Predicting OCT2/MATEs‐Mediated Drug Interactions in Healthy Volunteers and Patients with Chronic Kidney Disease: Insights from Extended Clearance Concept, Endogenous Biomarkers, and In Vitro Inhibition Studies (Perspectives from the International Transporter Consortium)

Abstract

Organic cation transporter (OCT) 2 and multidrug and toxin extrusion (MATE) transporters play significant roles in the renal secretion of organic cations and drug–drug interactions (DDIs). Recent in vitro studies indicate that the K _i values for OCT2 exhibit substrate dependency and increase in potency with pre‐incubation. However, consensus is lacking on whether these factors should be considered in predicting in vivo inhibition. Physiologically based pharmacokinetic models, combined with the extended clearance concept, have been used and are discussed here for OCT2/MATEs probes. In addition to modeling, early clinical studies use endogenous biomarkers to evaluate transporter‐mediated DDI risk, with the aim of avoiding unnecessary clinical DDI studies. Identified biomarkers for OCT2/MATEs, such as creatinine, N¹‐methylnicotinamide, and N¹‐methyladenosine, have proven useful in confirming clinically relevant OCT2/MATEs‐mediated DDIs when renal clearance (CLr) is used as an endpoint; their application is discussed further. From a clinical perspective, the intact nephron hypothesis (INH), which postulates that the decrease in CLr in chronic kidney disease (CKD) is proportional to that in nephron numbers, has been proposed. However, reports suggest that the secretion clearance of creatinine and substrates of organic anion transporters (OATs) does not follow this proportionality in patients with CKD. This state‐of‐the‐art review highlights key developments in predicting OCT2/MATEs‐mediated DDIs in healthy volunteers and explores the prediction of clinical OCT2/MATEs DDI risk in patients with CKD by comparing substrate‐dependent changes in secretion clearance for substrates of OCT2/MATEs and OATs. Recommendations for the prediction of OCT2/MATEs‐mediated DDI risk, together with the current knowledge gaps and future directions, are discussed.

Renal clearance (CLr) of drugs typically involves glomerular filtration and may also involve secretion into the renal tubules and reabsorption back into the blood. Only unbound drugs undergo glomerular filtration in the glomerulus, and glomerular filtration clearance depends on glomerular filtration rate (GFR) and unbound fraction of drugs in the plasma. The reabsorption process also involves not only passive diffusion but also transporters. ¹ , ² , ³ In general, compounds with high CLr have a greater contribution of tubular secretion, resulting in CLr that is more dependent on secretion clearance. It has also been reported that most basic and acidic compounds are susceptible to renal tubular secretion. ⁴ Renal secretion involves organic cation transporter (OCT) 2 and organic anion transporter (OAT) 1/3 in uptake from renal blood into renal epithelial cells and multidrug and toxin extrusion (MATE) transporters, P‐glycoprotein (P‐gp) and multidrug resistance protein (MRP) in excretion from epithelial cells into the tubular lumen. ⁵ , ⁶ The coordinated function of uptake by OCT2 and excretion by MATEs (MATE1 and MATE2‐K) is responsible for the secretion of cationic compounds. Metformin, a typical substrate for OCT2/MATEs, serves as a probe substrate for OCT2/MATEs‐mediated DDI risk assessment. Increase in exposure in plasma and decrease in CLr of metformin have been reported with the combination of MATEs inhibitors such as cimetidine, trimethoprim, and pyrimethamine. ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ Moreover, plasma exposure of metformin increased in the co‐administration of dolutegravir, which has a lower inhibition constant (K _i ) for OCT2 than MATEs. ¹⁶ , ¹⁷ These results highlight the significance of OCT2/MATEs‐mediated DDIs. ⁷ , ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ , ¹⁸

Transporter‐mediated DDIs are generally evaluated using a static or dynamic model using K _i for each transporter obtained from in vitro studies to assess clinical DDI risk for new molecular entities (NMEs). However, there may be false‐negative cases where the predicted DDI risk is lower than the actual one ¹⁹ , ²⁰ , ²¹ due to the discrepancy of apparent K _i values between in vitro prediction studies and actual in vivo values, or false‐positive cases where the DDI risk is predicted to be high, ¹⁹ , ²⁰ , ²¹ leading to unnecessary clinical studies. ¹⁸ Some endogenous compounds can serve as biomarkers to determine early clinical transporter‐mediated DDI risk. Therefore, assessing the effect of newly developed drugs on the plasma concentrations or CLr of endogenous biomarkers in early clinical studies has been proposed as a useful approach. ¹¹ , ¹⁴ , ¹⁷ , ²² , ²³ , ²⁴ , ²⁵ , ²⁶ , ²⁷ , ²⁸ , ²⁹ , ³⁰ , ³¹ , ³² , ³³ The International Council for Harmonization of Technical Requirements for Pharmaceuticals for Human Use (ICH) has also recommended consideration of endogenous biomarker coproporphyrin I (CP‐I) to support early evaluation of OATP1B1 DDI risk. ³⁴

Creatinine, N¹‐methylnicotiamide (NMN), and N¹‐methyladenosine (m¹A) have been reported as endogenous substrates of OCT2/MATEs. Their CLr is reduced with the administration of clinically used MATEs inhibitors, cimetidine, pyrimethamine, and trimethoprim. This effect is correlated to changes observed for a probe substrate, metformin. ⁷ , ⁹ , ¹¹ , ¹⁴ , ¹⁵ , ¹⁷ Based on evidence so far, creatinine and NMN are currently proposed as Tier 2 endogenous biomarkers (not fully validated biomarkers, but recommended to collect these biomarker data in clinical studies for DDI risk assessment) to support the evaluation of OCT2/MATEs‐mediated DDI risk. ³⁰ Clinical OCT2/MATEs inhibition risk for NMEs can be assessed according to the ratio of the maximum unbound concentration in plasma (I _max,u) to K _i (I _max,u/K _i ), with cutoff criteria specified in the DDI guidance/guidelines (0.1 for OCT2 and MATE1/2‐K in United States, 0.02 for OCT2 and MATE1/2‐K in Europe, and 0.1 for OCT2 and 0.02 for MATE1/2‐K in Japan and ICH M12). ³⁴ , ³⁵ , ³⁶ , ³⁷ To avoid unnecessary clinical DDI studies using probe substrates, clinical DDI risk can be evaluated based on the change in the pharmacokinetics of an endogenous biomarker in early clinical studies, particularly if the ratio exceeds the cutoff value. ¹⁹ , ²⁴ , ²⁸ Moreover, physiologically based pharmacokinetic (PBPK) models considering renal physiological mechanisms have been employed for analyzing OCT2/MATEs‐mediated DDI with metformin ³⁸ , ³⁹ or endogenous substrates. ⁴⁰ , ⁴¹ , ⁴² , ⁴³ This approach contributes to elucidating the mechanism of OCT2/MATEs‐mediated DDIs. To predict OCT2/MATEs‐mediated DDIs more accurately, it is necessary to estimate in vivo‐relevant inhibitory potencies from in vitro data. Various properties of OCT2/MATEs, such as substrate‐dependent and pre‐incubation effects on the extent of inhibition, have been reported. ¹⁹ , ⁴⁴ , ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ , ⁴⁹ A better understanding of these properties is expected to enhance the accuracy of predicting clinical OCT2/MATEs‐mediated DDIs.

In addition to the activities of renal drug transporters, the CLr of a drug is significantly influenced by the degree of kidney function. Approximately 10% of the world’s population consists of patients with chronic kidney disease (CKD), ⁵⁰ which may impact the clearance of a drug not only in the kidney but also in the liver. ⁵¹ , ⁵² , ⁵³ , ⁵⁴ In patients with CKD, DDI risk can differ relative to healthy individuals due to the concomitant use of multiple medications. ⁵⁵ Accurate prediction of drug clearance in patients with kidney failure is essential for determining appropriate dosage. GFR is commonly used as a surrogate marker to assess kidney function. ⁵⁶ Creatinine clearance (CL_cre) or estimated GFR, which is calculated using creatinine concentration in the serum, is employed in current clinical practice. ⁵⁷ , ⁵⁸ Currently, Chronic Kidney Disease Epidemiology Collaboration (CKD‐EPI) 2021 equation is the preferred method for estimating GFR in the United States and Europe. ⁵⁹ Dose adjustment of a drug in patients with CKD is generally recommended based on data supporting the assumption that the CLr declines proportionally with a decrease in GFR in kidney failure based on the intact nephron hypothesis (INH). The INH assumes that any loss of glomerular/tubular function in a nephron represents a proportional loss of the number of intact nephrons; thus, tubular secretion and re‐absorption should also decrease proportionally to GFR in CKD. ⁶⁰ , ⁶¹ However, recent studies have identified scenarios when secretion clearance changes beyond INH (defined as “non‐INH scenario”). ⁴² , ⁵³ , ⁶² This trend is particularly evident for OAT transporters where the activity of these transporters was estimated to decline by an additional 50% relative to the GFR decline in severe CKD. ⁶² However, it has not been clarified yet whether OCT2/MATEs substrates also follow INH scenario. Additionally, patients with impaired kidney function can take multiple medications, posing a latent risk for DDIs. ⁶³ While some studies have employed PBPK models to predict the pharmacokinetics of drugs in patients with kidney failure, ⁴² , ⁵⁴ , ⁶⁴ , ⁶⁵ reports predicting DDIs in these patients, accounting for the non‐INH scenario, are limited. ⁶⁶

The goal of this state‐of‐the‐art review is to summarize published research on predicting OCT2/MATEs‐mediated DDIs in healthy volunteers and to predict clinical DDIs in patients with CKD by considering the contribution of secretion clearance for substrates of OCT2/MATEs and OATs. Analysis of available clinical data for OCT2/MATEs substrates in patients with CKD is performed to investigate the effect of CKD on activity of these transporters in the patients. Subsequently, prediction of OCT2/MATEs‐mediated DDIs in patients with CKD was performed. Additionally, the findings from both in vitro and in vivo studies have been summarized, together with analyses using the extended clearance concept (ECC) and PBPK models related to OCT2/MATEs‐mediated DDIs reported to date. ECC specifies the rate‐limiting step in drug elimination, facilitating comprehension of DDIs ⁶ , ³⁹ , ⁶⁷ , ⁶⁸ (See the section “OCT2/MATEs‐mediated DDI prediction using a static model based on the extended clearance concept (ECC)”). Factors contributing to the prediction of OCT2/MATEs‐mediated DDI risks in healthy volunteers, including endogenous biomarkers and characteristics for OCT2/MATEs are discussed.

HISTORICAL OVERVIEW OF OCT2/MATEs IN THE RENAL SECRETION OF DRUGS

Table 1 provides an overview of the primary research on OCT2/MATEs. Cimetidine was historically recognized to decrease CLr of metformin by strongly inhibiting so‐called “organic cation transport systems”. Early studies identified OCT2 as an important renal organic cation transporter and later studies identified also MATE1 and MATE2‐K as renal transporters for various organic cations. ⁶⁹ , ⁷⁰ In Oct2/Oct1 knockout mice, increased plasma concentrations and decreased CLr of tetraethylammonium (TEA) ⁷¹ and metformin ⁷² after intravenous administration were observed, suggesting the importance of OCT2‐mediated renal uptake. Furthermore, studies with Mate1 knockout mice demonstrated increased plasma concentrations and reduced CLr of multiple cationic compounds including metformin, after its intravenous bolus or infusion administration, indicating the involvement of MATE1 in the renal excretion of these drugs in vivo. ⁷³ , ⁷⁴ The International Transporter Consortium (ITC) has recommended a framework for assessing transporter‐mediated DDI of NMEs, ⁵ and current DDI guidance/guidelines include in vitro transporter inhibition studies for OCT2/MATEs to assess the DDI risk of NMEs. ³⁴ , ³⁵ , ³⁶ , ³⁷

Table 1.

Historical overview of the research on OCT2/MATEs

Year	Non‐clinical study	Clinical study	DDI	PBPK	Biomarker	Contents	Reference
1996–	✓	–	–	–	–	Identification of OCT2	[136, 137]
2003–	✓	–	–	–	–	Decrease in CLr and increase in exposure of TEA and metformin in Oct2/Oct1 knockout mice	[71, 72]
2005	✓	–	–	–	–	Identification of MATE1	[69]
2005–	✓	–	–	–	–	Accumulation of platinum‐based anticancer agents in the kidneys by OCT2/MATEs‐mediated transport	[86, 87, 88, 89, 90]
2006–	✓	–	–	–	–	Identification of MATE2‐K	[70, 138]
2009	✓	–	✓	–	–	In vitro inhibition study of cimetidine using double‐transfected OCT2/MATE1 cells, suggesting MATEs‐specific in vitro inhibition by cimetidine	[75]
2009	✓	–	–	–	–	Decrease in CLr and increase in exposure of metformin in Mate1 knockout mice	[73]
2010	✓	–	✓	–	–	Identification of pyrimethamine as a potent MATEs inhibitor	[76]
2010	–	–	✓	–	✓	White paper on transporters in drug development by ITC	[5]
2011	–	✓	✓	–	–	Decrease in CLr of metformin by in the co‐administration of MATEs‐specific inhibitor, pyrimethamine in human	[9]
2012	✓	✓	✓	–	✓	Decrease in CLr of an endogenous biomarker, NMN, correlated with metformin in the co‐administration of pyrimethamine in human	[29]
2012	✓	–	✓	–	–	Inhibition of MATEs is the mechanism of DDIs caused by cimetidine in the kidney.	[77]
2013	✓	–	✓	–	–	Identification of trimethoprim as a more potent inhibitor for MATE2‐K compared to MATE1 and OCT2	[139]
2013	–	–	–	✓	–	Establishment of Mech KiM (interplay of GFR, OCT2 and MATEs)	[119]
2013	–	–	✓	–	✓	Proposal of the decision tree for DDI evaluation of OCT2/MATEs substrates or inhibitors by ITC	[122]
2013–	✓	✓	✓	–	✓	Decrease in CLr of NMN correlated with metformin in the co‐administration of trimethoprim in human	[10, 11]
2016	–	✓	✓	–	✓	Increase in AUC of metformin in the co‐administration of dolutegravir possibly by OCT2 inhibition in human	[16]
2016	✓	–	–	–	–	Quantitative handling of bidirectional transport of OCT1 substrates driven by electrochemical membrane potential	[94]
2016	–	–	✓	✓	–	PBPK model analysis of DDI between metformin and cimetidine, considering OCT2/MATEs‐mediated interaction processes	[38]
2016	✓	–	–	–	–	Diminishment of OCT2‐mediated uptake by TKIs through a mechanism that suppresses the OCT2 tyrosine phosphorylation	[140]
2018	–	–	✓	✓	✓	Review of the endogenous biomarkers of transporters by ITC	[22]
2019	–	–	✓	✓	–	PBPK model analysis of DDI between metformin and cimetidine, indicating the MATEs inhibition plays a major mechanism of in vivo inhibition	[39]
2019–	✓	–	✓	–	–	Pre‐incubation effect of inhibitors in vitro on OCT2/MATEs‐mediated inhibition	[47, 48]
2020	–	–	✓	✓	✓	PBPK model analysis of creatinine in healthy volunteers and patients with CKD	[40, 41, 42]
2021	✓	✓	✓	–	✓	A clinical DDI study between metformin or the endogenous substrates of OCT2/MATEs (creatinine, NMN, and m1A) and pyrimethamine	[14]
2021	✓	–	✓	–	–	No significant difference in the inhibitory effect of MATE1 inhibitors between the uptake and efflux transport directions	[44]
2022	–	–	✓	✓	✓	PBPK model analysis of NMN considering diurnal variation of OCT2 activity, GFR, renal blood flow and biosynthesis rate and inhibition of NMN biosynthesis	[43]
2024	–	–	✓	✓	✓	ITC recommendation of endogenous biomarkers to measure in Phase 1 study for DDI risk assessment	[30]
2024	✓	✓	✓	–	✓	NMN may be more effective for detecting OCT2 inhibition, while m1A may be better suited for identifying MATEs inhibition	[17]

CKD, chronic kidney disease; CLr, renal clearance; DDI, drug–drug interaction; GFR, glomerular filtration rate; ITC, International Transporter Consortium; IVIVE, in vitro–in vivo extrapolation; m¹A, N¹‐methyladenosine; Mech KiM, Mechanistic Kidney Model; NMN, N¹‐methylnicotinamide; PBPK, physiologically based pharmacokinetics; TEA, tetraethylammonium; TKI, tyrosine kinase inhibitor.

Since the discovery of MATEs, extensive analyses have been conducted to clarify the mechanism of OCT2‐ and MATEs‐mediated renal DDIs. Several studies have shown that the inhibition of CLr of organic cations by cimetidine and pyrimethamine is mainly attributed to the inhibition of MATEs rather than OCT2 by comparing unbound plasma concentration and the in vitro K _i values. ⁷⁵ , ⁷⁶ , ⁷⁷ Kusuhara and Ito et al. reported increase in plasma concentration and decrease in CLr of metformin when co‐administered with pyrimethamine in humans. ⁹ These findings, together with much higher I _max,u/K _i for MATEs than OCT2, suggested that pyrimethamine inhibited MATEs rather than OCT2. ⁹ Trimethoprim also exhibited a reduction in CLr of metformin in humans, with stronger inhibition for MATEs than OCT2. ¹⁰ , ¹¹ PBPK model analysis suggested cimetidine caused DDI with metformin to primarily inhibit MATEs compared to OCT2 ³⁹ (See the section “Prediction of OCT2/MATEs‐mediated DDIs of metformin caused by typical inhibitors using PBPK model”). Dolutegravir increased AUC and decreased CLr of metformin and the I _max,u/K _i for OCT2 was higher than that for MATEs, ¹⁶ , ¹⁷ suggesting OCT2 inhibition may play a predominant role in the DDI. ¹⁶ , ¹⁷ , ⁷⁸ ICH M12 and U.S. Food & Drug Administration (FDA) DDI guidance identify dolutegravir as a more potent inhibitor of OCT2, distinguishing it from MATE inhibitors such as pyrimethamine. ³⁴ , ⁷⁹ (See the section “Clinical OCT2/MATEs‐mediated DDIs and characterization of OCT2/MATEs inhibitors”).

NMN was initially suggested as a metabolite of niacin or nicotinamide which is a possible endogenous biomarker to assess the renal secretion, ⁸⁰ and was later confirmed as a substrate for OCT2/MATEs. ⁷⁰ , ⁸¹ Ito et al. reported that a decrease in CLr of NMN after the co‐administration of pyrimethamine corresponded to that of metformin in human, suggesting that change in CLr of NMN may serve as a clinical endogenous biomarker for OCT2/MATEs. ²⁹ Similarly, m¹A has been reported as an endogenous substrate of OCT2/MATEs. ¹⁴ , ⁸² However, m¹A is currently considered as an exploratory biomarker due to the limited clinical evidence ³⁰ (See the section “Comparison of the apparent in vivo K _i values for metformin and creatinine, NMN and m¹A”). Recent PBPK analyses of DDI for creatinine and NMN have also contributed to elucidating the mechanism of renal DDIs with the concurrent use of OCT2/MATEs inhibitors (See the section “Prediction of OCT2/MATEs‐mediated DDIs of endogenous biomarkers using PBPK model”). Moreover, advancements in understanding the properties of active transport by OCT2/MATEs, such as the pre‐incubation effect, ⁴⁵ , ⁴⁶ , ⁴⁷ , ⁴⁸ and the effect of serum protein on inhibitory potency ⁸³ (defined as “protein‐mediated inhibitory effect” in this review), have been reported. These advancements are expected to enhance the accuracy of in vitro‐in vivo extrapolation (IVIVE) by obtaining more in vivo‐relevant K _i values (See the section “Advances of in vitro studies for prediction of OCT2/MATEs‐mediated DDIs and the effect of protein binding on DDI prediction”). Recently, metabolomic analyses of human plasma and urine during co‐administration of cimetidine have been conducted to identify novel potential biomarkers for OCT2/MATEs with consideration for sensitivity and specificity (e.g. 5‐amino valeric acid betaine). ⁸⁴

From a toxicological perspective, cisplatin, a platinum‐based anticancer agent, exhibits nephrotoxicity. ⁸⁵ Inui and coworkers suggest that OCT2/MATEs are involved in the accumulation in the kidney, resulting in nephrotoxicity. ⁸⁶ , ⁸⁷ A study using Oct2/Oct1 knockout mice showed reduced cisplatin‐induced nephrotoxicity with lowering urinary excretion of cisplatin, ⁸⁸ suggesting the importance of OCT2 on the renal toxicity of cisplatin. Unlike cisplatin, other platinum‐based anticancer agents may exhibit reduced nephrotoxicity because they are either more transported from renal epithelial cells to urine by MATEs or less transported from blood to epithelial cells by OCT2. ⁸⁹ , ⁹⁰ Therefore, the selectivity of OCT2/MATEs transport plays a crucial role in the toxicology and/or pharmacology of a drug. Inui’s group has also reported that the interaction between metformin and cimetidine is primarily caused by MATE inhibition using double‐transfected hOCT2/hMATE1 cells. ⁷⁵ A recent study showed that IC₅₀ values in permeability assays with hOCT2/hMATE1 cells enable more accurate OCT2/MATEs‐mediated DDI predictions compared to single‐transfected OCT2 or MATE1 systems. ⁹¹ These findings suggest the potential role of double‐transfected OCT2/MATEs cells, and further research in this field is expected.

CHARACTERISTICS OF OCT2‐ AND MATEs‐MEDIATED TRANSPORT

Characteristics of OCT2‐mediated transport

OCT2 is expressed in the basolateral membrane of renal epithelial cells and plays a predominant role in the uptake of organic cations from the renal blood into the epithelial cells (Figure 1 ). The driving force behind OCT2‐mediated transport is the electrochemical gradient based on the relative concentrations between drugs in the intra‐ and extracellular spaces and the difference of membrane potential between the intracellular and extracellular space, ⁹² leading to the bidirectional transport between renal blood and epithelial cells. However, owing to the strong negative charge within the cells, transport from the extracellular to intracellular space predominates over that from intracellular to extracellular space, leading to the accumulation of organic cations within the cells. The electrochemical gradient based on the difference of membrane potential is defined by the Nernst equation, ⁹³ and the velocity of apparent transport of the ionized cation into the intracellular space (J _app,ionized) is expressed as Eqs. (1) and (2). The parameters in all equations presented in the main text are described in Supplementary Method S1 .

$$J_{\mathrm{app},\text{ionized}}=P_{d,\text{ionized}}\times \mathrm{SA}\times \frac{N}{e^N-1}\times \left(C_{\text{out}}-C_{\text{in}}\times e^N\right)$$ (1)

$$N=\frac{z\times \varphi \times F_c}{R\times T}$$ (2)

Figure 1.

The mechanism of OCT2 and MATEs‐mediated transport of organic cations in the kidney. OCT2 exhibits bidirectional transport of substrates into and out of cells, driven by the electrochemical gradient across the cell membrane. The negative membrane potential in cells (approximately −70 mV) leads to predominant influx transport by OCT2. MATEs transport substrates from renal cells into tubules, utilizing the pH difference between cells and tubules as a driving force. OCT2 and MATEs exhibit broadly overlapping substrate specificity and coordinate in tubular secretion. ¹²⁰

Chien et al. extended this membrane potential‐dependent transport model to the OCT1‐mediated active transport. ⁹⁴ Eq. (3) illustrates bidirectional transport by OCT2 driven by an electrochemical gradient in the renal epithelial cells.

$$\begin{array}{c}{J}_{\mathrm{OCT}2}=\frac{N}{e^N-1}\times V_{\max ,\mathrm{OCT}2,\mathrm{int}}\hfill \\ \times \left(\frac{C_{r,\text{blood}}}{K_{m,\mathrm{OCT}2}+C_{r,\mathit{\text{blood}}}}-\frac{e^N}{R_{\mathrm{OCT}2,\inf /\mathrm{eff}}}\times \frac{C_{r,\text{cell}}}{K_{m,\mathrm{OCT}2}+C_{r,\text{cell}}}\right)\hfill \end{array}$$ (3)

Moreover, transport by OCT1/2 driven by electrochemical gradient has been applied to PBPK model analyses for DDIs of transporter substrates, metformin and creatinine, effectively explaining the DDIs. ³⁸ , ³⁹ , ⁴⁰ , ⁴¹ , ⁴²

Characteristics of MATEs‐mediated transport

Unlike OCT2, which is a transporter driven by the electrochemical gradient, MATEs are electroneutral proton anti‐porters. MATE substrates move in one direction in exchange for a proton, which moves in the opposite direction. MATEs are expressed in the brush‐border membrane of renal epithelial cells ⁷⁰ (Figure 1 ). Organic cations are transported from renal epithelial cells to tubular lumens primarily by MATEs in a pH‐dependent manner, utilizing the exchange with protons from the renal tubular lumen to the intracellular space. ⁶⁹ , ⁷⁰ , ⁷¹ , ⁷² , ⁷³ , ⁹⁵ To account for pH‐dependent transport, in vitro inhibition assays for MATEs are generally performed in the uptake direction from the extracellular to the intracellular space, with a lower pH in the intracellular space than in the extracellular one. However, there have been concerns about potential disparities between in vitro and in vivo conditions due to the opposite transport direction used in vitro relative to in vivo conditions. To address this concern, Saito et al. evaluated the inhibitory effects of various OCT2/MATEs inhibitors using MATE1‐overexpressing cells in the efflux direction from intracellular to extracellular space. ⁴⁴ They adjusted the pH to be higher in the intracellular space for in vitro studies in the efflux direction and compared the difference of IC₅₀ values in both uptake and efflux directions. ⁴⁴ The results showed that the IC₅₀ values obtained in the efflux direction were not significantly different from the IC₅₀ values in the uptake direction (Figure S1 ). These results suggest that the difference in the transport direction of a substrate does not substantially impact the inhibitory effect on MATE1. Consequently, the finding supports that the conventional methods of in vitro inhibition assays are suitable to assess the clinical DDI risk associated with MATEs inhibition.

OCT2/MATEs‐MEDIATED DDI PREDICTION USING A STATIC MODEL BASED ON THE EXTENDED CLEARANCE CONCEPT (ECC)

The ECC in the kidney describes renal tubular secretory clearance as an integrated model encompassing elementary processes such as tubular uptake, basolateral and apical efflux. ³⁹ This concept enables the identification of rate‐limiting steps in overall tubular secretion and allows for the quantitative assessment of the impact of DDIs on tubular secretory clearance through inhibition of specific transport processes. Based on the ECC, the overall intrinsic secretion clearance (CL_int,sec) involves the uptake from renal blood to renal epithelial cells (CL_int,r,inf), efflux from the epithelial cells to renal blood (CL_int,r,eff), and the excretion from epithelial cells to tubular lumen (CL_int,urine) (Eq. 4) (Figure 2 a ). Here, β _kidney is defined in Eq. (5), and crucial to determine the rate‐limiting process in renal elimination.

$${\mathrm{CL}}_{\mathrm{int},\sec }={\mathrm{CL}}_{\mathrm{int},r,\inf }\times \frac{{\mathrm{CL}}_{\mathrm{int},\text{urine}}}{{\mathrm{CL}}_{\mathrm{int},\mathrm{r},\mathrm{eff}}+{\mathrm{CL}}_{\mathrm{int},\text{urine}}}={\mathrm{CL}}_{\mathrm{int},r,\inf }\times {\beta }_{\text{kidney}}$$ (4)

$${\beta }_{\text{kidney}}=\frac{{\mathrm{CL}}_{\mathrm{int},\text{urine}}}{{\mathrm{CL}}_{\mathrm{int},r,\mathrm{eff}}+{\mathrm{CL}}_{\mathrm{int},\text{urine}}}$$ (5)

Figure 2.

Prediction of DDIs of metformin with OCT2/MATEs inhibitors in healthy volunteers using the ECC. (Panel a) The ECC in the kidney. CLr: renal clearance, CL_int,r,inf: the intrinsic renal clearance via influx into renal tubular cells, CL_int,r,eff: the intrinsic renal clearance via efflux out of renal tubular cells, CL_int,urine: the intrinsic efflux clearance from renal tubular cells to the urinary lumen; CL_int,OCT2,inf: the OCT2‐mediated intrinsic clearance via influx into the renal tubular cells, CL_int,OCT2,eff: the OCT2‐mediated intrinsic clearance via efflux out of the renal tubular cells, CL_pass,r,inf: the intrinsic passive clearance via influx into renal tubular cells, CL_pass,r,eff: the intrinsic passive clearance via efflux out of renal tubular cells, CL_int,MATEs: the MATEs‐mediated intrinsic clearance via efflux out of the renal tubular cells, CL_{pass,urine,eff}: the intrinsic passive clearance via efflux out of the renal tubular cells. (Panels b–g) The β _kidney‐dependent impact of OCT2 and/or MATEs inhibition on AUC (Panels b–d) and on CLr (Panels e–g) of metformin. Red, blue, and purple columns represent the AUC_ratio or CLr_ratio of metformin with OCT2 inhibition, MATEs inhibition, and both OCT2 and MATEs inhibition, respectively. CLr_ratio and AUC_ratio with and without OCT2 and/or MATEs inhibition can be described using Eqs. S1 and S32, respectively. The values of β _kidney were set from 0.1 (uptake/excretion‐limited process), 0.5, and 0.9 (uptake‐limited process). The K _puu of the inhibitor between renal blood and epithelial cells and I _max,u/K _i were set to 1 and 10, respectively. Inhibition for MATEs was not considered (I _max,u/K _i,MATEs = 0) in the simulation of OCT2 inhibition, and inhibition for OCT2 was not considered in the simulation of MATEs inhibition (I _max,u/K _i,OCT2 = 0). Other used parameters were described in Table S14 .

When CL_int,r,eff < CL_int,urine (β _kidney ≈ 1), CL_int,sec equals CL_int,r,inf, where CL_int,sec is governed by the uptake from renal blood to cells (uptake‐limited process). Conversely, when CL_int,r,eff > CL_int,urine (β _kidney ≈ 0), CL_int,sec equals CL_int,r,inf × CL_int,urine/CL_int,r,eff. Here, CL_int,sec is limited not only by CL_int,r,inf but also CL_int,r,eff and CL_int,urine (uptake/excretion‐limited process) (Figure 2 a ).

Understanding of the rate‐determining process based on β _kidney concept has implications on the predicted change in CLr of OCT2/MATEs substrates caused by the inhibition of OCT2/MATEs. To assess the extent of DDIs by OCT2 and/or MATEs inhibition, simulations of OCT2/MATEs‐mediated DDIs in healthy volunteers (AUC_ratio_healthy and CLr_ratio_healthy) for metformin were conducted using a static model in a β _kidney‐dependent manner (Equations are shown in Supplementary Method S1 ) (Figure 2 b–g ). CL_int,sec was fixed for all β _kidney values, and OCT2‐mediated intrinsic uptake clearance into the renal tubular cells (CL_int,OCT2,inf), OCT2‐mediated intrinsic efflux clearance into renal blood (CL_int,OCT2,eff), and MATEs‐mediated intrinsic excretion clearance into the renal tubules (CL_int,MATEs) were calculated separately for each β _kidney scenario. In these DDI simulations, it is assumed that OCT2 and/or MATEs inhibition does not affect intestinal absorption or hepatic elimination.

In an uptake/excretion‐limited process (β _kidney = 0.1), MATEs inhibition significantly increases AUC (Figure 2 b ) and decreases CLr (Figure 2 e ). In an uptake‐limited process (β _kidney = 0.9), MATEs inhibition results in a minor increase in AUC (Figure 2 d ) and a minor decrease in CLr (Figure 2 g ). CL_int,sec is affected by not only OCT2 transport but also MATEs excretion when β _kidney is smaller, thus leading to a stronger DDI with MATEs inhibition. Conversely, as β _kidney value increases and approaches the uptake‐limited process, the effect of MATEs inhibition on CL_int,sec becomes less pronounced. In an uptake‐limited process, OCT2 inhibition significantly increases AUC (Figure 2 d ) and decreases CLr (Figure 2 g ), whereas in an uptake/excretion‐limited process, OCT2 inhibition results in a minor increase in AUC (Figure 2 b ) and a minor decrease in CLr of metformin (Figure 2 e ). This finding can be explained by the fact that both influx and efflux transport by OCT2 affect CL_int,sec of metformin at low β _kidney, and therefore, OCT2 inhibition results in mild net changes in CLr and AUC. In contrast, basolateral efflux transport by OCT2 does not strongly affect CL_int,sec at higher β _kidney. In this case, inhibition of OCT2‐mediated basolateral influx transport mainly affects a decrease in CL_int,sec, causing a significant decrease in CLr and an increase in AUC. Therefore, considering ECC is crucial for accurate predictions of OCT2/MATEs‐mediated DDIs.

Inhibition of OCT2/MATEs depends on the unbound concentration of an inhibitor in the renal blood and epithelial cells (for OCT2) and in the epithelial cells (for MATEs). Specifically, inhibition of OCT2‐mediated influx transport is determined by the unbound concentrations of an inhibitor in renal blood, while OCT2‐mediated efflux and MATEs‐mediated excretion depend on the unbound concentrations in renal epithelial cells. For accurate estimation of inhibition of OCT2‐mediated efflux and MATEs‐mediated excretion, the unbound tissue‐to‐blood partition coefficient (K _puu) between renal epithelial cells and blood should be estimated. To assess the influence of unbound concentrations of an inhibitor in the renal epithelial cells on OCT2/MATEs‐mediated DDIs, DDI simulations were performed with varied K _puu of an inhibitor (Figure S2 ). In these simulations, β _kidney of a substrate was set to 0.1, representing the scenario of uptake/excretion‐limited elimination. For OCT2‐mediated inhibition, a larger K _puu of an inhibitor resulted in a smaller increase in plasma AUC and a decrease in CLr of a substrate because OCT2‐mediated basolateral efflux transport was more inhibited. Conversely, for MATEs‐mediated inhibition, a larger K _puu resulted in a more pronounced increase in AUC and a decrease in CLr of a substrate. These results indicate that K _puu of an inhibitor strongly influences OCT2/MATEs‐mediated DDIs and should be estimated accurately for predicting OCT2/MATEs‐mediated DDIs. Various methods for estimating K _puu in the hepatocytes have been reported, ⁹⁶ including a method using the ratio of the sum of active transport clearance and passive diffusion clearance (CL_pass) to CL_pass ⁹⁷ or the cell‐to‐medium concentration ratio at steady‐state in uptake assays conducted at 37°C and on ice. ⁹⁸ Cationic inhibitors can accumulate inside renal epithelial cells due to the strong negative charge within the cells, and K _puu changes more if these inhibitors are substrates for OCT2/MATEs. Further work is needed to develop a method for estimating the unbound concentration of inhibitors in renal epithelial cells to predict OCT2/MATEs‐mediated renal DDIs more accurately.

ADVANCES OF IN VITRO STUDIES FOR PREDICTION OF OCT2/MATEs‐MEDIATED DDIs AND THE EFFECT OF PROTEIN BINDING ON DDI PREDICTION

In vitro substrate‐dependency for OCT2/MATEs‐mediated inhibition

It has been previously reported that OATP1Bs, which are responsible for transporting organic anions into hepatocytes, exhibit substrate‐dependent differences in K _i values. ⁹⁹ , ¹⁰⁰ , ¹⁰¹ These substrate‐dependent differences in K _i values can impact the decision‐making process for the assessment of clinical DDI risks, and therefore likely co‐medications are often considered as substrates of in vitro OATP1B inhibition assays. ¹⁰⁰ , ¹⁰¹ Some of the previous studies have suggested that OCT2/MATEs exhibit minimal substrate‐dependent differences in in vitro K _i values. ¹⁹ , ⁴⁴ , ⁴⁹ To investigate this further, we performed a comprehensive analysis of reported in vitro K _i values for OCT2, MATE1, and MATE2‐K. Generally, IC₅₀ values are obtained as an index of inhibitory potency. However, they approximate K _i values because the substrate concentration is sufficiently lower than the Michaelis–Menten constant (K _m) for transporter inhibition assays. We compared geometric mean K _i values of 52 inhibitors using metformin vs. creatinine and NMN as substrates. The results showed that the differences in K _i values for OCT2‐ and MATEs‐mediated inhibition independent of the substrate generally remained within fourfold (Figure 3 a–i , Table  S1 ), in agreement with the previous reports. ¹⁹ , ⁴⁴ , ⁴⁹ Consequently, the substrate‐dependent difference in in vitro K _i values for OCT2/MATEs generally seems to be relatively low compared with OATP1B substrates. ⁹⁹ Yoshikado et al. proposed that substrate‐dependency should be considered for the prediction of OATP1B‐mediated DDIs with the probe substrates, and that in vivo K _i values obtained from the change in pharmacokinetics of CP‐I should be corrected by using the ratio of in vitro K _i values between probe substrates and CP‐I. ¹⁰² Conversely, for OCT2/MATEs‐mediated DDIs, in vitro IC₅₀ values using metformin as a substrate can provide sufficient risk assessment. These findings suggest that determining IC₅₀ values for each substrate is unnecessary for the prediction of OCT2/MATE‐mediated DDIs, reducing unnecessary non‐clinical studies.

Figure 3.

Substrate‐dependency of K _i values for OCT2, MATE1 and MATE2‐K‐mediated transport. Correlation between in vitro K _i values of inhibitors on metformin (K _{i,metformin,vitro}) and those on creatinine (K _{i,creatinine,vitro}, Panels a–c), NMN (K _i,NMN,vitro, Panels d–f) or TEA (K _i,TEA,vitro, Panels g–j) is shown. Red symbols represent the geometric mean of K _i values of each inhibitor based on IC₅₀ values evaluated in the previous reports (Mathialagan et al. ¹⁹ for correlation between metformin and creatinine or NMN, and Sandoval et al. ⁴⁹ for that between metformin and TEA). Blue symbols represent unevaluated inhibitors in the reports. K _i values are shown in Table S1 . Solid and dotted lines show the line of unity and fourfold error, respectively. K _i values are calculated using the Cheng‐Prusoff equation, K _i  = IC₅₀/(1 + S/K _m ), where S represents substrate concentration. Regarding the K _m values of each substrate for OCT2, MATE1, and MATE2‐K, geometric mean values obtained from the multiple reports are used (Table S13 ). In panel j, individual K _{i,metformin,vitro} and K _i,TEA,vitro for dolutegravir are plotted. Blue triangle, square, and circles represent individual K _{i,metformin,vitro} values in Human Embryonic Kidney (HEK) 293, ¹⁷ , ¹⁹ , ⁴⁷ Madin‐Darby canine kidney (MDCK)II ¹⁴¹ and Chinese hamster ovary (CHO) cells, ¹¹⁴ respectively. Open symbols represent K _{i,metformin,vitro} values from Tátrai et al., ⁴⁷ which reported the effect of pre‐incubation of OCT2/MATEs inhibitors on the inhibitory potency (shown in Table S1 ). The K _i values shown in these figures are those without pre‐incubation of an inhibitor. Abe, abemaciclib; Ama, amantadine; Ami, amiodarone; Amt, amitriptyline; Ana, anastrozole; Ato, atomoxetine; Atp, atropine; Ben, benztropine; Bup, bupropion; Bus, buspirone; Cep, cephalexin; Cim, cimetidine; Clo, clonidine; Cob, cobicistat; Cor, corticosterone; Des, desipramine; Dol, dolutegravir; Dro, dronedarone; Fam, famotidine; Flu, fluoxetine; Imi, imipramine; Iri, irinotecan; Isa, isavuconazole; Lan, lansoprazole; Lev, levofloxacin; Mex, mexiletine; Mit, mitoxantrone; Nal, naloxone; Nil, nilotinib; Niz, nizatidine; Ond, ondansetron; Phe, phenformin; Pim, pimagedine; Pro, procainamide; Pyr, pyrimethamine; Qui, quinidine; Rab, rabeprazole; Ran, ranolazine; Rit, ritonavir; Rnt, ranitidine; Rux, ruxolitinib; Sir, sirolimus; Tac, tacrine; Tel, telaprevir; Ten, teneligliptin; Tri, trimethoprim; Tro, tropicamide; Trp, tripelennamine; Tuc, tucatinib; Van, vandetanib; Ver, verapamil.

The effect of pre‐incubation on K _i values for OCT2/MATEs‐mediated inhibition

For certain transporters, such as OATP1B1/1B3, the inhibitory effect depends on the pre‐incubation time of an inhibitor. The strongest evidence of this pre‐incubation effect has been reported for cyclosporin A, ⁹⁹ , ¹⁰¹ , ¹⁰³ but also for its metabolite AM1. ¹⁰⁴ The proposed inhibition mechanism involves both competitive inhibition from the extracellular space (cis‐inhibition) and non‐competitive inhibition from the intracellular space (trans‐inhibition). ¹⁰³ Tátrai et al. evaluated the pre‐incubation effect on IC₅₀ values for various transporters, showing that inhibitors with higher lipophilicity (LogD_7.4) exhibit a stronger pre‐incubation effect on K_i values. ⁴⁷

Pre‐incubation effects for OCT2/MATEs have also been reported for tyrosine kinase inhibitors (TKIs). Crizotinib, a TKI, demonstrated a lower K _i value for OCT2 and MATE1 following pre‐incubation compared to co‐incubation. ⁴⁵ , ⁴⁶ In cases where both pre‐ and co‐incubation were performed, the K _i value was lower than those in either pre‐ or co‐incubation alone. ⁴⁵ , ⁴⁶ Similarly, ibrutinib showed persistent inhibitory effects on TEA uptake in MATE1‐expressing cells even after a 480‐minute washout period. ¹⁰⁵ Analysis of 17 inhibitors showed that the extent of the pre‐incubation effect on OCT2/MATEs inhibition is correlated with LogD_7.4 ⁴⁷ or unbound fraction in plasma (f _p) (Figure S3 and Table S2 ), suggesting that lipophilicity (i.e. LogD_7.4) and f _p may guide the decision on performing pre‐incubation studies and assessing the pre‐incubation effect for OCT2/MATEs. The exact mechanism for such correlations remains to be resolved. Additionally, the extent of the pre‐incubation effect for OCT2 may be stronger than for MATEs (Figure S3 and Table S2 ), suggesting that in vitro inhibition studies with pre‐incubation may be necessary, especially for OCT2 inhibition. Currently, ICH M12 recommends obtaining unbound IC₅₀ values that account for non‐specific binding (NSB) of inhibitors. ³⁴ NSB affects IC₅₀ values and consequently the assessment of DDI risk, especially for lipophilic compounds that are susceptible to high NSB. For accurate OCT2/MATEs‐mediated DDI risk evaluation, it is recommended to obtain IC₅₀ values that consider both pre‐incubation effects and corrections for NSB of inhibitors.

The effect of protein binding of inhibitors on prediction of OCT2/MATEs‐mediated DDIs

We assessed the effect of protein binding of inhibitors on CLr_ratio of metformin with and without an inhibitor in healthy volunteers using Eq. S1 (assuming the uptake/excretion‐limited process, β _kidney = 0.1). The predicted CLr_ratio_healthy values were notably higher than the observed values for highly plasma protein‐bound inhibitors (f _p <0.1), including abemaciclib (f _p = 0.037 ¹⁰⁶ ), bictegravir (f _p = 0.0030 ¹⁰⁷ ), dolutegravir (f _p = 0.010 ¹⁰⁸ ), tucatinib (f _p = 0.030 ¹⁰⁹ ), vandetanib (f _p = 0.070 ¹¹⁰ ) and fedratinib (f _p = 0.045 ¹¹¹ ) (Figure 4 ). The results suggest that for highly plasma protein‐bound inhibitors, OCT2/MATEs‐mediated DDIs may be underpredicted based on the assumption of “free inhibitor hypothesis” and there might be a need to scrutinize in vitro assay conditions for OCT2/MATEs‐mediated DDI prediction. An inhibition study in human serum performed by Jahic et al. ⁸³ showed protein‐mediated inhibitory effect for dolutegravir. ⁸³ The unbound fraction in serum or protein containing buffer (f _u)‐adjusted IC₅₀ for OCT2 by dolutegravir (highly protein‐bound) was sixfold lower compared to that obtained in the protein‐free buffer (Table S3 ). However, Koishikawa et al. reported that there was no apparent OCT2‐mediated IC₅₀ change of dolutegravir dependent on human serum albumin concentrations. ¹¹² Kikuchi et al. conducted in vitro inhibition studies for OCT2, MATE1, and MATE2‐K using pyrimethamine (f _p = 0.13 ⁹ ) in a buffer with 4% bovine serum albumin, and reported that the f _u‐adjusted IC₅₀ value was within the twofold of that obtained in protein‐free buffer ¹¹³ (Table S3 ). Therefore, there is currently no consensus regarding protein‐mediated inhibitory effect on assessment of inhibition for OCT2/MATEs. Based on the existing evidence, future OCT2/MATEs inhibition studies with highly protein‐bound inhibitors are advised to explore the protein‐mediated inhibitory effect and implications on prediction of clinical OCT2/MATEs‐mediated DDIs.

Figure 4.

Effect of plasma protein binding of OCT2/MATEs inhibitors on prediction of CLr_ratio_healthy of metformin with and without inhibitors. The red, blue, green, and brown circles represent the predicted and observed CLr_ratio_healthy of metformin with/without co‐administration of OCT2/MATEs inhibitors with weak to moderate protein binding, cimetidine (f _p = 0.81 ⁷ ), pyrimethamine (f _p = 0.13 ⁹ ), trimethoprim (f _p = 0.49 ¹⁴² ) and ondansetron (f _p = 0.27 ¹⁴³ ), respectively. The black, gray, purple, green, orange, and red triangles describe those with/without co‐administration of OCT2/MATEs inhibitors with high protein binding, vandetanib (f _p = 0.070 ¹¹⁰ ), abemaciclib (f _p = 0.037 ¹⁰⁶ ), bictegravir (f _p = 0.0030 ¹⁰⁷ ), fedratinib (f _p = 0.045 ¹¹¹ ), tucatinib (f _p = 0.030 ¹⁰⁹ ) and dolutegravir (f _p = 0.010 ¹⁰⁸ ), respectively. CLr_ratio_healthy was calculated using Eq. S1. The solid and dotted lines represent the predicted CLr_ratio_healthy values where the rate of change is 1‐fold and 2(0.5)‐fold, respectively. The contribution of MATE1 (f _t,MATE1, 0.974) and MATE2‐K (f _t,MATE2‐K, 0.026) was calculated by using the difference in the expression levels between human kidney cortex and overexpressing cells and uptake clearance in the overexpressing cells. ¹⁴⁴ Geometric K _{i,metformin,vitro} values shown in Table S1 were used except for bictegravir. For bictegravir, K _i values for OCT2 and MATE1‐mediated inhibition using TEA as a substrate were used because K _{i,metformin,vitro} values were not available. It was assumed that bictegravir does not inhibit MATE2‐K due to the lack of reported in vitro K _i values for MATE2‐K. Other parameters used were described in Table S14 .

CLINICAL OCT2/MATEs‐MEDIATED DDIs AND CHARACTERIZATION OF OCT2/MATEs INHIBITORS

The I _max,u/K _i value for OCT2 for cimetidine, pyrimethamine and trimethoprim are approximately 0.1. However, the values for MATE1/2‐K exceed 1 (Figure 5 and Table S4 ). The CLr_ratio_healthy of metformin when co‐administered with these inhibitors was predicted considering the rate‐limiting process (Equations are shown in Supplementary Method S1 ). The predicted CLr_ratio_healthy are higher than observed CLr_ratio_healthy when β _kidney equals 0.9 (uptake‐limited process). However, as β _kidney decreases to 0.3–0.5, where MATEs inhibition significantly affects DDI, the predicted CLr_ratio_healthy values decrease and closely match the observed values (Figure S4 and Table S5 ). Nishiyama et al. demonstrated that the interaction between metformin and cimetidine was well predicted based on in vitro K _i values of cimetidine for MATEs when β_kidney values are set to lower values (0.1 or 0.3) using mechanistic PBPK modeling ³⁹ (See the section “Prediction of OCT2/MATEs‐mediated DDIs of metformin caused by typical inhibitors using PBPK model”). These results support that CLr changes of metformin during co‐administration with these inhibitors are primarily caused by inhibition of MATEs rather than OCT2.

Figure 5.

Correlation between I _max,u/K _i for OCT2, MATE1 and MATE2‐K and AUC_ratio (panels a–c) or CLr_ratio (panels d–f). The I _max,u/K _i values of inhibitors for OCT2, MATE1, and MATE2‐K were plotted against AUC_ratio (a–c) and CLr_ratio (d–f) of metformin, comparing AUC or CLr with and without co‐administration of inhibitors in healthy volunteers. The symbols of open downward triangle, open circle, open diamond, plus symbol, cross symbol, open square, open upward triangle, open diamond, and pentagon represent I _max,u/K _i of abemaciclib, cimetidine, tucatinib, fedratinib, ondansetron, pyrimethamine, trimethoprim, tucatinib, and vandetanib, which are classified as potent MATEs inhibitors. The gray triangle and gray square represent I _max,u/K _i of bictegravir and isavuconazole (inhibitors with clinically unknown mechanism), respectively. In the calculation of I _max,u/K _i , the geometric mean of K _i values using metformin as a substrate was used except for dolutegravir and bictegravir. The light blue and blue closed circles represent the I _max,u/K _i of dolutegravir using K _i values with TEA and metformin as a substrate, respectively. For dolutegravir, K _i values on TEA and metformin from each in vitro experiment were used in the figures. For bictegravir, K _i values on TEA were used. The calculated I _max,u/K _i values for OCT2, MATE1, and MATE2‐K are shown in Table S4 . Solid and dotted lines represent predicted CLr_ratio_healthy or AUC_ratio_healthy using Eqs. S1 and S32 with β _kidney of 0.1 and 0.9, respectively. K _puu was set to 1 in this calculation. The red vertical and horizontal dotted lines represent the criteria for judging clinically relevant DDI (AUC_ratio > 1.25 or CLr_ratio < 0.8) and the cutoff values specified in ICH M12 (0.1 for OCT2, and 0.02 for MATE1 and MATE2‐K), respectively. Used parameters for the prediction were described in Table S14 . Dol, dolutegravir; GM, geometric mean; Met, metformin; TEA, tetraethlammonium.

Besides the aforementioned three inhibitors, abemaciclib, fedratinib, ondansetron, tucatinib, and vandetanib exhibited higher I _max,u/K _i values for MATEs compared to OCT2, with ratios for MATEs above 0.1, exceeding the cutoff value in DDI guidance ³⁴ , ³⁵ , ³⁶ , ³⁷ (Figure 5 and Table S4 ). Furthermore, I _max,u/K _i values of these inhibitors for MATEs are close to the values required to account for decreased CLr and increased AUC of metformin (Figure 5 ). These results suggest that DDIs with metformin by co‐administration of these inhibitors are primarily attributed to MATEs inhibition. Conversely, for dolutegravir, bictegravir, and isavuconazole, I _max,u/K _i for OCT2 was equivalent to or greater than that for MATEs (Figure 5 and Table S4 ). For dolutegravir, there is large variability in K _i values for OCT2 inhibition, ranging from 0.207 to 22.7 μM when using metformin as a substrate (Figure 3 i and Table S1 ). I _max,u/K _i using the lowest K _i value (0.207 μM) approximates the value necessary to explain the CLr decrease and AUC increase of metformin, though the geometric mean of K _i values (1.42 μM) cannot explain the DDI (Figure 5 a,d ). Regarding MATEs, I _max,u/K _i remains lower than 0.1 even when considering the reported range of K _i values (0.83–9.05 μM for MATE1 and 2.97–12.5 μM for MATE2‐K) (Figure 5 b,c,e,f ). While dolutegravir showed predominant OCT2 inhibition with a geometric mean OCT2 K _i value approximately 1/3–1/5 times that of its MATE1/2‐K K _i value, inhibitors such as pyrimethamine, cimetidine, and trimethoprim showed the opposite trend with OCT2 K_i values approximately 10–40 times higher than their MATE1/2‐K K _i values, demonstrating strong predominance of MATEs inhibition (Figure S5 and Table S1 ). The in vitro K _i values of dolutegravir for OCT2 were consistently lower than those for MATE1/2‐K within the same studies from different laboratories ¹⁷ , ¹⁹ , ¹¹⁴ (Figure S5 ). Furthermore, since dolutegravir has high membrane permeability and is not a substrate for uptake transporters but is for P‐gp, the intracellular unbound concentration of dolutegravir would likely not exceed the plasma unbound concentration. ¹⁷ These results indicate that the mechanism of DDI by dolutegravir can primarily be attributed to OCT2 inhibition, and that the lower reported K _i value for OCT2 is necessary to explain the clinical DDI with metformin. Interestingly, the K _i value for OCT2 using TEA as a substrate is lower (0.0642 μM, Figure 2 g and Table S1 ) than the geometric mean K _i value using metformin as a substrate (1.42 μM, Figure 2 g and Table S1 ) and I _max,u/K _i exceeds 1 (Figure 5 a,d ). These findings suggest that OCT2 inhibition by dolutegravir may be substrate‐dependent, although the mechanism remains unknown. Additionally, it has been reported that IC₅₀ values for OCT2 are 6–11 times lower when pre‐incubating with dolutegravir ⁴⁷ (Table S2 ). Current DDI guidance states that IC₅₀ (K _i ) of a NME should be determined following a pre‐incubation step for transporters. ³⁴ , ³⁵ , ³⁷ Existing data suggest that it may also be necessary to obtain IC₅₀ (K _i ) values with pre‐incubation for accurately predicting OCT2‐mediated DDIs. Most recently, Koishikawa et al. found that in vitro OCT2‐mediated IC₅₀ value by dolutegravir was substrate‐ and uptake time‐dependent and the lowest in vitro IC₅₀ value was comparable to the in vivo K _i value. ¹¹² Further studies on the experimental conditions for the OCT2 inhibition study are expected to refine K_i values and improve prediction of OCT2‐mediated clinical DDIs.

The I _max,u/K _i cutoff values for OCT2 and MATEs in ICH M12 are conservatively set to detect clinically relevant DDIs. The lower cutoff value for MATEs compared to OCT2 considers the possibility of inhibitor accumulation in renal tubular cells. ¹¹⁵ However, I _max,u/K _i required to demonstrate clinically relevant OCT2/MATEs‐mediated DDI (AUC_ratio > 1.25 or CLr_ratio < 0.8) is close to 1 based on the theoretical line in Figure 5 , and most MATEs inhibitors show I _max,u/K _i values similar to the theoretical value. These results suggest that the current plasma concentration‐based cutoff value for MATEs could be increased. However, as intracellular inhibitor concentrations are crucial in evaluating MATEs inhibition, establishing K _puu evaluation methods is critical to enable appropriate cutoff values for assessing MATEs‐mediated DDI risk.

A further challenge is to obtain the clinical evidence and delineation of OCT2 and MATEs‐mediated inhibition because of the lack of specific in vivo probe substrates for individual transporters, although in vitro MATEs‐specific substrates ¹⁷ , ¹¹⁶ and inhibitors ¹¹⁷ , ¹¹⁸ have been suggested. Currently, OCT2/MATEs inhibitors are mainly characterized by comparison of in vitro K _i values for OCT2 and MATEs. To address this issue, clinical DDI studies using OCT2 or MATEs‐specific in vivo probes are necessary. Further studies are expected to identify the specific probes and conduct DDI studies with OCT2/MATEs inhibitors.

PREDICTION OF OCT2/MATEs‐MEDIATED DDIs OF METFORMIN CAUSED BY TYPICAL INHIBITORS USING PBPK MODEL

Studies using PBPK models have been reported for DDI between metformin and OCT2/MATEs inhibitors (Table S6 ). Mechanistic Kidney Model (Mech KiM), which mimics the physiological structure of the kidney, has been introduced. ¹¹⁹ Mech KiM divides the kidney’s structure into glomerulus, proximal tubule, loop of Henle, distal tubule, cortical collecting duct, and medullary collecting duct. Each section has a corresponding renal blood, renal epithelial cells, and tubules, creating a detailed model that replicates the physiological structure of the kidney. Burt et al. attempted to elucidate the mechanism of DDI between metformin and cimetidine using PBPK model with Mech KiM. ³⁸ In Burt’s model, OCT2 functions as an electrogenic transporter, mediating net flux between renal blood and epithelial cells. The driving force for the transport depends on both the membrane potential and the substrate concentration gradient across the cell membrane. ³⁸ The authors showed that cimetidine in vivo K _i values for OCT2 and MATEs needed to be lower by 8‐fold and 18‐fold, respectively, compared to the in vitro K _i values in order to capture two individual clinical DDI studies between metformin and cimetidine. ³⁸ Nishiyama et al. incorporated bidirectional (both influx and efflux) OCT2‐mediated transport of metformin for PBPK model analysis of DDIs between metformin and cimetidine. ³⁹ Their results demonstrated that the DDIs between metformin and cimetidine were primarily attributed to inhibition of MATEs rather than OCT2, and that changes in plasma concentration, CLr, and the concentration in renal epithelial cells of metformin were dependent on the β _kidney value of metformin. ³⁹ The optimized in vivo K _i value for MATEs fell within the range of in vitro K _i values with a metformin β _kidney of 0.1 or 0.3 (Figure S6 ). Koepsell ¹²⁰ and Hillgren et al. ¹²¹ also suggested that DDIs between metformin and inhibitors (cimetidine and pyrimethamine) are mainly caused by inhibition of MATEs rather than OCT2. There have been reports of utilizing PBPK models for analysis to predict clinical DDI studies, with the results documented in regulatory submissions. ¹²² For instance, in the case of apalutamide, PBPK model analyses of OCT2/MATEs‐mediated DDIs with metformin have been described in the new drug application form, which showed that the AUC increase of metformin with co‐administration of apalutamide was small. ¹²³ It is envisaged that the number of PBPK analyses for prospective prediction of OCT2/MATEs‐mediated DDIs prior to clinical DDI studies will increase in the drug development of NMEs.

APPLICATION OF ENDOGENOUS BIOMARKERS FOR OCT2/MATEs FOR PREDICTION OF CLINICAL DDIs OF METFORMIN

Comparison of the apparent in vivo K _I values for metformin and OCT2/MATEs endogenous biomarkers

Recently, ITC proposed classification of endogenous biomarkers for several hepatic and renal transporters and provided recommendations for their application in drug development. ³⁰ Multiple studies have reported that, for endogenous OCT2/MATEs biomarkers, including creatinine, NMN, and m¹A, CLr decreases upon the administration of OCT2/MATEs inhibitors (Table S7 ), as summarized by the ITC. ³⁰ Creatinine has been reported as a substrate for OAT2 as well as OCT2. ¹²⁴ , ¹²⁵ , ¹²⁶ PBPK model analysis based on proteomics‐informed IVIVE of transporter clearance by Scotcher et al. showed that OCT2 and OAT2 contributed equally to the uptake of creatinine in the uptake model (when OCT2 was assumed to operate as uptake only), while in the OCT2 bidirectional model, OCT2 contribution was dependent on epithelial cell concentration. ⁴⁰ , ⁴¹ A recent report by Mathialagan et al. indicated that decinium‐22 (a potent OCT2 inhibitor) inhibited creatinine uptake, while indomethacin (a potent inhibitor of OATs) did not inhibit creatinine uptake in human primary renal tubular cells. ¹²⁷ Clinical DDI studies to confirm the change in CLr and plasma concentrations of metformin, creatinine, NMN, and m¹A were conducted during the co‐administration of pyrimethamine, ¹⁴ cimetidine ¹⁷ and dolutegravir. ¹⁷ The results showed that the change in CLr was more sensitive than the change in AUC. CLr of NMN and m¹A changed in an exposure‐dependent manner of pyrimethamine, cimetidine, and dolutegravir; these changes were correlated to the changes observed for metformin (Figure 6 ). In contrast, the decrease in CLr of creatinine was not prominent by the inhibitors compared with NMN and m¹A (Figure 6 ). Miyake et al. showed that the apparent in vivo K _i value (K _i,app,vivo) for pyrimethamine based on creatinine data was approximately 10 to 20‐fold lower than that for metformin, NMN, and m¹A. ¹⁴ Notably, K _i,app,vivo values of metformin, NMN, and m¹A were consistent with in vitro K _i values for MATEs, which were obtained in the same study, while for creatinine, the K _i,app,vivo value was significantly lower than the in vitro K _i value, ¹⁴ reflecting minimal contribution of active secretion of creatinine to its renal clearance compared with other probes.

Figure 6.

CLr and AUC change of metformin, creatinine, NMN and m¹A by oral administration of pyrimethamine, cimetidine and dolutegravir. CLr (left) and AUC (right) ratio of NMN, m¹A, or creatinine with co‐administration of pyrimethamine (10, 25 and 75 mg), ¹⁴ cimetidine (400 and 800 mg) ¹⁷ and dolutegravir (50 mg after 1^st and 5^th administration). ¹⁷ Symbols and error bars represent median and. 90% confidence interval, respectively. Cim, cimetidine; Dol, dolutegravir; Pyr, pyrimethamine.

Further analysis was conducted using CLr_ratio data from multiple reports to examine whether K _i,app,vivo values differ between metformin and the biomarkers (Table S9 ). The K _i,app,vivo ratio (R_Ki,app,vivo) between metformin and the biomarkers (creatinine and NMN) was calculated using Eqs. (6) and (7).

$$\begin{array}{c}{\mathrm{CL}}_{\mathrm{r}\_}{\text{ratio}}_{\text{healthy}}\left(\text{metformin}\right)=1\hfill \\ -\frac{\left(1-{\mathrm{CL}}_{\mathrm{r}}\_{\text{ratio}}_{\text{healthy}}\left(\mathrm{BM}\right)\times \frac{f_{\sec ,\text{metformin}}}{f_{\sec ,\mathrm{BM}}}\right)}{\left({\mathrm{R}}_{\mathrm{Ki},\mathrm{app},\text{vivo}}-\frac{\left(1-{\mathrm{CL}}_{\mathrm{r}}\_{\text{ratio}}_{\text{healthy}}\left(\mathrm{BM}\right)\right)}{f_{\sec ,\mathrm{BM}}}\times \left({\mathrm{R}}_{\mathrm{Ki},\mathrm{app},\text{vivo}}-1\right)\right)}\hfill \end{array}$$ (6)

$$R_{\mathrm{Ki},\mathrm{app},\text{vivo}}=\frac{K_{\mathrm{i},\mathrm{app},\text{vivo},\text{metformin}}}{K_{\mathrm{i},\mathrm{app},\text{vivo},\mathrm{BM}}}$$ (7)

The results showed the ratio of the apparent in vivo K _i value between metformin and the biomarker (R _Ki,app,vivo) was approximately 3.6 between metformin and creatinine, while R _Ki,app,vivo between metformin and NMN approached 1 (Figure S7 ). This analysis shows that metformin and NMN have comparable in vivo K _i values, confirming the suitability of NMN as an endogenous OCT2/MATEs biomarker for predicting DDIs of metformin. K _i,app,vivo for 7 inhibitors based on creatinine data was lower than that for metformin. It is important to note that creatinine is predominantly eliminated through glomerular filtration, and the fraction of renal secretion (f _sec) of creatinine is much smaller than that of metformin and NMN, which translates to a smaller CLr change of creatinine in the co‐administration of OCT2/MATEs inhibitors compared to metformin and NMN. Subsequently, this may result in a greater disconnect in K _i,app,vivo between creatinine and metformin compared to that between NMN and metformin. Although f _sec of creatinine is lower than that of metformin and the other biomarkers, creatinine is still included as a Tier 2 endogenous biomarker ³⁰ because it is routinely measured to monitor kidney dysfunction and interpret the extent of OCT2/MATEs‐mediated inhibition.

Recent clinical studies identified m¹A as a potential OCT2/MATEs biomarker. ¹⁴ , ¹⁷ , ⁸² Reports of clinical DDI studies for m¹A are still very limited compared to creatinine and NMN, and therefore this biomarker is currently not recommended for monitoring in clinical phase I studies. ³⁰ Plasma concentration of m¹A increased with the co‐administration of OCT2/MATEs inhibitors and maintained relatively constant throughout the day due to minimal diurnal changes. ¹⁴ , ¹⁷ , ⁸² In contrast, both diurnal change (potentially due to fluctuations in its biosynthesis rate) and a decrease in plasma concentration with OCT2/MATEs inhibitors were observed for NMN. ¹¹ , ¹⁴ , ¹⁷ , ⁸² These findings suggest that m¹A, unlike NMN, has the potential as a biomarker not only when using changes in CLr, but also based on changes in plasma concentrations of this biomarker. Additionally, it has been suggested that NMN is more effective for OCT2 inhibition detection, while m¹A is better suited for identifying MATEs inhibition. ¹⁷ Further clinical studies using a variety of OCT2/MATEs inhibitors are required to explore the potential of m¹A as an endogenous biomarker for OCT2/MATEs‐mediated DDIs and verify its utility.

Prediction of OCT2/MATEs‐mediated DDIs of endogenous biomarkers using PBPK model

Several PBPK model analyses for endogenous OCT2/MATEs biomarkers, creatinine, and NMN have been reported (Table S8 ). Nakada et al. developed a simple model based on the biosynthesis rate and renal clearance of creatinine to predict OCT2/MATEs‐mediated DDIs by trimethoprim and simulated the changes in serum creatinine and renal clearance during and after the administration of trimethoprim. ¹²⁸ Galetin’s group constructed a PBPK model for creatinine, considering biosynthesis of creatinine and representing the physiological structure of the kidney, including renal blood, epithelial cells, and lumen of the proximal tubule. This model predicted the DDIs with creatinine during the co‐administration of multiple OCT2/MATEs inhibitors in both healthy volunteers ⁴⁰ , ⁴¹ and patients with CKD. ⁴² The authors simulated the change in serum creatinine concentration in patients with CKD during the co‐administration of OCT2/MATEs inhibitors using a PBPK model in which the transport activity of OAT2 and OCT2/MATEs in CKD was contrary to INH. In that model, decrease in OAT2 exceeded changes in GFR in corresponding CKD stage, whereas changes in OCT2/MATEs were less pronounced than decrease in GFR. ⁴² Lehr’s group has also conducted a PBPK model analysis for creatinine and NMN, considering diurnal variations in transport activity by OCT2, GFR, renal blood flow rate, and the biosynthesis rate of NMN. ⁴³ Plasma concentrations of NMN decreased with the administration of trimethoprim or pyrimethamine. ¹⁴ , ¹⁵ They incorporated the inhibition of NMN biosynthesis into their PBPK model, resulting in a successful simulation of the change in plasma concentration of NMN during the co‐administration of trimethoprim and pyrimethamine. ⁴³ Consequently, they proposed that the decrease in AUC of NMN during the co‐administration of OCT2/MATEs inhibitors is due to the inhibition of NMN biosynthesis, though there is a lack of direct evidence regarding the inhibitory effects of these inhibitors on NMN biosynthesis. Therefore, further insights in this regard are eagerly anticipated.

CHANGES IN SECRETION CLEARANCE AND RENAL DDIs IN PATIENTS WITH CHRONIC KIDNEY DISEASE (CKD): INTACT NEPHRON HYPOTHESIS (INH) OR NON‐INH?

Change in secretion clearance for OAT1/3 and OCT2/MATEs substrates in patients with CKD

INH proposes that declining kidney function results from a reduction in nephron number, with remaining nephrons functioning normally (Figure S8 ). INH assumes that the decrease in secretory clearance is proportional to disease‐related changes in glomerular filtration clearance. ⁶⁰ Eqn (8) describes F _x , the ratio of CL_sec ratio (between healthy volunteers and patients with CKD) to the GFR ratio (between the same populations). ⁶²

$$F_x=\frac{R_{\text{CLsec}}}{R_{\mathrm{GFR}}}=\frac{{\mathrm{CL}}_{\sec ,\mathrm{CKD}}/{\mathrm{CL}}_{\sec ,\text{healthy}}}{{\mathrm{GFR}}_{\mathrm{CKD}}/{\mathrm{GFR}}_{\text{healthy}}}$$ (8)

F _x  = 1 indicates that the change in CL_sec matches the change in GFR (INH scenario). F _x  < 1 indicates that the decrease in CL_sec exceeds the change in GFR (non‐INH scenario). In mild to moderate CKD, OAT1/3 substrates exhibited F _x values close to 1 (16% difference in moderate). However, severe CKD showed a marked decrease and inter‐drug variability in F _x value ⁶² (Figure 7 a,c ). These results suggest that OAT1/3‐mediated CL_sec decreases in parallel with GFR in mild and moderate CKD, but exceed GFR decline in severe CKD. Notably, uremic toxins may inhibit OAT1/3. ⁶² , ¹²⁹ , ¹³⁰ Hsueh et al. reported that some uremic toxins such as hippuric acid and phenylacetic acid inhibit OAT1/3‐mediated transport in vitro, with their unbound plasma concentrations exceeding the K _i values in CKD, suggesting the inhibitory potential for OAT1/3 (Table S10 ). ¹²⁹ Additionally, the 5/6 nephrectomy rat model exhibited reduced OATs expression. ¹³¹ These findings all support a greater CL_sec decrease than a GFR decrease (non‐INH scenario) in CKD, particularly in severe CKD.

Figure 7.

F _x values and simulations of DDIs for OAT1/3 and OCT2/MATEs substrates in patients with CKD. (Panel a–c) Correlation between F _x values and GFR ratio (R _GFR) in patients with CKD for OAT1/3 and OCT2/MATEs substrates. (Panels a, b) The magnitude of deviation of secretion clearance from the decline of GFR (F _x ) of OAT1/3 (Panel a) and OCT2/MATEs (Panel b) substrates. Blue, orange, and red symbols represent the data in patients with mild, moderate, and severe CKD, respectively. Dashed lines (F _x  = 1) represent the INH scenario, where secretion clearance declines proportionally to GFR. F _x was calculated as Eq. (8) in the text. The data for OAT1/3 substrates was taken from the previous reports by Tan et al. ⁶² The following criteria were used to select OAT1/3 and OCT2/MATEs substrates for the calculation according to Tan et al. ⁶² (1) In vitro studies showed that compounds are substrates for OAT1/3 or OCT2/MATEs, and (2) CLr is at least 1.5 times greater than the filtration clearance calculated by multiplying GFR and f _p. Drugs included in panel a are excluded from panel b. F _x values of OAT1/3 substrates in patients with severe CKD are as follows: acamprosate (0.27), adefovir (1.08), amoxicillin (0.67), avibactam (0.10), baricitinib (0.37), bumetanide (0.75, 0.73), cefazolin (0.78), cefmetazole (0.64), cefotaxime (0.60, 0.48), cefuroxime (0.32), ciprofloxacin (0.33), entecavir (0.31), ertapenem (0.38), famotidine (0.52, 0.04), furosemide (0.70, 0.58), ganciclovir (0.33), hydrochlorothiazide (0.05), lamivudine (0.25), meropenem (0.07), olmesartan (0.94), oseltamivir carboxylate (0.23), pemetrexed (0.17), penciclovir (0.68, 0.30), piperacillin (0.13), rivaroxaban (1.10), rosuvastatin (0.62), sitagliptin (0.94), sulbactam (0.52), tazobactam (0.53), tenofovir (0.38), zalcitabine (0.29). ⁶² F _x values of OCT2/MATEs substrates in patients with severe CKD are as follows: amantadine (1.01), guanfacine (1.15, 0.70), imeglimin (1.11), lefamulin (2.09, 1.82), memantine (1.13), metformin (0.77), remoxipiride (3.38), terbutaline (0.95), tiotropium (0.90) (Table S11 ). (Panel c) Median and 90% prediction interval (PI) of F _x in patients with mild, moderate, and severe CKD for OAT1/3 and OCT2/MATEs substrates. The 90% PI was estimated assuming the log‐normal distribution. (Panel d–k) Changes in CLr and AUC by the co‐administration of OAT1/3 and OCT2/MATEs inhibitors in healthy volunteers and patients with CKD. (Panels d–g) CLr_ratio of OAT1/3 (d), OCT2 (e), MATEs (f) or OCT2/MATEs (g) substrates in the co‐administration of OAT1/3 or OCT2/MATEs inhibitors in healthy volunteers and patients with CKD. Blue and orange lines represent calculated CLr_ratio with and without inhibitors under I _max,u/K _i  = 3 and 10, respectively. Solid and dotted lines represent calculated CLr_ratio using median and 90% PI of F _x (shown in Figure 7 c ), respectively. CLr_ratio_CKD was calculated using Eq. S40 for OAT1/3 and OCT2/MATEs substrates. (Panels h–k) Impact of DDI with an inhibitor of OAT1/3 (h), OCT2 (i), MATEs (j) or OCT2/MATEs (k) on AUC in healthy volunteers and patients with CKD. Orange and gray boxes represent calculated AUC with and without inhibitors under I _max,u/K _i  = 10, respectively. AUC in healthy volunteers and patients with CKD for OAT1/3 and OCT2/MATEs substrates were calculated using Eq. S45. Dose, F, and CL_h were assumed to be 500 (mg), 0.5, and 9.53 (L/h), respectively, which are the same as metformin (Table S14 ). In the simulations of AUC, it was assumed that OAT1/3 and OCT2/MATEs inhibition affected only renal secretion, not absorption and hepatic elimination. In each box, the upper, middle, and lower lines represent the AUC using the F _x value of the upper 90% PI, the median F _x , and the F _x value of the lower 90% PI, respectively. In the simulations through panels d–k, CL_{int,sec,control} for the OAT1/3 and OCT2/MATEs substrates were assumed to be the same, with β _kidney of 0.9 (uptake‐limited process) for the OAT1/3 substrate and 0.1 (uptake/excretion‐limited process) for the OCT2/MATEs substrates. Other parameters used are shown in Table S14 . Inhibition for MATEs was not considered (I _max,u/K _i,MATEs = 0) in the simulation of OCT2 inhibition, and inhibition for OCT2 was not considered in the simulation of MATEs inhibition (I _max,u/K _i,OCT2 = 0).

In contrast to OAT1/3, evidence of the effect of CKD on OCT2/MATEs is so far limited. Takita et al. reported that the CL_cre/GFR ratio increased with deteriorating kidney function. ⁴² However, as noted above, creatinine is excreted via a combination of transporters (OAT2, OCT2 and MATEs) and this increase mainly reflects a decrease in OAT2 beyond INH (similar decrease as reported for OAT1/3) and a compensatory change (less pronounced decrease, comparable to GFR) in OCT2/MATEs. ⁴² Our analysis expanded on this work, focusing on multiple OCT2/MATEs substrates, all with CLr that is at least 1.5 times greater than the glomerular filtration clearance. F _x values were not significantly different from 1 regardless of kidney function (Figure 7 b,c ), which is consistent with the previous reports. ⁴² , ¹³² Our analysis indicates that, unlike OAT1/3 substrates, the OCT2/MATEs‐mediated CL_sec decrease is less than or comparable to the GFR decrease. No significant deviation in F _x from 1 for OCT2/MATEs substrates agrees with the in vitro study that reported unlikely inhibition of OCT2 by uremic toxins (Table S10 ). ¹³²

While the F _x value of OAT1/3 substrates was lower than 1 in severe CKD, wide F _x variability (0.04–1.10) was observed (Figure 7 a ). No clear correlation was found between F _x value and physicochemical parameters (molecular weight, LogP and LogD_7.4) (Figure S9 ). We further investigated the relationships between F _x and GFR, f _p, and R _{act,OATs,healthy} (the ratio of intrinsic uptake clearance by OATs (CL_int,OATs) to CL_pass in healthy volunteers). In CKD, post‐translational guanidinylation in albumin can reduce its binding capacity, ¹³³ and the accumulation of uremic toxins can displace binding between albumin and drugs. ¹³⁴ , ¹³⁵ The f_p values increase in CKD, particularly for OATs substrates with f _p ≦ 0.2 ⁶² (Figure S10 ). F _x expressed by Eq. (8) can be transformed to intrinsic clearance basis (Eq. 9), assuming that secretion clearance is low and passive diffusion clearance decreases proportionally to GFR according to INH. ⁶²

$$F_x=\frac{f_{\mathrm{p},\mathrm{CKD}}}{f_{\mathrm{p},\text{healthy}}}\times \frac{\frac{1}{R_{\mathrm{act},\text{OATs},\text{healthy}}}\times \frac{{\mathrm{CL}}_{\text{pass},\mathrm{CKD}}}{{\mathrm{CL}}_{\text{pass},\text{healthy}}}+\frac{{\mathrm{CL}}_{\mathrm{int},\text{OATs},\mathrm{CKD}}}{{\mathrm{CL}}_{\mathrm{int},\text{OATs},\text{healthy}}}}{\left(\frac{1}{R_{\mathrm{act},\text{OATs},\text{healthy}}}+1\right)\times \frac{{\mathrm{CL}}_{\text{pass},\mathrm{CKD}}}{{\mathrm{CL}}_{\text{pass},\text{healthy}}}}$$ (9)

The blood‐plasma partition coefficient is assumed to be unchanged in both populations. The theoretical lines in Figure S11B,C suggest that OAT1/3 substrates with a high contribution of active secretion (large R_{act,OATs,healthy}) exhibit a small F _x value when the decrease in CL_int,OATs relative to healthy volunteers (CL_int,OATs,CKD/CL_{int,OATs,healthy}) exceeds the change in GFR (non‐INH scenario). Additionally, F _x values theoretically increase for OAT1/3 substrates with high protein binding (Figure S11C ). We investigated the correlation between the calculated R_{act,OATs,healthy} and observed F _x in severe CKD (Figure S11A ). F _x values tended to be lower for drugs with a higher contribution of active secretion, comparable with the theoretical lines in Figure S11B,C . This analysis suggests a more pronounced decrease in CL_sec in CKD for OAT1/3 substrates with a higher contribution of active transport, explaining the reported inter‐drug F _x variability in severe CKD. The analysis above highlights another important consideration in CKD, i.e., the change in the relative contribution of active and passive (GFR) processes in the disease due to differential effects on GFR and OAT1/3.

OAT1/3‐ and OCT2/MATEs‐mediated DDI prediction in patients with CKD

Prediction of OAT1/3‐ and OCT2/MATEs‐mediated DDIs with co‐administration of OAT1/3 or OCT2/MATEs inhibitors in patients with CKD (CLr_ratio_CKD and AUC_ratio_CKD) is described by Eqs. S40 and S45 for CLr_ratio and AUC_ratio, respectively, assuming constant bioavailability and that the hepatic clearance is unaffected by OAT1/3 and OCT2/MATEs inhibition. For OAT1/3 substrates, CLr_ratio is higher in patients with CKD than in healthy volunteers due to a lower f _sec, reflecting a more pronounced disease‐related decrease in active secretion relative to GFR decrease (Figure 7 d ). Conversely, for OCT2/MATEs substrates, CLr_ratio is similar regardless of kidney function (Figure 7 e–g ). The predicted AUC_ratio with co‐administration of inhibitors decreases as kidney function deteriorates for both substrates (Figure S12 ) because it is assumed that CLr is solely affected by kidney failure and its contribution to total clearance is reduced in CKD. However, despite a smaller fold increase in AUC in patients with CKD, absolute AUC (resulting from drug‐disease interaction) increases with deteriorating kidney function for both OAT1/3 (Figure 7 h ) and OCT2/MATEs substrates (Figure 7 i–k ). This combined effect of disease and drug interaction highlights DDI risk in patients with CKD, as illustrated recently in PBPK modeling of probenecid‐adefovir interaction in different stages of CKD. ⁶⁶ Currently, drug label information on such DDIs in patients with CKD is limited. Further clinical DDI studies in the patients, combined with PBPK modeling, are expected to improve our confidence in prospective DDI prediction for these transporters.

CONCLUSIONS

This review has highlighted the latest findings regarding OCT2/MATEs inhibitors and DDIs in both healthy volunteers and patients with CKD, offering the following key insights and recommendations. Advancing research in OCT2/MATEs‐mediated DDIs will enhance predictive accuracy and improve clinical decision making for these populations.

1. Improved prediction accuracy for highly lipophilic and protein‐bound inhibitors: To minimize false‐negative predictions, K _i values for OCT2 should be assessed after 60 minutes or longer of pre‐incubation. This may be critical for accurately evaluating the I _max,u/K _i value for highly lipophilic and protein‐bound inhibitors.

2. Early biomarker evaluation for clinical DDI studies: Changes in CLr of NMN should be investigated during early clinical studies to evaluate OCT2/MATEs‐mediated DDI risks. Further studies on m¹A are essential to establish its utility as a reliable OCT2/MATEs biomarker by assessing changes in its plasma concentration and CLr.

3. Prediction of CLr changes in patients with CKD: OCT2/MATEs substrates demonstrate a decrease in CL_sec comparable to changes in GFR, in line with the INH scenario. This trend is opposite to OAT1/3 transporters, which decrease to a greater extent than GFR in severe CKD, highlighting a transporter‐dependent disease effect. This difference is substantiated by the inverse relationship between F _x values and the ratio of active transport to passive diffusion clearance. Estimating F _x values from in vitro transport studies would facilitate the prediction of changes in CLr in patients with CKD.

4. Simulation of DDIs in patients with CKD: OAT1/3 substrates experience a larger CLr ratio (+ inhibitor/control) than OCT2/MATEs substrates reflecting more pronounced disease‐drug–drug interactions. Incorporating F _x values into predictions and adjusting dosages accordingly is recommended to optimize treatment strategies for these patients.

FUNDING

No funding was received for this work.

CONFLICT OF INTEREST

The authors declared no competing interests for this work. As Deputy Editor‐in‐Chief of Clinical Pharmacology & Therapeutics, Kathleen Giacomini was not involved in the review or decision process for this paper.

DISCLAIMER

The opinions expressed in this manuscript are those of the authors and should not be interpreted as the position of the U.S. Food and Drug Administration.

Supporting information

Appendix S1

Appendix S2

Appendix S3

Appendix S4