A PBPK model describing the pharmacokinetics of γ-HBCD exposure in mice

Abstract

The brominated flame retardant, hexabromocyclododecane (HBCD), is added—but not bound—to consumer products and is eventually found in the environment and human tissues. Commercial-grade HBCD mixtures contain three major stereoisomers, alpha (α), beta (β), and gamma (γ), that are typically at a ratio of 12%:6%:82%, respectively. Although HBCD is widely used, the toxicological effects from its exposure in humans are not clearly understood. Using a physiologically based pharmacokinetic (PBPK) model could help improve our understanding of the toxicity of HBCD. The aim of this work was to develop a PBPK model, consisting of five permeability limited compartments (i.e., brain, liver, adipose tissue, blood, and rest of the body), to evaluate the pharmacokinetics of γ-HBCD in C57BL/6 mice. Physiological parameters related to body size, organ weights, and blood flow were taken from the literature. All partition coefficients were calculated based on the log K_ow. The elimination in urine and feces was optimized to reflect the percent dose eliminated, as published in the literature. Compared with data from the literature for brain, liver, blood, and adipose tissue, the model simulations accurately described the mouse data set within 1.5-fold of the data points. Also, two examples showing the utility of the PBPK model supplement the information regarding the internal dose that caused the health effects observed during these studies. Although this version of the PBPK model expressly describes γ-HBCD, more efforts are needed to clarify and improve the model to discriminate between the α, β, and γ stereoisomers.

1. Introduction

Hexabromocyclododecane (HBCD) is a major brominated flame retardant that has been widely used for more than 50 years in expanded and extruded polystyrene foams and many other manufactured products (Alaee et al., 2003; Sjodin et al., 2003; Kajiwara et al., 2009). The global annual production of HBCD reached 16,700 tons in 2001 and rose to nearly 22,000 tons in 2006 (Birnbaum and Staskal, 2004; ECHA, 2009a), with the European market accounting for the largest portion of the global demand (ECHA, 2009a; ECHA, 2009b).

Due to the wide use of HBCD, as well as its high degree of lipophilicity (log K_ow = 5.6) (Liagkouridis et al., 2015) and stability, HBCDs are found in humans and wildlife (Kabir et al., 2015), occur throughout the environment and in biota, and are persistent organic pollutants (POPs) that bioaccumulate (Marvin et al., 2011; Koch et al., 2015). Technical mixtures of HBCD are dominated by three stereoisomers, alpha (α), beta (β), and gamma (γ), each present as a pair of enantiomers, with relative amounts of 1–12% for α-HBCD, 10–13% for β-HBCD, and 75–89% for γ-HBCD (Heeb et al., 2007; Wu et al., 2012). Humans are exposed to HBCD through the diet, occupational exposure, and ingestion of indoor dust (Covaci et al., 2006; Abdallah and Harrad, 2009; Schecter et al., 2012). For adults, the largest contributor of HBCD exposure is the diet (72%); for toddlers, it is dust (75%) (EFSA, 2011a; EFSA, 2011b). The daily intake of HBCD is estimated to be 400 to 1500 ng/day for toddlers compared with 130 to 330 ng/day for adults (UK-Agency, 2006; Abdallah et al., 2008).

HBCD is a neurotoxicant in adult and neonatal rodents. One study found that developmental exposure to technical-grade HBCD alters learning and memory in mice (Eriksson et al., 2006). Also, a developmental toxicity study of α-HBCD observed lower body weight (BW) of rat pups, some neuromotor developmental delays, a reduction in anogenital distance, and an impairment of sexual copulatory behavior in male rats, suggesting that α-HBCD is an endocrine disruptor (Eriksson et al., 2002; Birnbaum and Staskal, 2004; Olivier et al., 2016). In a 28-day rat toxicity study, the lowest observed effect level was 125 mg of the HBCD/kg of BW based on increased liver weight; however, the study did not address endocrine toxicity (Chengelis, 1997). In humans, the Comparison of Exposure-Effect Pathways to Improve the Assessment of Human Health Risks of Complex Environmental Mixtures of Organohalogens (COMPARE) study showed that early life exposure to HBCD is linked with effects on coordination and verbal intelligence in children (Roze et al., 2009).

Because HBCD is ubiquitous in the environment, it is crucial to understand its toxicokinetics in mammals. In humans, HBCD is detected in blood, breast milk (European Chemicals, 2008), hair (Barghi et al., 2018), fetal liver tissue, the placenta, and adipose tissue (Rawn et al., 2014). Unfortunately, only a few pharmacokinetic studies are available that focus on HBCD in humans. In contrast, a series of studies has examined the pharmacokinetics of HBCD in mice (Szabo et al., 2010; Szabo et al., 2011b; Sanders et al., 2013; Hakk, 2016) and one in rats (Hakk, 2016). In general, β- and γ-HBCD are more extensively metabolized, have shorter half-lives, and has less potential to accumulate than α-HBCD. Indeed, in the commercial mix, γ-HBCD is the isomer with the highest concentration compared with α-HBCD. γ-HBCD is stereoisomerized to α-HBCD in vivo; with the remainder of γ-HBCD metabolized and excreted. To improve our knowledge about how chemicals adversely affect humans, we created physiologically based pharmacokinetic (PBPK) model (Emond et al., 2019). The aim of this current work was to further develop a PBPK model for γ-HBCD in mice based on the limited data in the literature. This work provides the opportunity to generate a hypothesis to understand the biology influencing the distribution of γ-HBCD in mammals, particularly in the brain, which has been identified as a major target organ.

2. Materials and methods

2.1. PBPK model structure and physiological parameters

The intrinsic physicochemical and biological properties of chemicals such as HBCD are responsible for a large portion of their kinetic behavior in vivo. For persistent and lipophilic xenobiotics such as HBCD, the high log K_ow of 5.6 reflects the high lipid affinity and bioaccumulation in adipose tissue, which reduces the availability of chemicals for distribution and elimination. This assumption is valid if serum protein binding, organ sequestration, and binding to lipoprotein in the extracellular compartment or in the lymphatic vessels are of limited influence on the pharmacokinetic properties of a chemical (Andersen, 2003; Emond et al., 2004; Jackson et al., 2017). The data supporting the PBPK model for HBCD is from a series of publications from our laboratory (Szabo et al., 2010; Szabo et al., 2011a; Szabo et al., 2011b; Sanders et al., 2013; Szabo et al., 2017) that examined the pharmacokinetics of α-, β-, and γ-HBCD in mice and identified the important properties that must be described to study the pharmacokinetics. Since our presentation of the first version of the PBPK model at SOT’s Annual Meeting and ToxExpo in 2019 (Emond et al., 2019), we improved the model to consider the impact of the deep fat fraction (also called diffuse adipose tissue) storing HBCD more strongly and another pseudo compartment that describes extracellular and lymphatic circulation as contributing to the storage hypothesis for this chemical. The deep fat pseudo compartment was previously described for D4 and D5 chemicals (Reddy et al., 2003). The ELPLC pseudo compartment describes extracellular and lymphatic circulation as contributing to the storage hypothesis for this chemical.

This mouse PBPK model consists of five compartments: liver, adipose tissue, blood, brain, and rest of the body (Fig. 1). All compartments were described as permeability limited because of the high log K_ow and low adipose tissue concentration of γ-HBCD. Liver is described because it is a major metabolic compartment and accumulates γ-HBCD. Adipose tissue is included because the high log K_ow of γ-HBCD suggests that it should concentrate in adipose tissue. Brain is a compartment because it is a sensitive target organ in rodents and humans (Lilienthal et al., 2009; Ibhazehiebo et al., 2011). Blood is included because of the protein binding and distribution role of the systemic circulation. The rest of the body compartment, which represents 69% of the body, corresponds to organs and tissues not explicitly described but are required for mass balance. From this volume fraction, approximately 36% consists of muscles and 33% represents tissues or specific organs not described as an individual compartment in the model (e.g., skin, kidney, heart, lungs, spleen, gastrointestinal tract [GIT]).

Fig. 1.

Conceptual representation of the PBPK model to describe γ-HBCD exposure in mice.

The ELPLC pseudo compartment is not perfused by blood, but by lymphatic circulation (Trevaskis et al., 2005; Hjelmborg et al., 2008; Offman et al., 2016). This subcompartment has two inputs and two outputs: one from the blood systemic circulation and the other from the gastrointestinal lumen. The ELPLC subcompartment stores γ-HBCD bound to lipoprotein structures such as high-density lipoprotein (HDL), low-density lipoprotein (LDL), or chylomicrons not specifically describe in this PBPK model. The γ-HBCD that leaves the ELPLC and enters the systemic circulation or the liver compartment takes a similar path as cholesterol, with reverse circulation involving the HDL and LDL lipoprotein circuit (Ouimet et al., 2019; Eckardstein, 2020).

In mice, γ-HBCD is sequestered in the blood and liver by proteins or lipoproteins, but γ-HBCD is also sequestered in deep adipose tissue based on its lipophilicity, as previously shown for D4 and D5 (Reddy et al., 2003; Sarangapani et al., 2003; McMullin et al., 2015). There is some evidence that a percentage of adipose tissue can strongly sequester contaminants (Yu et al., 2011; Barrett, 2013); therefore, the deep fat subcompartment was described inside the adipose tissue compartment. This current structure of the model assumes that a significant amount of γ-HBCD is accumulated in such adipose tissue but was not measured because of the collecting method, which only samples perirenal fat.

It is important to note that the model described herein is for lifetime (from neonate to 2 years in rodent) exposures and describes the changing physiological parameters from birth through adulthood. The volumes and blood flows of each compartment were calculated using published polynomial equations (Luecke et al., 2007). In addition, each sex (male and female) was evaluated separately, depending on the experimental data available. The partition coefficients were calculated based on literature (Lukacova et al., 2009; GastroPlus, 2018); however, optimization was required based on the experimental data (Szabo et al., 2010) and the binding interactions affecting partition coefficients. Oral absorption is described with a first order parameter. Fecal and urinary elimination are described as first order processes.

Several important observations were made regarding Szabo’s disposition study to support our hypothesis (Szabo et al., 2010). First, it appears that γ-HBCD is sequestered more in the liver than in the adipose tissue. Therefore, for this current work, in addition to the partition coefficient, a non-specific protein binding (either albumin or lipoprotein) was incorporated in blood (as fraction blood binding) because blood concentrations are barely lower than adipose tissue, suggesting that protein binding traps HBCD (Fig. 1). This binding in blood is supported by a low concentration of γ-HBCD in adipose tissue compared with blood. The ratio of fat to blood was 1.1, 1.6, 2.3, and 2.9 at 3, 10, 30, and 100 mg/kg, respectively, for 96 h after the initial exposure (Szabo et al., 2010). While the liver sequesters available γ-HBCD, the results of the study showed no evidence of dose-dependent induction of a binding protein in the liver (Szabo et al., 2010). The results of the study also showed that the total of the percent dose accounted for in the major tissues and the total γ-HBCD excreted in urine or feces corresponds to approximately 80% of the dose (Szabo et al., 2010). Based on this finding, the model assumes that a large portion of the missing 20% of γ-HBCD is in the ELPLC regions.

2.2. Software, algorithms, model code, and statistics

All simulations and parameter fits were conducted by using acslX version 3.1.5.1 (AEgis Technologies, Huntsville, AL). The Gear algorithm was used to integrate double-precision variables. Parameter fitting was driven by using the relative-error model estimation. Maximization of the logarithm likelihood function was the decisive factor for fitting. A limitation of using this approach was that although maximum likelihood estimates may be reasonable, error estimates will be incorrect, resulting from the assumption of independent data.

2.3. Sensitivity analysis

A sensitivity analysis (SA) identifies the ways in which a model response changes under the influence of individual parameter changes (Easterling et al., 2000), and the results are expressed as a magnitude of change in the endpoint of interest (Krishnan and Andersen, 2008). Each parameter in the PBPK model was tested individually for sensitivity by using an exposure scenario of a single oral dose of 3 mg of HBCD/kg of BW, and the brain area under the curve (AUC) was determined at 96 h post-exposure. For the SA test, each parameter at the time was varied by ±2% expressed in percentage optimized value to determine the influence of small changes in the parameters. This SA was used to record the influence of individual parameter-driven variations in the AUC in the brain compartment based on Eq. 1, as follows:

$$SA\%=\frac{AU{C}_{\pm 2\%}-AU{C}_{\mathit{\text{optimized}}}}{AU{C}_{\mathit{\text{optimized}}}}\times 100\%$$ (1)

An independent t-test with equal variance was performed to compare the experimental γ-HBCD concentration in compartments (i.e., blood, liver, adipose tissue, and brain) to the single-dose concentrations simulated for 3, 10, 30, and 100 mg/kg, followed by 96 h post-exposure.

2.4. Model evaluation

A major challenge is that there is a lack of pharmacokinetic studies about γ-HBCD, and the ones that are available are all from the same laboratory (Szabo et al., 2010; Szabo et al., 2011b; Sanders et al., 2013). For the current work, the PBPK model was optimized by using the single oral dose of 3 mg of γ-HBCD/kg of BW with a time point of 4 days (96 h) post-exposure.

The model was then assessed with the single oral doses of 10, 30, and 100 mg of γ-HBCD/kg of BW for a single time point of 4 days (96 h), and with three repeated doses once per day for 10 consecutive days. The concentrations were measured 4 days (96 h) later. The single dose of 3 mg of γ-HBCD/kg of BW for which eight sequential time points were available (i.e., 1, 3, and 8 h and 1, 2, 4, 7, and 14 days) was also evaluated (Szabo et al., 2010). In addition, an intravenous dose of 3 mg of γ-HBCD/kg of BW was also compared, in which the γ-HBCD tissues concentrations were measured after 4 days post-exposure. Szabo et al. (2010) demonstrated that γ-HBCD is eliminated more slowly in developing mice compared with adult mice. The PBPK model was used to identify parameters that differ between the age groups to better understand the mechanism of the age-dependent pharmacokinetics of γ-HBCD (Szabo et al., 2010; Szabo et al., 2011b).

2.5. Model applications

Several studies in the literature focused on the neurotoxic effects of HBCD in rats (Chengelis, 1997; Ibhazehiebo et al., 2011) and mice (Genskow et al., 2015; Pham-Lake et al., 2017; Rasinger et al., 2018; Reffatto et al., 2018). Male mice exposed daily to 25 mg/kg of BW/day for 30 days and terminated 24 h after the last dose (Genskow et al., 2015) or for 6 weeks and terminated 6 weeks after the last dose (Pham-Lake et al., 2017) displayed alterations in dopaminergic circuitry in the striatum and the mesohippocampus, respectively. Using this dose and dosing paradigm, the PBPK model was used to estimate peak blood and brain concentrations associated with these dopaminergic circuitry changes. Genskow et al. (2015) also evaluated the effects of HBCD on catecholaminergic SK-N-SH neuroblastoma cells and primary cultures of mesencephalic neurons from postnatal mice (Days 1–3). In the SK-N-SH cells, the no observed adverse effect level (NOAEL) for viability was 5 μM, and in the mesencephalic cells, the no observed effect level (NOEL) for cell loss and neurite length was 2.5 μM. Mariussen and Fonnum (2003) observed that HBCD inhibited dopamine uptake in brain synaptosomes with an IC50 (half inhibitory concentration) value of 4 μM. Genskow et al. (2015) and Mariussen and Fonnum (2003) were evaluated by using the PBPK model to estimate the oral equivalent dose to attain peak blood and brain concentrations equivalent to the reported NOELs and IC50s.

3. Results

3.1. PBPK model development and calibration

The model discussed herein consists of five compartments (i.e., liver, adipose tissue, blood, brain, and rest of the body) and two subcompartments (i.e., ELPLC and deep fat tissue). The ELPLC consists of extravascular lipoprotein and lymphatic vessels that come from lymphatic drainage of the GIT or from the blood. The deep fat subcompartment is described within the adipose tissue compartment. Subcompartments for the brain and liver were not included in this model as they were for a D4 and D5 study (Dobrev et al., 2008) because the brain and liver concentrations were predicted with the partition coefficients, which adequately describe γ-HBCD concentrations.

All of the physiological parameters used in our current work came from the literature (Table 1). We optimized values such as first order transfer into the deep fat from adipose tissue and from deep fat into the adipose tissue or the exchange in the ELPLC. We also optimized urinary and fecal elimination, the oral absorption parameter, and tissue permeability for brain, adipose tissue, liver, and blood based on the experimental data (Szabo et al., 2010). To initially parameterize the PBPK model, we used an exposure scenario that consisted of a single oral dose of 3 mg of γ-HBCD/kg of BW (Szabo et al., 2010) and measured tissue concentrations at 96 h (4 days) post-exposure.

Table 1.

Physiological parameters used in the PBPK model for mice (Brown et al., 1997; Emond et al., 2004; Luecke et al., 2007; USEPA, 2010).

Parameter name	Value
Body weight (BW) (g)	Variable
Cardiac output (Qcc) (ml/min/kg)	275
Tissue volume (fraction of BW)
Blood (Vb0)	0.080
Brain (Vbr0)	0.016
Liver (Vli0)	0.055
Adipose tissue (Vf0)	0.080
Rest of the body (Vre0)	0.679
Tissue blood volume (fraction of tissue)
Brain (Vbrb0)	0.030
Liver (Vlib0)	0.310
Adipose tissue (Vfb0)	0.020
Rest of the body (VreB0)	0.040
Tissue blood flow (fraction of Qc)
Brain (Qbr0)	0.033
Liver (Qli0)	0.161
Adipose tissue (Qf0)	0.070
Rest of the body (Qre0)	0.736
Tissue permeability (fraction of Qt)
Brain (PaBrF)	4.0 × 10−4
Liver (PaliF)	9.0 × 10−1
Adipose tissue (PafF)	2.0 × 10−3
Rest of the body (PaReF)	1.0 × 10−2
Apparent partition coefficient
Brain (Pbr:b)	1
Liver (Pli:b)	9.6
Adipose tissue (Pf:b)	45
Rest of the body (Pre:b)	19
Oral absorption parameter (Ka, h−1)	0.15
Gastric empty parameter (Kst, h−1)	0.11
Kmet (metabolism rate) (h−1)	0.01
MW (g/mol)	641.7
Percentage of γ-HBCD binding to blood protein (PB)	0.05
Urinary elimination (Cluri) (ml/h)	2.5
Absorption in portal circulation (A)	0.35
Hepatic binding
Binding capacity (LibMax) (nmol/ml)	5.0 × 10−2
Binding affinity (LibMax) (nmol/ml)	1.0 × 10−5
Extravascular lipoprotein
KLIPO_I (Cte in extracellular/lymphatic from the blood) (h−1)	1.0 × 10−4
KLIPO_O (Cte out from extracellular/lymphatic to the blood) (h−1)	4
KLIPO_OLI (Cte out from extracellular lymphatic to the liver (h−1)	1
KLIOUT (Cte out from tissue liver to extracellular/lymphatic) (h−1)	1
Deep fat compartment
KFDO (in the deep fat) (h−1)	1.0 × 10−7
KFDI (out from the deep fat) (h−1)	5.0 × 10−2
Fraction of deep compartment (unitless)	0.2

The in vivo study in mice (Szabo et al., 2010) showed that from 4 to 14 days post-exposure, the tissue concentrations of γ-HBCD were in a pseudo steady state and the γ-HBCD tissue to blood concentration ratio varied between 1 and 3, suggesting that an important (possibly between 15% to 20% of the percent dose) amount of γ-HBCD was distributed in the ELPLC or deep fat tissue subcompartments (Table 2).

Table 2.

The γ-HBCD concentration after a single dose of 3 mg of γ-HBCD/kg of BW. All data are mean ± standard deviation, represented as the concentration of nanograms per gram of tissue (ng/g) (Szabo et al., 2010).

Time	Liver (ng/g)	Blood (ng/g)	Adipose tissue (ng/g)	Brain (ng/g)
1 h	2309 ± 220.0	34 ± 8.5	15 ± 2.5	35 ± 5.2
3 h	1862 ± 86.0	113 ± 19.0	40 ± 8.0	101 ± 2.8
8 h	1335 ± 59.0	105 ± 8.7	96 ± 7.2	55 ± 10.0
1 day (24 h)	631 ± 26.0	65 ± 9.3	108 ± 11.0	29 ± 17.5
2 days (48 h)	233 ± 15.0	32 ± 3.1	29 ± 3.0	28 ± 3.7
4 days (96 h)	160 ± 15.0	24 ± 1.5	27 ± 5.4	24 ± 3.0
7 days (168 h)	139 ± 9.3	21 ± 2.4	20 ± 1.9	23 ± 1.7
14 days (336 h)	76 ± 2.4	16 ± 2.1	15 ± 2.0	22 ± 3.7

To optimize the PBPK model to the data, we had to either decrease the apparent partition coefficient for adipose tissue or find a biological mechanism to reduce access in this compartment. It is important to note that reducing the apparent partition coefficient still did not provide a good simulation of the tissue concentrations (data not shown). To correct for this poor fit, we described a deep adipose tissue subcompartment in the model and included an ELPLC subcompartment and protein binding in blood. In addition, we included hepatic protein binding in the model to obtain a better simulation of hepatic concentrations of γ-HBCD to increase the amount of HBCD in the liver compartment (Fig. 2).

Fig. 2.

Simulation of HBCD in tissues over time, following a single oral exposure of 3 mg/kg of BW in adult female C57BL/6 mice of (≈20 g) (Szabo et al., 2010). The points correspond to the measured tissue concentrations (Exp_), and the line simulation (Sim_) of each compartment for liver (LI), blood (B), adipose tissue (F), and brain (BR).

Another important observation of Szabo et al. (2010) was that fecal elimination of γ-HBCD accounted for approximately 50% and urine for approximately 30% of the elimination during the first 24 h. The PBPK model accurately reproduced fecal and urinary elimination of γ-HBCD (Fig. 3). Part of the fecal elimination of γ-HBCD was from unabsorbed γ-HBCD.

Fig. 3.

The cumulative elimination of γ-HBCD in feces simulated (Sim_Feces) and measured (Exp_Feces) and urine simulated (Sim_Urine) and measured (Exp_Urine) after a single oral exposure of 3 mg of γ-HBCD/kg of BW in mice for 14 days (336 h).

3.2. Model predictions compared with experimental data

The PBPK model was assessed by simulating measured concentration in the blood of female C57BL/6 mice exposed to single oral doses of 3, 10, 30, and 100 mg of γ-HBCD/kg of BW and measured 4 days (96 h) post-exposure (Szabo et al., 2010) (Table 3). The model accurately expressed the concentration (in ng/g) in tissues for all compartments (i.e., blood, brain, liver, and adipose tissue) (Fig. 4). A comparison of the linkage between experimental data and simulation data at all doses (3, 10, 30, 100 mg of γ-HBCD/kg of BW) for all tissues suggested linearity for each tissue with a determinant coefficient higher than (R² = 0.99). In addition, a statistical t-test method for equal variance comparing the experimental and simulations for the four compartments showed that the p value ranged between 0.267 and 0.435 (Fig. 4). These observations suggest that there was no protein induction for sequestration in hepatic tissue and blood protein binding at the experimental doses used. In addition, increasing the dose did not significantly increase the urinary and fecal elimination across doses expressed in percent doses.

Table 3.

Concentration expressed in ng/g of tissue 96 h after single exposure of doses of 3, 10, 30, and 100 mg of γ-HBCD/kg of BW. “Exp” corresponds to measured data and “Sim” to the simulated data obtained with the PBPK model.

Tissue	3 mg/kg		10 mg/kg		30 mg/kg		100 mg/kg
	Exp (±SD) (ng/g)	Sim (ng/g)	Exp (±SD) (ng/g)	Sim (ng/g)	Exp (±SD) (ng/g)	Sim (ng/g)	Exp (±SD) (ng/g)	Sim (ng/g)
Liver	160 ± 15	232	440 ± 90	770	2015 ± 213	2319	6811 ± 1233	7684
Blood	24 ± 1.5	28	90 ± 9.5	94	270 ± 26	283	867 ± 93	938
Adipose tissue	27 ± 5.4	42	146 ± 73	142	625 ± 262	425	2580 ± 890	1418
Brain	24 ± 3.0	34	73 ± 10	113	248 ± 28	338	808 ± 101	1129

Fig. 4.

The correlation of (a) liver, (b) blood, (c) adipose tissue, and (d) brain concentrations (measured and simulated) 4 days (96 h) post-exposure to a single oral dose of 3, 10, 30, and 100 mg of γ-HBCD/kg of BW expressed in ng/g of tissue. The results from the t-test comparing the match between simulated and measured data showed a range from p = 0.267 to 0.435, suggesting no significant differences between simulated and measured for equivalence of variances.

We simulated an exposure to a single oral dose of 3 mg of γ-HBCD/kg of BW for mice on Postnatal Day (PND)10 and a serial sample at different time points (Szabo et al., 2011b). The objective of this task was to identify which parameters influence the discordance between pups (on PND10) and adult mice (C57BL/6 strain) exposed to γ-HBCD (Szabo et al., 2010; Szabo et al., 2011b). As expected, the concentrations (expressed in percent doses) in the pups were higher than in the adults (Fig. 5a). However, significantly reducing the excretion clearance parameter and the metabolism fraction parameter improved the simulation for liver, but still significantly underestimated neonatal brain concentrations (Fig. 5b). These results suggest that other physiological or pharmacokinetic mechanisms not described in this model are required to capture the HBCD concentration in brain concentrations. The blood brain barrier is under developed in the neonates compared to the adult animals and this may explain the greater concentrations in the neonatal brain compared to the adult (Dobbing and Sands, 1979; Chuang et al., 2011; Saunders et al., 2012).

Fig. 5.

Pharmacokinetic profile of tissue concentrations expressed in percent dose over time (in h) after exposure to a single oral dose of 3 mg of γ-HBCD/kg of BW for mice on PND9. The graph represents the simulation of the liver (LI), blood (B), adipose tissue (F), and brain (BR). The dots represent the experimental data and the standard deviation (SD) obtained for the same tissues from literature (Szabo et al., 2011b). The graphics correspond to (a) the simulation using the adult parameters and (b) the simulation reducing the urinary elimination and metabolic elimination parameters.

3.3. Model evaluation and sensitivity analysis

We conducted an SA of the PBPK model for the 35 parameters, using the brain AUC as the dose metric. We assessed the parameters one at a time to ensure the most sensitive model. This analysis showed that the most sensitive parameters were the PBr (the partition coefficient for brain/blood) and A (fraction in the portal vein), with an influence on the AUC in the brain by less than 1.5%. Parameters such as Kabs (the absorption rate), Cluri (urinary elimination), Kst (unabsorbed rate parameter), PaBrF (permeability fraction in the brain), Pre (the partition coefficient of the rest of the body/blood), and QBrF (the fraction of blood flow in the brain) (see Fig. 6) influenced the variation in the AUC in the brain by more than 0.5% when we changed the parameters by ±2%. All other parameters had negligible impacts on the AUC of γ-HBCD in the brain when a 2% change was applied to the parameters.

Fig. 6.

Results of the percentage of sensitivity analysis (SA%) of parameters used in the PBPK model for γ-HBCD in mice. Regarding the SA%, only those percentages higher than 0.5% are presented in this graph, where A is the fraction of portal absorption from the GIT, CLURI is clearance urinary elimination, KABS is oral absorption, KST is the transit of unabsorbed γ-HBCD in the GIT, PABRF is the fraction of permeability coefficient from tissue blood to cellular matrix in the brain, PBR is the partition coefficient for brain/blood, and QBRF is the fraction of blood flow in the brain.

3.4. Examples of utilization of the PBPK model

3.4.1. Example 1 (Mariussen and Fonnum, 2003)

Mariussen and Fonnum (2003) concluded from in vitro studies that the alterations in dopaminergic circuitry IC50 was 4 μM of a commercial mixture of HBCD of which γ-HBCD would make up 78%, or approximately 3 μM, of this mixture. Using the PBPK model with a reverse dosimetry approach and Eq. 1 (to convert the mixture of HBCD to a γ-HBCD part in the in vitro media), the oral equivalent dose needed to attain blood concentrations equivalent to γ-HBCD to the media concentration of 3 μM (or 1926 ng/ml) was estimated, along with the estimated brain concentration simulated to 2.94 nmol/ml, considering a density of ≈1.

$$\frac{4\mu mol}{L}X0.785\times \frac{1L}{1000ml}=\frac{3\mathit{\text{nmol}}}{ml(org\mathit{\text{blood}})}=\frac{3\mathit{\text{nmol}}}{ml(org\mathit{\text{blood}})}\times \frac{642ng}{\mathit{\text{nmol}}}=\frac{1926ng}{ml(org\mathit{\text{blood}})}$$ (2)

The PBPK model estimated that a dose of 14.3 mg of γ-HBCD/kg/day is required to attain the corresponding blood concentrations used in the in vitro study. Because the brain and blood concentrations are near unity, this exposure would result in brain concentrations that are similar to the blood concentration. In fact, the oral simulation results in a blood concentration of 3 nmol/ml and a brain concentration of 3.01 nmol/g tissue (1926 ng/ml). Example 2: (Pham-Lake et al., 2017) Estimated blood and brain concentrations of HBCD from in vivo neurotoxicity studies in mice.

This in vivo study in mice exposed male and female C57BL/6 J to 25 mg of HBCD/kg of BW/day for 30 days and terminated 24 h after the last dose (Genskow et al., 2015) or 6 weeks, followed by 6 weeks of recovery (Pham-Lake et al., 2017). Genskow et al. (2015) and Pham-Lake et al. (2017) did not mention using a specific isomer; therefore, we assumed that the HBCD was a commercial mixture. Similar to the previous example in Section 3.5.1, γ-HBCD represents a mean of 78.5%, so the actual exposure to HBCD would be 19.6 mg of γ-HBCD/kg of BW. Using the PBPK model, we simulated the maximum blood and brain concentrations (C_max) reached after the last dose for each study. The concentrations corresponded to 6.0 μM of γ-HBCD for blood (3852 ng/ml) and 4.6 μM of γ-HBCD for brain tissue (2953 ng/g), with a 6-week post-exposure concentration of 0.015 μM γ-HBCD for blood (10.0 ng/g) and 0.016 μM γ-HBCD for brain tissue (10.27 ng/g) (Fig. 7). A repetitive exposure increased the C_max, which reached a plateau at approximately 15 days of daily exposures. The C_mean concentration was 4.3 nmol of γ-HBCD/g of brain tissue (2760 ng/g) at steady state; however, no information was available in the article to determine when the health effects appeared.

Fig. 7.

The pharmacokinetic profile after daily exposures of 19.6 mg of HBCD/kg of BW in mice for 6 weeks, followed by 6 weeks of recovery. HBCD concentration is expressed in nmol/ml of blood (black line) and brain tissue (gray line) concentration profiles over time expressed in days.

4. Discussion

The three diastereoisomers (i.e., α-, β, and γ-HBCD), which are part of commercial-grade HBCD, result in different pharmacokinetic profiles and serum metabolomic responses, suggesting different effects and mechanisms of action (Szabo et al., 2017). This observation supports the importance of modeling each stereoisomer separately before developing a mixture PBPK model. The objectives of this current study were to develop a PBPK model for γ-HBCD in mice and to use it to simulate the pharmacokinetics. This new PBPK model for γ-HBCD adequately describes the major tissue concentrations involved in the pharmacokinetics in mice and includes estimates of urinary and fecal elimination rates from early life exposures through adulthood to aid in the development of potential biomarkers of exposure.

In rats, γ-HBCD is rapidly absorbed (t_1/2 ≈ 2 h; absorption parameter Ka = 0.35 h) (KEMI, 2008). In blood, C_max is reached at 4 h post-exposure. Rats exposed to a single ¹⁴C-labeled oral dose of γ-HBCD ranging between 7 mg and 9 mg total HBCD/kg of BW in a mixture of acetone/oil resulted in the major tissue depots of a ¹⁴C-labeled dose of γ-HBCD being adipose tissue (20%), muscle (14%), and the liver (7%). At 72 h, 14% of γ-HBCD was still found in adipose tissue, 2% in muscle, and 0.3% in the liver (Yu and Atallah, 1980). These findings are in contrast with what has been observed in mice (Szabo et al., 2010). These observations can be due to species-dependent variations and/or by the rats’ doses being 3-times higher than in the studies of mice. (Yu and Atallah, 1980).

The first major observation from this current work and supported by our group’s previous publications in mice, suggests that the tissue distribution of γ-HBCD does not follow a classical pattern for neutral lipophilic chemicals having a high log K_ow. We expected that adipose tissue would be a major storage compartment. Nevertheless, as reported in mice (Szabo et al., 2010), the partition coefficient using the ratio of adipose tissue to blood concentration at 96 h post-exposure (pseudo-steady state) was approximately 1:3, compared with calculated partition coefficients based on the log K_ow of approximately 120 (Lukacova et al., 2009). In fact, the simulation, using partition coefficients based on the neutral lipid equivalent ratio or lipid content (Emond and Krishnan, 2006; Lukacova et al., 2009), did not predict the observed experimental data and overestimated the tissue concentration. The amount absorbed and distributed in compartments and the amount eliminated suggested that other sites of deposition might be implicated. Dobrev et al. (2008) described a similar situation for D4 and D5 and speculated that lipoprotein and deep fat could be responsible for the storage (Dobrev et al., 2003; Dobrev et al., 2008). These two chemicals are also lipophilic, but they behave differently than classical lipophilic chemicals that partition into the lipid compartment. Chlordecone and dichlorodiphenyltrichloroethane (DDT) are chemicals having similar behavior with lymphatic and lipoprotein sequestration (De Winne, 1979; Soine et al., 1982). Regarding γ-HBCD, the initial PBPK model was unable to predict γ-HBCD blood concentrations.; therefore, we added lipoprotein binding in blood to sequester a fraction of the γ-HBCD. However, this change could not account for all of the experimental observations. Consequently, we described trapping γ-HBCD in ELPLC subcompartment (Fig. 1). Part of the γ-HBCD sequestered in the ELPLC subcompartment comes from the lymphatic circulation during oral absorption, in which γ-HBCD tracks with the chylomicrons. The lymphatic circulation was previously described for 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) (Emond et al., 2005; Emond et al., 2006). We also added a deep adipose tissue subcompartment as was described in a previous publication (Reddy et al., 2003) (Fig. 1). The sum of the amount of chemicals for each major organ measured, including feces and urine, only resulted in ≈ 80% total dose, suggesting that 20% was not measured. Thus, it is reasonable to presume that other tissues or biological structures may be implicated.

In one study that mentioned the use of the deep fat, the researchers referred to the concept of visceral adipose tissue (VAT) and subcutaneous adipose tissue (SCAT), suggesting a possible difference in the accumulation of persistent organic pollutants (POPs) (Malarvannan et al., 2013). However, others reported that the levels and the patterns of POPs distribution in humans were not significantly different between VAT and SCAT (Jackson et al., 2017). It is possible that the lipid composition in the various adipose tissue stores is reflected in the differential POP solubility in various adipose depots (Malarvannan et al., 2013; Jackson et al., 2017). Comparing the VAT and SCAT, several authors pointed out dissimilarities in lipid content between these two types of adipose tissues (Alessi et al., 1997; Freedland, 2004; Pestana et al., 2014).Alternatively, depending on the type of adipose tissue and anatomical location, the blood flows might be different between the different adipose compartments(Frayn and Karpe, 2014). Observations using microdialysis suggested that blood flow varies depending on adipose tissue region or type (Karpe et al., 2002). It is thus reasonable to hypothesize that the blood perfusion of each type of adipose tissue is different and thus the sequestration or the retention of POPs might be different. We also not that in rodents, metabolically less active SCAT shows lower density blood vessels compared with VAT (Mrzilkova et al., 2020). The difference between these adipose tissues is age dependent; specifically, SCAT increases with age, whereas VAT decreases with age (Saely et al., 2012). Overall, the age, vascularization, metabolic activity, the characteristic of the tissue composition are important differences to consider when characterizing their storage and release of POPs. More effort is required to incorporate more physiologically accurate adipose tissue compartments into PBPK models. Better characterization and sample collection of different adipose tissue may help to understand and further improve PBPK models for lipophilic chemicals. However, we did not identify any literature that suggests making these changes specifically for HBCD.

Moreau and Nong published a PBPK model for α-HBCD, which is another stereoisomer of HBCD that is more stable in biological tissues, but is present at much lower percentage in technical mixtures of HBCD than γ-HBCD (Moreau and Nong, 2019). There are several differences between the two models for these different isomers based, in part, on their different pharmacokinetics. The α-HBCD model does not estimate the feces and urinary elimination of α-HBCD, so potential uncertainty can come from the distribution of the chemical into different compartments such a lipoprotein, deep fat, and liver, but this model offers an interesting possibility for understanding the pharmacokinetic behavior. In contrast, the present γ-HBCD model describes different pharmacokinetic behavior and explores different hypotheses such as hepatic protein binding and blood in the compartments. We also explore the brain as a target organ, which is not in the Moreau and Nong model. In addition, the γ-HBCD PBPK model does not exchange with subcompartment lipoprotein but directly with lymphatic distribution. This current mouse PBPK model for γ-HBCD is designed to support the next version of the model by including a mixture of HBCD lifetime exposure dose.

Accumulation of HBCD in the brain induced excitotoxic insults by dysregulation of tightly controlled homeostasis of calcium and zinc (Reffatto et al., 2018). This finding may be relevant for risk assessments of dietary POPs in children and teenagers because disruptions during brain development may be associated with neurological disorders in adults (Rasinger et al., 2014). HBCD also disrupts zinc-dependent calcium ion (Ca²⁺) signaling, resulting in a Ca²⁺ signal reduction, with 86% inhibition at only 1 μM of HBCD (Reffatto et al., 2018). These authors concluded that low concentrations of HBCD affect neural signaling in the brains of mice and act through the dysregulation of Ca²⁺ and zinc ion (Zn²⁺) homeostasis (Reffatto et al., 2018). Another study demonstrated that low doses of HBCD (10⁻¹⁰ M) significantly suppressed thyroid hormone (TH) (Ibhazehiebo et al., 2011) and demonstrated that TH induced dendrite arborization of Purkinje cells in primary cerebellar culture derived from newborn rats, suggesting that HBCD can interfere with TH action in target organs, including the developing brain (Ibhazehiebo et al., 2011). TH is also implicated in many processes of neural development, including neuronal migration in the cerebral cortex; cerebellar cell proliferation, migration, and apoptosis; and hippocampal and cortical synaptogenesis (Miller-Rhodes et al., 2014).

The PBPK model was applied in two different manners. The first was for in vitro to in vivo (IVIVE) extrapolation, where we assume the media concentrations in the in vitro assays are equivalent to serum or blood concentrations of a chemical at steady-state (Bell et al., 2018; Wambaugh et al., 2018). The application of IVIVE to the of Mariussen and Fonnum (2003) study and resulted in estimates of HBCD intake of approximately 14 mg/kg to attain an equivalent blood concentration to the IC50 of 4 μM HBCD. Both authors used a commercial mixture of HBCD without specifying the isomers, we assumed that 78% of the mixture was γ-HBCD. To understand how this estimate compares to in vivo studies, we then estimated the blood concentrations of HBCD from the neurotoxicity studies of Genskow et al. (2015) and Pham-Lake et al. (2017).

We used the PBPK model to predict the steady-state blood concentrations from these studies The in vivo studies involved a daily exposure of 25 mg of HBCD/kg of BW/day (estimated at 19.6 mg of γ-HBCD/kg of BW/d) for 6 weeks followed by 6 weeks of recovery, during which the authors observed several modifications in the brain, including alterations to the dopamine synapse 6 weeks later (Pham-Lake et al., 2017). Genskow et al. (2015) exposed 3 month-old male mice to 25 mg of HBCD/kg of BW for 30 days (Genskow et al., 2015). The researchers observed significant reductions in the expression of the striatal dopamine transporter and vesicular monoamine transporter 2. Both studies focused their strategies on the observed health effect but did not measure tissue brain concentration in either case. Using the PBPK model, we were able to estimate blood and concentrations (Fig. 7). The PBPK model estimated that HBCD will attain steady-state concentrations by 15 days of daily exposure. The PBPK model estimated steady-state blood concentrations at the last exposure of approximately 4–5 μM. It is of note that the IVIVE concentrations and the in vivo blood concentrations are similar. However, these comparisons must be viewed with caution, this analysis assumes similar magnitudes of effect and translating some of the in vitro effects to the in vivo effects is not straight forward. In addition, the in vitro disputation between media and cells may not be similar to that of the disposition between blood and brain tissue.

The SA performed for the PBPK model was based on the percentage of brain AUC change for ±2% of the parameters optimized after single exposure and following 96 h post exposure. Seven parameters had a percentage of 0.5% and higher. Three of them were linked to the brain compartment, which suggests the importance of thoroughly describing the physiology, and the four remaining were related to the absorption or elimination of γ-HBCD. Regarding the brain compartment, there was good confidence in the QBRF parameter (the fraction of blood flow in the brain) because it was relatively well documented in the literature (Brown et al., 1997; Luecke et al., 2007) (Fig. 6). Regarding the parameters of PBR (the partition coefficient for brain/blood) and PABRF (the permeability fraction in the brain), there is more uncertainty, especially for PABRF because there is some uncertainty related to diffusion in the brain compartment. The PBR parameter is usually driven by lipophilic characteristics of the chemicals, but in this case, more physiology seems to limit HBCD in crossing the blood–brain barrier. Regarding the parameters of A (fraction in the portal vein), Cluri (urinary elimination), Kabs (the absorption rate), and KST (unabsorbed rate parameter) (see the legend in Fig. 6), there is some confidence in these parameters because they were optimized based on the time-course data for elimination (Fig. 3) and blood concentrations (for more information, see Section 2.4, Optimization).

5. Conclusion

At the SOT Annual Meeting and ToxExpo in 2019, we described the first PBPK model for γ-HBCD in mice that represents 75% to 89% of the total HBCD in commercial mixtures (Emond et al., 2019). The γ-HBCD is eliminated (metabolized as different stereoisomers) relatively quickly in rats. The PBPK model for HBCD describes the major tissues involved in the pharmacokinetics of γ-HBCD, and it can be used in mice to describe early life exposures up through adulthood. Based on the findings of this current study, the PBPK model describes the data set for mice with good accuracy. Although this version of the PBPK model only describes γ-HBCD, more effort is required to clarify and improve the model to discriminate between the γ, α, and β isomers on the same model using our data (Szabo et al., 2010; Szabo et al., 2011a; Sanders et al., 2013). Therefore, future efforts are needed to describe the metabolic isomerization of γ-, β-, and α-HBCD as mixtures in the body by using this framework model code.

Abbreviations:

α

alpha

β

beta

γ

gamma

α-HBCD

alpha-hexabromocyclododecane

β-HBCD

beta-hexabromocyclododecane

γ-HBCD

gamma-hexabromocyclododecane

AUC

area under the curve

BW

body weight

Ca²⁺

calcium ion

COMPARE

Comparison of Exposure-Effect Pathways to Improve the Assessment of Human Health Risks of Complex Environmental Mixtures of Organohalogens

D4

octamethylcyclotetrasiloxane

D5

decamethylcyclopentasiloxane

DDT

dichlorodiphenyltrichloroethane

ELPLC

extracellular lipoprotein and lymphatic circulation

GIT

gastrointestinal tract

HBCD

1,2,5,6,9,10-hexabromocyclododecane

HDL

high-density lipoprotein

IC50

half maximal inhibitory concentration

LDL

low-density lipoprotein

log K_ow

log of octanol/water partition coefficient

NIH

National Institutes of Health

NOAEL

no observed adverse effect level

NOEL

no observed effect level

PBPK

physiologically based pharmacokinetic

PND

Postnatal Day

POP

persistent organic pollutant

SA

sensitivity analysis

SA%

percentage of sensitivity analysis

SCAT

subcutaneous adipose tissue

SD

standard deviation

T

thyroxine hormone

TCDD

2,3,7,8-tetrachlorodibenzo-p-dioxin

TH

thyroid hormone

VAT

visceral adipose tissue

Zn²⁺

zinc ion