The National Research Council of the National Academies report, Toxicity Testing in the 21st Century: A Vision and Strategy, highlighted the importance of quantitative exposure data for evaluating human toxicity risk (Hubal, 2009). The report supports the use of chemical exposure data to provide critical information on the magnitude, timing, and duration of a biologically relevant dose delivered to target tissues (Sheldon and Cohen Hubal, 2009; Cohen Hubal et al., 2010). Direct measurement of chemical exposures using personal monitoring provides the most accurate estimation of a subject’s true exposure (Nieuwenhuijsen et al., 2006). In this context, biomonitoring is a critical tool for quantitatively evaluating exposure from both environmental and occupational settings to a wide range of pollutants, including pesticides (Borzelleca and Skalsky, 1980; Nigg and Wade, 1992; Friberg and Elinder, 1993; Christensen, 1995; Angerer et al., 2006, 2007; Barr et al., 2006). Non-invasive methods have also been advocated for quantifying the pharmacokinetics and bioavailability of drugs and xenobiotics, and the use of saliva has been suggested as an ideal body fluid that can be substituted for blood in biomonitoring (Pichini et al., 1996; Timchalk et al., 2004; Timchalk et al., 2007b). In this regard, a broad range of drugs, organic chemicals, metals, and pesticides are readily secreted in saliva (Borzelleca and Skalsky, 1980; Drobitch and Svensson, 1992; Nigg and Wade, 1992; Lu et al., 1997, 1998; Silva et al., 2005). For many of these xenobiotics, saliva concentration readily correlates with blood concentration; hence, it is feasible to utilize pharmacokinetic models to accurately estimate systemic dose based upon a saliva measurement.

The mechanism of acinar cell activation, electrolyte transport and saliva formation illustrated in Figure 1 is derived from an extensive literature that has characterized these processes in detail (Nauntofte, 1992). Salivary glands are comprised of a number of major (parotid, submandibular, and sublingual) and minor glands that primarily consist of acinar cells that collectively produce saliva (for review, see Hold et al., 1995b). Salivary glands are highly perfused and blood flow is in a countercurrent direction to salivary flow (Figure 1B; Davenport, 1977). The primary mechanism by which xenobiotics leave the blood and enter saliva is thought to involve paracellular transport, passive transcellular diffusion, or transcellular active transport (Figure 1A; Landon and Mahmod, 1982). Paracellular transport (i.e., ultrafiltration) favors small (MW ∼300 Da) polar lipid insoluble molecules that generally have a low (i.e., <1.0) saliva/plasma (S/P) ratio. Whereas, transcellular diffusion or active transport are favored by lipid soluble materials that can readily cross cell membranes (Hold et al., 1995a). The majority of drugs and xenobiotics are cleared from plasma into saliva by a passive diffusion process that is a function of concentration gradient, surface area, membrane thickness, and diffusion constants (Hold et al., 1995b). This focused review will describe how the modeling of this passive diffusion process in both rats and humans was accomplished utilizing a pharmacokinetic model that were initially developed for the organophosphorus insecticide chlorpyrifos (CPF) and its major metabolite trichloropyridinol (TCPy; Smith et al., 2010, 2012).

It is well recognized that biomonitoring offers one of the best strategies for quantifying human dosimetry and assessing risk associated with both occupational and environmental chemical exposures (Friberg and Elinder, 1993; Christensen, 1995). As noted above, saliva has been suggested as an ideal non-invasive body fluid, yet there are a number of limitations which have hampered saliva’s use in biomonitoring. First, to utilize saliva for biomonitoring the relationship between blood and saliva concentrations for target biomonitoring, analytes needs to be well established (Timchalk et al., 2004). Second, to more broadly exploit saliva as a matrix it is critical to a priori identify which chemicals are readily cleared in saliva at levels that can be quantified analytically. The current manuscript reviews the initial saliva model development focused on in vivo uptake and clearance studies in the rat as well as computational prediction in humans (Smith et al., 2010, 2012). In addition, a partitioning coefficient algorithm that readily accommodates species-dependent differences in plasma protein-binding and pH is presented. This is of particular importance since changes in binding and/or pH appear to have a significant impact on plasma:saliva partitioning thereby contributing to variable saliva analyte concentrations. Finally, a brief perspective on the utility of cell based in vitro assay systems and computational modeling are presented as a potential strategy for broad based screening of chemicals for their salivary clearance potential (Weber et al., 2015).

Initial in vivo studies in rats focused on the comparison of blood and saliva TCPy concentration time-course following intravenous (IV) administration of the CPF over a broad dose range (0.5–5 mg/kg). These in vivo experiments focused on exposing rats to CPF by IV administration and quantifying the time-course of CPF and TCPy in both blood and saliva over a range of CPF doses. Previous studies have characterized route-dependent pharmacokinetic differences in rats, and IV administration was chosen for these studies to minimize first-pass hepatic metabolism thereby maximizing the concentration of CPF in blood available for salivary clearance (Busby-Hjerpe et al., 2010). These previous studies demonstrated that the blood:saliva ratios were not impacted by a range of factors including: CPF or TCPy dose, pilocarpine dose, or the timing of saliva collection (Smith et al., 2010, 2012). In Figures 3A–C where the time-course of CPF (blood) and it major metabolite TCPy (blood and saliva) are illustrated following a 5 mg/kg IV dose. These results are very consistent with previous observations that CPF undergoes rapid metabolism to TCPy (Poet et al., 2003; Timchalk et al., 2007a; Busby-Hjerpe et al., 2010), although CPF was detectable in three saliva samples (data not shown). Results (Figure 3D) were also consistent with previous TCPy dosing experiments where the ratio of TCPy in saliva/blood (ratios ∼0.0 4–0.06) demonstrated AUC linearity consistent with parallel pharmacokinetics of TCPy in blood and saliva. These initial in vivo studies in rats clearly demonstrate that saliva analyte concentrations are highly proportional to blood concentrations and PBPK model simulations accurately predict saliva/blood concentrations over a range of doses, consistent with a diffusion based transport mechanism.

Although in the rat in vivo model system saliva/blood TCPy concentration ratios did not change over varying experimental conditions (Figure 3D), as previously mentioned normal differences in saliva pH values for humans (range 5.6–7.9) have been noted in the literature (Hold et al., 1995b). Hence, to adequately accommodate differences in partitioning due to variable pH, and to enable accurate rat to human extrapolation it was critical to simulate changes in analyte ionization states particularly when the pKa was bracketed by the pH range of saliva. In this regard, the Schmitt algorithm provided a good model prediction of the saliva/blood partitioning (Smith et al., 2010). These initial finding do suggest that inter- and intra-species differences in blood:saliva partitioning may be of importance. To further evaluate the potential impact of these differences a model sensitivity analysis was conducted.

Physiologically based pharmacokinetic models have been developed that incorporated saliva clearance as a computational tool to quantify non-invasive biomonitoring of heavy metals and pesticides (Timchalk et al., 2001, 2007b; Barry et al., 2009; Smith et al., 2010, 2012). As previously noted, transcellular diffusion (Figure 1) is thought to be the primary method by which many drugs, chemicals, and pesticides, including the CPF metabolite TCPy transfer from blood to saliva (Haeckel, 1993; Jusko and Milsap, 1993; Lu et al., 1997, 1998, 2003; Smith et al., 2010, 2012). As noted in Figure 3D, in vivo partitioning of TCPy from blood to saliva in rats is very constant over a range of varying conditions which is entirely consistent with a partitioning mechanism associated with transcellular diffusion as proposed by Smith et al. (2010, 2012).

When considering transcellular or paracellular diffusion, it is of interest to note that a number of physiochemical properties including molecular size, lipid solubility, the dissociation constant of ionized compounds, and the extent of plasma-protein binding all contribute to modifying the diffusion across a concentration gradient from blood to saliva. In the current model simulation (Figure 3) the analyte concentration in saliva (based upon partitioning) is directly reflective of the concentration that is not bound to plasma proteins in blood, since protein-analyte complexes are too large for transcellular/paracellular diffusion (Haeckel, 1993; Jusko and Milsap, 1993; Lu et al., 1998; Smith et al., 2012).

Likewise, as shown in Figure 5, the pH of blood and saliva are important parameters modulating salivary clearance of compounds particularly where the chemical pKa is comparable to the pH of biological fluids (Haeckel, 1993; Jusko and Milsap, 1993). Although blood pH is reasonably consistent (pH 7.4), saliva pH can vary and is primarily controlled by the amount of bicarbonate present in saliva (Haeckel, 1993). Hence, due to variation in rat and human pH values (Smith et al., 2010), the Schmitt algorithm (Schmitt, 2008) has been utilized successfully to calculate plasma:saliva partitioning. This modification enables the extrapolation of the blood to saliva analyte partitioning coefficient to humans over a range of reported human salivary pH levels.

Although transcellular or paracellular diffusion across tight junctions has been noted as a key transport mechanism for many drugs and environmental contaminants (Haeckel, 1993; Jusko and Milsap, 1993), both active and carrier-assisted cellular transport may be of importance for some xenobiotics. In this regard, transport is dependent upon specific rate-limiting transport mechanisms. In salivary glands, unidirectional and saturable influx of amino acids, neurotransmitters, and drugs have been shown to follow classic Michaelis–Menten kinetics where affinity and velocity constants can be experimentally derived and computationally solved (Eq. 8). In this scenario, T_r is the transport rate, T_max the maximum rate of transport, C_u the unbound analyte concentration and K_m is the transport affinity constant. Hence, the current model structure can be readily modified to accommodate these active transport processes; thereby, enabling the models to simulate salivary clearance for a broad range of chemicals and drugs.

For saliva biomonitoring to be more broadly utilized there is a need to rapidly identify a comprehensive range of chemical and drugs that can readily be quantified in saliva and utilized to predict systemic dose based upon these saliva measurements. A major limitation of the current experimental and modeling strategy is the dependence upon in vivo animal model systems as a means of identifying and screening chemical/drug candidates for salivary clearance. In this regard, the current in vivo models are limited by relatively low throughput and experimental complexity. To address these limitations we are developing an in vitro salivary gland epithelial cell based Transwell^® assay to enable broad based screening of uptake and clearance mechanisms associated with both diffusional and active transport mechanisms (Figure 6; Weber et al., 2015). Initial characterization of the cell system has demonstrated that tight junctions are well formed and cells maintain functional characteristics (Figure 6C) associated with salivary acinar cells. The computational approach is based on a simplified four-compartment model Figure 6D) exploiting the Schmitt algorithm (Schmitt, 2008) to describe diffusional based processes along with active transport (Eq. 8). It is anticipated that this novel experimental and computational strategy will enable the prediction of chemical uptake and clearance in saliva for a broad range of chemicals based upon limited in vitro experiments which are integrated into the pharmacokinetic modeling framework.

Biomonitoring is a critical tool for quantitatively evaluating exposure from both environmental and occupational settings and saliva has been advocated as a potentially important non-invasive method that could be substituted for blood or urine. However, there are a number of limitations that have hampered saliva’s use in biomonitoring, including the need to initially identify which chemicals are readily cleared in saliva at concentrations that can be quantified. This review describes recent advances in the use of a computational modeling approach that is closely coupled to in vivo and in vitro experiments to predict salivary uptake and clearance of xenobiotics. The approach simulates transcellular or paracellular diffusion of unbound chemicals in plasma to saliva using a combination of compartmental and physiologically based computational models. Of key importance for determining the plasma:saliva partitioning is the utilization of a modified Schmitt algorithm (Schmitt, 2008) that calculates partitioning based upon the tissue composition, pH, chemical pKa, and plasma protein-binding. Sensitivity analysis of key model parameters specifically identified that both protein-binding and pH/pKa had the most significant impact on the determination of partitioning and that there were clear species dependent differences based upon physiological variance between rats and humans. Future research needs to focus on extending this modeling strategy to an in vitro salivary acinar cell based system that can be utilized to experimentally determine and computationally predict salivary gland update and clearance for a broad range of xenobiotics. Hence, it is envisioned that a combination of salivary biomonitoring and computational modeling will enable the non-invasive measurement of both environmental and occupational exposure in human populations.

Dr. Charles Timchalk and Dr. Jordan N. Smith have previously received funding from The Dow Chemical Company, the manufacturer of chlorpyrifos (CPF), to conduct research. However, the Dow Chemical Company had no input on the current research.