribution were ignored, we were able to achieve a better simulated fit to these data (Supplemental Figure 3).

Simulated tissue concentrations of BPA and BPS were determined by only a limited subset of model parameters. For example, the simulated toxicokinetic profile of conjugated BPA in fetal plasma following administration of conjugated BPA to the fetus is fully determined by the conjugated BPA partition coefficients in the fetus. For this model, three experimental plasma datasets were available for calibration. No dataset was individually able to fully determine unknown model parameters, so all datasets were collectively used for calibration. This strategy effectively omitted model validation in favor of a more robust model calibration. Additionally, previous studies reported low values for the BPA fraction unbound in plasma (Csanady et al. 2002). When constraining the BPA fraction unbound to a low value, we were unable to produce a well-calibrated BPA model. The only values of BPA fraction unbound that produced a model with meaningful correspondence to the experimentally obtained data were in the range of 90% to 100%. The discrepancy in the percentage of unbound BPA could potentially be explained by the low dose of BPA used in the original plasma protein binding study (Csanady et al. 2002) compared with the higher doses of BPA used in the calibration studies (Corbel et al. 2013; Gingrich et al. 2019), which was more likely to result in saturation of plasma protein binding capacity. This is also the case for emerging bisphenol, BPS.

There, additionally, exist a breadth of data on differences in toxicokinetics and toxicological exposure outcomes between BPA and BPS, including on adipose cell differentiation (Pu et al. 2017a), muscle cell differentiation (Jing et al. 2019), and placental development (Gingrich et al. 2018). A toxicokinetic study used in the development of both models shows that BPA and BPS toxicokinetics differ significantly in both maternal and fetal circulation (Gingrich et al. 2019). Importantly, the current p-PBPK models add to this knowledge base by suggesting, through parameterization, that BPS has a potentially higher subcutaneous bioavailablity than BPA. This is similar to the higher oral bioavailability of BPS than BPA observed in pigs (Gayrard et al. 2019). The differences in toxicological behavior between these two bisphenol analogs warrant future studies into the safe use of BPS.

One overarching goal of PBTK model development is to create human exposure predictions for risk assessment that are constrained by relevant physiological and biochemical parameters. Both BPA and BPS models accurately recapitulate BPA and BPS calibration data, which demonstrate chemical accumulation in the fetal compartment; the majority of which is simulated as the bisphenol conjugate for both BPA and BPS. Our models address the six major model evaluation criteria for applicability in risk assessment as proposed by the World Health Organization (WHO et al. 2010): purpose and scope, structure and mathematical representation, computer implementation, parametrization, performance, and model documentation. These criteria were reviewed by (Lautz et al. 2019) who provided four recommendations for farm animal models to be applied in risk assessment that we have included in this work: 1) inclusion of a generic and flexible model structure, 2) implementation in an open source environment and publicly available model and code, 3) use of input data independent on calibration or otherwise estimated data using established techniques, and 4) inclusion of tools to assess performance, e.g. sensitivity analysis. Even though our models represent the first step towards addressing the possibility of human risk assessment, they require additional refinement before they could be extrapolated in humans. Additional data on human exposures during pregnancy, bisphenol physicochemical characteristics, and biochemical parameters for human tissues, as well as adjustments due to differences in placental physiology between humans and sheep would be necessary. The collection of these data should be the focus of future studies.

Global sensitivity analysis of the fetal compartment beyond the expected results showed that the rest of body partition coefficient for unconjugated bisphenols (PR_{_f}) exhibits very little contribution to output variance in determining both BPA and BPS fetal plasma kinetics. This effect presumably arises due to quick fetal metabolism and efflux of the unconjugated bisphenol forms. Additionally, the importance of PR_{_f}^(c) in determining fetal plasma kinetics of conjugated bisphenols decreases over time, most likely because the deconjugation process ramps up and the conjugated compounds are cleared from the fetal tissues.

Importantly, when considering extrapolation to daily exposure patterns in sheep, we noted that the accumulation of bisphenols in the fetal compartment has been observed in humans (Pan et al. 2020), with glucuronide-conjugates being the predominant form detected (Andra et al. 2016). Using the U.S. Environmental Protection Agency’s reference dose for BPA (50 μg/kg/day), we simulated repeated maternal dosing over a two-week period for both BPA and BPS in sheep, to evaluate fetal plasma chemical burden. Here, our simulations predicted that a pseudo steady-state of 0.28 ng/ml unconjugated BPA would be reached, which falls within the range of detection for unconjugated BPA (0-53 ng/ml) in cord blood (Veiga-Lopez et al. 2018). For BPS, our simulations predict that a pseudo steady state of 0.45 ng/ml unconjugated BPS would be reached. Since biomonitoring of unconjugated BPS has not been reported for cord blood, a direct comparison to human exposure cannot be made. However, the simulated total BPS in the fetal compartment (12.5 ng/ml on day 14) is in excess of total BPS concentrations measured in cord blood (<0.03-0.12 ng/ml total BPS) from a Chinese cohort (Liu et al. 2017). Most of the BPS accumulated in the fetus is predicted to be in the form of BPS conjugated metabolites. Although these metabolites are generally considered non-bioactive, BPA-G has been shown to be bioactive, and has adipogenic potential in vitro (Boucher et al. 2015). Given the predicted accumulation potential of BPS-G in the fetal compartment, the bioactivity of BPS metabolites like BPS-G should be further examined. Our simulations also demonstrate that, given a steady maternal intake of BPA, unconjugated BPA quickly reaches a state where it no longer accumulates in fetal blood. Unconjugated BPS, on the other hand, continues to accumulate in fetal blood even after 14 days of daily administrations. These results highlight the need to further study the precise fetal toxicokinetics of BPS, as well as the fetal accumulation potential of other BPA analogs.

Despite exhibiting model stimulations of comparable values to human bisphenol exposures, these models were derived from sheep data in which too much uncertainty exists to accurately extrapolate these data for human risk assessment. The uncertainties likely relate to species differences in placenta type and cellularity and capacity to conjugate and deconjugate bisphenols. Nevertheless, similarities in other important physiological parameters like body weight and cardiac output between sheep and human make it an excellent animal model for capturing high quality longitudinal toxicokinetic data for use in PBTK model development.

To improve the development of future p-PBTK models we recommend, on the basis of post hoc analyses, the inclusion of additional information from toxicokinetic studies, like total urine and fecal volume, and concentration of unconjugated and conjugated bisphenols in urine and feces. Importantly, these samples should be taken at least for two time points, ideally between 6 and 24 h after administration. Additional relevant information, like the role of protein transporters in the uptake and/or clearance of both parent compounds and metabolites, an understudied area, will likely be necessary knowledge for further accuracy in extrapolation to humans.

Conclusions

This study presents novel p-PBTK models for BPA, and the second most common bisphenol in humans, BPS. We have demonstrated how these models can be used to evaluate, in silico, maternal and fetal bisphenol exposures during pregnancy. These models represent unique tools for experimental design and risk assessment for BPA and BPS during pregnancy.

Highlights.

• Two open-source pregnancy PBTK models for BPS and BPA were successfully developed

• A 2-week daily maternal exposure simulation with the EPA RfD was performed

• BPS, but not BPA, accumulates in the fetus after a 2-week maternal exposure

• Both BPA and BPS PBTK models revealed partitioning differences in mother and fetus

• These PBTK models revealed differences in feto-maternal kinetics between BPA and BPS

Abbreviations

A

blood

ADME

absorption, distribution, metabolism, and excretion

BPA

bisphenol A

BPA_conj

bisphenol A conjugate

BPA-G

bisphenol A-glucuronide

BPS

bisphenol S

BPA_conj

bisphenol S conjugate

BPS-G

bisphenol S-glucuronide

BW

body weight

C_A

chemical concentration in arterial blood

C_VF

chemical concentration in the venous blood exiting the fat.

C_VK

chemical concentration in the venous blood exiting the kidney.

C_VL

chemical concentration in the venous blood exiting the liver

C_VR

chemical concentration in the venous blood exiting the rest of body (maternal)

C_VPL

chemical concentration in the venous blood exiting the placenta

C_{VR_f}

chemical concentration in the venous blood exiting the rest of body (fetal)

EDC

endocrine disrupting chemical

F

fat

F_SC

bioavailability

GD

gestational day

K

kidney

K_a

subcutaneous first order absorption constant

KEL_R

renal excretion constant

KEL_L

biliary excretion

K_m

Michaelis-Menten constant

kt₁

placental to fetal blood diffusion

kt₂

fetal to placental blood diffusion

L

liver

p-PBTK

pregnancy-specific physiologically-based toxicokinetic

PBTK

physiologically-based toxicokinetic

P_F

blood/tissue partition coefficient for fat

P_K

blood/tissue partition coefficient for kidney

P_L

blood/tissue partition coefficient for liver

P_R

blood/tissue partition coefficient for rest of body (maternal)

P_PL

blood/tissue partition coefficient for placenta

P_{R_f}

blood/tissue partition coefficient for rest of body (fetal)

Q_CC

total cardiac output

Q_FC

fractional blood flow to the fat

Q_KC

fractional blood flow to the kidney

Q_LC

fractional blood flow to the liver

Q_PLC

fractional blood flow to the placenta

QSAR

quantitative structure-activity relationships

R

rest of body

RfD

reference dose

T

tissue

V_BC

fractional volume of blood

V_EFC

fractional volume of fetus

V_FC

fractional volume of fat

V_KC

fractional volume of kidney

V_LC

fractional volume of liver

V_max

maximum rate of enzymatic reaction

V_PLEFC

fractional volume of feto-placental unit