Pharmacokinetic modeling of the impact of P-glycoprotein on ondansetron disposition in the central nervous system

Abstract

Purpose:

Modulation of 5HT3 receptor in the central nervous system (CNS) is a promising approach for treatment of neuropathic pain. The goal was to evaluate the role of P-glycoprotein (Pgp) in limiting exposure of different parts of the CNS to ondansetron (5HT3 receptor antagonist) using wild-type and genetic knockout rat model.

Methods:

Plasma pharmacokinetics and CNS (brain, spinal cord, and cerebrospinal fluid) disposition was studied after single 10 mg/kg intravenous dose.

Results:

Pgp knockout resulted in significantly higher concentrations of ondansetron in all tested regions of the CNS at most of the time points. The mean ratio of the concentrations between KO and WT animals was 2.39–5.48, depending on the region of the CNS. Male and female animals demonstrated some difference in ondansetron plasma pharmacokinetics and CNS disposition. Mechanistic pharmacokinetic model that included two systemic disposition and three CNS compartments (with intercompartmental exchange) was developed. Pgp transport was incorporated as an efflux from the brain and spinal cord to the central compartment. The model provided good simultaneous description of all data sets, and all parameters were estimated with sufficient precision.

Conclusions:

The study provides important quantitative information on the role of Pgp in limiting ondansetron exposure in various regions of the CNS using data from wild-type and Pgp knockout rats. CSF drug concentrations, as a surrogate to CNS exposure, are likely to underestimate the effect of Pgp on drug penetration to the brain and the spinal cord.

Introduction

Neuropathic pain is a complex chronic pain condition affecting 7–10% of the general population (1). It is more prevalent in individuals with certain comorbidities; for example, 60–70% of diabetic patients also report peripheral neuropathy, and 30–40% of cancer patients receiving particular chemotherapeutic regimens develop symptoms of painful peripheral neuropathy (2). Available treatment options often do not meet the needs of the patients, where in many cases the therapies either do not achieve sufficient pain relief, or produce intolerable adverse events (3, 4). Serotonergic projections from the rostral ventromedial medulla (RVM) to the spinal cord play an important role in pain modulation, and data suggest that while under normal conditions these pathways mostly inhibit pain in neuropathic pain conditions this descending serotonergic pathway may contribute to pain facilitation (5–7). One of the hypotheses for this functional change in descending serotonergic projections is the overexpression of serotonin 5-HT₃ receptors in the spinal cord after nerve injury. While most serotonin receptor subtypes (5-HT₁, 5-HT₇) are G-protein coupled receptors (GPCRs) with mostly inhibitory function, 5-HT₃ is an ion channel with mostly excitatory function (8). The involvement of 5-HT₃ receptors in altered pain transmission in neuropathic pain makes it a potential pharmacological target in this context (6). Indeed, previous reports have demonstrated the potential utility of 5-HT₃ antagonists for the treatment of neuropathic pain (7, 9). For example, in a spinal cord injury model in male Wistar rats, a significant antinociceptive effect was obtained with increasing doses of intrathecal ondansetron (a 5-HT₃ receptor antagonist) compared to control animals (7).

Ondansetron is commonly used for the prevention and treatment of nausea and vomiting after surgery or emetogenic chemotherapy (10). Ondansetron has a strong binding to 5-HT₃ receptor (pK_i = 8.07), acts as a competitive antagonist to 5-HT₃ receptors, and has comparable potency to other 5-HT₃ receptor antagonist antiemetics (11–13). These drugs exert their pharmacologic effect through binding to the receptors in the chemoreceptor trigger zone in the area postrema, within the fourth ventricle of the brain, as well as peripheral locations such as the gastrointestinal tract (14–16). The blood-brain barrier (BBB) at the chemoreceptor trigger zone is incomplete (16), and therefore may not present a substantial barrier for ondansetron antiemetic action. However, when evaluating the therapeutic potential for ondansetron for neuropathic pain, understanding the factors that may impact achieving therapeutic concentrations at the site of action (in the CNS) becomes critical. For example, mouse data show that under normal conditions, 5HT₃ receptors are expressed in many areas of the brain, including the somatosensory cortex and amygdala, and partially in the spinal cord (17).

P-glycoprotein (Pgp) is expressed on the luminal membrane of brain endothelial cells thereby limiting CNS penetration of numerous compounds (18). Using an LLC-PK1 cell culture expressing human Pgp ondansetron was shown to be subject to Pgp efflux (18). Concentration of ondansetron in the brain of Pgp knock-out mice was 4-fold higher compared to wild type animals 30 minutes following intravenous (IV) injection of 0.2 mg/kg [¹⁴C]ondansetron (18).

In humans, ondansetron is administered IV (usually 4–8 mg every 8 hours) and orally (up to 8 mg three times daily) with bioavailability of 50–87%. The volume of distribution in adults is reported as 1.9 L/kg. The drug is metabolized by CYP1A2, CYP2D6, and CYP3A4 enzymes, and the half-life in adults is 3–6 hours and can be prolonged to 20 hours in severe hepatic impairment (19). Sex-dependent pharmacokinetics has been reported in humans and rats, with females showing a higher exposure (20–22). In rats, the area under the plasma concentration-time curve (AUC) of ondansetron after intravenous administration was 23% lower, and the systemic clearance was 26.8% higher in male rats compared to females (20). In humans, the AUC of ondansetron after oral intake was significantly lower in males compared to females (22).

The main goal of this study was to evaluate the role of Pgp in determining ondansetron exposure at different parts of the CNS (brain, spinal cord, and cerebrospinal fluid) using a genetic knockout rat model. An additional goal was to examine sex-dependent differences in plasma pharmacokinetics and CNS distribution. Mechanism-based pharmacokinetic model was constructed to quantify CNS disposition of ondansetron in all cohorts.

Materials and Methods

Materials

Ondansetron hydrochloride, N-benzylbenzamide, ethylenediaminetetraacetic acid tripotassium (K₃EDTA) were obtained from Sigma-Aldrich (St. Louis, MO, USA). Pooled male rat plasma and cerebrospinal fluid was purchased from BioIVT (Westbury, NY, USA). All solvents were of HPLC or higher grade and were purchased from Fisher Scientific (Fair Lawn, NJ, USA).

Animals

All animal studies were approved and conducted under guidelines of the Institutional Animal Care and Use Committee at Rutgers University. Wild-type Sprague-Dawley rats (WT) and P-glycoprotein knockout (KO) rats (Mdr1a(−/−), SD-Abcb1a^tm1sage) were purchased from Horizon Discovery (previously Sage Labs Inc., Boyertown, PA). Male WT and KO rats weighing 300–390 grams (10–12 weeks) and female WT and KO rats weighing 250–280 grams (12–14 weeks) were used in all studies. Animals were maintained in a vivarium with controlled temperatures and 12/12 hour dark and light cycle with free access to food and water. The jugular vein was cannulated to support intravenous drug administration and serial plasma sampling using polyethylene tubing (PE-50, Braintree Scientific, Braintree, MA) under light isoflurane anesthesia. After surgery, animals were allowed to recover for 24 hours and subcutaneous meloxicam and intradermal bupivacaine analgesia was provided.

Experimental design

For initial assessment of the effect of Pgp on plasma pharmacokinetics of ondansetron a single dose study with sequential blood sampling was conducted in male wild-type (WT-M) (n=3) and male Pgp knock-out (KO-M) (n=5) rats. Ondansetron (10 mg/kg, 10 mg/mL in water) was freshly prepared and administered intravenously through the jugular vein cannula followed by a saline flush (0.2 mL). Serial blood samples (300 µL) were collected at 0.25, 0.5, 0.75, 1, 1.5, 2, 2.5, 3, and 4 hours into EDTA tubes. Heparinized saline (20 IU/mL) was used for volume replacement after each sample. Plasma was separated by centrifugation (7 min, 13,000 RPM), transferred to a fresh tube, and stored in −20°C pending analysis.

Assessment of the effect of Pgp on CNS disposition of ondansetron was conducted next in male WT and KO animals. Animals previously used in the plasma pharmacokinetics study were used following a one-week washout period; and additional animals were included to achieve n=4–6 per time point. Freshly prepared ondansetron solution (10 mg/kg) was injected intravenously through the jugular vein cannula or through the tail vein. Animals were sacrificed at predetermined time points (0.16, 0.25, 0.5, 1, 1.5, 2, and 2.5 hours) under isoflurane anesthesia, and terminal samples were collected including plasma, cerebrospinal fluid (CSF), brain, and spinal cord and stored at −20°C (other tissues were also preserved for future analysis).

The study was expanded to include female rat cohorts (wild-type female (WT-F), Pgp knock-out female (KO-F)) using terminal sampling approach (n=4–6 per time point) as described above.

Sample analysis

The approach for analysis of ondansetron in plasma and CSF was based on our published method (23) that was developed with slight modifications to previously reported assays (24, 25). Briefly, 100 μL of plasma or CSF were mixed with 10 μL of a N-benzylbenzamide (100 μg/mL in acetonitrile - internal standard) and 300 μL of an alkalizing agent (0.1M NaOH for plasma or saturated sodium carbonate solution for CSF). Methyl-tert-butyl-ether (3 mL) was used for extraction; samples were vortexed for 10 min and centrifuged for 5 min at 1900g at 4°C (5810R centrifuge, Eppendorf, Hauppauge, NY, USA). Organic layer was transferred into a new test tube, and samples were then evaporated to dryness under nitrogen (TurboVap, Biotage, Charlotte, NC, USA). The residue was reconstituted with 100 μL of acetonitrile in water (3:7, v/v). Samples were vortexed at high speed for 5 min and transferred into HPLC vials.

Tissue homogenization was performed using the Bullet Blender 5E Gold (NextAdvance, Troy, NY, USA). Whole tissues were thawed over ice and washed with phosphate-buffered saline (PBS 1X) three times, or until residual blood was washed off. Tissues were then blotted dry and weighed. PBS 1X solution was added in a 2:1 v/w ratio to the tissues, and the homogenizing beads were added in a 1:1 w/w ratio to tissues. Speed and duration of homogenization cycle was optimized following protocols published by NextAdvance for each tissue (26). Homogenate was transferred to a fresh tube and frozen at −20°C for future analysis.

For extraction of ondansetron from tissue homogenates, Agilent Bond Elut solid phase extraction (SPE) cartridges were used (Agilent Technologies, Santa Clara, CA, USA) with a vacuum manifold. Homogenate (100 μL) was mixed with a 10 μL N-benzylbenzamide (100 µg/mL in acetonitrile - internal standard) and methanol (100 μL) and vortexed for 20s, then centrifuged at 13,000 rpm for 3 minutes. The supernatant was loaded onto SPE cartridges (preconditioned with 500 μL of methanol and equilibrated with 500 μL water). The column was first washed with 500 μL of 5% methanol in water, and the samples were eluted with two washes of 400 µL of methanol. Samples were evaporated to dryness under a stream of nitrogen gas. The residue was reconstituted with 100 μL of acetonitrile in water (3:7, v/v). Samples were vortexed at high speed for 5 min and transferred into HPLC vials.

Agilent 1260 Infinity HPLC (Santa Clara, CA, USA) equipped with a photodiode array detector was used in the study. The running conditions were based on our previously published method (23). Briefly, separation was achieved with a Phenomenex Gemini Twin Technology C18 column (particle size x length x diameter: 3 μm x 150 mm x 2 mm), protected by a SecurityGuard pre-column (Torrance, CA) maintained at 45°C. The mobile phase consisted of 5 mM ammonium acetate buffer (pH 4 adjusted with glacial acetic acid) and acetonitrile under gradient conditions that were published before (23), and the flow rate 0.6 mL/min. The injection volume was 40 μL. The detection wavelength was 310 nm and 275 nm with a retention time of 2.5 and 8.5 min for ondansetron and N-benzylbenzamide, respectively. The limit of detection was 10 ng/mL in plasma, 5 ng/mL in CSF, and 50 ng/mL in brain and spinal cord tissues.

Data analysis

A standard noncompartmental analysis was performed using individual plasma concentration-time profiles in WT-M and KO-M rats. The area under the plasma concentration-time curve from time zero to infinity (AUC_plasma), half-life (t_1/2,plasma), mean residence time (MRT), systemic clearance (CL) and volume of distribution at steady state (V_d,ss) were calculated for each animal using Phoenix WinNonlin version 8.1 (Certara, Princeton, NJ) and reported as mean ± standard deviation (SD). Pharmacokinetic parameters were compared between WT-M and KO-M groups using t-test, and p<0.05 was considered significant.

For plasma profiles in WT-F and KO-F rats and for all tissue samples ondansetron concentration is reported as mean ± SD for each time point; and noncompartmental analysis was conducted using naïve-averaged data. For brain, spinal cord, and CSF the area under the tissue concentration-time curve (AUC_tissue) and tissue half-life (t_1/2,tissue) are reported.

To provide an overall assessment of the effect of Pgp status (WT or KO) on ondansetron exposure, an average fold increase in concentration for each tissue was calculated by obtaining a ratio of mean concentrations for KO and WT animals at each time point and then calculating the mean (and SD) of these ratios across all time points. Furthermore, ondansetron partition coefficient (Kp) for each tissue was calculated as a ratio between the AUC_tissue (for brain, spinal cord, and CSF) and the AUC_plasma for each of the cohorts (WT-M, KO-M, WT-F, KO-F).

Pharmacokinetic modeling

Mechanistic pharmacokinetic model for describing ondansetron systemic disposition and distribution into various parts of the CNS was developed. Mean concentration time profiles for plasma, brain, spinal cord, and CSF were used. The model was first developed using data for male rats only and then evaluated for female rats. Initially WT-M and KO-M plasma datasets were used to construct systemic disposition model for ondansetron. Following IV administration ondansetron exhibited polyexponential plasma profiles; therefore, two- and three-compartment systemic disposition model with linear elimination were tested. Two-compartment model was able to satisfactorily describe the data, and inclusion of an additional peripheral compartment did not improve model fits. A semi-physiological CNS distribution model was developed using WT-M and KO-M data in a step-wise approach beginning with the brain, followed by spinal cord, and CSF compartments. Initially, the parameters estimated for systemic disposition were fixed while CNS model was developed (under the assumption that the total amount of the drug in the CNS is minimal compared to the rest of the body).

Interconnectivity of CNS compartments followed animal physiology, and previously published CNS model were consulted (27–29). The volume of the brain (V_brain) was fixed to the observed tissue size values in this work, which agreed with previously reported value of 1.8 mL (30, 31). The volume of the CSF compartment (V_CSF) was previously reported as 0.25 mL for a 250 g rat (31); these value was scaled based on the mean body weight of animals in this work. The volume of the spinal cord (V_spinal) in rats has not been reported before and the parameter was fixed to values measured in this study. A density of 1 was assumed for all tissues. All model parameters are presented in Table 4.

Table 4.

Final model estimated parameters for plasma pharmacokinetics and CNS disposition of ondansetron in male and female wild-type and Pgp knock-out rats

			Male		Female
Parameter	Parameter Description	Units	Estimate	CV%	Estimate	CV%
V1	Central volume of distribution	mL	14.6	10	18.9	5.6
kel	Systemic elimination rate constant	h−1	4.80	17	2.48	9.4
k12	Distribution rate constants to/from peripheral distribution compartment	h−1	3.01	31	1.28	21
k21		h−1	2.37	10	1.16	23
k13	Distribution rate constants to/from brain compartment	h−1	39.1	35	3.76	12
k31		h−1	73.2	32	14.6	13
k14	Distribution rate constants to/from spinal cord compartment	h−1	39.1	36	1.41	6.7
k41		h−1	188	34	31.7	9.1
k15	Distribution rate constants to/from CSF compartment	h−1	0.158	37	0.168	19
k51		h−1	55.9	50	28.5	26
k35=k45	Rate constant for brain/spinal cord and CSF exchange	h−1	0.481	37	0.842	14
k53=k54		h−1	24.4	40	74.4	9.7
kPgp	Rate constant for Pgp-mediated efflux from the CNS	h−1	133	34	47.9	12
Vbrain_WT	Brain volume in WT rats	mL	1.76*	-	1.62*	-
Vbrain_KO	Brain volume in KO rats	mL	1.81*	-	1.67*	-
Vspinal_WT	Spinal cord volume in WT rats	mL	0.65*	-	0.63*	-
Vspinal_WT	Spinal cord volume in KO rats	mL	0.55*		0.57*	-
VCSF	CSF volume	mL	0.28*	-	0.20*	-
var_P	Variance		0.05	18	0.03	19

^*Parameters were fixed to physiological values

Schematic of the final model is presented in Figure 1. All CNS compartments were directly connected to the central disposition compartment (with the volume V₁) with first-order rate constants (k₁₃, k₃₁, k₁₄, k₄₁, k₁₅, k₅₁) to describe a bidirectional passive permeability. Physiologically, the CSF and CNS extracellular fluid have areas for nutrient and compound exchange, therefore distributional terms between CSF and the brain, CSF and spinal cord were included (k₃₅, k₅₃, k₄₅, k₅₄) (32, 33). Pgp efflux at the BBB was described using a first-order term (k_Pgp). Pgp efflux was previously shown to occur at the barrier between the blood and the spinal cord tissue, similar to the BBB; therefore, k_Pgp term was also included to describe Pgp efflux of ondansetron from the spinal compartment to plasma (34, 35). The following differential equations were used to describe the model:

$$\frac{d{A}_1}{dt}=k_{21}\cdot A_2+k_{31}\cdot A_3+k_{41}\cdot A_4+k_{51}\cdot A_5-\left(k_{el}+k_{12}+k_{13}+k_{14}+k_{15}\right)\cdot A_1+k_{Pgp}\cdot A_3+k_{Pgp}\cdot A_4$$ (1)

$$\frac{d{A}_2}{dt}=k_{12}\cdot A_1-k_{21}\cdot A_2$$ (2)

$$\frac{d{A}_3}{dt}=k_{13}\cdot A_1+k_{53}\cdot A_5-\left(k_{31}+k_{35}+k_{Pgp}\right)\cdot A_3$$ (3)

$$\frac{d{A}_4}{dt}=k_{14}\cdot A_1+k_{54}\cdot A_5-\left(k_{41}+k_{45}+k_{Pgp}\right)\cdot A_4$$ (4)

$$\frac{d{A}_5}{dt}=k_{15}\cdot A_1+k_{35}\cdot A_3+k_{45}\cdot A_4-\left(k_{51}+k_{53}+k_{54}\right)\cdot A_5$$ (5)

where A1, A2, A3, A4, A5 represent the amount of ondansetron in the central, peripheral, brain, spinal cord, and CSF compartments. All parameters were shared between WT-M and KO-M datasets, except for k_Pgp term that was estimated for the WT-M and set equal to zero for the KO-M. During final model runs, all model parameters were estimated simultaneously using WT-M and KO-M data for plasma, brain, spinal cord, and CSF. At the next stage, the final model structure established for male animals was applied for female cohorts (WT-F and KO-F) and a separate set of parameters was estimated.

Figure 1.

Schematic of the semi-physiological pharmacokinetic model used to capture systemic disposition and CNS distribution of ondansetron in male and female wild type and Pgp knockout rats. Model equations and parameters are described in Methods and Table 4. k_Pgp was included only for wild type animals.

R (version 3.31) and Rstudio (version 1.2.5001, Boston, MA, USA) with Ubiquity package were used for model development and estimation of the parameters (36). Nelder-Mead Optimization method was used and a variance model was defined as: $VA{R}_i={\left({\text{σ}}_1\cdot \text{ϒ}\left(\theta ,t_i\right)\right)}^2$ where $VA{R}_i$ is the variance of the ith data point, ${\text{σ}}_1$ is the variance model parameter, and $\text{ϒ}\left(\theta ,t_i\right)$ is the ith estimated value from the pharmacokinetic model. The model performance was evaluated by visual inspection of fitted curves, system convergence, Akaike Information Criterion, and objective function value.

Results

Plasma pharmacokinetics and disposition of ondansetron to various regions of the CNS were investigated in male and female WT and Pgp KO rats after IV administration of a single 10 mg/kg dose. Physiological measurements of the size of the brain and spinal cord were obtained and used in pharmacokinetic model development (Table 1). Initially, the concentration of ondansetron in upper and lower parts of the spinal cord in WT-M and KO-M rats was determined separately. The concentrations were not statistically different, and a single concentration measurement was adopted for the rest of the study.

Table 1.

Weight of brain and spinal cord in male and female wild type and Pgp knockout rats, mean (SD)

Tissue	WT-M (n=28)	KO-M (n=25)	WT-F (n=31)	KO-F (n=43)
Brain	1.76 (0.14)	1.81 (0.19)	1.62 (0.20)	1.67 (0.21)
Spinal cord	0.650 (0.15)	0.550 (0.08)	0.636 (0.06)	0.577 (0.08)

Observed concentration-time profiles of ondansetron in plasma of WT-M and KO-M rats were almost superimposable (Figure 2). Results of the noncompartmental analysis of individual profiles are presented in Table 2. While there was a statistically significant difference in t_1/2,plasma, the rest of the parameters were similar. Observed concentration-time profiles of ondansetron in plasma of WT-F and KO-F rats were also superimposable. The results of the noncompartmental analysis of plasma naïve averaged data are presented in Table 2. Due to limitations in the limit of detection for ondansetron in the CNS, plasma pharmacokinetic study in female rats and tissue disposition study in both male and female rats were not conducted beyond 2.5 hours. Comparison of noncompartmental parameters between WT-M and WT-F cohorts showed that V_d,ss and CL were higher in female rats, similar to previous reports (20, 21).

Figure 2.

Concentration-time profiles of ondansetron in plasma, brain, spinal cord and cerebrospinal fluid in male WT and Pgp KO animals following IV bolus 10 mg/kg dose of ondansetron. Symbols are observed data (shown as mean±SD) and lines are model fits.

Table 2.

Noncompartmental plasma pharmacokinetic parameters for ondansetron following IV dosing of 10 mg/kg in wild-type and Pgp knockout rats

	Male		Female*
Parameter	Wild-type	Knock-out	Wild-type	Knock-out
t1/2,plasma (h)	0.52 (0.01)	0.65 (0.04)	0.54	0.64
AUCplasma,0−∞ (h·μg·mL−1)	3.83 (0.64)	3.70 (0.49)	2.52	2.57
MRT (h)	0.47 (0.04)	0.47 (0.10)	0.76	0.72
Vd,ss (mL)	417 (98)	422 (124)	802	766
CL (mL·h−1)	869 (132)	892 (120)	1050	1029

^*In female rats, concentrations were obtained from terminal sampling; therefore, individual profiles were not available, and SD could not be calculated.

CNS tissue disposition study showed that in both male and female rats Pgp genetic knockout resulted in significantly higher concentrations of ondansetron in all tested regions of the CNS at most of the time points (Figures 2 and 3). On average, ondansetron CNS concentrations were 2.39–5.48 higher in KO animals, depending on the region of the CNS (Table 3). Tissue-to-plasma partition coefficients (Kp) were less than 1 for brain and spinal cord in WT animals and were increased to 1.34 – 3.04 in KO animals. CSF:plasma partition coefficient was lower than partition to the brain or spinal cord.

Figure 3.

Concentration-time profiles of ondansetron in plasma, brain, spinal cord and cerebrospinal fluid in female WT and Pgp KO animals following IV bolus 10 mg/kg dose of ondansetron. Symbols are observed data (shown as mean±SD) and lines are model fits.

Table 3.

Noncompartmental pharmacokinetic parameters for ondansetron following IV dosing of 10 mg/kg for brain, spinal cord, and cerebrospinal fluid in wild-type and Pgp knockout rats

	Brain				Spinal Cord				CSF
Parameter	WT-M	KO-M	WT-F	KO-F	WT-M	KO-M	WT-F	KO-F	WT-M	KO-M	WT-F	KO-F
t1/2,tissue (h)	0.57	0.50	0.56	0.63	0.62	0.65	0.92	0.59	0.25	0.51	0.71	0.59
AUCtissue,0−∞ (h·μg·mL−1)	1.78	9.33	1.78	7.83	1.50	6.62	1.18	3.45	0.25	0.71	0.27	0.59
KP,AUC	0.46	2.52	0.71	3.04	0.39	1.78	0.47	1.34	0.06	0.19	0.11	0.22
Average ratio of KO/WT concentrations, mean (SD*)	5.48 (1.7)		4.37 (0.6)		4.52 (1.5)		3.61 (1.4)		2.91 (1.3)		2.39 (0.7)

^*For the rest of parameters, SD could not be calculated because naïve-averaged data was used.

t_1/2,tissue (half-life in tissue), AUC_{tissue,0−∞} (Area under the curve in tissues from time zero to time infinity), K_P,AUC (drug partitioning based on the ration of AUC values between the tissue and the corresponding plasma profile)

Comparison of pharmacokinetic profiles between male and female rats for both strains is shown in Figure 4. For plasma and CSF the concentrations were similar in two sexes at most time points; however, there was a trend for lower initial concentrations and higher concentration at later time points in females. For the brain and especially for the spinal cord the concentrations in males were higher compared to female rats of both strains.

Figure 4.

Comparison of the observed pharmacokinetic profiles between male and female rats in WT (filled symbols) and Pgp KO (open symbols) strains following IV bolus 10 mg/kg dose of ondansetron. Data are shown as mean±SD

Mechanistic pharmacokinetic model was successfully developed to simultaneously describe plasma pharmacokinetics and CNS disposition of ondansetron. The final model included two compartments to describe systemic disposition and three physiological CNS compartments (i.e., brain, spinal cord, and CSF) – Figure 1. All CNS compartments were connected to the central distribution compartment with bidirectional exchange processes. In addition, in WT animals the model included Pgp-mediated efflux from brain and spinal cord compartments. The rest of the parameters were shared between WT and KO strains; and parameter estimation was performed separately for male and female rats. The final model provided good simultaneous description of all 4 tested tissues (plasma, spinal cord, brain, and CSF) in WT and KO strains for male (Figure 2) and female (Figure 3) rats. In the final model, for simplicity and to improve the precision of parameter estimated the rate constant for exchange between CSF and brain and CSF and spinal cord compartments were set to be equal (k₃₅=k₄₅ and k₅₃=k₅₄). All parameters were estimated with sufficient precision (Table 4).

Discussion

Neuropathic pain is a common and debilitating condition, with limited therapeutic options available for its management. Antagonism of 5HT₃ receptors is a promising approach for treatment of neuropathic pain, and direct delivery of 5HT₃ receptor antagonists to the site of action in the CNS was shown to be efficacious in preclinical models of painful neuropathy (7, 37). However, the results of several small clinical studies with systemic administration of 5HT₃ receptor antagonists were contradictory (9, 38). Pgp transporter is known to affect drug pharmacokinetics, and especially permeability to the CNS; furthermore, Pgp genetic polymorphism in humans is known to affect drug efficacy (18, 39, 40). We hypothesized that polymorphism in Pgp expression and the extent of Pgp-mediated efflux from the CNS in various patients may have contributed to contradictory outcomes of clinical trials with 5HT₃ receptor antagonists. Since a direct assessment of CNS disposition of drugs in humans is rarely feasible, in this preclinical study we performed a mechanistic assessment of the role of Pgp in exposure of various regions of the CNS to ondansetron using a genetic Pgp knockout rat model. Selection of the KO rat model offers an important advantage over KO mice by allowing for a significantly larger size of biological specimens, which facilitates bioanalytics. Previously, Pgp KO mice and rats have been shown to provide comparable results in brain disposition studies (41); and pharmacokinetics of other Pgp substrates (loperamide, paclitaxel) was shown to be altered in these Pgp KO rats (42, 43).

In this study, we found that Pgp KO has not significantly affected plasma pharmacokinetics of ondansetron, as plasma profiles in KO rats overlapped with profiles in both male and female WT animals (Figures 2 and 3). Our findings support a previous report in which no difference was detected in plasma concentrations 30 min after IV administration of 0.2 mg/kg ondansetron in wild-type and mdr1a KO mice (18, 40). In contrast, the concentration of ondansetron in all tested CNS regions in Pgp KO rats was increased on average 2.4–5.5-fold, compared to WT animals. These data are in line with a previously reported 4-fold increase in ondansetron concentration in the brain of Pgp KO mice compared to WT (only a single time point data were available) (18).

Differences in plasma pharmacokinetics of ondansetron have been reported between sexes in both rats and humans (20, 21). Plasma AUC was 23% lower in male rats following intravenous dose of 8 mg/kg (p<0.05) (20). In a clinical study, women were observed to have consistently higher AUC values after dosing of an oral or two suppository formulations (p < 0.05) (22). In this study in rats, we found little difference in plasma profiles of ondansetron between sexes (Figure 4); there was a slight trend for lower initial concentrations and higher concentrations at later time points. The differences are reflected in noncompartmental parameters (Table 2). The lack of a more significant differences compared to previously published results may be related to the experimental design (a shorter time frame for sample collection and terminal sampling). From the CNS data, the spinal cord concentrations were substantially higher and brain concentrations were a little higher in male rats. There was no significant difference for the CSF. The reason for differences in CNS distribution between sexes is currently unknown and should be investigated in future studies. Overall, male and female data could not be reliable fitted simultaneously using a shared set of parameters; therefore, parameters estimation was performed separately for two sexes (Table 4). Information comparing Pgp expression or other physiological differences at the level of BBB between males and females could not be found in the literature; and sex-dependent differences in CNS disposition of 5HT₃ receptor antagonists warrant further investigation.

In this study we evaluated ondansetron disposition into three different regions of the CNS. Partitioning into the spinal cord is usually not considered in pharmacokinetic analysis; and CNS disposition studies and CNS pharmacokinetic modeling (empirical, semi-physiological, or physiologically-based) are primarily focused on the brain and sometimes include the CSF (28, 29, 44). Only a single example including the spinal cord into a population pharmacokinetic model to describe an antisense oligonucleotide in a multi-compartmental CNS model could be found in the literature (45). Ondansetron can bind to 5-HT₃ receptors expressed in the brain and the spinal cord; and understanding the exposure in the spinal cord disposition is required to ultimately connect the pharmacokinetics to ondansetron efficacy in neuropathic pain studies, as the key site of action may be in the spinal cord dorsal horn. Ondansetron concentrations were shown to be similar in the upper and the lower parts of the same spinal cord (collected spinal cord was cut approximately in the middle at the thoracic level). To the best of our knowledge, such data have not been reported before. In this study, we showed that total ondansetron exposure (AUC_tissue) is different in various CNS matrices (highest in the brain, and lowest in the CSF). Pgp KO increased ondansetron concentration in all CNS regions; however, the magnitude of the effect was dependent on the region. Importantly, using CSF values for quantitative assessment may underestimate the role of Pgp on drug exposure in the brain or the spinal cord. These findings are important for translational research because CSF samples can often be used as a surrogate matrix to evaluate overall CNS drug disposition in humans (46, 47). CSF can be relatively easily sampled through cisterna magna puncture in preclinical species and through lumbar punctures or intrathecal catheter implantations in human subjects (46). We have previously reported CSF ondansetron partition coefficient of 0.15 in a population pharmacokinetic analysis in a human study (48), which is similar to a value for CSF partition found in this study (Table 3).

An important goal of the work was to construct a mechanistic pharmacokinetic model to quantitatively describe the role of Pgp efflux on ondansetron exposure in the CNS. The final model successfully captured the time-course of ondansetron in plasma, brain, spinal cord and CSF; and all parameters (except for K_Pgp) were shared between WT and KO animals. Model parameters were estimated with sufficient precision, and separate sets were needed for male and female rats. Successful fitting of the model was supported by simultaneous analysis of measurements in the plasma and three CNS compartments, and use of data from both strains (WT and KO). Alternative structural models were evaluated during data analysis, including models with a unidirectional transfer from the brain to the CSF or division of brain and/or spinal cord into two subcompartments. For example, models that included a “deep-tissue” compartment in the brain had been reported (44, 49). None of these models were able to provide superior data description based on model evaluation criteria. We have also attempted to include separate parameters for describing Pgp-mediated efflux in the brain and the spinal cord; however, these could not be estimated with sufficient precision, and the final model only included a shared parameter for this process. The final model allowed to estimation of parameters with sufficient precision; however, rate constants included in the model may represent a combination of several processes, and therefore, direct physiological interpretation may not yet be possible. For example, a product of V₁ and k₁₃ parameters is higher than the reported blood flow to the rat brain. It can be seen in Figures 2 and 3 that initial brain concentration is slightly overestimated by the current model (this may be related to the timing of the first samples in our studies). Several physiological models focusing on the CNS were published before, and mostly focused on the brain and the CSF (28, 29) and did not include the spinal cord. The weight of the rat spinal cord is reported in this study (Table 1). Development of more physiological-based models with additional tissue subcompartment will necessitate data with a higher spatial and temporal resolution. For example, microdialysis of selected regions of the CNS can be used and can also provide unbound drug concentrations.

Conclusion

In conclusion, the study provides important quantitative information on the role of Pgp in limiting ondansetron exposure in various regions of the CNS using data from WT and Pgp KO rats. CSF drug concentrations, as a surrogate to CNS exposure, are likely to underestimate the effect of Pgp on drug penetration to the brain and the spinal cord. Male and female animals demonstrated some difference in ondansetron plasma pharmacokinetics and CNS disposition. Mechanistic model was developed and successfully captured the data in all tissues in all study groups. Proposed modeling framework could serve as the base to further analysis of the potential use of Pgp inhibitors in enhancing delivery of 5HT₃ receptor antagonists to the CNS.