Interactions of Tyrosine Kinase Inhibitors with Organic Cation Transporters and Multidrug and Toxic Compound Extrusion Proteins

Accumulating evidence highlights the importance of drug transporters in drug disposition, efficacy, and safety, particularly with respect to drug–drug interactions (DDI; ref. 1). A recent review has underscored the need for drug development programs to perform relevant clinical studies of transporter-based DDIs as suggested from in vitro studies of drug-transporter interactions (1). In terms of transporter-based DDIs, an interacting drug may be an inhibitor that causes a DDI by inhibition of a transporter's function, or may be a substrate of the transporter. Methodologies and algorithms for predicting transporter-mediated DDIs based on in vitro experiments are under discussion (1). A comparison of the concentration of an inhibitor (I; generally, the maximum unbound plasma concentration) and its half maximal inhibitory concentration (IC₅₀) for a transporter as determined through in vitro studies is performed to determine whether a potential clinical DDI may occur (1). Obviously, high unbound plasma concentrations relative to the IC₅₀ of the drug is an indicator of a potential clinical DDI. To err on the side of caution, an I/IC₅₀ value ≥0.1 has been suggested as a criterion to use for consideration of conducting a clinical transporter-based DDI (1).

Tyrosine kinase inhibitors (TKI; Supplementary Fig. S1), a new class of anticancer drugs, are rationally designed to target specific tyrosine kinases that are fused, mutated, or overexpressed in cancer (2). In spite of their increasing use, only a few studies have examined the interactions of TKIs with influx transporters and to our knowledge there are no known influx transporter-based clinical DDI between TKIs and other drugs on the market. Most studies examining the interaction of TKIs with influx transporters have focused on substrate rather than inhibition interactions (3–10). Moreover, several TKIs have cationic charge and high lipophilicity (Supplementary Table S1), which are requirements of organic cation transporter, OCT inhibition (11), and therefore have potential to inhibit organic cation transporters, including OCT1 (SLC22A1), OCT2 (SLC22A2), OCT3 (SLC22A3), and MATE1 (SLC47A1) and MATE2-K (SLC47A2). However, to our knowledge, there have been no studies assessing the clinical DDI potential of TKIs.

Drugs from multiple therapeutic classes that are transported by organic cation transporters may be concomitantly used with TKIs. Metformin is recommended as the initial pharmacologic therapy for type 2 diabetes, which is one of the most common diseases in the world. Diabetes and cancer are common conditions, and their codiagnosis in the same individual is not infrequent (12). In humans, OCT1 on the basolateral membrane of hepatocytes is involved in the uptake of metformin from the blood into the liver, which is its major site of pharmacologic action (13, 14). Similarly, OCT2 on the basolateral membrane of renal tubular cells mediates the uptake of metformin from blood into the kidney, which is the major eliminating organ for the drug (15). Recently, MATE1 and MATE2-K on the apical membrane of renal tubular cells have been proposed to act as the final efflux transporters of metformin from the body, namely from renal cells to the urine (16). Several clinical DDIs between metformin and inhibitors of organic cation transporters, such as cimetidine, have been reported (1). Metformin, a typical substrate of OCTs and MATEs, has been extensively studied. Data obtained for metformin may be applicable to other substrates of OCTs and MATEs. Interestingly, metformin has recently been considered a potential candidate in novel treatment strategies for human pancreatic cancer (17) and in combination with lapatinib-based protocols in the treatment of breast cancer (18). Clinical studies of combination therapies of gemcitabine, erlotinib, and metformin in pancreatic cancer and of lapatinib in combination with metformin for various cancers are currently underway (Study ID NCT01210911 and NCT01087983 in http://clinicaltrials.gov/). Recently, a clinical case report regarding lactic acidosis induced by a DDI of erlotinib, a TKI, and metformin has been published, but lacks any mention of transporters (19). Because lactic acidosis associated with metformin treatment is a rare but potentially fatal adverse event, this clinical case report underscores the importance of the safe use of metformin in cancer patients. Given that drug transporters are a determinant of metformin pharmacokinetics and pharmacodynamics, possible risks of concomitant use of metformin and TKIs in terms of drug transporters need to be examined. Oxaliplatin, an anticancer platinum analog, is a substrate of organic cation transporters including OCT1, OCT2, and OCT3, which are thought to play a role in its anticancer effects by facilitating its entry into tumor cells (20, 21). Currently, TKIs are being tested in clinical trials in combination with oxaliplatin for the treatment of various tumors (22–25). Moreover, many trials involving oxaliplatin with TKIs are ongoing or completed (http://clinicaltrial.gov).

Here, we conducted in vitro studies comparing the DDI potential of TKIs on the transport of metformin by OCT1, OCT2, OCT3, MATE1, and MATE2-K to obtain a rationale for clinical DDI studies. We attempted to thoroughly assess the in vitro inhibitory effects of 8 small-molecule TKIs (imatinib, dasatinib, nilotinib, gefitinib, erlotinib, sunitinib, lapatinib, and sorafenib; Supplementary Fig. S1) on transport of metformin by the transporters. For the TKIs that potentially cause clinical DDIs, time-dependent inhibition, genetic variant-mediated alteration of inhibition potency, and inhibition by major metabolites of the drugs were also investigated. Inter-species differences in the transporter inhibition by TKIs were also examined in vitro to evaluate the utility of mice as possible surrogate in vivo models for assessing clinically relevant transporter-mediated DDI. Relationships between transporter inhibition and physicochemical properties of the TKIs were also determined. In addition, we examined the DDI potential of TKIs with oxaliplatin that may be used in combination with TKIs, and also an important substrate of OCTs and MATEs.

Imatinib, dasatinib, nilotinib, gefitinib, erlotinib, sunitinib, lapatinib, sorafenib, N-desmethyl imatinib, O-desmethyl gefitinib, and O-desmethyl erlotinib were purchased from Toronto Research Chemical. [¹⁴C]Metformin was purchased from Moravek Biochemicals and Radiochemicals. Metformin and oxaliplatin were purchased from Sigma-Aldrich. Other reagents and chemicals were purchased from commercial sources.

Human embryonic kidney (HEK293) Flp-In-293 cells stably transfected with the full-length reference human OCT1 cDNA (reference sequence, NM_003057; HEK-hOCT1), OCT2 cDNA (HEK-hOCT2), OCT3 cDNA (HEK-hOCT3), MATE1 cDNA (HEK-hMATE1), mutant OCT1 M420del cDNA inserts, and the empty vector (HEK-MOCK) were established previously in our laboratory (13, 21, 26, 27). Stable cell lines expressing recombinant MATE2-K (HEK-hMATE2-K), and mouse Oct1 (HEK-mOct1) were also established previously in our laboratory (unpublished reports). HEK293 cells stably transfected with the full-length reference mouse Oct2 cDNA (HEK-mOct2), Oct3 cDNA (HEK-mOct3), and Mate1 cDNA (HEK-mMate1) were established by transfection of pcDNA5/FRT vector (Invitrogen) containing the full-length transporter cDNA into HEK293 Flp-In-293 cells using Lipofectamine 2000 (Invitrogen) as per the manufacturer's instructions. Stable clones were selected using 75 μg/mL hygromycin B. The general characteristics of all transfected HEK293 cell lines were monitored by checking cell morphology microscopically, and measuring the doubling time of each cell line. The expression level of each transporter mRNA in each transfected HEK293 cell line was verified by real-time RT-PCR method. The transport of metformin and oxaliplatin in transporter-transfected HEK293 cells was compared to that in HEK-MOCK in every transport assay. Authentication of HEK293 Flp-In-293 cells was not performed by the authors.

Stably transfected HEK293 cells were maintained in Dulbecco's modified eagle medium (DMEM) H-21 medium supplemented with 10% fetal bovine serum, 100 units/mL penicillin, 100 units/mL streptomycin, and 75 μg/mL hygromycin B. All cell lines were grown at 37°C in a humidified atmosphere with 5% CO₂/95% air.

The uptake study was performed as described in our previous report with slight modifications (28). Briefly, HEK293 cells were grown on poly-d-lysine-coated 24-well plates in DMEM H-21 medium supplemented with 10% fetal bovine serum to at least 90% confluence (24 hours after seeding). For uptake studies, HEK-MOCK, HEK-hOCT1, HEK-hOCT2, HEK-hOCT3, HEK-mOct1, HEK-mOct2, and HEK-mOct3 were preincubated before the addition of the uptake buffer for 20 minutes with Hank's balanced salt solution (HBSS; 5.4 mmol/L potassium chloride, 0.44 mmol/L monobasic potassium phosphate, 4.2 mmol/L sodium bicarbonate, 137 mmol/L sodium chloride, 0.34 mmol/L dibasic sodium phosphate, 5.6 mmol/L d-glucose, 1.3 mmol/L calcium chloride, 0.49 mmol/L magnesium chloride, 0.41 mmol/L magnesium sulfate, pH 7.4), and then incubated in the uptake buffer (HBSS containing 10 μmol/L unlabeled and [¹⁴C]-labeled metformin). Uptake was performed at 37°C for the designated period of time for which linear uptake was observed, and then the cells were washed twice with ice-cold HBSS. The cells were lysed with 0.1 N NaOH and 0.1% SDS, and the lysate was used for scintillation counting (Beckman Coulter) and for the bicinchoninic acid protein assay (Pierce). For uptake by MATE, HEK-MOCK, HEK-hMATE1, HEK-hMATE2-K, and HEK-mMate1, cells were preincubated before the uptake study with HBSS (pH 7.4) in the presence of 30 mmol/L ammonium chloride for 20 minutes (16). The preincubation medium was then removed, and the cells were incubated with uptake medium containing [¹⁴C]metformin in the same manner as for OCTs.

The cellular accumulation of platinum was determined after 2-hour exposure to oxaliplatin as previously described (21), with slight modification. Briefly, HBSS (pH 7.4) was used as the uptake buffer, and the platinum content in the cells was determined by inductively coupled plasma mass spectrometry in the Analytical Facility at University of California, Santa Cruz. Cellular platinum accumulation was normalized to the protein content determined via bicinchoninic acid protein assay.

Half-maximal IC₅₀ values were calculated by simultaneously fitting the data in transporter-expressing cells and mock cells, without a weighting factor, to the following equations (29) using WinNonlin:

where CL_uptake is the uptake clearance (μL/min/mg protein) in the absence of inhibitor, CL_uptake,+inh is that in the presence of inhibitor, [I] is the inhibitor concentration, and P_dif is the component of passive diffusion (μL/min/mg protien). Given that the substrate concentration in the inhibition study was low compared with the Michaelis-Menten constant K_m, the IC₅₀ values are virtually equal to the inhibition constant K_i values, assuming competitive or noncompetitive inhibition (1). Each set of [I] consists of 0 and 6 concentrations, so that each IC₅₀ value was within the concentration range used in the studies.

Maximum clinical concentrations of unbound TKIs in systemic plasma (C_max,sys,p) and unbound fraction in plasma were obtained from the literature (Supplementary Table S1; refs. 30-38). Absorption rate constants (k_a) were calculated using the reported time to reach C_max,sys,p (T_max) and half life (t_1/2; ref. 30), and the following equation (39):

where k_el is the elimination rate constant, assuming a monoexponential decline of plasma levels of drug with first order absorption and k_a > k_el. If k_a was calculated to be greater than 0.1/min, then it was fixed at 0.1/min (39). Clinical maximum concentration of unbound TKI in portal plasma (unbound C_max,portal,p) was calculated using the following equation (39):

where f_u,p is the unbound fraction in plasma, R_b is the blood-to-plasma concentration ratio, F_a and F_g are bioavailability via the first-pass effect in absorption and intestinal metabolism, respectively, and Q_h is hepatic blood flow rate (1,500 mL/min; ref. 1). F_a, F_g, and R_b were assumed to be 1 in the present study.

Charge at pH 7.4 and lipophilicity parameters (calculated octanol-water partition coefficient, C log P, and calculated octanol-water distribution coefficient at pH 7.4, C log D_pH7.4) of TKIs were calculated using the software Marvin View (ChemAxon Kft.; Supplementary Table S1). Other physicochemical parameters, H-bond donor count, H-bond acceptor count, rotatable bond count, tautomer count, molecular weight, topological polar surface area, heavy atom count, and complexity, were obtained from PubChem (http://pubchem.ncbi.nlm.nih.gov). PubChem is an open repository for experimental data identifying the biological activities of small molecules, which includes computed properties of the compound record (40). The IC₅₀ values of TKIs in inhibiting metformin uptake via human OCTs and MATEs (Supplementary Table S2) were expressed in mol/L, transformed into logarithm, and then multiplied by −1, i.e., negative log₁₀(IC₅₀) values, so that higher values equate to higher inhibitory effects. In cases in which the IC₅₀ value was >30 μmol/L, negative log₁₀(IC₅₀) was assumed to be –log₁₀(3×10⁻⁶). Pearson's correlation coefficient (R) and P value (two-tailed) between physicochemical parameters and negative log₁₀(IC₅₀) values were calculated using GraphPad Prism (GraphPad Software, Inc.), after confirmation of normality of data distribution by D'Agostino and Pearson omnibus normality test.

The IC₅₀ values of the 7 TKIs against, OCT1, OCT2, OCT3, MATE1, and MATE2-K are shown in Supplementary Table S2. Of note, most of the TKIs including imatinib, dasatinib, nilotinib, gefitinib, and sunitinib were potent inhibitors of the transporters, with IC₅₀ values generally in the low micromolar range. Lapatinib and sorafenib, in contrast, were poor inhibitors of [¹⁴C]metformin uptake by the transporters. Comparison of the unbound C_max,sys,p (Supplementary Table S1) and IC₅₀ values of TKIs against human OCT1, OCT2, OCT3, MATE1, and MATE2-K (Supplementary Table S2) showed that imatinib (for OCT1, hMATE1, and MATE2K), nilotinib (for OCT3), gefitinib (for MATE2K), and erlotinib (for OCT1 and MATE2K) exerted potent inhibitory effects at clinically relevant concentrations, with [I]/IC₅₀ ≥ 0.1 using unbound C_max,sys,p as [I] (Fig. 1). If unbound maximum concentrations in the portal vein, C_max,portal,p (Supplementary Table S1) were used as [I] for the liver transporter (hOCT1), several of the TKIs (imatinib, dasatinib, gefitinib, and sunitinib) were predicted to inhibit metformin uptake into the liver at concentrations in the hepatic sinuosoids that may be observed after oral doses of the drugs.

Because erlotinib was a potent inhibitor of OCT1 at clinically relevant concentrations, we studied its potency as an inhibitor of a common polymorphism of OCT1, M420del, which has reduced function with respect to metformin uptake. Our findings showed that erlotinib exerted a more potent inhibitory effect on the polymorphism than it did on the reference OCT1 (Fig. 2). Further, erlotinib showed an even more potent inhibitory effect on metformin transport by the reference OCT1 after preincubation with cells before initiating the uptake reaction (Fig. 3).

N-desmethyl imatinib and O-desmethyl gefitinib, which are major metabolites of imatinib and gefitinib, respectively, showed inhibition properties similar to those of imatinib (potent and selective inhibition of MATE1) and gefitinib (potent and selective inhibition of MATE2-K; Supplementary Table S3). In contrast, O-desmethyl erlotinib, a major metabolite of erlotinib, was not a potent inhibitor of OCT1 (Supplementary Table S3) as was the parent compound (Supplementary Table S2).

We noted large differences in IC₅₀ values and selectivity of inhibition of TKIs between human and mouse transporter orthologs (Fig. 4). In particular, imatinib, nilotinib, and erlotinib exerted less potent inhibition of mouse transporters mMate1, mOct3 and mOct1, respectively, in comparison to the human orthologs.

To determine whether erlotinib and nilotinib would inhibit oxaliplatin uptake at concentrations in the range of clinically achieved unbound concentrations in plasma, we tested 2 μmol/L as the highest concentration. Cellular platinum uptake after incubation of HEK-hOCT1, -hOCT2, and -hOCT3 with oxaliplatin (2 μmol/L for 2 hours) was decreased in the presence of erlotinib and nilotinib in a concentration-dependent manner (Fig. 5). Erlotinib and nilotinib showed potent uptake inhibition of HEK-hOCT1 and HEK-hOCT3, respectively, between transporter-expressing HEK cells.

Lipophilicity parameters, C log P and C log D_pH7.4, correlated negatively with negative log₁₀(IC₅₀) of TKIs for human OCT1 and OCT2 (Fig. 6). With regard to human MATE1, the charge at pH7.4 correlated positively with negative log₁₀(IC₅₀) values (Fig. 6). For OCT3 and MATE2-K, lipophilicity or charge did not significantly correlate with negative log₁₀(IC₅₀) [data not shown]. In addition, for all tested transporters, there was no significant correlation between negative log₁₀(IC₅₀) values and the parameters of H-bond donor count, H-bond acceptor count, rotatable bond count, tautomer count, molecular weight, topological polar surface area, heavy atom count, and complexity (data not shown).

Here, we investigated the possibility of clinical DDIs via inhibitory effects of 8 small-molecule TKIs on the transport activity of human OCT1, OCT2, OCT3, MATE1, and MATE2-K. Our results showed that imatinib, nilotinib, gefitinib, and erlotinib exerted selectively potent inhibitory effects with unbound C_max,sys,p/IC₅₀ ≥ 0.1 on MATE1, OCT3, MATE2-K, and OCT1, respectively. In particular, unbound C_max,sys,p/IC₅₀ was greater than 1 for the inhibition of MATE1 by imatinib and of OCT1 by erlotinib, suggesting a high probability of a DDI if the compounds are administered with metformin.

Because drug concentrations at the inlet of the liver after oral administration are generally expected to be higher than in the systemic circulation (39), actual risks of DDI may be underestimated by unbound C_max,sys,p/IC₅₀. Therefore, unbound C_max,portal,p values (Supplementary Table S1; ref. 39) may be used as [I] to predict potential DDIs with transporters in the liver (hOCT1) to avoid false-negative predictions. Prevalence of type 2 diabetes ranges from 2% to 10% in industrialized countries, with rates on the rise (41). Because cancer and type 2 diabetes are 2 common diseases that may share risk factors (41), many present and future patients will likely undergo concomitant treatment with TKIs and metformin. Importantly, because metformin appears to have some efficacy in cancer management, clinical studies with TKIs and metformin are ongoing. Though relatively safe, at high concentrations, metformin can cause lactic acidosis, a life-threatening adverse reaction. DDIs with TKIs may result in increased levels of metformin and therefore, increase patient susceptibility to lactic acidosis. Such DDIs may be particularly problematic in patients with widespread metastases who are at increased risk for lactic acidosis. Our study strongly supports the need for clinical DDI studies between metformin and TKIs.

Methods used in the in vitro assessment of drug transporter interactions are important to obtain accurate predictions of potential clinical DDIs. The current standard in vitro method of evaluating a drug as an inhibitor of an uptake transporter involves simultaneous addition of the test drug inhibitor and model substrate and determination of uptake of the substrate. In real clinical situations, however, a drug may be metabolized into several distinct chemical species, and the transporter may be exposed to the inhibitor drug for a long period of time before a substrate is administered (28). We found that preincubation of erlotinib with the cells before initiation of metformin uptake enhanced the inhibitory effects of erlotinib on OCT1 (Fig. 3). This suggest that the interaction of erlotinib with the lipophilic and negatively charged cell membrane, which is supposed to be crucial step in achieving OCT1 inhibition (11), or other steps required for erlotinib to reach its binding domain on OCT1 may be slow. We also observed that the inhibition potency of erlotonib was increased for the M420del (Fig. 2), a common polymorphism of OCT1 found at allele frequencies of about 20% in white Americans (13). Our data are consistent with a previous study showing that M420del is more sensitive to drug inhibition by glibenclamine, simvastatin, and verapamil, with IC₅₀ values up to 8 times lower for the variant than those observed for the reference OCT1 (28).

Major metabolites of TKIs showed IC₅₀ values similar to those for unchanged TKIs. Plasma concentrations of O-desmethyl gefitinib in CYP2D6 extensive metabolizers are higher than for unchanged gefitinib (42). Therefore, O-desmethyl gefitinib may contribute to potential clinical DDIs that may occur following concomitant metformin and gefitinib administration. Given that plasma concentrations of N-desmethyl imatinib are approximately 10-fold lower than those of the unchanged imatinib (43), and the unbound fractions of N-desmethyl imatinib and imatinib are nearly equal (44), the inhibitory effects of N-desmethyl imatinib on the tested transporters may be negligible at therapeutic doses of imatinib. O-desmethyl erlotinib is a 10-fold weaker inhibitor of OCT1 than erlotinib (Supplementary Table S3), and the metabolite concentration in plasma is more than 10-fold lower than that of the unchanged erlotinib (45), suggesting little inhibitory effects at normal doses.

To determine whether mice could be used as animal models to study transporter-based human DDIs between TKIs and OCTs and MATEs, we investigated the inhibition potency of TKIs for mouse orthologs of OCTs and MATEs. Our data showed striking species differences in the interaction of erlotinib, nilotinib, and imatinib, with mouse orthologs of OCT1, OCT3, and MATE1, respectively, in comparison to the human orthologs (Fig. 4). All 3 compounds had weaker interactions with the mouse transporters in comparison to the human orthologs. This finding suggests that mice are unlikely to be suitable animal models for in vivo DDI studies of TKIs and OCTs/MATEs, The structural basis for the reduced potency of interaction of the TKIs with the mouse orthologs of the transporters is not known and needs investigation. Because high resolution crystal structures of OCTs and MATEs are not available for construction of molecular models, it is difficult to explain the structural basis for the striking interspecies differences that were observed in the interaction of TKIs with OCTs and MATEs.

In this study, we observed that the cellular uptake of platinum was reduced after oxaliplatin treatment in the presence of clinically relevant concentrations of erlotinib and nilotinib (Fig. 5). These experiments suggest that combination therapy involving oxaliplatin and these TKIs may not be optimal, particularly for tumors in which OCTs are involved in the uptake of oxaliplatin. Coadministration of erlotonib or nilotinib may reduce the anticancer effects of oxaliplatin by inhibiting the drug's uptake into such tumors. Nilotinib appears to be a more potent inhibitor of oxalplatin transport by hOCT1, hOCT2, and hOCT3 than of metformin transport. This may suggest that the inhibition potency of TKIs on OCTs may be different depending on the substrate and/or that experimental condition such as incubation time may affect the results of in vitro assessment.

Negative correlations between lipophilicity (C log P and C log D_pH7.4) and negative log₁₀(IC₅₀) of the TKIs for human OCT1 and OCT2 (Fig. 6) suggest that excessive lipophilicity weakens the inhibitory effects of TKIs on metformin uptake by OCT1 and OCT2. These data are not in agreement with studies demonstrating that high lipophilicity is one of the key physicochemical properties for OCT1 inhibition (11). Excessive lipophilicity may make it difficult for a drug to access the substrate binding region of OCT1 that is in contact with the aqueous phase (46). C log P values of the tested TKIs (Supplementary Table S1) met the minimum requirement of lipophilicity needed for OCT1 inhibition (11). Our study also suggests that positive charge may be one of key factors for MATE1 inhibition (Fig. 6). Because the number of tested compounds is small, further studies are needed for accurate drug-structure-based prediction of transporter inhibitory effects.

In the present study, we investigated the possibility of clinical DDIs in response to the inhibitory effects of 8 small-molecule TKIs on the transport activity of human OCT1, OCT2, OCT3, MATE1, and MATE2-K. Results showed the potential for clinical DDIs when certain TKIs were used concomitantly with drugs that are transported by organic cation transporters such as metformin or oxaliplatin. The results of this study provide the basis for further clinical studies investigating the transporter-based DDI potential of TKIs.

No potential conflicts of interest were disclosed.

This work was supported by National Institutes of Health (NIH) Grants GM36780 and GM074929.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.