In Silico Absorption Analysis of Valacyclovir in Wildtype and Pept1 Knockout Mice Following Oral Dose Escalation

Abstract

Purpose

We developed simulation and modeling methods to predict the in vivo pharmacokinetic profiles of acyclovir, following escalating oral doses of valacyclovir, in wildtype and Pept1 knockout mice. We also quantitated the contribution of specific intestinal segments in the absorption of valacyclovir in these mice.

Methods

Simulations were conducted using a mechanistic advanced compartmental absorption and transit (ACAT) model implemented in GastroPlus™. Simulations were performed for 3 hours post-dose in wildtype and Pept1 knockout mice following single oral doses of 10, 25, 50 and 100 nmol/g valacyclovir, and compared to experimentally observed plasma concentration-time profiles of acyclovir.

Results

Good fits were obtained in wildtype and Pept1 knockout mice. Valacyclovir was primarily absorbed from duodenum (42%) and jejunum (24%) of wildtype mice, with reduced uptake from ileum (3%) and caecum/colon (1%), for a total of 70% absorption. In contrast, the absorption of valacyclovir in Pept1 knockout mice was slow and sustained throughout the entire intestinal tract in which duodenum (4%), jejunum (14%), ileum (10%) and caecum/colon (12%) accounted for a total of 40% absorption.

Conclusion

The ACAT model bridged the gap between in situ and in vivo experimental findings, and facilitated our understanding of the complicated intestinal absorption processes of valacyclovir.

INTRODUCTION

The proton-coupled oligopeptide transporter PEPT1 is a membrane-bound protein responsible for the uptake of di-/tri-peptides and a broad range of peptidomimetics in mammals. PEPT1 has been extensively studied in regard to its structural characteristics, transport mechanism, substrate specificity, tissue distribution, regulation, and physiological and pharmaceutical relevance (1–5). In humans and mice, PEPT1 is abundantly expressed at the apical membrane of epithelial cells in small intestine, the main site of absorption for orally administered drugs. It exhibits low affinity and high capacity in facilitating the transcellular uptake of substrates across the intestinal cell barrier. Because of its favorable transport characteristics and localization, intestinal PEPT1 has long been viewed as a promising target for drug delivery, and PEPT1-targeting prodrugs have been synthesized to improve the intestinal permeation of otherwise poorly absorbable parent compounds (6–9).

Valacyclovir is an L-valyl ester prodrug of the potent antiviral agent acyclovir and is viewed as a model PEPT1-targeting prodrug. Intensive effort has been focused on delineating the quantitative importance of PEPT1 in the intestinal absorption of valacyclovir, which understandably lays a solid basis for promoting the rational design of prodrugs that selectively target intestinal PEPT1. In our laboratory, we evaluated the role and relevance of PEPT1 by performing in situ intestinal permeability and in vivo pharmacokinetic studies of valacyclovir in wildtype and Pept1 knockout mice (10,11). The in situ intestinal perfusion studies demonstrated that valacyclovir was actively transported by intestinal PEPT1, which contributed approximately 90% of the permeability of valacyclovir in mouse small intestine (10). Moreover, during the in vivo pharmacokinetic studies of orally administered valacyclovir, 5-fold and 2-fold differences were observed between wildtype and Pept1 knockout mice in acyclovir peak plasma concentrations (C_max) and area under the plasma concentration-time curves from time 0 to 180 min (AUC_0–180), respectively (11). Based on the in situ and in vivo results, we concluded that PEPT1 had a major role in mediating the intestinal absorption of valacyclovir. However, these methods were rather descriptive and lacked a systematic means to interpret the marked differences observed in acyclovir plasma concentration-time profiles between genotypes and the quantitative importance of each intestinal segment in valacyclovir absorption.

The intestinal absorption of valacyclovir, using in situ and in vivo experimental findings, was integrated into a mechanism-based absorption model to predict a priori the involvement of PEPT1 in the intestinal absorption of orally administered valacyclovir and the systemic exposure of acyclovir. The intestinal absorption of valacyclovir is influenced by numerous physical-chemical, physiological and biochemical variables such as pH, absorption surface area, gastric emptying, intestinal transit times and transporters. Therefore, modeling approaches offer a unique opportunity to assess the impact of changes in these specific variables on the rate and extent of valacyclovir intestinal absorption.

Several mechanism- or physiology-based absorption models have been published, such as the Grass (12), compartmental absorption and transit (CAT), gastrointestinal transit absorption (GITA), advanced compartmental absorption and transit (ACAT), and advanced dissolution, absorption, and metabolism (ADAM) models (13). All these examples utilize compartment models that take into account fluid movement along the digestive tract and absorption over time from each gastrointestinal segment. Simulation analyses using these models have greatly facilitated our understanding of the impact of various formulation-related, and physicochemical and physiological factors on oral drug absorption (14–17). The ACAT model, modified from the original CAT model (18,19) to include first-pass metabolism and colonic absorption, is flexible enough to account for the intestinal processes of drug release, dissolution, precipitation, luminal degradation, active and passive uptake, and gastrointestinal transit (20). GastroPlus implements the ACAT model and has been successfully used in a number of recent simulation studies investigating the oral absorption of various drugs, including valacyclovir, via transporter and/or metabolizing enzyme pathways (21–23).

There were two major objectives in performing this analysis. The first was to develop simulation and modeling methods for predicting the in vivo pharmacokinetic profiles of acyclovir, following escalating oral doses of valacyclovir, in wildtype and Pept1 knockout mice. This was accomplished by integrating experimental findings regarding the in situ intestinal permeability and degradation of valacyclovir, expression levels of PEPT1 in the intestines, and in vivo disposition of acyclovir into the ACAT model. The second major objective was to quantitate the contribution of specific intestinal segments in the absorption of valacyclovir in PEPT1-expressing mice and during PEPT1 ablation.

MATERIALS AND METHODS

Computer Hardware and Software

Simulations were conducted on a standard Dell desktop computer using GastroPlus™ (version 7, Simulations Plus, Lancaster, CA, USA).

ACAT Model

The ACAT model implemented in GastroPlus was used to simulate in silico the pharmacokinetics of valacyclovir following escalating oral administration in wildtype and Pept1 knockout mice (11). The intestinal fate of orally dosed valacyclovir solution was governed by three simultaneous kinetic processes in the gastrointestinal (GI) tract including GI transit, intestinal absorption and luminal degradation, based on findings from the previous in situ perfusion studies (10). Simulations were then performed by coupling the three intestinal processes of valacyclovir (i.e., GI transit, absorption and degradation) to a pharmacokinetic compartment model for intravenously administered acyclovir.

Using GastroPlus, the GI transit of dissolved valacyclovir was modeled as a first-order process across a sequence of 9 consecutive compartments corresponding to the stomach and 8 different regions of the small and large intestines. The luminal degradation of valacyclovir was extensive in mouse (24) and modeled as a first-order pH dependent process. Valacyclovir GI transit and luminal degradation processes were modeled the same way for wildtype and Pept1 knockout mice based on the assumption that these two processes were not affected by Pept1 ablation. The key difference in the intestinal handling of valacyclovir between genotypes resides in the absorption of drug-related species across the apical membrane of enterocytes. As reported in our perfusion studies (10), permeability of valacyclovir in the small intestine of wildtype mice was quite high and attributable primarily to PEPT1-mediated and, to a lesser extent, passive absorption. Moreover, permeability of the luminal degradation product acyclovir was found to be insignificant in the small intestine. Therefore, only the absorption of intact valacyclovir, via active and passive pathways, was modeled in wildtype mice. In contrast, intestinal permeability of valacyclovir was similarly low to that of acyclovir during PEPT1 ablation. As a result, the intestinal absorption of both valacyclovir and acyclovir were modeled in Pept1 knockout mice. Mathematically, the PEPT1-mediated absorption of valacyclovir was modeled in wildtype mice using a Michaelis-Menten equation while absorption, via passive diffusion, was modeled as a first-order process in both wildtype and Pept1 knockout mice. Once absorbed into epithelial cells, valacyclovir was assumed to undergo instantaneous hydrolysis via biphenyl hydrolase-like protein to generate its active species so that only acyclovir was available to the systemic circulation. This assumption was reasonable based on experimental evidence that only acyclovir was measurable in the portal vein of blood samples obtained from both genotypes (10). All the kinetic processes described above were expressed by a system of coupled linear and non-linear differential equations that can be integrated numerically in GastroPlus.

Input Parameters

Model-dependent input parameters had to be specified in order to perform the pharmacokinetic simulations. Thus, GastroPlus included parameters associated with dose, the physical-chemical properties of valacyclovir and acyclovir, mouse GI physiology including intestinal PEPT1 protein expression, valacyclovir luminal degradation, valacyclovir and acyclovir permeability, and the in vivo pharmacokinetics of intravenous bolus acyclovir.

Physical-Chemical, Formulation and Dose Information

The physical-chemical properties, formulation and dose information for valacyclovir and acyclovir are summarized in Table I. Default values in GastroPlus were used when no other information was available.

Table I.

Physical-Chemical, Formulation and Dose Information of Valacyclovir and Acyclovir

Property	Valacyclovir	Acyclovir	Source
Molecular formula	C13H20N6O4	C8H11N5O3
Molecular weight	324.3	225.2
Predicted log P (neutral)	−0.95		Bolger et al (21)
pKa1	1.9		Balimane and Sinko (23)
pKa2	7.47		Balimane and Sinko (23)
pKa3	9.43		Balimane and Sinko (23)
Aq. solubility (pH 7)	174 mg/mL		Bolger et al (21)
Dosage form	Valacyclovir solution
Doses (nmol/g)	10.0	6.9
	25.0	17.4
	50.0	34.7
	100.0	69.4
Dose volume	0.2 mL
Mean precipitation time	900 sec		GastroPlus default value
Diffusion coefficient	7.50×10−6 cm/sec		GastroPlus default value
Drug particle density	1.2 g/mL		GastroPlus default value
Particle size	25 μm		GastroPlus default value

Mouse GI Physiology

Information regarding mouse GI transit times, segmental volumes, length (L), radius (R), and luminal pH profiles along the GI tract were provided by GastroPlus as the default values in fasted mice. These physiological parameters were used internally by GastroPlus for the calculation of first-order transit rate constants governing the movement of dissolved valacyclovir along consecutive GI compartments. Luminal pH profiles were also used in the estimation of first-order degradation rate constants characterizing the luminal hydrolysis of valacyclovir.

Absorption scale factors (ASF) were needed to convert intestinal permeability (denoted as P_eff) of each segment to a corresponding absorption rate constant (denoted as k_a), based on the relationship between P_eff and k_a:

$$k_a=\frac{2\ast P_{\mathit{eff}}}{R},\mathit{ASF}=\frac{2}{R}$$ (1)

Note that other models for the calculation of ASF were also available in GastroPlus. However, the above relationship was chosen because it offered the best model fit.

Intestinal expression of PEPT1 in mice was extracted from a study reported by Jappar et al (25). In this study, it was found that PEPT1 protein levels were highest in the jejunum, followed closely by levels in the duodenum and ileum. PEPT1 was not detected in the colon of mice. Thus, in our simulations PEPT1 was assumed to be uniformly distributed along the small intestine but absent from the large intestine of wildtype mice. Not surprising, PEPT1 protein was not expressed in the small or large intestine of Pept1 knockout mice.

Valacyclovir Luminal Degradation

The luminal degradation of valacyclovir was modeled as a first-order process characterized by the degradation rate constant k_de, which was estimated by reanalyzing in situ intestinal perfusion results obtained in the jejunum of wildtype mice at pH 5.5, 6, 6.5, 7 and 7.5 (26), as well as those perfusion results obtained in different intestinal segments of wildtype mice at pH 6.5 (10). The derivation of k_de is described below.

During the intestinal perfusion studies, absorption of luminally-formed acyclovir was assumed negligible when estimating the in situ permeability of valacyclovir. This assumption was valid considering the marked difference in permeability between valacyclovir and acyclovir in wildtype mice. Based on this assumption, the following mass balance equations hold for each intestinal segment under steady-state perfusion conditions:

$$\mathit{Rate}\mathit{of}\mathit{valacyclovir}\mathit{degradation}=\mathit{Rate}\mathit{of}\mathit{acyclovir}\mathit{formation}$$ (2)

$$\mathit{Rate}\mathit{of}\mathit{valacyclovir}\mathit{degradation}=k_{\mathit{de}}\ast V\ast C_{\mathit{ave}}$$ (3)

$$\mathit{Rate}\mathit{of}\mathit{acyclovir}\mathit{formation}=Q\ast C_{\mathit{out},\mathit{acyclovir}}$$ (4)

where V is the volume of a given intestinal segment, C_ave is the logarithmic mean of valacyclovir luminal concentration within the intestinal segment, Q is the perfusion flow rate (0.1 mL/min) and C_{out,acyclovir} is the acyclovir concentration exiting the perfused segment. C_ave was calculated using the measured inlet (C_{in,valacyclovir}) and outlet (C_{out,valacyclovir}) concentrations of valacyclovir:

$$C_{\mathit{ave}}=\frac{(C_{\mathit{in},\mathit{valacyclovir}}-C_{\mathit{out},\mathit{valacyclovir}})}{\ln (C_{in,\mathit{valacyclovir}}/C_{\mathit{out},\mathit{valacyclovir}})}$$ (5)

By combining equations 2–4, k_de can be calculated as:

$$k_{\mathit{de}}=\frac{Q\ast C_{\mathit{out},\mathit{acyclovir}}}{V\ast C_{\mathit{ave}}}$$ (6)

Using equation 6, k_de was estimated in jejunum at different pH values. Subsequently, the relationship between k_de or log-transformed k_de (log-k_de) and pH was modeled using a linear function, and goodness of fit was assessed by the coefficient of determination r². The better-fitting function was then used to estimate k_de at pH values found in the GI tract of fasted mice. k_de was also estimated in different intestinal segments at pH 6.5 to assess whether or not regional differences existed at the same pH value.

In Situ Intestinal Permeability of Valacyclovir in Wildtype Mice

Total permeability of valacyclovir in the small intestine of wildtype mice was comprised of a major PEPT1-mediated component (denoted as P_eff,PEPT1) and a minor passive component (denoted as P_eff,passive). P_eff,passive alone accounted for the total permeability of valacyclovir in large intestine due to the absence of PEPT1 in this segment. For the pharmacokinetic simulations in wildtype mice, P_eff,passive was assumed to be equal in both the small and large intestines, and initially set at 0.27×10⁻⁴ cm/sec, the in situ P_eff of valacyclovir measured in the colon of wildtype mice during previous perfusion studies (10).

P_eff,PEPT1 was described by the Michaelis-Menten equation:

$$P_{\mathit{eff},\mathit{PEPT}1}=\frac{V_{\mathit{max}}\ast C_{\mathit{luminal}}}{K_m+C_{\mathit{luminal}}}$$ (7)

where the maximum transport velocity V_max was estimated to be 1.4 nmol/cm²/sec and the apparent Michaelis constant K_m estimated to be 10 mM in single-pass concentration-dependent perfusion studies performed in the jejunum of wildtype mice (10). Note that the in situ V_max estimate was reported for completeness; however, this in situ value could not be directly converted to the in vivo V_max needed in the simulations. Instead, the in vivo V_max was obtained through an optimization step as described later in the “Parameter Optimization” section. C_luminal was the time-varying luminal concentration of valacyclovir generated by GastroPlus. In the simulations, V_max and K_m were assumed to remain constant over the entire small intestine. P_eff,PEPT1 was calculated internally by GastroPlus using equation 7 after considering the regional expression levels of PEPT1 in the small intestine of wildtype mice.

In Situ Intestinal Permeability of Valacyclovir and Acyclovir in Pept1 Knockout Mice

As described previously in the ACAT Model section above, the absorption of both valacyclovir and acyclovir needed to be simulated in Pept1 knockout mice because of extensive luminal hydrolysis of valacyclovir to acyclovir, and the similarly low intestinal permeability of both species. However, the version of GastroPlus used for the present pharmacokinetic simulations did not have the capability of modeling simultaneously the intestinal absorption of two compounds. A solution to circumvent this limitation was to assume the same passive permeability of valacyclovir and acyclovir so that these two compounds were treated equally in terms of their intestinal absorption process. This assumption was reasonable based on previous perfusion studies demonstrating the similar and low intestinal permeabilities of valacyclovir and acyclovir in Pept1 knockout mice (10).

In the simulations, initial input values of valacyclovir (and acyclovir) permeability in Pept1 knockout mice were based on four approaches using results obtained from previous in situ perfusion studies (10): Method #1, corresponding to the mean P_eff of valacyclovir in small intestine; Method #2, corresponding to the mean P_eff of valacyclovir in both small and large intestines; Method #3, corresponding to the mean P_eff of valacyclovir in duodenum; and Method #4, corresponding to the mean P_eff of acyclovir in jejunum. In doing so, the four methods could be evaluated separately to assess which P_eff estimate resulted in the best model predictions. The four P_eff values, along with their 90% confidence intervals, are summarized in Table II. Although the intestinal P_eff estimates of valacyclovir and acyclovir were numerically different in Pept1 knockout mice, the differences were not statistically significant.

Table II.

In Situ Intestinal Permeability (P_eff) of Valacyclovir or Acyclovir in Pept1 Knockout Mice

Peff (estimation method)	Mean value (×10−4 cm/sec)	90% CI (×10−4 cm/sec)
Valacyclovir Peff in small intestine (#1)	0.18	(0.04, 0.32)
Valacyclovir Peff in entire intestine (#2)	0.23	(0.11, 0.35)
Valacyclovir Peff in duodenum (#3)	0.27	(−0.02, 0.57)
Acyclovir Peff in jejunum (#4)	0.074	(−0.023, 0.72)

In Vivo Pharmacokinetics of Intravenous Acyclovir

Individual plasma concentration-time curves of acyclovir in wildtype and Pept1 knockout mice, following a 25 nmol/g intravenous bolus dose of acyclovir, were fitted to one-, two- and three-compartment models using the PKPlus™ module of GastroPlus. Akaike’s Information Criteria (AIC) and Schwarz’s Bayesian Criteria (SBC) were used to select the most appropriate model. Since these acyclovir plasma concentrations were virtually superimposable over time, as reported previously in wildtype and Pept1 knockout mice (11), the estimated values from both genotypes were combined and then used to calculate mean pharmacokinetic parameters for initial input in the simulations. The unbound fraction of acyclovir in plasma was set at 0.87, which was obtained from the literature (27).

Modeling and Simulation Procedure

The plasma concentration-time profiles of acyclovir were simulated in wildtype and Pept1 knockout mice following single oral doses of 10, 25, 50 and 100 nmol/g valacyclovir. The simulations followed three sequential steps, including parameter sensitivity analysis (PSA), parameter optimization, and comparison of the simulated pharmacokinetic profiles of acyclovir with previously reported experimental data (11). Acyclovir results from the 25 nmol/g oral dose of valacyclovir in wildtype mice served as a “training set” in evaluating the PSA and optimization procedures, while the remaining results (i.e., other three oral dose groups) were used as external validation datasets in the final comparison step.

Parameter Sensitivity Analysis

Inaccurate input parameters could impair the predictive performance of the specified ACAT model. Therefore, PSA was performed to evaluate which input parameters most influenced the simulated plasma concentrations of acyclovir, as well as C_max and AUC_0–180. Parameters evaluated in the PSA step were associated with the in situ PEPT1-mediated and passive intestinal permeability of valacyclovir, as well as the in vivo pharmacokinetics of intravenous acyclovir. The initial input values were varied by multiplying them with 10 scaling factors in the range of 0.1–10 to allow an order of magnitude increase or decrease. Acyclovir C_max and AUC_0–180 were then evaluated at each of the 10 scaling factors for each of the parameters studied.

Parameter Optimization

Following PSA, parameter optimization was performed to fine-tune the most influential input parameters for improving the model fit of the training set. In particular, the in vivo value for PEPT1 V_max had to be obtained through optimization since the in situ V_max value cannot be directly scaled to the in vivo value needed in GastroPlus. Parameter optimization was, therefore, performed using a built-in optimization module in GastroPlus.

Comparison of Simulated and Experimental Acyclovir Pharmacokinetic Profiles

Once parameter optimization (if performed) was completed, the parameter values were fixed and then used for predicting the plasma concentration-time profiles of acyclovir after the escalating oral doses of valacyclovir in wildtype and Pept1 knockout mice. Note that when performing the simulations in Pept1 knockout mice, PEPT1-mediated permeability was turned off and four initial passive P_eff values were used separately. The quality of simulations was evaluated by the following three criteria: 1) visual inspection of the simulated and observed plasma concentration-time curves of acyclovir, 2) comparison of the predicted and observed C_max and AUC_0–180 values, and 3) inspection of the coefficient of determination r².

Model Application

Once the predictive performance of the specified ACAT model was fully validated, it was used to quantify the contribution of each intestinal segment to the overall oral absorption of valacyclovir in wildtype and Pept1 knockout mice. The fraction of oral dose absorbed in each segment was calculated as the ratio of amount of prodrug disappearing from that segment to the administered dose. These fractional absorption values were computed directly and reported by GastroPlus after the simulations. The ACAT model was also used to evaluate the influence of luminal degradation on extent of valacyclovir absorption by comparing the predicted total fraction absorbed in the presence and absence of luminal degradation.

RESULTS

Valacyclovir Luminal Degradation

Figure 1 depicts the estimated k_de (panel A) or log-k_de (panel B) of valacyclovir as a function of pH, both of which are overlaid with regression lines from linear models. The plots clearly demonstrate an increase in the luminal hydrolysis of valacyclovir with increasing pH values. Given the better fit of the log-transformed k_de relationship, this equation was used to predict the k_de of valacyclovir at physiological pH along the entire length of intestine in fasted mice. The predicted k_de, degradation half-life (calculated as 0.693/k_de), and intestinal pH values in fasted mice are listed in Table III. These k_de predictions were used as input values in the pharmacokinetic simulations. In addition, analysis of the estimated valacyclovir k_de in different intestinal segments of wildtype mice at pH 6.5 suggested there were no regional differences in the luminal degradation of valacyclovir (data not shown).

Figure 1.

Relationship between the luminal degradation rate-constant k_de of valacyclovir and pH (A) or log-transformed k_de and pH (B) when fitted by linear regression.

Table III.

In Situ Degradation Rate Constants of Valacyclovir at Physiological pH Values in the Intestinal Lumen of Fasted Mice

Segment	pHa	kde (min−1)b	Half-life (min)
Duodenum	4.74	0.031	22
Jejunum	5.01	0.038	18
Ileum	5.24	0.045	15
Caecum	4.63	0.028	24
Colon	5.02	0.038	18

^aGastroPlus default values.

^bDetermined using the equation: log k_deg = 0.771 × pH – 7.61.

In Vivo Pharmacokinetics of Intravenous Acyclovir

A three-compartment model achieved the smallest AIC and SBC values when fitted to the plasma concentration-time profiles of acyclovir following an intravenous bolus injection of 25 nmol/g acyclovir in wildtype and Pept1 knockout mice. The pharmacokinetic parameters needed to fully characterize this model included volume of central compartment V₁, clearance CL, first-order transfer rate-constants between the central and shallow peripheral compartment k₁₂ and k₂₁, and first-order transfer rate-constants between the central and deep peripheral compartment k₁₃ and k₃₁. Mean pharmacokinetic parameter estimates along with standard deviation (SD) for this three-compartment model were used as initial input values in the simulations and are summarized in Table IV.

Table IV.

Pharmacokinetic Parameters of Intravenous Acyclovir Derived From a Three- Compartment Model

Parameter (unit)	Fitted (Mean ± SD)	Optimized	Optimized/Fitted
V1 (L/kg)	0.38 ± 0.08	0.545	1.4
k12(hr−1)	6.54 ± 3.25	_	_
k21(hr−1)	13.2 ± 7.4	_	_
k13(hr−1)	2.91 ± 0.68	2.221	0.8
k31(hr−1)	0.44 ± 0.36	0.908	2.1
CL (L/hr/kg)	1.0 ± 0.5	1.078	1.1

PSA Results

As shown in Figure 2 for wildtype mice, PSA suggested that the predicted AUC_0–180 of acyclovir was most sensitive to changes in CL, followed by changes in V₁, k₁₃, k_31, V_max and K_m (panel A). The predicted C_max of acyclovir was most sensitive to changes in V₁, followed by changes in V_max and K_m (panel B). In contrast, the predicted AUC_0–180 and C_max values were insensitive to changes in P_eff,passive, showing less than 30% and 60% changes, respectively, when P_eff,passive was varied over a 100-fold range. Since the predicted C_max was similarly sensitive to changes in V_max and K_m, and since V_max had to be obtained through optimization, only V_max was selected for subsequent optimization. Initial in situ-derived values for K_m were used directly without further refinement. The pharmacokinetic parameters V₁, CL, k₁₃ and k₃₁ were also optimized based on the PSA results.

Figure 2.

Sensitivity of predicted acyclovir AUC_0–180 (A) and C_max (B) to input parameters after oral administration of 25 nmol/g valacyclovir in wildtype mice. Parameters were changed by multiplying the initial input values with scaling factors in the range of 0.1–10.

Parameter Optimization

The maximum rate of PEPT1-mediated transport, V_max, was optimized at 0.000726 mg/sec for a best fit of simulations to the observed data. In addition, the parameters V₁, CL, k₁₃ and k₃₁ were also fine-tuned using the selected training set. These optimized pharmacokinetic parameters are summarized in Table IV. Most optimized parameters showed <1.5-fold change relative to the initial input values. The optimized value for k₃₁ had an acceptable 2.1-fold change compared with its initial estimate.

Comparison of Simulated and Experimental Acyclovir Pharmacokinetic Profiles in Wildtype Mice

Figure 3 depicts the simulated and experimental plasma concentration-time profiles of acyclovir in wildtype mice following escalating oral doses of valacyclovir. Figure 3A demonstrates that the simulated curves are contained within the range of individually observed data for the four dose groups studies, while Figure 3B demonstrates good agreement between the simulated and observed mean data with slight under-prediction at the terminal time points for the two higher dose groups (i.e., 50 and 100 nmol/g). As shown in Table V, adequate model predictions were further confirmed in wildtype mice by the high r² values (≥ 0.875) across four doses and the close agreement between simulated and observed C_max (< 12% difference) and AUC_0–180 (< 19% difference) values of acyclovir.

Figure 3.

Model predicted (solid lines) and observed (open diamonds) plasma concentration-time profiles of acyclovir after escalating oral doses of valacyclovir in wildtype mice using individual (A) and mean (B) results. The observed data in panel B were expressed as mean ± SD (n=4–7).

Table V.

Simulation Results for Acyclovir Following Escalating Oral Doses of Valacyclovir in Wildtype and Pept1 Knockout Mice

Genotype	Dose (nmol/g)	Cmax (μM)		AUC0–180 (min·μmol/L)		r2
		Observed	Predicted	Observed	Predicted
Wildtype	10	3.8	3.9	278	279	0.898
	25	9.6	9.7	663	691	0.969
	50	20.9	19.0	1675	1364	0.902
	100	40.6	35.9	2923	2636	0.875
Pept1 Knockout	10	0.9	1.0	128	131	0.897
	25	1.7	1.8	235	238	0.877
	50	5.0	5.2	743	687	0.814
	100	8.8	8.8	1233	1176	0.885

Comparison of Simulated and Experimental Acyclovir Pharmacokinetic Profiles in Pept1 Knockout Mice

Figure 4 depicts the simulated and mean observed pharmacokinetic profiles of acyclovir in Pept1 knockout mice. Using a passive P_eff of 0.074 × 10⁻⁴ cm/sec (method #4) in the simulations, a pronounced under-prediction was observed at the early time points (i.e., < 45 min) but a reasonable fit at later time points (i.e., 45 to 180 min). Model predictions obtained using the other P_eff values listed in Table II (i.e., methods #1, #2 and #3) all over-predicted the observed pharmacokinetic profiles of the four dose groups (data not shown). Simulations were also performed using the upper and lower limits corresponding to the 90% CI of the mean P_eff value of valacyclovir estimated in the small intestine of Pept1 knockout mice (method #5). As shown in Figure 5, at a typical dose of 25 nmol/g oral valacyclovir, the observations fell within the 90% CI enclosed by the two simulations. However, it was still challenging to satisfactorily match the observed observations using this method. Similar results were observed for the other dose groups (data not shown).

Figure 4.

Model predicted (solid lines) and observed (open diamonds) plasma concentration-time profiles of acyclovir after escalating oral doses of valacyclovir in Pept1 knockout mice. Model simulations were obtained using a P_eff input value of 0.074 × 10⁻⁴ cm/sec for valacyclovir (method #4). The observed data were expressed as mean ± SD (n=4–7).

Figure 5.

Comparison of mean observations (open diamonds) with simulations (solid lines) using 0.04 ×10⁻⁴ cm/sec (lower line) and 0.32 × 10⁻⁴ cm/sec (upper line) as the valacyclovir P_eff input values following oral administration of 25 nmol/g valacyclovir in Pept1 knockout mice. These two P_eff values represent the upper and lower limits of the 90% confidence interval for the estimated P_eff of valacyclovir in the small intestine of Pept1 knockout mice (method #5). The observed data were expressed as mean ± SD (n=7).

In a subsequent ad hoc parameter optimization procedure, P_eff was optimized separately for each dose level (method #6) in order to improve the model fit. The optimized P_eff values were 0.096 × 10⁻⁴ cm/sec, 0.070 × 10⁻⁴ cm/sec, 0.101 × 10⁻⁴ cm/sec and 0.086 × 10⁻⁴ cm/sec for the 10, 25, 50 and 100 nmol/g dose groups, respectively. These optimal values were similar to the jejunal P_eff estimate of acyclovir in Pept1 knockout mice (method #4) and fell within the 90% CI of mean P_eff estimates listed in Table II. As shown in Figure 6, the model simulations using method #6 significantly improved the prediction of experimental data, although some degree of model misfit was still visible during the early absorption phase. The need to use optimized P_eff values for obtaining reasonable model fits suggests a high sensitivity to the P_eff input values in the simulation of poorly permeable compounds. Even statistically insignificant differences in P_eff estimates may lead to poor in silico pharmacokinetic predictions. Table V summarizes the predicted and observed C_max (< 12% difference) and AUC_0–180 (< 8% difference) values of acyclovir in Pept1 knockout mice using optimized method #6. The r² values (≥ 0.814) further support a reasonable fit between the simulated and observed plasma concentration-time profiles of acyclovir in Pept1 knockout mice using this final method.

Figure 6.

Model predicted (solid lines) and observed (open diamonds) plasma concentration-time profiles of acyclovir after escalating oral doses of valacyclovir in Pept1 knockout mice using individual (A) and mean (B) results. Model simulations were obtained using separately optimized P_eff values for each dose level (method #6). The observed data in panel B were expressed as mean ± SD (n=4–7).

Model application

Figure 7 displays the contribution of specific regions of the small and large intestines toward the oral absorption of 25 nmol/g valacyclovir in wildtype and Pept1 knockout mice. Similar results were observed in the other three dose groups (data not shown). In wildtype mice, the duodenum was found to be the most important absorption site, accounting for 42% of the dose absorbed (Figure 7A). The two jejunal segments accounted for 24% of dose absorbed, and the ileum, caecum and colon combined for the remaining 4%, resulting in a total dose absorbed of 70%. In contrast, in Pept1 knockout mice, the duodenum accounting for only 4% of the dose absorbed (Figure 7B). However, the two jejunal segments accounted for 14% of dose absorbed, the ileum 10%, and the caecum and colon combined for 12%, resulting in a total dose absorbed of 40%. Finally, when the simulations were performed for wildtype mice in the absence of luminal hydrolysis (i.e., valacyclovir k_de = 0), nearly 100% of the prodrug was absorbed within 3 hr of oral dosing (Figure 7C). Thus, it appears that the incomplete oral absorption of valacyclovir was due to competing luminal degradation rather than to insufficient permeability or intestinal residence times.

Figure 7.

Contribution of specific intestinal regions in the oral absorption of 25 nmol/g valacyclovir in wildtype (A) and Pept1 knockout (B) mice, and in wildtype mice in the absence of luminal degradation (C).

DISCUSSION

In the present study, we used an in silico analysis to investigate the role and relevance of intestinal PEPT1 on the oral absorption of the model prodrug valacyclovir, and the systemic exposure of the active species acyclovir in wildtype and Pept1 knockout mice. This was accomplished by optimizing the input parameters (e.g., in situ intestinal permeabilities of valacyclovir and acyclovir, luminal degradation of valacyclovir, PEPT1 intestinal expression levels, in vivo disposition kinetics of acyclovir) when simulating the plasma concentration-time profiles of active drug. Upon integrating these simulations into an ACAT model, and comparing them with the experimental data, we achieved excellent model predictions for both wildtype and Pept1 knockout mice over a 10-fold range of oral doses. From this analysis, three major observations were made: 1) in wildtype mice, the primary absorption sites of valacyclovir were the duodenum (42%) and jejunum (24%) in which 66% of the total 70% absorption of prodrug occurred; 2) in Pept1 knockout mice, the absorption of valacyclovir was slow but sustained across the entire intestinal tract in which the duodenum (4%), jejunum (14%), ileum (10%) and caecum/colon (12%) accounted for the total 40% absorption of prodrug; and 3) in the absence of luminal degradation, the intestinal uptake of valacyclovir was essentially complete (99%), thereby, revealing that permeability was not a rate-limiting factor in the reduced in vivo oral absorption of prodrug. Collectively, these findings are novel in reporting for the first time the quantitative contribution of specific intestinal segments on the in vivo absorption of valacyclovir in the presence and absence of PEPT1. Moreover, the findings demonstrate conclusively that the reduced in vivo oral absorption of valacyclovir is solely the result of luminal degradation.

A mechanistic model that adequately captures the in vivo intestinal processes of valacyclovir is a key determinant in successfully characterizing the plasma concentration-time profiles of the active drug acyclovir. Provided that the ACAT structural model is correct, the quality of the in silico simulations depends mainly upon the accuracy of input parameters. Thus, parameters pertaining to the intestinal permeability and luminal degradation of valacyclovir, and the in vivo disposition of acyclovir, had to be reliably estimated and when necessary optimized. The luminal degradation of valacyclovir was examined by re-analyzing the results from a previous intestinal perfusion study in mice (10). These results represented our best estimate of valacyclovir degradation in vivo since gastrointestinal physiology remains intact presumably under in situ conditions. In this analysis, the luminal degradation of valacyclovir was found to increase with increasing pH, a finding consistent with another study showing that valacyclovir was stable under acidic conditions but metabolically labile under basic conditions in rats and humans (28). However, valacyclovir degradation rate constants determined in situ in mice were much higher than values in buffer solutions of the same pH, suggesting that membrane-bound enzymes accelerated the hydrolysis of valacyclovir in the intestinal lumen. The best-fitting function between log-transformed degradation rate constants and pH was then used to predict degradation rate constants of valacyclovir in mouse intestinal lumen under physiological pH conditions (Figure 1 and Table III), and applied to the current simulations.

Accurate parameter estimates were also needed for the active and passive permeability of valacyclovir, along with the in vivo pharmacokinetics of acyclovir. However, substantial inter-subject variability was observed for some of these parameters. Since we used parameter values collected from one group of mice to predict the pharmacokinetics in other group of mice (i.e., animals were not crossed-over), parameter sensitivity analysis was performed to evaluate the influence of population variability on the prediction results. Given the dominance of PEPT1-mediated intestinal uptake of valacyclovir in wildtype mice, it is conceivable that predictions were sensitive to changes in parameters characterizing this active transport process (i.e., V_max and K_m). Conversely, the predicted AUC_0–180 and C_max values of acyclovir were almost invariant to 100-fold changes in valacyclovir passive permeability (Figure 2), suggesting an insignificant role of passive diffusion during the intestinal absorption of prodrug in wildtype mice. The AUC_0–180 and C_max values of acyclovir were highly sensitive to variability in the in vivo kinetics of acyclovir (e.g., CL and V₁), suggesting that minor inaccuracies in acyclovir disposition would significantly bias the results. As a result, optimized values were used for these parameters (Table IV). Most parameters were not optimized for a variety of reasons. For example, the default physiological parameters in the ACAT model were not optimized since we assumed these values were correct in the absence of our own experimental findings. In addition, degradation rate constants were not optimized because GastroPlus did not allow this option. It should be noted, however, that parameter optimization is only feasible during retrospective studies where experimental pharmacokinetic data are available as the training dataset. In the absence of such data, approaches taking into account population variability in input parameters would be more suitable to generate the simulations.

Using the ACAT model, in silico plasma concentration-time profiles of acyclovir were simulated following a 10-fold oral dose range of valacyclovir (i.e., 10–100 nmol/g). For wildtype mice, there was good agreement between the simulations and experimental observations, with slight under-prediction of acyclovir plasma concentrations being observed at two terminal time points for the 50 and 100 nmol/g dose groups (Figure 3). One possible explanation for the slight lack of fit is that some input parameters are still not accurately parameterized for these two cohorts of mice. It is also possible, that at the two higher doses, our assumption of linear disposition of acyclovir is invalid because of the involvement of organic anion and cation transporters limiting the renal secretion of drug (29). For Pept1 knockout mice, the in silico simulations showed high sensitivity to the low intestinal permeability values of valacyclovir and acyclovir in the absence of PEPT1. By inspecting the optimized permeability values for the four dose groups, it was obvious that slight numerical changes in passive permeability could significantly impact the model predictions of Pept1 knockout mice. As a result, ad hoc optimization of passive permeability had to be performed at each dose level (i.e., method #6) to obtain a reasonable model fit (Figure 6). This finding, along with previous PSA results showing that wildtype mice were highly sensitive to K_m and V_max, suggests the importance of intestinal permeability in affecting the pharmacokinetic profiles of drugs with either high or low permeability.

Overall, the quality of simulations was somewhat worse for Pept1 knockout mice (r² ≥ 0.814) than for wildtype animals (r² ≥ 0.875) even when using optimized permeability values (Table 5). Although speculative, this may be due to several factors. For example, the assumption of equal intestinal permeabilities in different intestinal regions of Pept1 knockout mice may not be sufficiently accurate since differences in passive permeability, although not statistically significant (10), were observed in the in situ studies (Table II). While it is possible that the use of optimized P_eff values for different intestinal segments may help to better understand the source of minor misfits in the simulated results, we decided not to pursue this course since it is unlikely that the approach would significantly enhance the predictive ability of the ACAT model. It is also possible, as noted before for wildtype animals, that population variability in input factors can confound the quality of simulations as compared to the observed results.

Our final ACAT model was used to examine the effect of hydrolysis as well as the contribution of specific intestinal regions on the absorption of valacyclovir. We found that luminal hydrolysis caused an incomplete oral availability of valacyclovir in wildtype mice even though its small intestinal permeability was sufficiently high (Figure 7C). The results also indicated that the duodenum played a major role in the absorption of valacyclovir in wildtype mice but that more distal intestinal segments were important for valacyclovir absorption in the absence of PEPT1 (Figure 7A–B). Similar findings were reported by Hironaka et al (17) where they found in rats that PEPT1 accounted for about 50% of the total oral absorption of cephalexin (another PEPT1 substrate). However, their simulation study indicated an 83% bioavailability of cephalexin in the absence of PEPT1, which they attributed to greater absorption of drug in the lower intestinal segments.

The simulation analysis provided a pivotal mechanistic explanation for our previous findings in which mouse genotypes exhibited a greater magnitude of change in the in situ intestinal permeability of PEPT1 substrates than in their in vivo oral absorption. For example, even though glycylsarcosine (25,30) and valacyclovir (10,11) had jejunal permeabilities that were 10-fold lower during PEPT1 ablation, their systemic availabilities were only 2-fold and 2- to 3-fold lower, respectively, after oral dosing. At that time, we proposed that in Pept1 knockout mice glycylsarcosine and valacyclovir may take advantage of the residual length and prolonged residence times in more distal regions of the small and large intestines, resulting in a greater than expected passive absorption to partially compensate for the loss of PEPT1. The quantitative findings from the present simulation study supports this contention in demonstrating that even in the absence of intestinal PEPT1, passive absorption provides a consistent 5–10% absorption of valacyclovir along the entire length of the intestines.

The oral dose range of valacyclovir (i.e., 10 to 100 nmol/g) was chosen so that plasma concentration of acyclovir in mice would be similar to that produced in humans during commonly prescribed oral doses. Thus, peak plasma concentrations of acyclovir ranged from 3.8 – 40.4 μM over this 10-fold dose range (11), whereas in adult human subjects, acyclovir had peak plasma concentrations of 10 – 37 μM following 250 to 2000 mg oral doses of valacyclovir (31,32). It was also estimated that, at the highest oral dose given, the initial concentration of valacyclovir in the gastrointestinal tract would be 10 mM, which approximated the value observed in mice during in situ jejunal perfusions (10).

A more complete plasma concentration-time profile would have been preferred for acyclovir in wildtype and, even more so, in Pept1 knockout mice. However, the simulations were based on results from our previous paper (11) and, as a result, sampling was limited to 0–3 hr for both genotypes. In this previous paper, no attempt was made to measure plasma concentrations of valacyclovir in the systemic circulation. This was reasonable since in situ intestinal perfusions of valacyclovir had indicated that only acyclovir was present in the portal vein of samples obtained from wildtype and Pept1 knockout mice (10). Moreover, ≥ 90% of valacyclovir was converted to acyclovir by presystemic metabolism in rats after a 90 nmol/g dose (33), which was similar to the 100 nmol/g oral dose of valacyclovir given to mice in the present study.

CONCLUSIONS

In concluding, these simulations integrated the physical-chemical and pharmacokinetic properties of valacyclovir and acyclovir, along with mouse physiology and biological determinants, in predicting the prodrug’s intestinal absorption (and subsequent systemic availability of acyclovir) in wildtype and Pept1 knockout mice. Results from this study were particularly noteworthy in delineating the quantitative in vivo contribution of PEPT1 in distinct regions of the small and large intestines, and the significance of luminal hydrolysis in reducing the oral absorption of valacyclovir. Moving forward, a similar analysis should be performed with valacyclovir in other transgenic models such as humanized PEPT1 mice (34) and biphenyl hydrolase-like (Bphl) knockout mice. Ultimately, future studies should be aimed at translating these animal models to human in order to better predict a priori oral drug performance, drug-drug interactions, and the development of PEPT1 targeted drugs or prodrugs.

ABBREVIATIONS

ACAT

advanced compartmental absorption and transit

AUC_0–180

area under the plasma concentration-time curve from time 0 to 180 min

C_max

peak plasma concentration

GI

gastrointestinal

P_eff

intestinal permeability

PEPT1

peptide transporter 1

PSA

parameter sensitivity analysis