Physiologically based pharmacokinetics modeling and transporter proteomics to predict systemic and local liver and muscle disposition of statins

Abstract

Statins are used to reduce liver cholesterol levels but also carry a dose‐related risk of skeletal muscle toxicity. Concentrations of statins in plasma are often used to assess efficacy and safety, but because statins are substrates of membrane transporters that are present in diverse tissues, local differences in intracellular tissue concentrations cannot be ruled out. Thus, plasma concentration may not be an adequate indicator of efficacy and toxicity. To bridge this gap, we used physiologically based pharmacokinetic (PBPK) modeling to predict intracellular concentrations of statins. Quantitative data on transporter clearance were scaled from in vitro to in vivo conditions by integrating targeted proteomics and transporter kinetics data. The developed PBPK models, informed by proteomics, suggested that organic anion–transporting polypeptide 2B1 (OATP2B1) and multidrug resistance–associated protein 1 (MRP1) play a pivotal role in the distribution of statins in muscle. Using these PBPK models, we were able to predict the impact of alterations in transporter function due to genotype or drug–drug interactions on statin systemic concentrations and exposure in liver and muscle. These results underscore the potential of proteomics‐guided PBPK modeling to scale transporter clearance from in vitro data to real‐world implications. It is important to evaluate the role of drug transporters when predicting tissue exposure associated with on‐ and off‐target effects.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Statins are substrates of membrane transporters, for example, OATP2B1 and MRP1, that are expressed in skeletal muscle and have been found to influence statin intracellular concentration and myotoxicity in an in vitro model. Still, these pathways are disregarded by statin PBPK models currently available in the public domain due to limited in vitro kinetic data to inform model development and challenges in translatability.

• WHAT QUESTION DID THIS STUDY ADDRESS?

Can an integrated, quantitative, proteomics‐informed PBPK modeling approach be successful to scale transporter clearance from in vitro data? What is the impact of OATP2B1 and MRP1 transporters on plasma and local muscle concentrations?

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

This study provides novel in vitro kinetic and proteomics data for membrane transporters in skeletal muscle that for the first time enable the prediction of intracellular muscle exposure of statins by quantitative proteomics‐informed PBPK modeling. Results indicate that OATP2B1 contributes up to a 50‐fold increase in statin muscle exposure and that MRP1 efflux decreases statin muscle exposure by up to 13‐fold, whereas plasma concentration changes were minimal.

• HOW MIGHT THIS CHANGE DRUG DISCOVERY, DEVELOPMENT, AND/OR THERAPEUTICS?

This study highlights the importance of quantitatively considering the role of drug transporters when predicting tissue distribution and associated concentration‐driven on‐ and off‐target effects. Thus, a proteomics‐informed PBPK model could aid in predicting intracellular concentrations and thereby contribute to understanding of their implications on efficacy and safety in different scenarios and populations.

INTRODUCTION

Statins are widely prescribed cholesterol‐lowering drugs for the treatment of hypercholesterolemia and prevention of cardiovascular events. ¹ This class of drugs is generally well tolerated and safe, but a dose‐dependent risk of skeletal muscle toxicity, leading to conditions ranging from myalgia to potentially lethal rhabdomyolysis, has been reported. ² Plasma exposure alone does not fully predict the risk of statin myopathy, ³ indicating that systemic concentration may not be an adequate indicator of local toxicity‐driving concentrations. Statins are substrates of membrane transporters that are expressed in skeletal muscle and that have been found to influence statin intracellular concentrations in an in vitro model of human skeletal muscle cells. ⁴ Consequently, understanding of the actual tissue exposure profile, and hence the predictability of statin‐induced myopathy may be improved by approaches that account for the interplay between passive permeability and transport of statins as governed by uptake and efflux transporters in vivo. ⁴

Physiologically based pharmacokinetic (PBPK) models strive to describe the mechanistic relationship between drug concentrations in plasma and relevant local tissues over time. As such, PBPK modeling provides a suitable platform to predict the impact of membrane transporters on muscle exposure to statins and to shed light on the risk of statin‐associated myotoxicity. For example, the risk of myotoxicity has been associated with enzyme‐ and transporter‐mediated drug‐gene interactions (DGI) and drug–drug interactions (DDI). ⁵ Several transporters expressed by human skeletal muscle cells, including organic anion–transporting polypeptide 2B1 (OATP2B1) (uptake) and multidrug resistance‐associated protein 1 (MRP1), MRP4, and MRP5 (efflux), ⁴ are known to influence the transport of statins across cellular membranes. Still, to our knowledge these pathways are disregarded by currently available statin PBPK models. ⁶ , ⁷ , ⁸ Likely reasons include limited in vitro kinetic data to inform model development ⁷ ; poor translatability of in vitro transporter kinetic data generated in, for example, overexpressing cell lines ⁹ ; and lack of human muscle exposure to validate model assumptions.

The use of integrated quantitative proteomics and PBPK modeling may help to overcome some of these challenges. In quantitative proteomics, liquid chromatography–tandem mass spectrometry (LC–MS/MS) is used to quantify protein levels with selected surrogate peptides. ¹⁰ , ¹¹ Over the last decade, LC–MS/MS has become a powerful technique in clinical pharmacology for the quantification of enzymes and transporter proteins because of its sensitivity, reproducibility, and high‐throughput efficiency. ¹² Hence, differences between tissue transporter levels in human and in vitro systems can be used to extrapolate intrinsic transporter clearance from in vitro to in vivo systems to aid in the development of proteomics‐informed PBPK models. ¹³ Studies integrating in vitro transporter kinetics and proteomics‐informed scaling approaches in PBPK models have been successful in predicting in vivo transporter hepatic uptake and biliary clearance without applying an empirical scaling factor. ¹⁴ , ¹⁵ Nevertheless, the field currently lacks a comprehensive PBPK approach that can be used to quantify the impact of transporters, in intestine, liver, kidney, and muscle, on statin systemic and local tissue exposure by using the latest available proteomics data.

Simvastatin and pravastatin are widely prescribed statin drugs whose contrasting physicochemical profiles influence their respective dispositions after oral administration. ¹⁶ Simvastatin is a lipophilic statin ¹⁶ that is administered as a prodrug, simvastatin lactone (SVL), and is converted to the active drug, simvastatin acid (SVA), mainly in plasma and liver in a reversible reaction. ¹⁷ Although it is a prodrug, SVL shows significant systemic exposure after oral administration. ¹⁷ The main route for simvastatin elimination is via the metabolism, mediated by CYP3A4. ¹⁸ Active intestinal efflux and hepatic uptake transporters, predominately breast cancer resistance protein (BCRP) and OATP1B1, also influence systemic disposition (Figure 1). ¹⁹ , ²⁰ In contrast, pravastatin is a hydrophilic statin ¹⁶ that is administered in its active acid form. Because lactonization of pravastatin is slow, exposure to the lactone form is marginal. ²¹ Pravastatin intestinal absorption is influenced by uptake (OATP2B1) and efflux (MRP2) transporters. ²² , ²³ Pravastatin elimination takes place predominantly via the biliary and renal routes ²¹ and is mediated by the transporters MRP2 and OAT3, respectively. ²⁴ , ²⁵ Although the contribution of metabolism to the total systemic clearance of pravastatin is minimal, ²⁶ active hepatic uptake occurs and is mediated mainly by OATP1B1 (Figure 1). ²⁷ OATP1B1 DGIs have been reported resulting in an increased plasma exposure for both simvastatin and pravastatin in individuals with reduced transporter function genotypes. ²⁰ , ²⁸ Clinical DDIs have also been observed when the statins are coadministered with inhibitors or inducers of the aforementioned enzymes and transporters (e.g., gemfibrozil, clarithromycin, and rifampicin). ²⁹

FIGURE 1.

Main absorption, metabolism, excretion, and distribution pathways for simvastatin lactone (SVL), simvastatin acid (SVA), and pravastatin (PRV) included in the PBPK models.

The purpose of this study was to develop novel PBPK models for simvastatin and pravastatin based on quantitative in vitro to in vivo scaling by integration of transporter‐targeted proteomics data and in vitro kinetic measurements. The models were then applied to predict liver and muscle tissue time‐concentration profiles for different scenarios with reduced transporter function (e.g., DGI or DDI) to assess the implication of these transporter's functionality on plasma and local tissue exposure.

METHODS

Materials

Cell lines with mock‐transfected HEK Flp‐In cells and cells stably expressing organic anion‐transporting polypeptide 2B1 (OATP2B1; SLCO2B1) were developed at Uppsala University (Uppsala, Sweden). ³⁰ Spodoptera frugiperda–derived (Sf9) membrane vesicles overexpressing MRP1 (ABCC1) and the respective control vesicles were purchased from Charles River Laboratories (SOLVO Biotechnology, Budapest, Hungary). SVA and pravastatin were purchased from Sigma‐Aldrich (St. Louis, MO, USA). All other chemicals were purchased from commercial sources.

Transporter kinetic experiments

OATP2B1 transport experiments in HEK293 cells were performed as previously described by Karlgren et al. ³⁰ In brief, control and OATP2B1‐expressing HEK‐293 cells were incubated for 2 min at 37°C with a range of substrate concentrations (SVA, 2–500 μM; pravastatin, 0.01–100 μM). MRP1 efflux was assessed with inverted Sf9 membrane vesicles, using the rapid filtration technique. ³¹ A range of substrate concentrations (SVA, 2–500 μM; pravastatin, 2.5–50 μM) were incubated for 5 min at 37°C. In both uptake and efflux transporter experiments, drug accumulation in cells or vesicles was quantified by LC–MS/MS. Active transport was calculated by subtracting the passive diffusion rate (accumulation in mock cells or vesicles) from the active uptake rate in the transporter‐expressing cells or vesicles. Michaelis–Menten kinetic parameters were determined by nonlinear regression in Prism, version 8.03 (GraphPad Software, La Jolla, CA, USA), using the equation:

$$V=\frac{V_{\max }\times \left[S\right]}{K_m+\left[S\right]}$$ (1)

where V is the transport rate, V _max is the maximal transport rate, [S] is the substrate concentration, and K _m is the substrate concentration at which the transport rate is half of V _max. A more detailed description of the transporter experiments is provided in Section S1.

Protein quantification

The absolute amount of MRP1 transporter in overexpressing Sf9 membrane vesicles was determined by using a targeted proteomics method with LC–MS/MS. ³² Details are provided in Section S2. The absolute amount of OATP2B1 transporter in the overexpressing stable HEK293 cell line used in this study had been previously determined. ³⁰

Meta‐analysis of quantitative transporter proteomics in human tissues

The PubMed database (https://pubmed.ncbi.nlm.nih.gov) was searched by combining keywords (e.g., hepatic/renal/muscle transporter quantification, abundance, expression). The titles of the articles returned were screened and filtered based on relevance. Abstracts and full text of the remaining articles were reviewed, and additional relevant articles were identified by searching the list of collected articles. Only studies reporting quantitative transporter protein abundance (pmol/mg protein or fmol/μg protein) in human tissue from Caucasian adults with targeted proteomic LC–MS/MS methods were included. Studies that did not report number of samples, variability in measured abundance, or specific laboratories were excluded. A random‐effects meta‐analysis to calculate the overall mean transporter tissue abundance from studies reporting a single mean was performed by using the inverse variance method for pooling. ³³ Heterogeneity between studies was assessed using the I ² method. ³⁴ All data analyses were performed using R (version 4.0.5; The R Foundation for Statistical Computing, Vienna, Austria) with the meta package (version 5.2–0). ³⁵

PBPK modeling

The novel PBPK models for simvastatin and pravastatin were built in PK‐Sim and MoBi, version 9.1, as part of the Open Systems Pharmacology suite. ³⁶ For the DDI analysis, the available models of enzyme and transporter inhibitors and inducers were applied and minor modifications were made (see Section S3). Model parameters were estimated by the Monte‐Carlo optimization method. Simulation settings are described in Section S4 (Tables S1–S3). A total of 100 healthy individuals were simulated to assess interindividual variability in statin pharmacokinetics (PK). The predictive performance of each model was evaluated by overlaying the observed concentration‐time data with the model predicted profile (geometric mean and 90% prediction interval) and by a predefined criterion of majority of the predicted/observed ratio within 2‐fold range, based on the high inter‐ and intra‐individual variability (above 30% and 50% respectively) in simvastatin and pravastatin PK. ³⁷ , ³⁸ Additionally, increases in statin PK less than 2‐fold are not considered clinically relevant from a safety perspective. ³⁸ , ³⁹ All PK‐Sim population simulations, PK analyses, and graphics were carried out in R with the ospsuite‐R package (version 3.6.0).

Virtual population

A virtual healthy population specific for this analysis was built in PK‐Sim, based on a previously developed virtual healthy population ⁴⁰ using as reference the European (P‐pg modified, CYP3A4, 36 h, EHC) individual. Only relevant transporters and metabolizing enzymes were included in this population. Transporter tissue expression and variability were informed by the meta‐analysis performed in this study. All system‐dependent default parameters, including reference protein concentrations, protein degradation half‐lives, and tissue expression profiles of metabolizing enzymes and membrane transporters, were used as described in Table S4. In addition, modifications to the healthy population were performed to implement the BCRP, OATP1B1, and MRP2 polymorphisms (see Section S5).

Membrane transporter kinetics model input: proteomics‐informed scaling

The PBPK model input for transporter‐mediated clearance parameters (K _m and V _max) were scaled from in vitro experiments in HEK293 cells or vesicles by accounting for the transporter abundance in the in vitro system and the fraction unbound in the incubation (f _u,inc). In addition, for efflux transporters, the fraction of membrane vesicles that were inverted was taken into account. A detailed description of the equations used and in vitro data on f_u,inc, and transporter abundance are provided in Section S6 and Tables S5 and S6.

Model development and validation strategy

The full PBPK models for simvastatin and pravastatin were developed in a stepwise manner (Figure 2). In step 1, predicted statin PK was validated against those in DGI clinical studies. In step 2, the models were validated against a set of clinical PK data in oral dose ranges of 20–80 mg for simvastatin and 20–60 mg for pravastatin. In step 3, the final models were validated against DDI ratios when statins were co‐administered with inhibitors/inducers of the enzymes and transporters involved in the disposition of each statin. The pathways included are graphically summarized in Figure 1. Detailed clinical trial information, including original references and the simulated trial designs used for development and validation, are provided in Tables S7 and S8. The predictive performance of each model was evaluated as described previously. ⁴⁰ Published plasma concentration‐time profiles and variability were digitized in WebPlotDigitizer (version 3.6, GitHub).

FIGURE 2.

Stepwise modeling strategy for simvastatin lactone (SVL) and simvastatin acid (SVA) and pravastatin. Data sets used are indicated in brackets (SVi for simvastatin and Pi for pravastatin); detailed information on dose regimen and population demographics is provided in Tables S2, S3, S7, and S8. DDI, drug–drug interaction; DGI, drug‐gene interaction; PK, pharmacokinetics; Perm., permeability.

Simvastatin PBPK model

The starting point for the simvastatin PBPK model used herein was recently published. ⁴⁰ Briefly, the model captures BCRP transporter involvement in the absorption of SVL, and SVL elimination is described by CYP3A4 enzyme kinetics. Conversion of SVL to SVA is mediated by PON1 esterase in plasma and CES1 in tissue, mainly in liver. The reconversion of SVA to SVL is mediated by PON1 esterase in plasma and UGT1A1 in liver. SVA is an OATP1B1 substrate, and its elimination is described by an unspecific liver enzyme. The model was modified to incorporate the muscle transporter kinetics for OATP2B1 and MRP1 that were generated in this study. OATP1B1 and BCRP V _max values were adjusted to the updated reference tissue transporter concentrations. The final simvastatin input parameters are listed in Table S9.

Pravastatin PBPK model

Intestinal absorption of pravastatin was modeled by adopting a solution formulation and the default multicompartmental transit and absorption model. Tissue distribution, described by tissue partition constants, and cellular permeabilities were calculated by the standard method provided in PK‐Sim. ⁴¹ Pravastatin is eliminated by renal and biliary clearance (~50% each route), whereas metabolism is of negligible importance. ²¹ , ²⁶ The transporters involved and considered in the PBPK model for the absorption, distribution, and elimination of pravastatin are described in Figure 1. Metabolic processes were not considered for modeling purposes. The final model input parameters are summarized in Table S10.

RESULTS

Transporter kinetic experiments

Transporter activity in OATP2B1‐expressing HEK293 cells was confirmed by uptake studies, using estrone 3‐sulfate (E3S) as a probe substrate (Figure 3a). The OATP2B1‐mediated uptake of simvastatin followed Michaelis–Menten kinetics and the parameters are summarized in Figure 3b. At physiological conditions (pH 7.4), pravastatin was not a substrate of OATP2B1 (Figure 3c). Transport activity in membrane vesicles was confirmed by using estradiol‐17β‐D‐glucuronide as a probe substrate (Figure 3d). Both simvastatin and pravastatin were identified as substrates of MRP1 (Figure 3e,f). The kinetic parameters describing vesicular uptake mediated by the efflux transporter MRP1 are summarized in Figure 3.

FIGURE 3.

(a) Representative uptake of 1 μM E3S (probe substrate) in HEK293 cells stably expressing OATP2B1 compared with passive uptake in mock‐transfected cells. (b) Concentration‐dependent uptake of simvastatin acid in HEK293 cells overexpressing OATP2B1. (c) Representative uptake of pravastatin (at 1 μM) in HEK293 cells stably expressing OATP2B1 compared with passive uptake in mock‐transfected cells. (d) Representative uptake of 1 μM β‐estradiol 17‐(β‐D‐glucuronide) (E17bG) (probe substrate) in membrane vesicles expressing MRP1 compared with passive uptake in control vesicles. Concentration‐dependent uptake of simvastatin acid (e) and pravastatin (f) in membrane vesicles overexpressing MRP1 is shown. Simvastatin acid and pravastatin uptake followed Michaelis–Menten kinetics; the kinetic parameters are summarized in each graph.

Targeted proteomics: MRP1 quantification in vesicles and meta‐analysis of human tissue

The protein abundance of MRP1 transporter in vesicles was 139 ± 2 pmol/mg, which is in line with other MRP‐overexpressing vesicles. ⁴² , ⁴³ A total of 17 studies met the meta‐analysis inclusion criteria. Quantitative human tissue abundance data for OATP1B1, OATP2B1, BCRP, MRP2, OAT3, and MRP1 were collected from five different laboratories. A summary of mean abundance data and confidence intervals is shown in Figure 4. The studies included and values of meta‐analysis are listed in Figure S1. A heterogeneity test (I²) was performed to assess interstudy variability. For all transporters, the I² value was higher than 80, indicating high heterogeneity for the abundance data (Figure S1).

FIGURE 4.

Meta‐analysis of literature‐reported quantitative transporter abundance in human tissue. Lines indicate weighted means and colored areas indicate 95% confidence intervals from the meta‐analysis. Muscle tissue abundance was back‐calculated from reported measurements in liver and lung according to PK‐Sim relative tissue abundance (details described in Table S4). m, number of studies; n, number of samples.

Statin PBPK model validation

DGI validation

Observed PK changes in individuals with reduced function OATP1B1 genotype (c.521CC) versus wild‐type (c.521TT) genotype were well captured by the model for both SVA and pravastatin incorporating the reported 90% reduction in transport function (Figure 5). Similarly, the observed SVL PK change associated with reduced function BCRP genotype (c.421AA) versus wild‐type (c.421CC) genotype was also well captured based on the reported 83% reduction in BCRP function (Figure 5). A simulated change in pravastatin PK as a result of 2‐fold increased MRP2 expression in individuals with MRP2 c.1446C<G genotype was also in agreement with clinical observations (Figure S2). Overall, the final PBPK models predicted the maximum concentration (C _max) and area under the concentration‐time curve (AUC) ratio for individuals being homozygous or heterozygous for the reduced function transporter genotype versus wild‐type genotype within 2‐fold of the observed DGIs for OATP1B1, BCRP, and MRP2 ¹⁹ , ²⁰ , ²² , ²⁸ (Figures S9 and S10).

FIGURE 5.

(a) Predicted simvastatin lactone (SVL) PK profiles (lines) versus clinical observations reported as means (circles) for the BCRP wild‐type c.421CC genotype (dark colors) and the reduced function c.421AA genotype (light colors). ¹⁹ (b) Predicted (lines) versus observed (circles) simvastatin acid (SVA) PK profiles for the OATP1B1 wild‐type c.521TT genotype and the reduced function c.421CC genotype. ²⁰ (c) Predicted (lines) versus observed (circles) pravastatin (PRV) PK profiles for the OATP1B1 wild‐type c.521TT genotype and the reduced function c.421CC genotype. ²⁸ In (a–c) solid lines represent predicted geometric means, and dashed lines represent the predicted 5–95% quantiles for virtual populations. (d–f) Predicted simvastatin lactone (d), simvastatin acid (e), and pravastatin (f) PK profiles versus clinical observations shown as study mean (dots) for a 40‐mg dose (data sets SV9–SV27 and P8–P13). Solid lines represent predicted geometric means, and the shaded area indicates the predicted 5–95% quantiles for virtual populations. Conc, concentration; details on the implementation of the reduced function genotypes in the PBPK model can be found in Section S5.

PK validation

Simulated and observed plasma concentration‐time profiles for SVL and SVA after administration of a single dose of 20, 40, 60, and 80 mg of simvastatin were in good agreement (Figure 5, Figures S3–S5). Simulated and observed plasma concentration‐time profiles for pravastatin after administration of a single dose of 20, 40, and 80 mg were also in good agreement (Figure 5, Figures S6–S8). The final PBPK models predicted 90% of AUC values within 2‐fold and 50% within 1.25‐fold of the observed clinical data at different doses for both simvastatin and pravastatin (Figures S9 and S10). Similarly, 63% and 93% of observed clinical C _max for simvastatin and pravastatin, respectively, were predicted within 2‐fold and 40% to 33% within 1.25‐fold (Figures S9 and S10).

DDI validation

The predicted AUC and C _max ratios for the interaction between simvastatin (SVL and SVA) strong and moderate CYP3A4 and/or transporter inhibitors/inducers were within 2‐fold for 90% and 80% of the observed values (Figure 6). The predicted AUC and C _max ratios for the interaction between pravastatin and transporter inhibitors and/or inducers were within twofold for 100% of the observed values (Figure 6).

FIGURE 6.

Predicted versus observed simvastatin and pravastatin AUC and C _max ratios when co‐administered with CYP3A4 and transporter inhibitors for nine and three clinical DDI studies, respectively (data sets SV35–SV43 and P15–P17). Each study is represented by a circle (lactone) or triangle (acid); details on references, study design, and dose regimens can be found in Tables S2, S3, S7, and S8. The solid line represents the line of unity, and the shaded area represents the twofold error.

PBPK model application: systemic and local tissue exposure

The qualified PBPK models that incorporate the generated muscle transporter data were applied to predict unbound plasma, liver, and muscle exposure under different conditions. Simulations were performed for a 40‐mg oral dose in virtual populations with normal and 90% reduced OATP1B1, OATP2B1, or MRP1 function to assess the implication of transporter activities on plasma and local tissue exposure. In the normal population, liver‐to‐plasma (L:P) and muscle‐to‐plasma (M:P) ratios for SVA were larger than for pravastatin (Figure 7). For both SVA and pravastatin, the L:P ratio decreased by 5‐ to 8‐fold with reduced OATP1B1 function, increased by 15% with reduced MRP1 function, and remained unchanged with reduced OATP2B1 function (Figure 7). The M:P ratio for SVA remained unchanged with reduced OATP1B1 function, decreased by 8.5‐fold with reduced OATP2B1 function, and increased by 5‐fold with reduced MRP1 function (Figure 7). For pravastatin, the M:P ratio remained unchanged with reduced OATP1B1 and OATP2B1 function but increased by 2‐fold with reduced MRP1 function (Figure 7).

FIGURE 7.

Predicted unbound plasma (black), liver (red), and muscle (purple) concentrations for simvastatin acid and pravastatin after a 40‐mg oral dose in different scenarios: normal population, 90% reduced OATP1B1 function, 90% reduced OATP2B1 function, and 90% reduced MRP1 function. Solid lines represent the predicted geometric mean, and the shaded area indicates the predicted 5–95% quantiles for virtual populations. conc, concentration; L:P, liver‐to‐plasma ratio; M:P, muscle‐to‐plasma ratio, reported as mean (range).

DISCUSSION

Quantification of intracellular drug concentrations in humans remains challenging. Plasma concentration is consequently often used as a surrogate, overlooking the potential influence of drug transporter‐mediated distribution. Preclinical measurements are valuable in a qualitative manner but are not appropriate to quantitatively assess tissue distribution, due to large interspecies differences in membrane transporter orthologs, expression, and activity, as in the case of OATPs and MRPs. ⁴⁴ , ⁴⁵ In this context, PBPK modeling predictions based on in vitro data and quantitative proteomics is an attractive approach to understand the impact of drug transporters in humans. This study provides novel in vitro kinetic and proteomics data for skeletal muscle transport that for the first time enable the prediction of statin intracellular muscle exposure by quantitative proteomics‐informed PBPK modeling.

We demonstrated that, similar to other statins such as rosuvastatin and atorvastatin, SVA is an OATP2B1 and MRP1 substrate in vitro (Figure 3). ⁴ , ³⁰ , ⁴⁶ Our in vitro data additionally shows that pravastatin at physiological pH is a substrate for MRP1, but not for OATP2B1 (Figure 3). These results are consistent with those of previous in vitro studies in which pravastatin was a substrate of MRP1 ⁴⁷ and OATP2B1 at pH 5.5 but not at pH 7.4, ²³ as well as with reports of pH‐dependent OATP2B1 transport of other statins. ⁴⁸ Consequently, OATP2B1 was incorporated in the small intestine only in the pravastatin PBPK model (Table S10).

Quantitative proteomics, by which differences in enzyme and transporter expression can be accounted for in a quantitative manner, has emerged as a state‐of‐the‐art approach to translate activity between in vitro and in vivo settings or to disease populations. ¹¹ , ¹² , ¹³ Lack of protein standards and harmonized methodology to quantify surrogate peptides, which leads to interlaboratory variability in abundance levels, remains a challenge, however. ¹¹ Our meta‐analysis combined abundance data from five different laboratories applying different methods and confirmed high heterogeneity between studies (I ² > 80%). Further investigations showed that heterogeneity can be explained in part by different laboratory methodologies. By‐laboratory analysis showed I ² < 75 in some cases (Figures S11–S14), in line with inter‐laboratory variability reported previously. ⁴⁹ , ⁵⁰ Given these caveats, it can be argued that an aggregated assessment based on meta‐analysis currently provides a more robust approach to inform PBPK protein tissue expression.

Low expression of OATP2B1 and MRPs on the cellular membrane of skeletal muscle cells with marginal influence on disposition has been used as a justification for not considering these transporters in published statin PBPK models. ⁷ In contrast, our proteomics‐informed PBPK models indicate that OATP2B1 contributed to a 50‐fold elevation of SVA muscle exposure despite low muscle abundance (0.29 pmol/mg protein). Notably, the impact of increased muscle distribution on plasma concentration was minimal, indicating that plasma exposure is an inadequate surrogate for assessing local concentrations. Interestingly, a recent clinical study reported no difference in atorvastatin muscle‐to‐plasma concentration ratio in relation to the OATP2B1 genotype (c.935G>A). ⁵¹ However, this specific genotype has been reported to have minimal impact on OATP2B1 activity for several substrates, for example, E3S and rosuvastatin, among others. ⁵²

Knauer et al. ⁴ identified MRP1, MRP4, and MRP5 as potential efflux transporters for atorvastatin and rosuvastatin. Our meta‐analysis results, considering also tissue relative expression, showed that MRP1 expression in muscle is higher than that of MRP4 (10‐fold) or MRP5 (60‐fold). In addition, previous in vitro data indicate that simvastatin and pravastatin are not MRP4 substrates. ⁵³ Thus, we assumed that MRP1 is the main efflux transporter in skeletal muscle. Our results indicate that MRP1 efflux resulted in 13‐ and 2‐fold lower muscle exposures for SVA and pravastatin, respectively, compared with passive distribution to plasma. The importance of MRP1 efflux in statin muscle exposure and toxicity has been demonstrated in an in vivo DDI study between rosuvastatin and probenecid (MRP1 inhibitor) in rats. ⁵⁴ Rosuvastatin muscle toxicity was exacerbated with co‐administration of probenecid. Overall, our proteomics‐informed PBPK modeling approach indicates that OATP2B1 and MRP1 are membrane transporters with a significant impact on statin disposition, particularly on intracellular muscle concentrations and the associated risk of statin‐related myotoxicity.

We used our PBPK models to simulate the implications of these transporters' functionality on statin plasma and local tissue exposure. Statin concentrations in liver and muscle are especially relevant, as the pharmacological target of statins is located in the liver ¹ and myotoxicity originates in the muscle. ² Accumulation in liver tissue (L:P ratio >1) for both simvastatin and pravastatin (Figure 7) was predicted with normal transporter function. The significance of OATP‐mediated transport has been demonstrated clinically in a study where measured liver or biliary concentrations exceeded plasma concentrations for rosuvastatin. ⁵⁵ , ⁵⁶ A higher L:P ratio for SVA (L:P = 34) than for pravastatin (L:P = 4.8) can be explained by accumulation due to intrahepatic conversion of the more permeable lactone form, which is less prevalent for pravastatin. ²¹ Another reason is the higher biliary excretion of pravastatin into bile. ²¹ As a substrate of OATP2B1, simvastatin is also predicted to have greater distribution into muscle (M:P = 0.33) than pravastatin (M:P = 0.001). This prediction is consistent with observations in clinical studies in which calculated pravastatin M:P ratio was the lowest among evaluated statins. ⁵⁷

Virtual populations with normal and 90% reduced function of OATP1B1, OATP2B1, or MRP1 allowed for the reproduction of clinical scenarios with normal and compromised transporter activity, driven, for example, by DGIs or DDIs. The level of transporter impairment being simulated is in agreement with the 90% reduced transporter function reported for the OATP1B1 c.521CC genotype, ⁵⁸ and the 80–90% in vitro transporter inhibition observed for OATP2B1 and MRP1 by gemfibrozile and probenecid, respectively. ⁴ , ⁵⁹ Reduced OATP1B1 function predicted a lower L:P ratio but a conserved M:P ratio for both statins. Thus, reduced OATP1B1 function leads to an increase in plasma concentrations but only a small decrease in liver concentrations, as previously observed by PBPK modeling of simvastatin and rosuvastatin. ⁷ , ⁸ This could explain why clinical studies in individuals with OATP1B1 c.521CC genotype observed a significant increase in statin plasma concentrations but a small decrease in efficacy (LDL reduction) compared with individuals with wild‐type OATP1B1 genotype. ⁶⁰ , ⁶¹ A conserved M:P ratio is in agreement with observed plasma and muscle exposure for atorvastatin in individuals with OATP1B1 c.521CC genotype. ⁵¹ Interestingly, reduced OATP1B1 function leads to an increase in muscle exposure (Figure S15) which could explain the higher risk of myotoxicity associated with the OATP1B1 c.521CC genotype. ⁶² Reduced OATP2B1 function did not significantly influence the L:P ratio for any of the statins (Figure 7), as expected from the limited contribution of OATP2B1 in relation to OATP1B1, to the overall hepatic uptake. The exact addition by OATP2B1 is compound specific, ³⁰ , ⁴⁶ but in general, this contribution is low, as OATP2B1 liver expression is lower than that of OATP1B1 (Figure 4). On the contrary, reduced OATP2B1 activity had a significant impact on the SVA M:P ratio, a finding that supports the importance of OATP2B1 for intracellular muscle concentrations of certain statins. ⁴ With reduced MRP1 function, muscle exposure to both simvastatin and pravastatin was increased compared to plasma (Figure 7). As previously described, these predictions are consistent with in vivo data in rats showing that inhibition of MRP1 exacerbated statin‐related myotoxicity, ⁵⁴ demonstrating the protective effect of MRP1 in statin‐related myotoxicity.

The transporter‐mediated DDI validation strengthens the model's ability to capture the transporter‐mediated distribution and clearance mechanisms, as well as the in vitro‐in vivo extrapolation approach using quantitative proteomics. Quantitative validation of the predicted intracellular statin concentrations is not feasible, however, due to the lack of clinical observations. Thus, intracellular predictions could only be partially validated through relative comparison with liver, muscle, or biliary concentrations of other statins, PD readouts with reduced function OATP1B1 (c.521CC) genotype, or through the inhibition of MRP1 transporter in preclinical species. Future measurements of transporter abundance directly in muscle and availability of quantitative clinical intracellular data would further strengthen the current model validation and enhance our understanding of statins transporter‐mediated disposition, particularly in muscle. Additionally, future model development linking the predicted intracellular statin concentrations to their efficacy and myotoxicity biomarkers could further enhance model validation and application.

In summary, this study highlights the importance of quantitatively considering the role of drug transporters when predicting tissue distribution and associated concentration‐driven on‐ and off‐target effects. It also demonstrates the utility of proteomics‐informed PBPK modeling to scale transporter clearance from in vitro data. Future applications of the reported simvastatin and pravastatin PBPK models include the establishment of relationships between PK and pharmacodynamics for lipid‐lowering effects as well as myotoxicity. Thus, a mechanistic PBPK model could aid in the prediction of intracellular concentrations of statins, thereby contributing to our understanding of the implications of statin exposure on efficacy and safety in different scenarios and populations.

AUTHOR CONTRIBUTIONS

L.P.G. wrote the manuscript. A.V., T.T., C.A., P.N., H.L., and E.S. revised the manuscript. L.P.G., A.V., C.A., P.N., H.L., and E.S. designed the research. L.P.G., A.V., and E.S. performed the research. L.P.G., A.V., A.B.M., and T.T. analyzed the data.

FUNDING INFORMATION

This study was supported in part by the Academy of Finland GeneCellNano Flagship and PROFI6.

CONFLICT OF INTEREST STATEMENT

L.P.G., A.V., P.N., and C.A. are employees of AstraZeneca with stock ownership and/or stock options or interests in the company. All other authors declared no competing interests for this work.

Supporting information

Data S1.

Data S2.

Prieto Garcia L, Vildhede A, Nordell P, et al. Physiologically based pharmacokinetics modeling and transporter proteomics to predict systemic and local liver and muscle disposition of statins. CPT Pharmacometrics Syst Pharmacol. 2024;13:1029‐1043. doi: 10.1002/psp4.13139