Novel insights in drug transporter sciences: the year 2021 in review

Abstract

On behalf of the team I am pleased to present the second annual ‘novel insights into drug transporter sciences review’ focused on peer-reviewed articles that were published in the year 2021. In compiling the articles for inclusion, preprints available in 2021 but officially published in 2022 were considered to be in scope. To support this review the contributing authors independently selected one or two articles that were thought to be impactful and of interest to the broader research community. A similar approach as published last year was adopted whereby key observations, methods and analysis of each paper is concisely summarized in the synopsis followed by a commentary highlighting the impact of the paper in understanding drug transporters’ role in drug disposition. As the goal of this review is not to provide a comprehensive overview of each paper but rather highlight important findings that are well supported by the data, the reader is encouraged to consult the original articles for additional information. Further, and keeping in line with the goals of this review, it should be noted that all authors actively contributed by writing synopsis and commentary for individual papers and no attempt was made to standardize language or writing styles. In this way, the review article is reflective of not only the diversity of the articles but also that of the contributors. I extend my thanks to the authors for their continued support and also welcome Diane Ramsden and Pallabi Mitra as contributing authors for this issue (Table 1).

Table 1. Selected articles for this review.

Title	Area	Authors	Source
1. Atorvastatin-associated rhabdomyolysis in a patient with a novel variant of the SLCO1B1 gene: a case report	Polymorphism	Kiage et al.	J Clin Lipidol. 2022;16(1):23–27
2. Quantitation of plasma membrane drug transporters in kidney tissue and cell lines using a novel proteomic approach enabled a prospective prediction of metformin disposition	Proteomics	Kikuchi et al.	Drug Metab Dispos. 2021;49(10):938–946
3. Proteomics-informed prediction of rosuvastatin plasma profiles in patients with a wide range of body weight	Proteomics	Wegler et al.	Clin Pharmacol Ther. 2021;109(3):762–771
4. Performance of plasma coproporphyrin I and III as OATP1B1 biomarkers in humans	Biomarker	Neuvonen et al.	Clin Pharmacol Ther. 2021 Dec;110(6):1622–1632
5. Identification of appropriate endogenous biomarker for risk assessment of multidrug and toxin extrusion protein-mediated drug-drug interactions in healthy volunteers	Biomarker	Miyake et al.	Clin Pharmacol Ther. 2021;109(2):507–516
6. Population pharmacokinetic modeling and simulation to support qualification of pyridoxic acid as endogenous biomarker of OAT1/3 renal transporters	Biomarker	Ahmad et al.	CPT Pharmacometrics Syst Pharmacol. 2021 May;10(5):467–477
7. Acute and chronic effects of rifampin on letermovir suggest transporter inhibition and induction contribute to letermovir Pharmacokinetics	Induction/Biomarkers	Robbins et al.	Clin Pharmacol Ther. 2022;111(3):664–675
8. Exploring the use of serum-derived small extracellular vesicles as liquid biopsy to study the induction of hepatic cytochromes P450 and organic anion transporting polypeptides	Induction/Biomarkers	Rodrigues et al.	Clin Pharmacol Ther. 2021;110(1):248–258
9. Unraveling pleiotropic effects of rifampicin by using physiologically based pharmacokinetic modeling: assessing the induction magnitude of P-glycoprotein-cytochrome P450 3A4 dual substrates	Induction	Pan et al.	CPT Pharmacometrics Syst Pharmacol. 2021;10:1485–1496.
10. Quantification of CYP3A and Drug transporters activity in healthy young, healthy elderly and chronic kidney disease elderly patients by a microdose cocktail approach	Clinical pharmacology	Rattanacheeworn et al.	Front Pharmacol. 2021;12:726669.
11. Quantitative prediction of P-glycoprotein-mediated drug–drug interactions and intestinal absorption using humanized mice	DDI	Miyake et al.	Br J Pharmacol. 2021;178(21):4335–4351.
12. Regulation of OATP1B1 function by tyrosine kinase-mediated phosphorylation	Regulation	Hayden et al.	Clin Cancer Res. 2021;27(15):4301–4310.

Key Words:

• Transporters
• biomarkers
• induction
• IVIVE
• OATP
• physiologically-based pharmacokinetic modeling
• polymorphism
• regulation
• proteomics

 

1. Atorvastatin-associated rhabdomyolysis in a patient with a novel variant of the SLCO1B1 gene: a case report

Source: J Clin Lipidol. 2022;16(1):23–27.

Synopsis

In this paper, Kaige J et al. present a unique case of an African American male with hypercholesterolemia who developed rhabdomyolysis. This patient was prescribed with atorvastatin at 40 mg/day for 2 years to treat his hypercholesterolemia, however, his lipid profile did not improve as cholesterol and triglyceride levels were still high (Table 2). So the dose of atorvastatin was increased to 80 mg/day however, the patient developed rhabdomyolysis. The blood levels of creatinine phosphokinase (CPK) (42,760 IU/L) and myglobin (254 ng/mL) were markedly elevated confirming the diagnosis of rhabdomyolysis. Therefore, the patient was immediately taken off atorvastatin and given intravenous hydration which helped in lowering CPK level. Due to rhabdomyolysis the treatment was changed from atorvastatin to arilocumab, a proprotein convertase subtilisin kexin 9 (PCSK9) inhibitor, however, this regimen did not help in lowering LDL and therefore, the patient was further treated with arilocumab along with ezetimibe (inhibitor of intestinal cholesterol absorption). This combination helped the patient in lowering LDL <100 mg/dl. The lipid profile of the patient over time is shown in Table 2.

To investigate the reason for rhabdomyolysis, patient’s genomic DNA was sequenced using next-generation sequencing (NGS) which identified two potential pathogenic variants in the SLCO1B1 gene, the c.481 + 1G>T splice donor single nucleotide variant (rs77271279) and c.1200C>G missense variant (rs59113707). The c.481 + 1G>T variant is loss-of-function mutation while c.1200C>G variant has indicated reduced protein synthesis of OATP1B1 in HEK cells (Zhang et al. 2021). Thus, it was concluded that having these two variants in the trans-configuration (on separate strands of DNA, compound heterozygote) may have negatively impacted SLCO1B1 function leading to increased risk for statin-induced myopathy. The actual mechanism of statin-induced rhabdomyolysis is unknown, however, multiple pathophysiological mechanisms have been proposed. The most common mechanism involves decrease in ubiquinone (coenzyme Q) produced by the HMG-CoA pathway leading to necrosis of skeletal muscle (Safitri et al. 2021).

Commentary

Statins are widely prescribed drugs across the globe for the management of hypercholesterolemia however, these drugs have been shown to cause rare but life-threatening side effects such as rhabdomyolysis and myalgia. Some factors that are characterized to contribute to the adverse effects of statins include lower body mass, female sex, hepatic or renal dysfunction and some genetic factors. The genetic variants of various genes involved in fatty acid metabolism, glycogen metabolism, purine metabolism and abnormal skeletal muscle relaxation have been shown to be associated with rhabdomyolysis. Further, drug-drug interactions (due to inhibition of either metabolic enzymes, transporters or both by comedication) have also been identified to increase statin exposure causing adverse effects (Sibley et al. 2021). Among all the genes that have been identified the hepatic Organic Anion Transporting Polypeptide (OATP), c521T>C (p.Val174Arg, SLCO1B1*5) variant has been widely studied and recognized as a risk factor for statin-induced rhabdomyolysis (Brunham et al. 2018) and myalgia (Voora et al. 2009). However, the effect of c521T>C OATP1B1 variant varies among statins in terms of exposure with rank order of, simvastatin acid (221%) > atorvastatin (144%) > pravastatin (91%) > rosuvastatin (65%) with no effect on fluvastatin (Lee E et al. 2005; Pasanen et al. 2006; Pasanen et al. 2007).

The patient reported in this study clearly experienced rhabdomyolysis shortly after the increase in dose of atorvastatin (80 mg/day) as indicated by elevated levels of CPK in blood and myoglobin in urine. In this study, genetic testing in connection with dyslipidemia helped to identify risk factors and manage the patient’s rhabdomyolysis and dyslipidemia. The two OATP1B1 variants, c.481 + 1G>T and c.1200C>G that were previously uncharacterized were shown to be associated with atorvastatin-induced rhabdomyolysis in this patient. The c.481 + 1G>T variant is loss-of-function mutation and patients with homozygous mutation of this site have indicated complete OATP1B1 deficiency (van de Steeg et al. 2012). This patient was also tested heterozygous for c.1200C>G which is known to reduce OATP1B1 function (50% activity compared to WT). Thus, the presence of these two variants in the trans configuration (on separate strands of DNA, compound heterozygote) has been highlighted as an increased risk factor of atorvastatin-induced myopathy.

This is the first clinical case suggesting potential role of these two OATP1B1 variants, c.481 + 1G>T c.1200C>G for statin-indued rhabdomyolysis. Further studies in a large cohort of patients with statin myopathy will shed more light on the pathogenicity of these variants. Nonetheless, this report certainly demonstrated the critical use of the candidate gene approach using next-generation gene sequencing for appropriate management of hypercholesterolemia in patients that are intolerant to statin drugs. Recent literature have indicated the use of coproporphyrin (CP)-I, and CP-III as the selective endogenous biomarkers of OATP1B as they can precisely measure the changes in OATP1B1 function due to DDI and polymorphism (Yee et al. 2019). The measurement of these biomarkers in this patient would have added value in confirming the role of OATP1B1 polymorphism in atorvastatin-induced rhabdomyolysis.

Table 2. Lipid profile of the patient before and after the lipid lowering therapy.

Timeline^a	Treatment	Cholesterol (mg/dl)	LDL-cholesterol (mg/dl)	HDL-cholesterol (mg/dl)	Triglyceride (mg/dl)	CPK (U/L)
Week-60	Atorvastatin 40 mg/day	262	138	43	247	–
Week-5	None	343	230	46	247	–
Week 0	Atorvastatin 80 mg/day	246	168	46	161	>42,760
Week 12	Arilocumab 150 mg q2wk	266	160	52	272	120
Week 24	Alirocumab 150 mg/q2wk	209	144	46	95	143
Week 40	Alirocumab 150 mg/q2wk  + Ezetimibe 10 mg/day	184	97	57	151	96

^aWeeks prior to or after hospitalization.

2. Quantitation of plasma membrane drug transporters in kidney tissue and cell lines using a novel proteomic approach enabled a prospective prediction of metformin disposition

Source: Drug Metabolism and Disposition, 2021;49(10):938–946.

Synopsis

Previously, quantification of the protein abundance of drug transporters on plasma membrane from tissues has not been achieved due to challenges associated with the isolation of plasma membrane fractions. In this study, authors quantified the protein levels of drug transporters, OCT2, MATE1 and MATE2-K on the plasma membrane from human kidney as well as in vitro cells using a novel methodology involving serial centrifugation and targeted proteomics. First, plasma membrane fractions from both human kidney tissues and HEK-293 cells expressing these transporters were prepared by using serial centrifugation steps. The recovery of apical and basolateral plasma membrane was examined by measuring the abundance of localization marker proteins, Na⁺-K⁺-ATPase and γ-glutamyl transpeptidase 1, respectively. These data were used to obtain Relative Expression Factors (REFs) for each transporter, OCT2, MATE1 and MATE2-K, involved in metformin disposition (Table 3). Subsequently, metformin kinetic data from the transporter expressing HEK-293 cells were extrapolated to in vivo using these REFs and later incorporated into the PBPK model to simulate metformin pharmacokinetics. The simulated PK profile correlated well (within 2-fold) with the clinical data of metformin (Table 4). Overall, the authors showed successful application of PBPK model incorporating plasma membrane abundance of OCT2, MATE1 and MATE2-K in the prospective prediction of metformin PK.

Commentary

Drug transporters influence PK properties of various drugs in the intestine, liver and kidney and, therefore quantitative information on their expression levels in these tissues is important to accurately predict the disposition of substrate drugs. In the past decade, targeted proteomics approach has shown some success in measuring the protein abundance of drug transporters in human liver, kidney and intestine to extrapolate the kinetic parameters of substrate drugs to in vivo using PBPK models (Harwood et al. 2016; Li et al. 2019; Sharma et al. 2020). This approach has truly facilitated the understanding of differences in the protein expression of drug transporters in in vitro and in vivo systems in a more quantitative way for successful IVIVE and PK predictions. However, the targeted proteomics approach has some challenges caused by significant variability (about 100-fold) from sample preparation methodologies (Wegler et al. 2017).

Generally, a transport activity is a true reflection of the amount of a transporter protein present on plasma membrane and therefore the quantification of the protein level of a drug transporter on plasma membrane is critical for successful IVIVE (Bosgra et al. 2014; Kumar et al. 2018). In this study, the targeted proteomic approach was used to simultaneously quantify drug transporter proteins on the plasma membrane fractions from in vitro cell models and kidney tissues. The protein levels of OCT2 at the basolateral membrane and, MATE1 and MATE2-K at the apical membrane of proximal tubules, respectively were quantified using targeted proteomics. It was interesting to see low recoveries in all samples across cell lines and kidney tissues however, it did not impact REFs as the recoveries were consistently low in all membrane preparations. Given the significant role of OCT2, MATE1 and MATE2-K in the active secretion of metformin, the protein abundance of these transporters in the plasma membrane fractions of HEK cells and kidney tissues was used to determine REFs to translate in vitro kinetics to in vivo clearance of metformin. This information was subsequently used in the PBPK model to successfully simulate metformin PK profile. Although these data look promising, further validation of this approach is needed using more substrate drugs of these transporters. It is also important to determine the translatability of this approach to other drug transporters in the kidney such as OAT1 and OAT3 for prospective predictions of transporter-mediated drug disposition.

Table 3. Transporter abundance and relative expression factors.

Transporter	Prior to correction			Following correction
	Transporter-overexpressing HEK-293 (fmol/10^6 cells)	Human kidney cortex (fmol/10^6 PTC)	REF	Transporter -overexpressing HEK-293 (fmol/10^6 cells)	Human kidney cortex (fmol/10^6 PTC)	REF
OCT2	771	4233	5.5	4064	37355	9.2
MATE1	471	6774	14	4276	58982	14
MATE2-K	391	789	2	5351	6947	1.3

PTC: Proximal tubule cells; REF: relative expression factor.

Table 4. Predicted and observed exposure of metformin.

Dose (mg)	250			500
Parameters	Predicted	Observed	Ratio (Obs/Pred)	Predicted	Observed	Ratio (Obs/Pred)
Cmax (µg/mL)	0.36	0.59	1.7	0.75	1.19	1.6
AUC0-24 (µg*hr/mL)	2.42	4.26	1.8	5.03	6.58	1.3

3. Proteomics-informed prediction of rosuvastatin plasma profiles in patients

Source: Clin Pharmacol Ther. 2021;109(3):762–771.

Synopsis

Wegler and co-workers developed a three compartment semi-mechanistic model to predict the pharmacokinetics (PK) of rosuvastatin in 54 patients (with body weight range between 47 and 171 kg) that takes into account the interindividual variability in hepatic uptake transporter expression. These patients received rosuvastatin orally and then liver biopsies were taken from each individual to determine the expression of the relevant uptake transporters including OATP1B1, OATP1B3, OATP2B1 and NTCP that are known to be involved in the hepatic uptake of rosuvastatin. The uptake kinetics of rosuvastatin was measured in HEK293 cells overexpressing individual transporters, OATP1B1, OATP1B3, OATP2B1 and NTCP while the protein quantification of these transporter was performed by LC-MS/MS. Later the hepatic uptake clearance of rosuvastatin was determined using relative expression factor approach. A semi-mechanistic model using bottom-up approach was then developed by incorporating in vitro uptake clearance and transporter abundance data to predict PK profile of rosuvastatin in 54 patients. The PK data was directly compared against the observed PK data of rosuvastatin in each individual. The model was able to capture the rosuvastatin pharmacokinetics in the majority of patients. The model also predicted the plasma exposure of rosuvastatin well in OATP1B1 521C homozygous patients when accounted for the reduced activity (10% of wild type OATP1B1) but not for the reduced protein expression. Overall, this study showed that the hepatic uptake clearance measured using transporter cell lines and targeted proteomics approach serve as useful tool in predicting individual difference in the PK profile of transporter substrate drugs.

Commentary

Protein quantification of drug metabolizing enzymes and transporters is an important factor for successful in vitro-to-in vivo extrapolation (IVIVE) of enzyme- and transporter-mediated clearances (Anoshchenko et al. 2021; Storelli et al. 2021) and incorporation of these data into PBPK model has significantly improved PK and DDI predictions of transporter substrates drugs (Taskar et al. 2020). Generally the scaling of in vitro transporter-mediated clearance is performed using empirical scaling factors or proteomic data from the literature but this practice does not really account for the differences in the expression levels of transporters in each individual in the target population (Jones et al. 2012; Kumar et al. 2021).

In this paper, authors used a semi-mechanistic model to predict the PK profile of rosuvastatin in 54 patients. The measured PK parameters of rosuvastatin varied considerably between the patients with a 13-fold difference in observed area under the curve (AUC), 30-fold difference in maximum plasma concentration (C_max) and a 12-fold difference in observed terminal half-life (t_1/2). AUC estimates by this model were within 2-fold in 78% of the patients, C_max estimates were within 2-fold in 76% of patients, and t_1/2 were within 2-fold in 98% of patients indicating reasonable predictions. Importantly, the model underpredicted rosuvastatin plasma clearance in patients that were carriers of OATP1B1 521C variant and this was attributed to the proteomic method used in this study. The authors measured the total OATP1B1 protein in liver tissues and not the protein in the plasma membrane of hepatocytes which is actually reduced in OATP1B1 521C homozygous patients. Current methods do not allow to quantify transporter proteins from isolated plasma membrane fractions due to practical difficulties in protein enrichment and significant protein loss during fractionation. However, when authors accounted for reduced activity (10% of wild type OATP1B1 activity) of OATP1B1 521C the measured PK parameters of rosuvastatin were close to the observed PK in OATP1B1 521C variant homozygous patients. Interestingly, in this analysis, the difference in the protein expression of transporters alone could not explain the observed variability in the rosuvastatin PK. This could be explained by other contributing factors such as the differences in the expression and function of other transporters including OATP2B1 and BCRP which are known to significantly influence the oral absorption of rosuvastatin (Costales et al. 2021; Takahashi et al. 2021).

Overall, Wegler and colleagues developed a semi-mechanistic model that takes into account the individual hepatic uptake transporter expression to predict rosuvastatin PK. Although the model captured the PK profile of rosuvastatin reasonably well in the majority of 54 patients it was not enough to completely explain the PK variability, therefore, similar approach needs to be applied to other transporter substrate drugs to further validate this approach.

4. Performance of plasma coproporphyrin I and III as OATP1B1 biomarkers in humans

Source: Clin Pharmacol Ther. 2021;110(6):1622–1632.

Synopsis

The objective of this study was to compare four endogenous compounds coproporphyrin I (CPI), CP III, glycochenodeoxycholate 3-O-glucuronide (GCDCA-3G) and glycodeoxycholate 3-O-glucuronide (GDCA-3G) as biomarkers of OATP1B1 (SLCO1B1) in healthy volunteers. Depending on the SLCO1B1 haplotype that they carried, subjects were categorized into normal OATP1B1 function, decreased function, poor function, increased function and highly increased function groups. Compared with the normal function group, plasma CPI, GCDCA-3G, and GDCA-3G concentrations were higher in the poor and decreased function groups, and lower in the increased function and highly increased function groups. Notably, concentration changes of GCDCA-3G and GDCA-3G in the altered OATP1B1 function groups were more substantial than those of CPI. In contrast, for CPIII, statistically significant differences in concentrations were observed only in the poor function group. The biomarkers were also compared for their ability to distinguish SLCO1B1 c.521C>C homozygotes from the T/C and T/T genotypes. In all analyses (entire cohort, men, women), GCDCA-3G had the highest efficiency and CPIII the lowest for identifying individuals with poor OATP1B1 function; CPI and GDCA-3G had similar performance. Overall, the results of the study indicated that CP I and GDCA-3G had similar efficiencies for identifying individuals with altered OATP1B1 function. However, GCDCA-3G was the most sensitive OATP1B1 biomarker for this purpose.

Commentary

Over the last decade, coproporphyrin I (CPI) and CPIII have gained prevalence as endogenous biomarkers of OATP1B1 and OATP1B3 (Lai et al. 2016; Yee et al. 2019). More recently, the endogenous compounds GCDCA-3G and GDCA-3G have been reported as sensitive and specific substrates of OATP1B1. Plasma concentrations of GCDCA-3G were used to identify reduced function phenotypes of the SLCO1B1 c.521T>C polymorphic variant with high efficiency (Neuvonen et al. 2021). This study evaluated the ability of CPI and CPIII to identify SLCO1B1 c.521T>C variants with increased and decreased function. Since the study was conducted in the same set of individuals evaluated in the Neuvonen et al. (2021) study, it allowed direct comparison of the four biomarkers for their ability to distinguish SLCO1B1 c.521C>C homozygotes from the T/C and T/T genotypes.

Study participants (356 Finnish healthy volunteers) were genotyped for four SLCO1B1 single nucleotide variants (SNVs), and haplotypes were computed as combinations of these SNVs. Haplotypes *1 and *37 were considered to be normal function alleles, *14 and *20 as increased function alleles, and *5 and *15 as decreased function alleles. Subjects were grouped as normal function (*1, *37 homozygotes and compound heterozygotes), decreased function (having one decreased function allele and one normal/increased function allele), poor function (carrying two decreased function alleles), increased function (having one normal function and one increased function allele), and highly increased function (carrying two increased function alleles).

In a genome-wide association analysis study, only the SLCO1B1 c521T>C SNV displayed significant associations with CPI plasma concentrations while none of the OATP1B1 SNVs were associated with CPIII plasma concentrations. In individuals with the SLCO1B1 c.521C/C genotype, plasma concentrations of CPI and CPIII were 1.74-fold and 1.28-fold higher compared to individuals with the reference allele. In comparison, plasma concentrations of GCDCA-3G and GDCA-3G were 9.2-fold and 6.4-fold higher, respectively (Neuvonen et al. 2021) suggesting a more significant role of OATP1B1 in the hepatic uptake of GCDCA-3G and GDCA-3G than CPI, CPIII. Previously, rifampicin (OATP1B inhibitor) increased plasma levels of CPI, GCDCA-3G and GDCA-3G by 14.6-fold, 4-fold, and 3.7-fold respectively (Mori et al. 2020). The smaller effect of the SLCO1B1 c.521C/C genotype on CPI plasma exposure compared to the effect of rifampicin, suggests a greater role of OATP1B3 in the uptake of CPI compared to GCDCA-3G and GDCA-3G.

In the haplotypes, individuals homozygous or compound heterozygous for *5 and *15, had the highest plasma concentrations of CPI, while those with various combinations of *14 and *20 the lowest. Compared with the normal function group, CPI concentrations were significantly higher in the poor and decreased function groups, and significantly lower in the increased function and highly increased function groups (Table 5). Similar trends, as seen for CPI, were observed for GCDCA-3G and GDCA-3G but to a much greater extent (Neuvonen et al. 2021) (Table 5). For CPIII, statistically significant differences were observed only in the poor function group.

Subsequently, the four biomarkers were compared for their ability to distinguish SLCO1B1 c.521C>C homozygotes from the T/C and T/T genotypes. Precision-recall and receiver-operating characteristic (ROC) were conducted to calculate values of areas under the precision-recall (AUPRC) and areas under the receiver operating characteristic (AUROC) curves. AUPRC demonstrates the relationship between precision and sensitivity, and AUROC shows the relationship between sensitivity and specificity. In the entire cohort of 356 individuals, GCDCA-3G was the best performer with the highest AUPRC and AUROC values. The mean difference of AUPRC values of CPI, GDCA-3G, and CPIII from GCDCA-3G values were −0.172, −0.389, and −0.697. The mean difference of AUROC values of CPI, GDCA-3G, and CPIII from GCDCA-3G were −0.112, −0.100, and −0.257. Similar results were seen when the groups were separated into males and females. Thus, in all analyses, GCDCA-3G had the highest efficiencies and CPIII the lowest for identifying individuals with poor OATP1B1 function. CPI and GDCA-3G had similar performance.

Collectively, the results demonstrated that CPI, GDCA-3G and GCDCA-3G but not CPIII performed well in identifying OATP1B1 polymorphic variants. GCDCA-3G was the most sensitive biomarker for identifying individuals with poor OATP1B1 function.

Table 5. Effect of OATP1B1 phenotypes on biomarker plasma concentration ratios (adapted from Table 2 in the manuscript).

OATP1B1 phenotype	Ratios of plasma concentrations in OATP1B1 altered function groups with respect to concentrations in the normal function group
	CPI	CPIII	GCDCA-3G	GDCA-3G
Highly increased function	0.772*	0.972ns	0.403*	0.675^ns
Increased function	0.896*	0.978ns	0.805*	0.668*
Normal function	1	1	1	1
Decreased function	1.07*	0.986ns	1.6*	1.65*
Poor function	1.68*	1.27	8.57*	5.76*

*Statistical significance. ^nsStatistical non-significance.

5. Identification of appropriate endogenous biomarker for risk assessment of multidrug and toxin extrusion protein-mediated drug-drug interactions in healthy volunteers

Source: Clin Pharmacol Ther. 2021;109(2):507–516

Synopsis

Three endogenous compounds (creatinine, N¹-methylnicotinamide (1-NMN), and N¹-methyladenosine (m¹A)) were evaluated as potential biomarkers of renal transporters MATE1 and MATE2-K. Metformin was chosen as the reference substrate of MATE function. The control group received metformin (500 mg) while treatment groups received metformin and three doses (10, 25, and 75 mg) of the MATE inhibitor pyrimethamine (PYR). With increasing doses of PYR, there were dose-dependent decreases in renal clearance (CL_r) of metformin, 1-NMN and m¹A. Creatinine CL_r did not decrease beyond the PYR 25 mg dose. There was good correlation of CL_r ratios (+PYR/controls) of metformin with those of 1-NMN and metformin and m¹A (r² > 0.5), but no correlation with creatinine CL_r ratios (r² = 0.11). These results suggested that the dose response of 1-NMN and m¹A to MATE inhibition by PYR is similar to that of metformin. Further, unbound in vivo K_i,app values were similar to in vitro K_i values of PYR for MATE1 (1-NMN as the substrate) and MATE2-K (1-NMN and m¹A as substrates). This suggested that CL_r decreases of 1-NMN and m¹A observed in the presence of PYR are due to in vivo inhibition of MATE1 and MATE2-K. In conclusion, the results of the study suggest that 1-NMN and m¹A can be considered to be suitable urinary biomarkers of in vivo inhibition of MATE1 and MATE2-K.

Commentary

The objective of this study was to evaluate the performance of three endogenous compounds (creatinine, 1-NMN, and m¹A) as biomarkers of in vivo inhibition of MATE1 and MATE2-K transporters. Creatinine and 1-NMN have been previously utilized as MATE biomarkers (Müller et al. 2015; Chu et al. 2016). m¹A has been recently identified as an endogenous substrate of OCT2 and MATE1/MATE2-K (Miyake et al. 2019) and has been evaluated as a novel potential biomarker in this study. Metformin was chosen as the reference substrate of MATE function (Stage et al. 2015) and PYR was selected as the MATE inhibitor (Kusuhara et al. 2011).

The study involved 12 healthy Japanese subjects in a four-way crossover design where the control group received metformin (500 mg) and treatment groups received metformin and three doses of PYR (10, 25, and 75 mg). The effect of PYR dose on the Clr and plasma exposure of metformin, 1-NMN and m¹A is shown in Table 6. These data clearly indicated the similar effect of PYR on metformin, 1-NMN and m¹A. Further, the urinary excretion does not appears to be the major elimination pathway of 1-NMN and m¹A. Based on these data 1-NMN and m¹A can be used as a urinary biomarkers of MATE inhibition but they still have limited applicability as plasma biomarkers.

CL_r ratios of metformin, 1-NMN, m¹A and creatinine as a function of average PYR plasma concentrations were fitted to obtain the in vivo inhibition constants (K_i,app) of PYR, K_i,app,u (K_i,app corrected for PYR plasma unbound fraction) and f_PYR (fraction sensitive to inhibition by PYR). The In vitro K_i values of PYR for OCT2, MATE1, and MATE2-K were also determined. K_i.app estimates of metformin, m¹A, 1-NMN were comparable (Table 7), and f_PYR values of all three compounds were ∼0.8, suggesting m¹A and 1-NMN to have similar responses to PYR as metformin. Further, K_i,app,u values of m¹A and 1-NMN were similar to in vitro K_i values of MATE1 (using 1-NMN as the substrate) and MATE2-K (using 1-NMN and m¹A as substrates), suggesting that m¹A and 1-NMN can be utilized as quantitative biomarkers of MATE1/MATE2-K mediated DDIs. Notably f_PYR of creatinine was low at 0.26 indicating a low proportion of tubular secretion to the total urinary excretion of creatinine, and thus suggesting limited utility of creatinine as a biomarker of MATE1/MATE2-K DDIs.

Taken together, this study demonstrated that 1-NMN and m¹A can be considered as urinary biomarkers of in vivo inhibition of MATE1 and MATE2-K. The low proportion of tubular secretion to the total urinary excretion of creatinine may limit its utility for this purpose.

Table 6. Geometric mean ratios (+PYR/control) of AUC (area under the plasma concentration time curve) and CL_r of the four substrates (adapted from Table 1 in the manuscript).

Condition	Geometric mean ratios
	Metformin		1-NMN		m^1A		Creatinine
	AUC (µM.h)	CLr (mL/min)	AUC (µM.h)	CLr (mL/min)	AUC (µM.h)	CLr (mL/min)	AUC (µM.h)	CLr (mL/min)
PYR (10 mg) vs. control	1.02	0.88	0.62	0.93	1.03	0.90	1.18	0.82
PYR (25 mg) vs. control	1.24	0.66	0.55	0.64	1.08	0.79	1.20	0.75
PYR (75 mg) vs. control	1.69	0.45	0.69	0.41	1.07	0.53	1.09	0.75

Table 7. Inhibition constants pyrimethamine (adapted from Table 2 in the manuscript).

Compound	In vivo		In vitro Ki (µM)
	Ki,app (µM)	Ki,app,u (µM)	OCT2	MATE1	MATE2-K
Metformin	1.05	0.137	37.9	0.154	0.094
m^1A	1.94	0.252	0.466	NDa	0.068
1-NMN	1.37	0.178	41.2	0.125	0.082
Creatinine	0.091	0.0661	1.57	ND^a	ND^a

^aND: Not determined.

6. Population pharmacokinetic modeling and simulation to support qualification of pyridoxic acid as endogenous biomarker of OAT1/3 renal transporters

Source: CPT Pharmacometrics Syst Pharmacol. 2021;10(5): 467–477

Synopsis

In this work, the authors demonstrated the qualification of pyridoxic acid (PDA) as a new biomarker for renal organic anion transporters OAT1/3. Population pharmacokinetic modeling of PDA and probenecid, a potent OAT1/3 inhibitor was developed using clinical plasma and urine pharmacokinetic data with and without probenecid administration to describe the baseline of PDA and the OAT1/3 inhibition effect by probenecid accurately. This model was verified adequately using an independent clinical dataset. PDA showed flat baseline without apparent circadian rhythm and high sensitivity for OAT1/3 inhibition. Power calculation based on the simulation for weak, moderate and strong OAT1/3 interactions indicated the broad utility of PDA as an OAT1/3 biomarker. Thus this model enabled optimal clinical DDI study design by leveraging biomarker data for OAT1/3 inhibition in early phase clinical trials.

Commentary

Utility of endogenous biomarkers for drug transporters is one of the emerging and powerful tools for early assessment of drug-drug interaction (DDI) liability, to understand complex interaction mechanisms where compounds inhibit multiple pathways. The majority of qualification of endogenous biomarkers for drug transporters and its clinical application in drug development has evolved for liver uptake transporters. The biomarker data suggesting no or low interaction with the transporters of interest could preclude and in some cases be used in lieu of dedicated DDI studies. This could provide enormous benefit by reducing the number of clinical studies in drug development. Few studies have been conducted to identify and qualify biomarkers of renal transporters due to the less clinical evidence of DDI via renal transporters and the lower magnitude compared to hepatic uptake transporters such as OATPs (Mochizuki et al. 2021). However, DDIs in renal transporters remain an area of concern in drug development and high sensitive biomarkers for renal transporters are desired to predict potential DDI risk with perpetrators.

Previously Shen et al identified PDA and homovanillic acid (HVA) as endogenous OAT1/3 biomarkers from metabolomics analysis in monkey plasma with and without probenecid treatment and confirmed their utility in humans (Shen et al. 2018, 2019). These studies provided qualification of PDA as the most sensitive endogenous biomarker for OAT1/3. Population modeling of the interaction between PDA pharmacokinetics and OAT1/3 inhibitors enables simulation of a wide variety of cases where the compounds have weak, moderate or strong inhibitory effect and are administered at different doses and regimens (single or multiple). Thus, if a compound of interest is capable of inhibiting OAT1/3 and an IC₅₀/K_i value is estimated, the ideal study size and sampling points can be determined considering the PK characteristics of the compound. In early DDI potential assessment, regardless of the approaches used; static or dynamic model, reliable determination of inhibition potency is an essential part of the assessment. However, the discrepancy of in vitro and in vivo IC₅₀ is a concern for accurate DDI potential assessment. Utility of biomarkers for DDI assessment can provide an in vivo IC₅₀ earlier and will support further risk assessment in dedicated DDI studies if interaction response is observed. In addition, this investigation also demonstrated that inhibition of synthesis rate of an endogenous biomarker underestimated the interaction or led to the opposite interaction. Therefore, one needs to be cautious while interpreting DDI data using endogenous biomarkers and ideally the counter assay of synthesis inhibition of biomarkers would be warranted whenever the rate limiting step in the synthesis of a particular endogenous biomarker is known.

7. Acute and chronic effects of rifampin on letermovir suggest transporter inhibition and induction contribute to letermovir pharmacokinetics

Source: Clin Pharmacol Ther. 2022;111(3):664–675

Synopsis

This study aimed to evaluate whether systemic exposure of OATP1B substrates including letermovir, and two endogenous biomarkers, coproporphyrin (CP) I, and glycochendeoxycholic acid-sulfate (GCDCA-S), could be changed following chronic administration of rifampin, a strong PXR inducer. Letermovir is a substrate of OATP1B, P-gp and UGT1A1/3 and the elimination is primarily regulated by hepatic OATP1B. In line with being a substrate of OATP1B, plasma exposure of letermovir was increased with single oral or IV doses of rifampin albeit to a lower extent compared to selective OATP1B substrates. When rifampin was administered chronically the magnitude of change was different on Day 28 (inhibition + induction) compared with Day 29 (induction only), with Day 29 data indicating strong induction. Moreover, endogenous biomarkers for OATP1B activity, CPI and GCDCA-S showed PK changes consistent to observation with letermovir. These results suggest that chronic administration of rifampin could induce OATP1B activity and the approach taken could be utilized to inform clinical study design for investigating OATP1B induction and inhibition by rifampin.

Commentary

Whether OATP is inducible or not remains a contentious topic (see commentary on Rodrigues et al. 2021). Although several reported clinical studies have indicated potential OATP1B induction by rifampin (Rodrigues et al. 2020), preclinical and clinical evidence of chronic rifampin effects on OATP1B induction is not yet fully clarified. There are several reasons why the role of PXR activation on OATP1B induction remains controversial including the lack of sensitive response in the gold standard in vitro assays and the lack of selectivity for clinical OATP1B probes which tend to be substrates for other transporters/enzymes regulated by PXR (e.g. P-gp). This overlap leads to interplay of induction and inhibition of multiple transporters/enzymes and complicates the interpretation of OATP1B induction liabilities.

Letermovir is an antiviral agent targeted to inhibit cytomegalovirus (CMV) DNA terminase complex indicated for CMV prophylaxis in hematopoietic stem cell transplant recipients. Letermovir is a substrate of OATP1B, P-gp and UGT1A1/3 and eliminated primarily as an unchanged parent drug in feces, suggesting hepatic uptake by OATP1B is the major rate determining step for clearance rather than metabolism. A clinical study evaluating letermovir as a substrate of P-gp showed a weak increase in exposure (∼30%) when co-dosed with itraconazole, a P-gp inhibitor. Therefore, the authors concluded that letermovir is an appropriate drug to delineate the impact of chronic rifampin effects for OATP1B substrate disposition without interference from P-gp. Moreover, OATP1B endogenous markers were measured in plasma for comparison with letermovir. The study included three single dose periods, with a fixed sequence design in n = 16 healthy adult women, where letermovir was administered alone (Period 1) followed by a 7-day washout, co-administered with 600 mg oral rifampin (Period 2) followed by a 7-day washout, or co-administered with 600 mg IV rifampin (Period 3) and a multiple dose period where letermovir was administered at 480 mg daily for 14 days followed by co-administration of 600 mg oral rifampin for 14 days in n = 14. Blood samples were collected for PK analysis on days 1, 14, 28 and 29 to separate acute inhibition from chronic induction. Biomarkers were analyzed from periods 1 and 2 and from the first 6 participants with complete PK in the multiple dose phase. Letermovir exposure was increased after both oral and intravenous single administration of rifampin. The oral rifampin increased letermovir AUC and C_max higher than intravenous rifampin by 30%. The results indicated P-gp might contribute intestinal absorption of letermovir moderately and P-gp function can be inhibited by oral rifampin. After chronic treatment of rifampin for 14 days, letermovir AUC_0–24h, C_max and C_24h were reduced by 85%, 73% and 91%, respectively. Since P-gp contribution for letermovir disposition is considered as minor, the dramatic decrease in exposure is postulated to be due primarily to OATP1B induction by rifampin. Moreover, plasma concentrations of CPI and GCDCA-S which are structurally distinct molecules as endogenous markers of OATP1B activity were measured to further investigate the possibility of OATP1B induction by rifampin (Barnett et al. 2019). AUC_0–24h of CPI and GCDCA-S were decreased by 53% and 48% after pretreatment of rifampin for 14 days, respectively, suggesting OATP1B activity is inducible by rifampin. Although induction of P-gp cannot be excluded as contributing to the magnitude of decrease for letermovir by rifampin, CPI is not a P-gp substrate suggesting that OATP1B induction is the major driver of clinical CPI changes. In conclusion, the data from this clinical study demonstrates OATP1B activity could be increased by chronic treatment of rifampin and the study design incorporated here may enable a roadmap to study DDI that is complicated by interplay of induction and inhibition of transporters and enzymes.

8. Exploring the use of serum-derived small extracellular vesicles as liquid biopsy to study the induction of hepatic cytochromes P450 and organic anion transporting polypeptides

Source: Clin Pharmacol Ther. 2021;110(1):248–258

Synopsis

The authors report the application of a novel two-step liquid biopsy protocol to isolate liver derived small extracellular vesicles (sEVs) from banked serum samples. Isolated sEVs were subjected to proteomic analysis to investigate changes in protein expression of hepatic OATPs and CYPs after administration of rifampicin (300 mg for 7 days and 600 mg for 14 days) to healthy male subjects and across trimesters during pregnancy. The liquid biopsy approach concluded that rifampicin treatment induced CYP3A4 protein levels to an extent that correlated with the AUCR change for oral midazolam from the same subjects. The data also convincingly demonstrate that the protein levels of hepatic OATP1B1 and OATP1B3 were unchanged following rifampicin treatment. These results offer additional direct evidence that hepatic OATPs are not induced via PXR. Moreover, the authors deployed liver sEVs to investigate OATP protein changes during pregnancy as there are multiple reports which indicate that OATP expression changes in placenta. The data herein demonstrate that there was no difference in protein levels of hepatic OATPs when comparing sEVs isolated from nonpregnant and women at varying stages of pregnancy (first, second and third trimester). Therefore, reported OATP changes in placenta appears to be independent of maternal hepatic OATPs levels. In conclusion, the use of serum-derived liver sEVs offers a valuable tool to support risk assessment of drug interaction and subject phenotyping including patient stratification without the need for invasive tissue sample collection.

Commentary

In general, phenotyping of human subjects is considerably more complex than genotyping since tissue sampling would be essential to evaluate absolute transporter and enzyme level changes in organs. To date, transporter selective probe drugs and endogenous biomarkers have been used as primary phenotyping tools. For example, statins are widely used to evaluate drug-drug interaction (DDI) via hepatic OATPs in clinical studies. However, a notable limitation of these probe drugs is that they are not selective towards OATPs but are also substrates of other transporters (e.g. MDR1 and BCRP) and drug metabolizing enzymes and do not selectively isolate hepatic dependent effects via OATPs. Therefore, data interpretation of clinical transporter DDI studies is complicated in cases where perpetrators inhibit multiple transporters. These current gaps could be resolved by employing liquid biopsies through isolation of tissue-derived sEVs because various sEVs are shed into blood and include cargo that is tissue specific (Rodrigues and Rowland 2019). In the present study, the authors validated the preparation of tissue-derived sEVs and elucidated whether hepatic OATPs were induced following treatment of two dose levels of rifampicin. This topic remains controversial, and the international transporter consortium recently concluded that further studies are needed to delineate OATP1B induction and additional mechanisms that lead to changes in exposure of OATP1B substrates (Zamek-Gliszczynski et al. 2021). In addition, protein expression changes of hepatic OATPs were evaluated from sEVs isolated from women at various stages of pregnancy where data is currently lacking. OATP protein expression in circulating liver sEVs was compared with whole human liver and the authors calculated that the OATP1B protein amount is 0.03–0.05% of total liver pool and equivalent to 500–800 mg of the liver. The results suggested isolation of sEVs using the liquid biopsy approach yielded higher recovery of liver fraction than conventional needle biopsies which obtain less than 40 mg of tissue. No induction of OATP1B1 and OATP1B3 protein was observed with 600 mg of rifampicin for 14 days, while protein levels of CYP3A4 were induced as expected. Compared with CYP3A4, CYP3A5 protein levels were only weakly induced. Further validating this approach, the authors showed correlation between liver sEV-catalyzed dextromethorphan O-demethylation to dextrorphan with CYP2D6 expression and demonstrated no induction of CYP2D6 by rifampicin, consistent with literature. These results add evidence that hepatic OATPs are unlikely to be induced via PXR. Finally, whether hepatic OATP protein levels change during pregnancy was investigated. There was no difference in levels of hepatic OATP1B1 and OATP1B3 during pregnancy although hormone profiles of estradiol, progesterone and prolactin were consistent with reported data. Therefore, pregnancy-related changes in OATP expression in placenta might be independent of maternal liver OATP expression.

In conclusion, liver-derived sEVs clarified hepatic OATPs protein levels are not induced following treatment with rifampicin and do not change during pregnancy. Since a limitation of the current analysis is the limited number of subjects in each group, further expanding the analysis by increasing the number of individuals would strengthen the conclusions drawn. Additionally, future analysis which compares the readout between different approaches including changes in clinical OATP probe substrates (e.g. statins) and endogenous biomarkers (e.g. CPI and CPIII), with the use of serum-derived liver sEVs would offer a powerful approach to interpret DDI risk and enable noninvasive subject phenotyping.

9. Unravelling pleiotropic effects of rifampicin by using physiologically based pharmacokinetic modeling: assessing the induction magnitude of P-glycoprotein-cytochrome P450 3A4 dual substrates

Source: CPT Pharmacometrics Syst Pharmacol. 2021;10:1485–1496

Synopsis

The authors applied a physiologically-based pharmacokinetic (PBPK) modeling approach to systematically evaluate the impact of simultaneous induction of P-glycoprotein and CYP3A4, using clinical data from 12 substrate drugs including three sensitive towards P-gp, seven dual P-gp and CYP3A substrates, and two dual substrate and inhibitor drugs. While PBPK modeling of induction is becoming more widely accepted, there is still hesitation in acceptance by regulatory agencies. Particularly, as example, there are observations of underprediction of rifampicin mediated induction observed in cases where both P-gp and CYP3A contribute to the disposition and elimination of the compound (Hariparsad et al. 2021). In the current paper the authors incorporated a 3.5-fold increase in P-gp protein level into the PBPK model. This value was based on literature data which reported this magnitude of protein change, measured with Western blot, from healthy volunteers following 10 day administration of 600 mg rifampicin. The modified PBPK model resulted in improved accuracy of prediction when comparing CYP3A induction alone, with RMSE decreasing from 0.11 to 0.05, and 0.15 to 0.10 for AUC and C_max ratios, respectively. The AFE were likewise reduced from 1.93 to 1.48 and 1.54 to 1.04 for AUC and C_max ratios, respectively. The authors propose that the PBPK modeling approach applied here, when informed by robust in vitro and clinical data, can serve as a framework to assess DDI potential for investigational drugs that are dual substrates of P-gp and CYP3A.

Commentary

CYP3A and P-glycoprotein have demonstrated importance in modulating oral absorption of multiple drugs. In the intestine they can function synergistically to limit drug absorption and systemic exposure of dual substrates, with P-gp-mediated extrusion of drug molecules from the enterocytes back into the lumen aiding further metabolism by intestinal CYP3A (Lown et al. 1997; Zhang et al. 1998). Both P-gp and CYP3A4 are regulated through PXR and CAR, and can be induced by agonists of these nuclear receptors including rifampicin (Elmeliegy et al. 2020). In the current work, the authors used a PBPK approach to simulate the effect of rifampicin induction on sensitive P-gp substrate, dual substrates of P-gp and CYP3A, and substrates and inhibitors of P-gp and CYP3A (Table 8). The substrates spanned a range of intestinal P-gp kinetic parameters (from 2.33 to 531 µL/min) and CYP3A contributions (fm_CYP3A from 0 to 0.99). First the existing rifampicin PBPK model, which only considers CYP3A induction mechanisms, was applied. This model has been well validated to characterize induction for sensitive CYP3A substrates when co-administered with rifampicin but tends to underpredict induction of dual CYP3A and P-gp substrates. P-gp induction was then incorporated by applying a 3.5-fold multiplier to the relative expression factor of intestinal P-gp kinetics. Importantly, the kinetic parameters for P-gp-mediated active transport were determined by modeling in vitro data using the Simcyp in vitro data analysis toolkit (SIVA). When in vitro data was not available, the P-gp mediated transport was estimated using in vivo pharmacokinetic data. Sensitivity analysis was conducted on a hypothetical sensitive P-gp substrate and a dual P-gp and CYP3A substrate to understand the impact of P_eff,man and intestinal P-gp CL_int,T on the substrate PK parameters and the interplay between P-gp CL_int,T, fm_CYP3A4, and the hepatic extraction ratio (E_H) on the magnitude of DDI. This assessment indicated that both fm_CYP3A4, and intestinal P-gp CL_int,T function as key substrate characteristics that determine the level of interaction with rifampicin. The results also highlighted the importance of robust and reliable estimates for intestinal P-gp activity. In general, incorporation of robust P-gp kinetic parameters into the modified rifampicin PBPK model, with the scaler of 3.5-fold based on clinical protein changes, improved the accuracy of the predictions. Furthermore, this analysis is considered to be sufficient to qualify the system model within the PBPK platform and can be used as a valuable aid to understand combined P-gp and CYP3A induction with investigational drugs (EMA 2018). Noted limitations of the study include that time dependency of the P-gp induction was not incorporated into the model and is an area highlighted for further exploration.

Table 8. Substrate models used in the PBPK model with rifampicin with and without incorporation of P-gp induction.

Drug	Dose (mg)	Designation	CLint,p-gp	fmCYP3A4	Fg	Fh	AUCR observed with rifampicin
Abemaciclib	200	Dual	35.1	0.91	0.84	0.67	0.05
Acalabrutinib	100	Dual	27.3	0.80	0.95	0.33	0.206
Bosutinib	500	Dual	116	0.99	0.97	0.66	0.076
Crizotinib	250	Dual	43.4	0.68	0.98	0.55	0.181
Naldemedine	0.2	Dual	12	0.72	0.998	0.93	0.169
Naloxegol	25	Dual	5	0.90	0.95	0.45	0.109
Olaparib	300	Dual	8.37	0.75	0.998	0.92	0.109
Quinidine	332	Substrate/Inhibitor	40.6	0.76	0.97	0.81	0.168
Verapamil	120	Substrate/Inhibitor	2.33	0.68	0.96	0.28	0.065
Digoxin	1	Sensitive	4.9	–	1.0	0.97	0.697
Dabigatran etexilate	150	Sensitive	531	–	0.66	0.79	0.33
Talinolol	100	Sensitive	16.8	–	1.0	0.90	0.647

10. Quantification of CYP3A and drug transporters activity in healthy young, healthy elderly and chronic kidney disease elderly patients by a microdose cocktail approach

Source: Front Pharmacol. 2021;12:726669

Synopsis

The study highlights the potential and challenges with using PK endpoints from a microdose cocktail to evaluate age-versus Chronic Kidney Disease (CKD)-related changes in the activity of cytochrome P450 3A (CYP3A) and drug transporters including, P-glycoprotein (P-gp), organic-anion-transporting polypeptide (OATP)1B and breast cancer resistance protein (BCRP). Probe substrate cocktails are routinely used to simultaneously evaluate the modulation of CYP enzyme and drug transporter activity, with recent data demonstrating applications for microdose cocktails in healthy volunteers (Prueksaritanont et al. 2017). Subsequent investigations with the same microdose cocktail, which included, midazolam (MDZ, CYP3A), dabigatran etexilate (DABE, intestinal P-gp), pitavastatin (PTV, OATP1B), rosuvastatin (RSV, BCRP and OATP1B), and atorvastatin (ATV, CYP3A, OATP1B, BCRP and P-gp), suggested that CKD reduced intestinal P-gp and BCRP activity (Tatosian et al. 2021). This study by Rattanacheeworn et al. aimed to evaluate age-versus CKD-related changes in the Thai population by comparing the activity of CYP3A and drug transporters (P-gp, BCRP and OATP1B) derived from microdose cocktail PK observations between young healthy participants (n = 20, ages 20–40 years, with estimated glomerular filtration rate (eGFR) ≥ 90 mL/min/1.73 m²) and elderly participants without (n = 16, ages ≥60 years with eGFR ≥60 mL/min/1.73 m²) and diagnosed with CKD (n = 17, ages ≥60 years with eGFR 15–60 mL/min/1.73 m²). While the authors concluded that CYP3A activity was decreased with aging and although intestinal P-gp activity appeared to be affected by CKD status, additional exploration is needed. Further, no conclusions could be drawn on the impact of age and CKD status on OATP1B and BCRP activity.

Commentary

Understanding and predicting the variability in pharmacokinetic parameters due to intrinsic and extrinsic factors is of clinical relevance. Chronic kidney disease (CKD) is a significant public health concern worldwide due to its high prevalence, cost of treatment and patient morbidity and mortality (Carney 2020). The enclosed study aimed to investigate the impact of aging and CKD status on the activity of CYP3A, P-gp, BCRP or OATP1B in the Thai population, using a microdose cocktail approach previously reported for healthy volunteers (Prueksaritanont et al. 2017). The studies were appropriately powered to evaluate differences in PK properties, considering the difference in AUC for midazolam between young and elderly participants would be 70%, with a common standard deviation of 11.84, as described using the same microdose cocktail (Prueksaritanont et al. 2017). Genotyping analysis of all participants was conducted and included characterization of known variants for SLC01B1 (OATP1B1), ABCG2 (BCRP) and CYP3A. Comparison of the dose levels administered across referenced studies is presented in Table 9. The objectives of the study period 2 (DDI with rifampicin) were consistent with those previously reported using healthy Caucasian subjects, as were the observations that CYP3A and intestinal P-gp activities decrease with aging and CKD. In contrast to observations made by Tatosian et al. 2021), the study conducted here found that only OATP1B activity in elderly patients with CKD was decreased. However, there was a commensurate change in PTV-lactone (and no difference in the metabolite to parent ratio) complicating the interpretation of the trend. Hence no firm conclusions could be drawn on the impact of aging and CKD status to OATP1B function. Reduced activity BCRP polymorphisms showed increased PK parameters for RSV, as expected, as did both aging and CKD status. Similar to findings from Tatosian et al. (2021), RSV PK was not associated with CKD, as it could not be dissociated from the changes observed in elderly and a more significant effect on C_max would be anticipated with reduction in intestinal BCRP activity. When both RSV and PTV changes are considered, a definitive conclusion on the effect of CKD on hepatic OATP1B transporter activity could not be drawn. The study had notable limitations, including the sparse population evaluated, the lack of stratification based on CKD severity, and the characteristics of the population including gender and body weight differences. The study highlights that changes in drug metabolizing enzymes and drug transporter activity are observed in elderly patients and in CKD and further exploration using PBPK modeling is underway.

Table 9. Comparison of microdose cocktail across studies.

Probe substrate	Total dose administered (µg)
	Pruekasaritanont	Tatosian	Rattanacheeworn
Midazolam	10	10	30
Dabigatran etexilate	375	375	750
pitavastatin	10	10	10
Rosuvastatin	25	50	50
atorvastatin	50	100	100

11. Quantitative prediction of P-glycoprotein-mediated drug–drug interactions and intestinal absorption using humanized mice

Source: Br J Pharmacol. 2021;178(21):4335–4351

Synopsis

This article is focused on the utility and application of human MDR1 mouse artificial chromosome (hMDR1-MAC) mice carrying human P-glycoprotein (P-gp), with a purpose to predict P-gp-mediated drug–drug interactions (DDIs) and non-linear absorption. For DDI prediction, known P-gp substrates such as aliskiren, betrixaban, celiprolol, digoxin, fexofenadine and talinolol were administered to wildtype, Mdr1a/b-KO and hMDR1-MAC mice. The oral area under the curve (AUC) of these six substrates in wild-type and hMDR1-MAC mice were calculated and compared to their AUCs in Mdr1a/b-KO. The human AUCR of these drugs (caused by P-inhibition) was obtained from the University of Washington Metabolism and Transport Drug Interaction Database (DIDB). The results clearly showed a correlation between AUCR_human and AUCR_hMDR1-MAC (R² = 0.88) but not between AUCR_human and AUCR_wild-type (R² = 0.25).

For prediction of P-gp mediated non-linear absorption, three P-gp substrates (aliskiren, betrixaban and celiprolol) were administered to hMDR1-MAC mice. Using default values for regional distribution of P-gp in the human intestine (applied to humanized hMDR1-MAC mice), GastroPlus (ACAT model) was then utilized for each substrate to determine K_m and V_max as a quantitative index of P-gp-mediated non-linear absorption. These data showed that both the K_m and V_max values (per kg of body weight) of each substrate in hMDR1-MAC mice were similar to those in humans (R² close to 1).

Commentary

P-gp is ubiquitously expressed in tissues and handles a wide variety of substrates. Because of its broad substrate specificity it has been widely recognized for its important role in drug disposition and DDIs. This was clearly evident from the recent review by Saravanakumar et al. that identified P-gp as the most studied among all the clinically important drug transporters in drug disposition for the list of FDA-approved drugs (Saravanakumar et al. 2019). However, while regulatory agencies suggest studying DDI risk assessments during drug development, there are still challenges to accurately and quantitatively predict P-gp mediated DDIs and non-linear absorption. Current practice for evaluating preclinical P-gp-mediated DDIs involves in vitro systems and basic prediction models, but do not offer quantitative assessment. Thus, hMDR1-MAC mice was evaluated in this study to improve the predictions of DDI and nonlinear absorption quantitatively.

The results of the PK study in wild-type and hMDR1-MAC mice demonstrated that the increase in maximum human AUC of drugs due to intestinal P-gp inhibition is better captured in hMDR1-MAC mice compared to wild-type mice. This was confirmed by the correlation analysis of AUC ratios (AUCR_hMDR1-MAC vs AUCR_human) suggesting similar role of P-gp to drug exposure in hMDR1-MAC mice and humans. Moreover, the kinetic parameters such as affinity (K_m) and maximum velocity (V_max) for betrixaban, celiprolol and aliskiren were similar between hMDR1-MAC mice and humans showing similar P-gp function in hMDR1-MAC mice and humans. However, this analysis relied on the scaling factor that was needed for scaling AUCR due to the differences in the protein expression of P-gp in humans and hMDR1-MAC mice. The species difference was evident as indicated by poor correlation of AUCR between wild type mice and humans which was attributed to differences in the kinetics parameters (K_m and V_max) and expression level of P-gp in the intestine.

It is important to note that genetically-modified rodent models, such as hMDR1-MAC mice, still have challenges in predicting oral absorption and DDIs. e.g. in the present study, the dynamic range of AUCR in hMDR1-MAC mice (0.9–2) was smaller than the AUCR range in human (1.5–6.3) potentially due to lower expression of human P-gp in hMDR1-MAC mice and this is evident by the non-linearity of absorption seen across the tested doses of drugs (aliskiren, betrixban and celoprolol). Interestingly, authors show that the oral absorption of celiprolol is decreased in Mdr1a KO mice and it is due to the lower expression of OATP2B1 which is involved in its absorption. Thus, further understanding of protein expression levels of other clinically important transporters (BCRP, OATP2B1, PEPT1 and ASBT) and enzymes (CYPs and UGTs) is critical for the potential application of this model in quantitative prediction of DDIs for drugs that involve enzyme-transporter interplay. Overall, this study showed the potential application of hMDR1-MAC mice to quantitatively predict both non-linear absorption and DDIs mediated by P-gp during early drug development.

12. Regulation of OATP1B1 function by tyrosine kinase-mediated phosphorylation

Source: Clin Cancer Res. 2021;27(15):4301–4310

Synopsis

In their paper, Hayden and co-workers examined the regulation of OATP1B1 function by tyrosine kinases. Using HEK-293 cells overexpressing OATP1B1 and 8-(2-[fluoresceinyl]-aminoethylthio)-adenosine-30,50-cyclic-monophosphate (8FcA) and estradiol-17β-glucuronide as substrates authors screened 46 tyrosine kinase inhibitors (TKIs) against OATP1B1 transport activity and determined that 29 of 46 TKIs were inhibitors of OATP1B1 with nilotinib being the most potent inhibitor. Nilotinib was determined to be a non-competitive inhibitor of OATP1B1 and was shown to reduce phosphorylation of tyrosine residues at positions 640 and 645. Site-directed mutagenesis of these residues also reduced OATP1B1 transport activity highlighting the importance of the phosphorylation of these residues in OATP1B1 function. The expression of OATP1B1 in the overexpressed HEK293 cells was quantified by mass spectrometry and the probability of OATP1B1 peptides to be phosphorylated was estimated using PhosphoRS. 31 tyrosine residues were identified as having a high probability of being phosphorylated and 23 of these were determined to be phosphorylated to some degree by mass spectrometry. Exposure of OATP1B1 overexpressed HEK293 cells to nilotinib significantly reduced the phosphorylation of the tyrosine residue at position 645 but to a lesser extent at position 640. The authors then performed uptake experiments in HEK293 cells expressing OATP1B1 where the tyrosine residues at 640 and 645 were mutated to phenylalanine and observed that the mutation at either of these residues significantly reduced OATP1B1 function.

To investigate the clinical relevance of this finding authors looked at the impact of nilotinib on the PK of OATP1B1 substrate drug, rosuvastatin. The data clearly indicated significant increase in the systemic exposure of rosuvastatin when OATP1B1 is inhibited by nilotinib. Consequently, authors determined the impact of siRNA-mediated knockdown of 779 kinases on OATP1B1 function in HEK-293 cells overexpressing OATP1B1 and identified involvement of Src-kinase, LYN in the regulation of OATP1B1 function. These data demonstrated that OATP1B1 is impacted by kinase activity and that dysregulation of this pathway by drugs can potentially result in adverse DDIs.

Commentary

Phosphorylation of proteins is a post-translational modification that is used by cells to regulate their function. Kinase activity is known to regulate many drug transporters including OAT1 and OAT3 which potentially could impact drug clearance when these cell signalling pathways became misregulated (Zhang J et al. 2021). Based on unexpected clinical DDIs in patients after co-administration of statins and TKIs authors evaluated how phosphorylation of OATP1B1 could impact its function and the clearance of statins.

Previously, some OATP inhibitors have been shown to have long-lasting inhibitory effect on OATP1B1 activity after pre-incubation compared to co-incubation (Panfen et al. 2019; Tatrai et al. 2019) and therefore, the regulatory agencies now require the effect of perpetrator drugs to be investigated on OATP1B transporters in pre-incubation (instead of coincubation) condition to predict potential DDI risk. However, the actual mechanism by which drugs inhibit OATP1B function has not been completely elucidated and it is an active area of research. Recently, OATP activity has been shown to be modulated by various post-translational modifications including N-glycosylation, phosphorylation and ubiquitination (Lee et al. 2020). These literature highlights the role diverse regulatory pathways in the regulation of OATP transporters and significant risk for DDI. In this context, Hayden E et al. identified novel regulatory pathway for OATP1B1 that involves LYN (Src-kinase) mediated phosphorylation which can be targeted by TKIs. This effect has direct clinical relevance as it gives potential explanation for the observed DDI between statins and TKIs. The results from this study certainly help in deciding drugs along with TKIs. Indeed, based on the data provided in this study, statin drugs need to be avoided or adjusted for dose in patients taking TKIs.

Acknowledgements

Authors would like to thank Michael Zientek (Takeda Development Center Americas Inc.) for his critical review and valuable feedback on this manuscript.

Disclosure statement

Paresh P. Chothe, Masanori Nakakariya, Charles J. Rotter, Philip Sandoval and Kimio Tohyama hold common stocks in Takeda Development Center Americas. Diane Ramsden holds common stocks in Takeda Development Center Americas and AstraZeneca.