当前位置: 首页 > 期刊 > 《毒物学科学杂志》 > 2005年第1期 > 正文
编号:11154582
PBPK Model for Radioactive Iodide and Perchlorate Kinetics and Perchlorate-Induced Inhibition of Iodide Uptake in Humans
http://www.100md.com 《毒物学科学杂志》
     ABSTRACT

    Detection of perchlorate () in several drinking water sources across the U.S. has lead to public concern over health effects from chronic low-level exposures. Perchlorate inhibits thyroid iodide (I–) uptake at the sodium (Na+)-iodide (I–) symporter (NIS), thereby disrupting the initial stage of thyroid hormone synthesis. A physiologically based pharmacokinetic (PBPK) model was developed to describe the kinetics and distribution of both radioactive I– and cold in healthy adult humans and simulates the subsequent inhibition of thyroid uptake of radioactive I– by . The model successfully predicts the measured levels of serum and urinary from drinking water exposures, ranging from 0.007 to 12 mg , as well as the subsequent inhibition of thyroid 131I– uptake. Thyroid iodine, as well as total, free, and protein-bound radioactive I– in serum from various tracer studies, are also successfully simulated. This model's parameters, in conjunction with corresponding model parameters established for the male, gestational, and lactating rat, can be used to estimate parameters in a pregnant or lactating human, that have not been or cannot be easily measured to extrapolate dose metrics and correlate observed effects in perchlorate toxicity studies to other human life stages. For example, by applying the adult male rat:adult human ratios of model parameters to those parameters established for the gestational and lactating rat, we can derive a reasonable estimate of corresponding parameters for a gestating or lactating human female. Although thyroid hormones and their regulatory feedback are not incorporated in the model structure, the model's successful prediction of free and bound radioactive I– and perchlorate's interaction with free radioactive I– provide a basis for extending the structure to address the complex hypothalamic-pituitary-thyroid feedback system. In this paper, bound radioactive I– refers to I– incorporated into thyroid hormones or iodinated proteins, which may or may not be bound to plasma proteins.

    Key Words: pharmacokinetics; human; perchlorate; radioactive iodide; inhibition; thyroid.

    INTRODUCTION

    Advances in detection sensitivity of ion chromatography have revealed widespread contamination of ground and drinking water with perchlorate () across the United States (Motzer, 2001; Urbansky and Schock, 1999; U.S. Environmental Protection Agency, 2003). The bulk of this contamination is associated with the use of ammonium perchlorate (NH4ClO4) as an oxidizing agent in missile and rocket fuel. Ammonium perchlorate is also used in pyrotechnics (fireworks) and air bag inflators. The salt is readily soluble in water, and its dissociation product, the perchlorate anion (), is very stable under environmental conditions and very mobile in most media (Motzer, 2001).

    Perchlorate is not metabolized in the body (Anbar et al., 1959; Yu et al., 2002). However, because has a similar hydrated ionic radius and carries the same charge as iodide (I–), it is able to affect biological systems by inhibiting I– uptake into the thyroid by the sodium-iodide symporter (NIS) (Anbar et al., 1959; Brown-Grant and Pethes, 1959). While it is known that competes with I– for NIS binding sites, whether is actually translocated into thyrocytes is the subject of debate (Eskandari et al., 1997; Riedel et al., 2001). The weight of evidence at this time, however, suggests is a competitive inhibitor of thyroid I– uptake, replacing I– as a substrate of NIS and crossing the basolateral membrane (Clewell et al., 2004, Van Sande et al., 2003). Reduced I– uptake may lead to a disturbance in the first stage of normal thyroid hormone genesis. Hence, there is reasonable concern that chronic exposure to low levels of in drinking water could induce thyroid hormone deficiencies and subsequent thyroid disorders.

    NIS resides in the basolateral membrane of thyroid epithelial cells and simultaneously transports two Na+ and one I– ion from extracellular fluid (plasma) into the thyroid epithelial cell (Spitzweg et al., 2000). NIS is expressed in the thyroid and other tissues including the GI tract, skin, mammary tissue, and placenta. However, only in the thyroid is I– organified to form thyroid hormones and iodinated proteins (Ajjan et al., 1998; Spitzweg et al., 1998). Thyroid hormone homeostasis is maintained through a complex feedback mechanism. A drop in circulating serum thyroid hormone levels signals the pituitary to produce more thyroid stimulating hormone (TSH), which in turn stimulates NIS expression.

    In rats, a decrease in free thyroxine (fT4) and subsequent increase in TSH occur quickly (within one day) after acute exposures (Wyngaarden et al., 1952; Yu et al., 2002). In humans, thyroid hormone conservation is more efficient, although thyroid hormone status is very dependant on the iodine status and life stage under consideration. Little or no significant change in T4, fT4, and TSH was seen in adults after 2 weeks of controlled exposure to via drinking water at 0.007 to 0.05 mg/kg/day (Greer et al., 2002) and 0.14 mg/kg/day (Lawrence et al., 2000), despite significant levels of thyroid I– uptake inhibition. However, significant drops in fT4, intrathyroidal iodine, as well as an increase in serum thyroglobulin (Tg) have been reported in humans after high levels of exposure (900 mg/day) for 4 weeks (Brabant et al., 1992). The dynamics of thyroid hormone homeostasis is very different for a late gestation fetus or neonate. Empirical measurement of intrathyroidal stores of thyroid hormone in human fetuses and neonates have shown that the amount of thyroid hormone stored in the colloid is less than that required for a single day (van den Hove et al., 1999). The extent of chronic low-level exposure required to cause significant hormone deficiencies in humans is not yet known. Thus, the question facing risk assessors and regulatory agencies is: what concentrations of perchlorate could be considered problematic? It is known that I– deficiency during the fetal and neonatal period affects physical and mental development (Laurberg et al., 2000; Porterfield, 1994). In the adult, effects of I– deficiency are less dramatic. Clinical and subclinical hypothyroidism is often overlooked due to the vague symptoms associated with the condition. Yet, hypothyroidism occurs in over 10% of older women and is associated with cognitive impairment (Volpato et al., 2002). The development of hyperthyroidism, especially in multinodular goiters with autonomous nodules, is also associated with long-term mild to moderate I– deficiency in adults. Hyperthyroidism is also often overlooked in the elderly and, if left untreated, may lead to cardiac arrhythmias, impaired cardiovascular reserves, osteoporosis, and other abnormalities (Laurberg et al., 2000). Therefore, perchlorate-induced I– deficiency may represent a public health concern not only during perinatal development, but also in the elderly and subpopulations with already compromised thyroid function.

    In order to better understand the effect of occupational and environmental exposure to perchlorate on the hypothalamus-pituitary-thyroid (HPT) axis, a few studies have been performed that directly correlate hormone changes to quantitative perchlorate doses. In two occupational health studies at U.S. production facilities, workers were exposed to ammonium perchlorate (NH4ClO4) dust in the air. Perchlorate exposure levels were estimated from monitoring breathing zone air over full work shifts (Gibbs et al., 1998; Lamm et al., 1999). Gibbs and coauthors categorized exposure groups based on job tasks and air monitoring results. Controls, selected from an associated plant, were not exposure free, but had exposures estimated to be several orders of magnitude below any of the ‘exposed’ groups. The researchers found no elevation in pre- and post-shift serum TSH, and no drop in serum free thyroxine (fT4) among any of the exposed workers. In the study by Lamm et al. (1999), control or ‘comparison group’ subjects worked at the same facility but at unrelated processes and were believed to have very low exposure to perchlorate-contaminated particulates. Daily perchlorate doses were estimated from breathing zone air monitoring of respirable particulates over full work shifts and by urinary measurements. No significant differences in triiodothyronine (T3) and T4 were reported between the exposure and comparison groups. However, the mean pre- and post-shift urine perchlorate measurements from the comparison group averaged, respectively, 64% and 22% of those from the lowest ‘exposed’ group.

    Ecological epidemiological studies on neonatal screening data from California, Nevada, and Arizona health departments have resulted in conflicting data. Studies by Lamm and Doemland (1999) and Li et al. (2000) showed no increase in incidence of congenital hypothyroidism or decrease in neonatal T4 associated with in drinking water up to 15 μg/l. In contrast, the studies of Schwartz (2001) and Brechner et al. (2000) both found effects on newborn thyroid hormones from exposures at similar environmental levels (1 to 15 μg/l). A retrospective study of school-age children and newborns in three Chilean cities with drinking water concentrations of <4, 5–7, and 100–120 μg revealed significantly higher fT4 but normal TSH in the two cities with highest concentrations (Crump et al., 2000), a result opposite to what might be expected. While dietary I– levels of the three Chilean populations were within normal range, their urinary excretion levels were increased. Hence, the increase in fT4 may represent an adaptive effect.

    These human epidemiological studies, while informative as to the specific populations in which they were performed, have been of limited utility in aiding the extrapolation across species, populations, or exposure scenarios. Furthermore, differences in route of exposure and lack of adequate adjustment for particle dosimetry (Gibbs et al., 1998; Lamm et al., 1999), ambiguities in level of exposure and exposure misclassification (Crump et al., 2000; Gibbs et al., 1998; Lamm et al., 1999) make these data sets of questionable use for predictive purposes (U.S. Environmental Protection Agency, 2003). Laboratory animal data, however, have indicated effects on developmental neurotoxicity, thyroid hormones, and thyroid histopathology that have raised concern for human health at various life stages (U.S. Environmental Protection Agency, 2003). Unfortunately, measurements of critical neuropsychological effects, such as IQ or physical development in children exposed to in drinking water, are not available. Therefore, to assist in evaluating the potential effects of in humans, a human PBPK model corresponding to those developed for the male, pregnant, and fetal rat (Clewell et al., 2003a,b; Merrill et al., 2003) is proposed. When used together, these models allow the entire body of literature (both animal and human) to be used to more effectively predict perchlorate-induced changes in thyroid I– uptake across species, life-stage, and exposure doses.

    The focus of this and ongoing modeling efforts with is to integrate animal and human data and to evaluate quantitatively the impact of chronic exposure to perchlorate-contaminated drinking water. This physiological model focuses on the first step in the process, perchlorate-induced inhibition of I– uptake in the thyroid. The model describes the kinetics and distribution of both radioactive I– and in healthy adult humans and simulates the subsequent inhibition of thyroid uptake of radioactive I– by . Distinct thyroid hormones and their regulatory feedback are not yet incorporated into the model structure, although free and organified I– (representing combined thyroid hormones and nonhormonal iodoproteins) in plasma and thyroid secretions are described separately. Hence, in addition to predicting perchlorate's ability to inhibit thyroid uptake of radioactive I–, the model represents significant work toward establishing a basis for quantifying the effect of on the total amount of circulating thyroid hormones.

    METHODS

    Supporting studies. As mentioned above, the proposed model was developed concurrently with other PBPK models, which describe several life stages in the rat, and shares several parameter values for which human data do not exist. However, whenever available, human data, obtained from Hays and Solomon (1965), Degroot et al. (1971), and Greer et al. (2002), were used for establishing I– and parameters. Hays and Solomon (1965) studied early radioactive I– kinetics in nine healthy males, who received an iv dose of 3.44 x 10–3 ng 131I–/kg. Frequent measurements of 131I– in the thyroid, aspirated gastric secretions, plasma, and cumulative urine samples were taken for 3 h post-dosing. To examine the effect of gastric uptake on I– kinetics, Hays and Solomon repeated these same measurements on the same subjects, without aspirating gastric juices. These data were used for establishing radioactive I– parameters describing early time-course behavior in the thyroid, stomach, urine, and plasma. Degroot et al. (1971) administered a tracer dose of 131I– to a health subject and measured thyroid 131I– uptake, total and protein-bound 131I– (PBI) in plasma, and urinary 131I– excretion for up to 11 days post-dosing. His data set further served in establishing I– parameters, describing thyroid production and secretion of incorporated radioactive I– into the systemic circulation.

    Data supporting the development of the portion of the model were obtained from Greer et al. (2002). In brief, Greer and colleagues administered 0.5, 0.1, or 0.02 mg/kg-day in drinking water for 14 days to twenty-four euthyroid subjects (n = 8, four males and four females at each level) in a ‘main study.’ Four equal portions of the daily dose were ingested approximately every 4 h from 8 A.M. to 8 P.M. Individual baseline serum and urine samples were collected 1 or 2 days prior to beginning the 14 days of dosing. During exposure, serum samples were collected at the following approximate times in the ‘main’ study: day 1 at 12 and 4 P.M., day 2 at 8 A.M., 12 and 5 P.M., day 3 at 9 A.M., day 4 at 8 A.M. and 12 P.M., day 8 between 8 and 9 A.M., day 14 at 8 A.M., 12 and 5 P.M. and on post-exposure days 1, 2, 3 and 14. Twenty-four-h urine collections were taken on exposure days 1, 2, 14, and post-exposure days 1 through 3. Eight- and 24-h thyroid radioactive I– uptake (RAIU) measurements were taken on the baseline day, exposure days 2 and 14, and post-exposure day 1, using an oral dose of 100 μCi of 123I–.

    An additional set of thirteen subjects were later exposed in a more limited ‘uptake’ study. Six women and one man were tested in this ‘uptake’ study at a lower dose of (0.007 mg/kg-day), and two additional subjects each were tested at the previous doses. In this ‘uptake study,’ serum samples were taken on days 8 and 14 and urine collected on day 14. RAIU measurements were also made in these subjects following 123I ingestion on the day prior to perchlorate exposure (baseline measurements) and at 9 A.M. on exposure day 14 and post-exposure day 1 (Greer et al., 2002). The age of the subjects in both the ‘main’ and ‘uptake’ studies ranged from 18 to 57 years with a mean of 38 (SD ± 12) years.

    The serum and urine samples were shipped to the Air Force Research Lab (AFRL/HEST) at Wright Patterson Air Force Base for analyses. The analytical method for the analyses was similar to that described in Narayanan et al. (2003). However, an important distinction between the analytical method used in this study and that of Narayanan and colleagues is the mobile phase concentration used. These concentrations were 80 mM NaOH for serum and from 60 to 120 mM NaOH for urine (depending on the sample). The range in mobile phase concentrations for the urine samples was required to attain proper sensitivity. Eight- and 24-h thyroid RAIU values were provided directly from the measurements of Greer et al.

    Model validation studies. Iodide model parameters were validated through predictions of protein-bound iodine (PBI) data from Scott and Reilly (1954). Data used in validating model predictions of serum were obtained from a recent unpublished drinking water study conducted by Drs. Georg Brabant and Holger Leitolf of the Medizinische Hochschule, Hanover, Germany. Seven healthy males ingested 12.0 mg/kg-day for 2 weeks. The daily dose was dissolved in 1 l of drinking water and divided into three equal portions, which were ingested at approximately 7 A.M., 12 P.M., and 7 P.M. each day for 14 days. Serum specimens were collected on exposure days 1, 7, 14, and post-exposure days 1 and 2. The serum samples from Brabant and Leitolf were also analyzed for perchlorate by AFRL/HEST. Data used in validating model predictions of cumulative urinary were obtained from Durand (1938), Eichler (1929), and Kamm and Drescher (1973). In these studies healthy males received a single oral dose of potassium perchlorate, ranging from 635 to 1400 mg .

    Predictions of ClO4-induced inhibition of thyroid I– uptake were validated with the RAIU data of Greer et al. (2002) and discharge data by Gray et al. (1972). Lastly, the model's ability to predict thyroid I– uptake and inhibition in hyperthyroid individuals was tested against data from Stanbury and Wyngaarden (1952).

    Model structure. The described PBPK model (Fig. 1) conforms to the structure of the concurrently developed model for the male rat (Merrill et al., 2003), with the exception of the newly added plasma subcompartment for bound I–, included for completeness. For both I– and , tissues containing NIS (thyroid, skin, and stomach) were described as compartments with nonlinear saturable uptake kinetics (Anbar et al., 1959; Chow et al., 1969; Kotani et al., 1998; Perlman et al., 1941; Slominski et al., 2002; Wolff, 1998). Other NIS-containing tissues, such as the salivary glands, choroid plexus, ovaries, mammary glands, and placenta (Brown-Grant, 1961; Honour et al., 1952; Spitzweg et al., 1998) were lumped with the slowly and richly perfused tissues, as either their anion pools are too small to significantly affect plasma levels (Cserr et al., 1980), or they are not applicable to the nonpregnant human. In development of the model, a quantitative sensitivity analysis showed no significant effect of either the gastric or thyroid uptake parameters on serum levels. Since the organ volumes are small and relative activity of NIS in the choroids plexus and salivary glands, etc. are even lower, we would expect these tissues to also have little to no effect on plasma concentrations.

    The stomach includes three subcompartments for capillary blood, stomach wall, and contents. Skin is described with two subcompartments for capillary blood and skin tissue. The thyroid is described with three subcompartments to describe the disposition of in the gland representing stroma, follicle, and lumen. The necessity for three subcompartments has been demonstrated elsewhere (Clewell et al., 2004; Merrill et al., 2003). Unlike , which eventually diffuses from the thyroid back into systemic circulation unchanged (Anbar et al., 1959; Wolff, 1998), most I– is quickly incorporated into monoiodothyrosine (MIT) and diiodothyrosine (DIT) in the thyroid follicle, which in turn combine to form thyroid hormones, triiodothyrosine (T3) and thyroxine (T4), which are then secreted into circulating blood. Endogenous iodine content in the normal thyroid is about 10,000 μg, of which greater than 90% is organic and 5–10% is free (Berman, 1967). Therefore, a fourth subcompartment, representing all ‘incorporated’ I– in the entire thyroid (i.e., total incorporated I– in the stroma, follicle, and lumen) is included. The incorporation of free I– into thyroid hormones and iodinated precursors, and the subsequent secretion of incorporated I– from the thyroid into venous blood were modeled with first-order rates. Passive diffusion is governed by the electrochemical gradients formed by the variation in anion concentration across thyroid subcompartments and wasdescribed with permeability area cross products and partition coefficients, (symbolized as small arrows in Fig. 1). Active uptake at NIS sites and between the thyroid follicle and lumen was describe using Michaelis-Menten (M-M) parameters (Fig. 1, bold arrows).

    Fecal elimination of the anions is minimal and therefore not included in the model. As mentioned earlier, is fully eliminated in the urine unchanged. Yu et al. (2002) reported 97% of eliminated in urine within 26 h after dosing with 3.0 mg/kg in rats. Fecal excretion of I– comes from the liver breaking down thyroid hormones and its secretion of the metabolic products into the bile, which enters the intestines. Therefore, when Hays (1993) administered 125I-T4 orally to seven healthy male subjects, she found that ‘the fraction of fecal radioactivity attributed to 125I was 0.55 ± 0.35’. But, administration of radioactive thyroid hormones is very different from administering radioactive I–. When radioactive I– is given orally, nothing is available for the liver to break down; virtually all of the radioactive I– is excreted by the kidneys. As reported in Elmer (1938), Scheffer (1937) measured 2.8–9.0% of the total radioactive I– excreted in the feces. They noted that this amount was highly variable. For example, after fasting fecal radioactivity may be undetectable. Braverman and Utiger (1991) stated ‘fecal excretion of dietary iodine is negligible’ (5 μg/day). Fecal elimination should be included in future expansion of the model, which will include thyroid hormone metabolism and homeostasis.

    The rapid urinary clearance of and radioactive I–, seen in both rats (Yu et al., 2002) and humans (Greer et al., 2002), was described with a kidney compartment. Urinary clearance could have easily been describe from the free anion portion in the blood; however, significant levels of deiodination occur in the kidneys, which would be critical in future model extrapolations to simulate thyroid hormone metabolism and clearance. Similarly, the liver compartment was maintained separately because it is the major site of extrathyroidal deiodination and it may be required in future pharmacodynamic elaborations of the model. At this point, the inclusions of these compartments did not add a great deal of complexity or uncertainty to the model.

    Because both anions are highly hydrophilic, fat acts as an exclusionary compartment. Given the large variability in human body fat, it was important to explore the contribution of this compartment to the overall anion kinetics. In addition, rapidly changing fat content during reproduction made the compartment important for extrapolation across life stages (Clewell et al., 2001, 2003a,b). The perfusion-limited compartments (e.g., kidney, liver, fat, slowly perfused, and richly perfused) were each described using partition coefficients and blood flows.

    The structure of the plasma compartments for and I– are similar. Both include passive diffusion between plasma and red blood cells (RBCs); however, reversible binding between plasma proteins and and I– are slightly different. In human serum, binding to plasma proteins has also been demonstrated (Hays and Green, 1973; Scatchard and Black, 1949). Approximately 95–98% of endogenous plasma iodine is reported as protein-bound iodide (PBI) in human serum (Rapport and Curtis, 1950; Underwood, 1977). However, unlike bound , the bound I– fraction primarily represents nonexchangeable, covalently bound, ‘incorporated I–’ secreted from the thyroid (e.g., hormonal iodine and iodinated proteins, including tri- and diiodothyronine) rather than simply inorganic I– bound to carrier proteins. In fact, it has been shown that approximately 90% of endogenous PBI represents T4 and 5% represents T3 (Berkow et al., 1977). Michaelis-Menten kinetics were used in the model to describe the association of the free and I– fractions to unspecific plasma binding sites, and first-order rates were used for their dissociations. In the case of I–, however, dissociation represents both deiodination of hormones and disassociation of I– from serum proteins. Hormone deiodination occurs at several sites throughout the body, but in the absence of better data, it was lumped together as one first-order rate. Although unlabelled I– is not included in the model, the behavior of tracer doses of radioactive I– is expected to follow that of endogenous I–. This is because the average levels of plasma inorganic iodine (PII) are expected to be around 0.40 ± 0.23 μg/dl (Elte et al., 1983), well below the NIS Km value (described below). Therefore, the NIS will not be saturated in the ‘average’ person.

    Model Parameters

    Physiological parameters. Tissue volumes and blood flows are presented in Table 1. Considerable variability was reported for some parameters, such as blood flow to the stomach (QG), which can increase 10-fold in response to enhanced functional activity (secretion and digestion) (Granger et al., 1985). The blood flows used represent estimates of resting values. Human data on the volume of the stomach capillary bed (VGBc) were not available. Therefore, a value derived from rat stomach data (Altman and Dittmer, 1971a) was allometrically scaled as described below. Mean values reported in the literature were used for all other physical parameters.

    A M-M affinity constant (KmTFi) for I– at the NIS of 4.0 x 106 ng/l was derived by Gluzman and Niepomniszcze (1983) from human thyroid slices. The authors noted little variation between normal and pathological thyroid specimens, or between specimens of different species. Wolff (1998) reported that I– Km(s) were similar across different NIS-containing tissues. This was supported by Kosugi et al. (1996), who reported a similar Km value of 4.4 x 106 ng/l for I– affinity for NIS in Chinese hamster ovary cells. Therefore, the value reported by Gluzman and Niepomniszcze was also used to describe iodide's affinity in the skin and stomach.

    Kosugi et al. (1996) also measured perchlorate's affinity for NIS and reported a Km of 1.5 x 105 ng/l, approximately ten times lower than the that for I–, indicating a greater affinity for perchlorate. Several other studies agree that perchlorate's affinity for NIS is approximately ten times greater than that of I– (Halmi and Stuelke, 1959; Harden et al., 1968; Lazarus et al., 1974; Wyngaarden et al., 1952). Based on this information, a KmTFp value was set to 1.6 x 105 ng/l, adjusting the literature value slightly based on the model fits to data in NIS-containing tissues. This values lies between that measured by Kosugi et al. and that required in the corresponding male rat model (1.8 x 105 ng/l).

    The apical membrane of the thyroid follicle also exhibits a selective I– channel, believed to be pendrin. Pendrin is a transmembrane glycoprotein, which facilitates both chloride and I– efflux across the apical membrane (Mian et al., 2001; Scott et al., 1999). Golstein et al. (1992) measured a Km of 4.0 x 109 ng/l, for I– transport from the bovine thyroid follicle into the lumen (KmTLi). However, as in the corresponding rat model, a slightly lower KmTLi of 1.0 x 109 ng/l was required to fit thyroid I– data at later time points (>8 h). Golstein et al. (1992) reported that this apical channel also appears sensitive to inhibition, suggesting a lower Km for (KmTLp) than for I–. A KmTLp value of 1.0 x 108 ng/l was derived from Chow and Woodbury's (1970) data, as described in Merrill et al. (2003).

    Maximum velocities, Vmax(s), for anion uptake vary between tissues and species (Bagchi and Fawcett, 1973; Wolff, 1998). Humans tend to have lower Vmax values than other species (Gluzman and Niepomniszcze, 1983; Wolff and Maurey, 1961) when expressed per gram of tissue. The Vmax(s) (ng/h/kg) for I– uptake in the thyroid and plasma were estimated by visually optimizing the clearance portion of the curves to respective time-course data of Degroot et al. (1971). This was accomplished by keeping all other parameters fixed, while the Vmax value was adjusted so that the model prediction adequately approximated the observed mean. It may be noted that Vmax values for the thyroid follicle and lumen differ by up to an order of magnitude from preliminary values, reported in Clewell et al. (2001). This was attributed to the availability of new data sets, which allowed improved parameterization.

    For tissues lacking time-course data, the Vmax(s) were estimated to yield kinetics similar to those described by the male rat model (Merrill et al., 2003). For example, for I– kinetics in the stomach and skin, VmaxGi and VmaxSki respectively, were visually optimized to resemble tissue:serum concentration ratios seen in the rat, while maintaining the fits to human serum data. Because data was only available in serum and urine, Vmax(s) for NIS-containing tissues were scaled from the I– Vmax(s), using the ratios between corresponding I– and Vmax(s) established in the male rat model.

    Diffusion, concurrent with active uptake in the stomach, thyroid, and skin, was described using permeability area cross products (PA) (l/h-kg) and effective partition coefficients (P). In general, PA values were visually fit to the uptake portion of the curves, prior to setting Vmax(s), with partition coefficients, and all other parameters were set to the values in Table 2 and held fixed. The early time-course data of I– in gastric aspirations from Hays and Solomon (1965) were used to estimate PAGJci, representing 131I– transfer from the gastric juice into the gastrointestinal plasma (l/h-kg). To simulate the removal of gastric aspirations, the amount of 131I– reabsorbed by the stomach wall had to be mathematically eliminated or set to zero. Once parameters were established using the aspiration session data, stomach reabsorption was reintroduced, and the permeability cross product for 131I– transfer between gastric blood and tissue (PAGci) was fit to the corresponding increase in 131I– in plasma, thyroid, and urine seen in the control session (where gastric juices were not aspirated). The permeability area cross product between the thyroid stroma and follicle, PATFci, was visually optimized to the uptake portion of the thyroid I– data.

    The first-order clearance rates describing the organification of I– shortly after it enters the thyroid follicle (Clhormci) and the secretion of organic I– into systemic circulation (Clsecrci) were visually optimized to the clearance portion of thyroid 131I– data, as well as the later time points of the plasma PBI data from Degroot et al. (1971). The first-order rate, describing the body's overall deiodination rate (Cldeiodci) was also estimated through visual optimization of the later PBI timepoints, while maintaining the model fit to total plasma iodine and keeping all other parameters fixed. Later plasma time points of PBI reflect the contribution of hormone secretion and deiodination rates due to sufficient lapse of time for radioactive I– incorporation into thyroid hormones and precursors. Therefore, earlier PBI time points were visually fit to establish parameters for binding of inorganic I– to plasma proteins (e.g., KmBi, VmaxBi). Similarly, reversible plasma binding of perchlorate was described using a first-order rate constant (Clunbp), which was visually optimized to available serum data. Urinary excretion rates for both anions (ClUi and ClUp) were visually fit to available cumulative urine data (Degroot, 1971; Greer et al., 2002; Hays and Solomon, 1965). Because cumulative urinary perchlorate data was available in the Greer et al. study, ClUp was visually fit to each individual's data, and the average value was then used for the model parameter.

    Allometric scaling and rate equations. Differential equations used to simulate radioactive I– and transport were written and solved in ACSLTM (Advanced Continuous Simulation Language) (AEgis Technologies, Austin, TX). To account for body-weight-dependent variables and species extrapolations, allometric scaling was applied to Vmax(s), PA(s), Cl(s), tissue volumes (V), and blood flows (Q). The variety of dosing regiments and routes were simulated using various pulse function codes in ACSL.

    Rate equations describing I– transport in ng/h in the thyroid stroma, follicle and lumen (colloid) (RATSi, RATFi, and RATLi, respectively), as well as the rate of change in bound thyroid iodine (RAbndi) are provided below. These equations demonstrate diffusion-limited uptake, using P(s) and PA(s), and saturable uptake and competitive inhibition using M-M parameters. Equations used in the other compartments are expressed similarly.

    Subscripts i and p identify the anion as either I– or , QT is thyroid blood flow (l/h), CAi is the arterial blood concentration (ng/l), CVTSi,p is the thyroid stroma concentration (ng/l), CTFi,p is the follicular concentration (ng/l) of I– or , and CTbndi is the concentration of incorporated I– in the entire thyroid. PTFi, PTLi, PATFi, and PATLi are the partition coefficients and permeability cross products describing passive diffusion of I– across the basal (follicle:stroma) and apical (lumen:follicle) membranes. Michaelis-Menten equations are used to describe the rates of active uptake of I– into the follicle by NIS and into the lumen by the apical I– channel (RupTFi and RupTLi, respectively), including inhibition by . VmaxTFi, VmaxTLi, KmTFi,p and KmTLi,p are the maximum velocities (ng/h/kg) and affinity constants (ng/l) for transport of I– or into the follicle and lumen. Clhormi and Clsecri are first-order clearance values (h–1) for the organification of I– into thyroid hormones and the secretion of organified I– into systemic circulation. Transport of through the thyroid is calculated similarly, except there are no terms for organification of (Clhormi and Clsecri). In addition, inhibition of uptake by I– is not included. As described earlier, due to the lower affinity of I– (10-fold higher Km) than that of , I– does not significantly inhibit sequestration in NIS-containing tissues. Example equations of other compartments are shown elsewhere (Merrill et al., 2003).

    Sensitivity analysis of parameters. To assess the relative impact of each parameter on model predictions, a sensitivity analysis was performed. After finalizing all model parameters, the model was run at a drinking water dose below NIS saturation (0.1 mg/kg/day) for 240 h (to ensure equilibrium was reached) to determine the average serum concentration [area under the curve (AUC)]. The model was then repeatedly rerun, using a 1% increase in each parameter to determine the resulting change in predicted serum concentration AUC, and sensitivity coefficients for each parameter were then calculated using the equation below.

    Where A equals serum AUC with 1% increased parameter value, B equals serum AUC using original parameter value, C equals parameter value increased 1% from original value, and D equals the original parameter value.

    RESULTS

    Model Parameterization

    Iodide kinetics. Effective partition coefficients (P) and affinity constants for the active transport mechanisms (Km), listed in Table 2, were kept consistent with those used in the male rat model. Parameterization of other I– parameters was obtained using available human time-course data as described in the Methods section. Figures 2A through 2D show the model simulations of the early distribution of 131I– in plasma, thyroid, gastric juice, and urine during both the control and gastric aspiration session by Hays and Solomon (1965). Degroot's plasma, urine, and thyroid time-course measurements extended nearly 11 days post-dosing, allowing greater calibration of parameters affecting both uptake and clearances in the thyroid and plasma. Model simulations of plasma inorganic I–, PBI, and total plasma iodine are presented in Figures 3A–3C. The model underpredicted urinary I– measured by Degroot et al. (1971) (Fig. 3D). However, considerable variations in I– excreted by humans exist, as it varies with thyroid function and dietary I– intake (NRC, 1996). Since increasing urinary clearance (ClUi) would result in underprediction of plasma iodine, the value for ClUi was not changed to fit this data set.

    DISCUSSION

    The validity of the model structure and its parameters are demonstrated by its ability to predict and I– in serum and urine data, as well thyroid I– inhibition data from various studies, which involve various dose levels and routes, while using a single set of parameters. The model adequately simulates serum and cumulative urine levels after drinking water exposure to spanning almost four orders of magnitude (0.02 to 12.0 mg/kg-day). Although serum levels were not available at 0.02 mg/kg-day, the model was able to simulate the cumulative urine from that dose group (Fig. 9). Comparison with parameters of the rat model indicates that humans have a considerably lower plasma binding capacity (VmaxcBp) for (approximately 20 times lower). Although binding of to plasma proteins has been directly measured in human serum, it is not surprising that it would occur to a lesser extent than in the rat. Carr (1952) tested the ability of to bind to various proteins in human blood, including albumin, pre-albumin and thyroxine binding globulin (TGB). They found that was able to bind to albumin and prealbumin, but not TBG. Thus, it would be expected that the binding capacity would be greater in the rat, whose primary binding protein is albumin, than the human, whose primary binding protein is TBG.

    Data available for calibrating and validating serum I– were limited to tracer dose levels. However, the kinetic behavior of I– is expected to be linear over a wide range of doses. Although the mechanism of transfer into the tissues with NIS is saturable, the value of Km (4.0 x 106 ng/l) indicates that very high doses would be required to saturate this mechanism. Additionally, similar parameter values and identical model structures in the corresponding rat models of various life stages have yielded validated serum predictions at dose levels ranging three or more orders of magnitude for both anions.

    Aspects of the model, which were supported in the literature or laboratory studies but could not be directly observed in humans, were incorporated if necessary to improve the fit of the model to available data. For example, active uptake of I– and into human skin and the relatively slow diffusion of both anions from skin back into systemic circulation were incorporated in spite of a lack of human time-course data. The literature supports this behavior, as NIS has been identified in human skin (Slominski et al., 2002), and slow diffusion has been noted with similar anions, such as pertechnetate. Hays and Green (1973) performed dialysis studies on intact human tissues with pertechnetate. They found skin had a relatively slow uptake of pertechnetate peaking at 18 h and, in fact, more retention after leaching dialysis than seen in brain, muscle, and serum.

    It is possible that the reason elevated I– in human skin has not been reported in clinical radioactive I– scans is the difficulty in differentiating skin radioactivity from background radioactivity. The large volume of the skin allows radioactive I– to be diffused over a large surface. However, this same property allows the skin to be an important pool for the storage and slow turnover of I–. Simulations with this model demonstrated that inclusion of active uptake in human skin was required to simulate serum data. Sensitivity analysis on the corresponding male rat model indicated that serum levels were highly sensitive to parameters of the skin and plasma binding and urinary excretion (Merrill et al., 2003).

    Kinetic data for establishing parameters for the gastric compartment were limited to the early I– data (3 h post-dosing) by Hays and Solomon (1965). Their gastric juice 131I– data indicated rapid transport of I– into the gastric mucosa (Fig. 2C). It is expected that uptake in the stomach would behave likewise, due to the similarity of the anions in size and charge. The time-course behavior of radioactive iodine in stomach contents of any species is complicated by the fact that it reflects more than sequestration of radioactive I– by NIS. Its appearance also reflects the accumulation of salivary radioactive I– that is swallowed involuntarily throughout the study. Several studies that examined sequestration of these anions in digestive juices have all shown high variability in the concentrations measured over time (Hays and Solomon, 1965; Honour et al., 1952). There is a tendency for the gastric juice:plasma ratio to be low when the rate of secretion of gastric juice and saliva is high (Honour et al., 1952). This is because the increase in secretions does not correspond with upregulation of NIS; therefore, the gastric juice concentration becomes diluted. Fluctuations in the secretion rate are probably the most important factor in determining the pattern of the concentration ratios in individuals. Therefore, variability in stomach or GI tract parameters between models is expected. However, the early rise in the gastric juice:plasma ratio mentioned earlier is a constant feature across these data sets, whether or not an attempt was made to eliminate contamination of gastric juices by dietary contents or saliva. Animal data that show both the anion uptake and clearance in the stomach (Yu et al., 2002), unlike the data in Figure 2C which only show uptake, indicate that the clearance portion is less rapid. This model, and the series of different life-stage rat models (Clewell et al., 2003a,b; Merrill et al., 2003;) consistently predict this same trend.

    Average urinary clearance values were found to be 0.11 l/h-kg for I– and 0.13 l/h-kg for . However, these values are not expected to be successfully applied to every euthyroid individual studied, though the use of these average values should provide a reasonable prediction of the euthyroid population. Individual differences in urinary I– are expected with variation in thyroid function and protein-bound I– in plasma. Iodide is ultimately removed or eliminated by two competing mechanisms, thyroidal uptake and urinary excretion. Thus, a higher amount of excreted I– in urine is indicative of reduced thyroid uptake. Historically, this relationship has been used to estimate thyroid function. A cumulative 24-h urinary clearance of less than 30% of a tracer dose is indicative of hyperthyroidism, whereas clearances exceeding 40% are often associated with normal or decreased thyroid function. However, a high degree of variability exists between human subjects. Such a significant degree of overlap exists in thyroid test results for normal, hyperthyroid, and hypothyroid patients, that it is often necessary to run several different additional screens in order to identify subclinical conditions (NRC, 1996).

    In addition to the expected variability in thyroid uptake parameters (VmaxcTFi values ranging from 5.0 x 104 to 5.0 x 105 ng/h-kg) between individuals, variability across data sets was also noted. However, the difference seen in the average VmaxcTFi obtained from the Greer et al. (2002) subjects (1.5 x 105 ng/h-kg) and those from Hays and Solomon (1965) (2.5 x 105 ng/h-kg) is easily explained by the difference in experimental conditions between the two studies. Hays and Solomon's subjects fasted 12 h prior to the administration of the 131I–, whereas Greer and coauthors' subjects had no dietary restrictions prior to 125I– administration. As a result, intrathyroidal I– levels would have been lower in the fasted individuals, and as anticipated, the average VmaxcTFi from Hays and Solomon (1965) would be increased.

    Dietary iodine and endogenous inorganic I– levels are clearly important in modeling I– and kinetics, because excessive I– levels cause the ion to inhibit its own uptake (Wolff and Chaikoff, 1948). The ability of the model to describe the bound and free I– fractions in the thyroid and serum provides the basis for subsequent modeling of hormone synthesis and regulation in humans. Measurements of tracer radioactive I– can be fitted to predict transfer rates. However, the use of these acute parameters is limited when attempting to describe long-term thyroid kinetics, unless the existing endogenous I– and the relationships between the regulating hormones are taken into consideration. Ultimately, such factors as preexisting thyroid conditions and regional dietary iodine might be addressed in a more comprehensive hormone feedback model. In its present state, our model is useful in predicting perchlorate's effect on thyroid I– uptake in what is considered the normal population: euthyroid individuals with adequate dietary I–.

    That the model is capable of predicting I– uptake in hyperthyroid subjects by increasing the VmaxTFi supports the usefulness of the current model structure. TSH increases the total amount of NIS in a membrane, thereby increasing VmaxTFi. Gluzman and Niepomniszcze (1983) reported elevated Vmax(s) in thyroid specimens from subjects with Graves' disease, toxic adenoma, and dishormonogenetic goiter. In future versions of the model, the increase of TSH and subsequent effect on this parameter can be described mathematically in order to predict the dose- and time-dependent response of the thyroid activity to various disease states. In specimens from nontoxic nodular goiter, Hashimoto's thyroid, or extranodular tissue from toxic adenoma, Vmax(s) were decreased. However, in all subjects there was little variation in the KmTi. Therefore, one would not expect the underprediction of thyroid inhibition in the subject with Graves' disease to be due to disease-induced lowering of Km, but rather the increased inhibition is mostly likely due to simple interindividual differences. Sensitivity analyses performed on the model for the rat indicates that model-predicted values of inhibition are highly sensitive to even small changes in Km for . Thus, it is quite possible that changing Km within the range of normal values would account for this apparent discrepancy in the model fit.

    The PBPK models developed for perchlorate-induced inhibition have been useful to the ongoing risk assessment of , and helped integrate the data from diverse data sets to evaluate the dose response of adverse effects from low levels of exposure (U.S. Environmental Protection Agency, 2003). The resulting parameters may be used in conjunction with those established for the male (Merrill et al., 2003) and perinatal rat models (Clewell et al., 2003a,b) to extrapolate to human gestational and lactational models (Clewell et al., 2001). In order to further assess model performance and to facilitate the use of these models in risk assessment, a more comprehensive statistical evaluation of model parameters may prove additionally useful. Sensitivity analysis provided insight into the relative importance of model parameters with respect to specific measures of dose. Comparing the highest sensitivity coefficients between the male rat and human models indicated that, at low doses, human serum levels are most sensitive to urinary clearance, whereas the rat's serum levels are more sensitive to plasma binding parameters (Fig. 15) (Merrill et al., 2003). The fact that data was available across multiple doses for establishing parameter values for urinary clearance and plasma binding adds confidence to the model's predictive ability. More useful to the application of the models, for human dosimetry predictions, is variability analysis that is performed with known distributions for model parameters. This tool can be applied to the model to allow the prediction of likely ranges of the dosimetrics within a human population.

    Modeled effects on hormone regulation are yet to be developed. Perturbations in hormones levels after exposure demonstrate complex differences in the hormone regulatory mechanisms between rats and humans, which are difficult to describe (Clewell et al., 2001; Merrill et al., 2001). However, the current model structures may provide a basis for evaluating thyroid effects from other environmental contaminants. For example, excessive exposure to other similarly behaving anions, such as sodium chlorate, thiocyanate, or nitrate, all found to also contaminate ground and surface waters, may contribute to environmental anti-thyroid effects in humans (Hooth et al., 2001; Wolff and Maurey, 1963). Further, the possibility of additive anti-thyroid effects to those of perchlorate from these cocontaminants may need to be considered (Kahn et al., 2004).

    NOTES

    The views expressed in this paper are those of the authors and do not necessarily reflect the views or policies of the U.S. Environmental Protection Agency. The U.S. Government has the right to retain a nonexclusive royalty-free copyright covering this article.

    2 Current address: CIIT Centers for Health Research, Research Triangle Park, NC 12137.

    3 Current address: Environmental Health Science, University of Georgia, Athens, GA 30602.

    ACKNOWLEDGMENTS

    The authors express special thanks to Drs. Monte Greer, Gay Goodman, Georg Brabant, and Holger Leitolf for supplying serum and urine samples from their studies for perchlorate analyses and Dr. Mel Andersen and Harvey Clewell for constructive comments. Also acknowledged are Lt. Col. Dan Rogers, Dr. Richard Stotts, and Dr. Dave Mattie for assistance in obtaining funding for this research from the U.S. Air Force, U.S. Navy, and NASA. Drs. Andrew Geller and Allan Marcus are thanked for their critical technical reviews of the draft manuscript. Lastly, this work would not have been possible without analytical support from Lt. Eric Eldridge, Latha Narayanan, Gerry Buttler, and SSgt. Paula Todd. Funding for this research was provided by the U.S. Air Force, U.S. Navy, and NASA.

    REFERENCES

    Wolff, J. (1998). Perchlorate and the thyroid gland. Pharmacol. Rev. 50, 89–105.

    Wolff, J., and Chaikoff, I. L. (1948). Plasma inorganic iodide as a homeostatic regulator of thyroid function. J. Biol. Chem. 174, 555–564.

    Wolff, J., and Maurey, J. R. (1961). Thyroidal iodide transport: II. Comparison with non-thyroid iodide-concentrating tissues. Biochim. Biophys. Acta 47, 467–474.

    Wolff, J., and Maurey, J. R. (1963). Thyroidal iodide transport: IV. The role of ion size. Biochim. Biophys. Acta. 69, 48–58.

    Wyngaarden, J. B., Wright, B. M., and Ways, P. (1952). The effect of certain anions upon the accumulation and retention of iodide by the thyroid gland. Endocrinology 50, 537–549.

    Yokoyama, N., Nagayama, Y., Kakezono, F., Kiriyama, T., Morita, S., Ohtakara, S., Okamoto, S., Morimoto, I., Izumi, M., Ishikawa, N., et al. (1986). Determination of the volume of the thyroid gland by a high resolutional ultrasonic scanner. J. Nucl. Med., 27, 1475–1479.

    Yu, K. O., Narayanan, L., Mattie, D. R., Godfrey, R. J., Todd, P. N., Sterner, T. R., Mahle, D. A., Lumpkin, M. H., and Fisher, J. W. (2002). The pharmacokinetics of perchlorate and its effect on the hypothalamus/pituitary-thyroid axis in the male rat. Toxicol. Appl. Pharmacol. 182(2), 148–159.(Elaine A. Merrill*,1, Reb)