Drug-Drug Interaction Potentials of Tyrosine Kinase Inhibitors via Inhibition of UDP-Glucuronosyltransferases

Tyrosine kinase inhibitors (TKIs) are anticancer drugs that may be co-administered with other drugs. The aims of this study are to investigate the inhibitory effects of TKIs on UDP-glucuronosyltransferase (UGT) activities, and to quantitatively evaluate their potential to cause drug-drug interactions (DDIs). Inhibition kinetic profiles of a panel of UGT enzymes (UGT1A1, 1A3, 1A4, 1A6, 1A7, 1A8, 1A9, 1A10, 2B4, 2B7, 2B15, and 2B17) by four TKIs (axitinib, imatinib, lapatinib and vandetanib) were characterized by using hepatic microsomes and recombinant proteins. Lapatinib exhibited potent competitive inhibition against UGT1A1 activity with a Ki of 0.5 μM. Imatinib was found to exhibit broad inhibition on several UGTs, particularly potent competitive inhibition against UGT2B17 with a Ki of 0.4 μM. The TKIs also exerted intermediate inhibition against several UGTs (i.e., UGT1A7 by lapatinib; UGT1A1 by imatinib; UGT1A4, 1A7 and 1A9 by axitinib; and UGT1A9 by vandetanib). Results from modeling for the quantitative prediction of DDI risk indicated that the coadministration of lapatinib or imatinib at clinical doses could result in a significant increase in AUC of drugs primarily cleared by UGT1A1 or 2B17. Lapatinib and imatinib may cause clinically significant DDIs when co-administered UGT1A1 or 2B17 substrates.

Previous in vitro and in vivo studies indicate that TKIs may alter the hepatic elimination of co-administered drugs by inhibiting their metabolism. For example, nilotinib and erlotinib inhibit UGT1A1 activity, and gefitinib inhibits UGT1A1, 1A7, 1A9 and 2B7 activities [12][13][14][15] . A clinical study also showed that co-administration of lapatinib with irinotecan led to a ~40% increase in the AUC of SN-38 (an active metabolite of irinotecan and a UGT1A1 substrate) 16 , suggesting the possible inhibition of UGT1A1 activity by lapatinib. However, whether these TKIs affect activities of others UGT isoforms and whether other TKIs affect UGTs remain unknown.
In this study, four commonly used TKIs− axitinib, imatinib, lapatinib and vandetanib ( Fig. 1)− were evaluated for their capabilities to inhibit UGT activities. The inhibition kinetics of each compound was further characterized, and the risks for clinically significant drug-drug interactions were estimated.

Results
Inhibition of UGT Activity by TKIs. As a preliminary study, we first examined whether TKIs inhibit different UGTs. To this end, axitinib, imatinib, lapatinib, or vandetanib (or vehicle control) was incubated with a UGT substrate (4-methylumbelliferone (4-MU) for all UGTs except for UGT1A4; trifluoperazine (TFP) was used for UGT1A4) and one of recombinant UGT enzymes (UGT1A1, 1A3, 1A6, 1A7, 1A8, 1A9, 1A10, 2B4, 2B7, 2B15, and 2B17). Then, the extent of glucuronide metabolite production was examined. The results showed that at 100 μ M concentration, TKIs inhibited the activity of UGT isoforms to varying extent (Table 1). For UGT isoforms whose activity is inhibited by >50% by individual TKIs, IC 50 values of TKIs were further estimated. The summary of IC 50 values is shown in Table 2.
UGT Inhibition Kinetics of TKIs. Except axitinib whose maximum plasma concentration (C max ) is in the sub-micromolar range 17 , the C max values of the remaining TKIs tested in this study are at the micromolar level at typical oral doses. For TKIs that exhibit IC 50 values lower or similar to the reported plasma  C max (i.e., imatinib against UGT1A1 and 2B17; lapatinib against UGT1A1, 1A4, and 1A7; and vandetinib against UGT1A9), their UGT inhibitory kinetics profiles were further characterized. The kinetic results were shown in Table 3. Imatinib inhibited UGT1A1 and 2B17 activities (Fig. 2) by a competitive inhibitory mechanism. K i values of imatinib were estimated to be 19 μ M and 0.4 μ M against the UGT1A1 and UGT2B17, respectively. Lapatinib (Fig. 3) exhibited a potent competitive inhibition against UGT1A1 with a K i of 0.5 μ M, a mixed inhibition against UGT1A4 with a K i of 36.4 μ M, a non-competitive inhibition with a K i of 4.5 μ M against UGT1A7. Vandetanib showed a mixed inhibition against UGT1A9 with a K i of 9.0 μ M (Fig. 4).

Inhibition SN-38 Glucuronidation by Lapatinib. Because UGT1A1 inhibition can exhibit
substrate-dependency 18 , the inhibition of UGT1A1 activity by lapatinib was further examined using SN-38, a drug mainly metabolized by UGT1A1 19 . IC 50 values were estimated for the inhibition of SN-38 glucuronidation by lapatinib in human microsomes and recombinant UGT1A1 proteins, and their kinetic parameters were estimated. Lapatinib showed strong inhibition against SN-38 glucuronidation, the IC 50 of which was 0.33 μ M in UGT1A1 and 0.42 μ M in human liver microsomes (HLMs) (Fig. 5a). The estimated K i values were 0.6 μ M and 1.6 μ M for UGT1A1 and HLMs, respectively, which were similar to the K i values for the inhibition of 4-MU glucuronidation by lapatinib. Interestingly, however, the kinetic profile of inhibition of SN-38 glucuronidation was different from that of 4-MU glucuronidation; while inhibition of 4-MU glucuronidation by lapatinib was competitive, the inhibition of SN-38 glucuronidation by lapatinib (Fig. 5b-e) was non-competitive in both UGT1A1 and HLMs (Table 3).

Quantitative Prediction of DDI Potential. The risk of DDIs due to the inhibition of UGTs by
TKIs was evaluated by predicting the ratio of area under the plasma drug concentration versus time curve (AUC) of a victim drug in the presence and absence of the TKI (i.e., AUC i /AUC). To this end, the hepatic or systemic concentrations of each TKI were first estimated using previously reported clinical pharmacokinetic data (Table 4). For the inhibition of UGT1A1, 1A9, and 2B17, the maximum unbound hepatic input concentrations of TKIs were estimated. For the inhibition of UGT1A7, which is mainly expressed in intestines, systemic drug concentration was estimated. The AUC ratios of each inhibitor   Table 3. The inhibition kinetic parameters and types of TKIs on UGTs.
were estimated based on K i values obtained in this study and the calculated hepatic or plasma concentrations of TKIs. The AUC changes of the UGT2B17 substrate by imatinib (at the time for C max ) were predicted to be significant. For example, UGT2B17 inhibition by imatinib was predicted to increase the AUC of 4-MU by 140% when the fraction of the UGT substrate metabolized by the inhibited enzyme (f m ) of a victim drug was 90%, which was much greater than the cut-off value for potential clinical significance (i.e., 25%) 20 . Also, lapatinib was predicted to cause ~25% increase in AUC of UGT1A1 substrates (4-MU and SN-38) when the f m is 1, indicating a potential DDI risk with clinical significance. On the other hand, axitinib could not cause the increase in AUC of UGT1A4, 1A7, or 1A9 substrates when the f m is 1, and the AUC i /AUC ratios were close to 1. UGT1A9 inhibition by Vandetanib, UGT1A1 inhibition by imatinib, and UGT1A7 inhibition by lapatinib were predicted to increase the AUC of substrate by about 2%, 3%, and 1%, indicating minor clinical impacts of the TKIs on the pharmacokinetics of other UGT substrates.

Discussion
In recent years, a number of TKIs of different classes have been approved for cancer treatment. Among these TKIs, several TKIs of the 4-anilinoquinazoline class (erlotinib and gefitinib) and pyridine class (sorafenib) have been shown to inhibit UGTs 12,13,15,21 . In the present study, we selected four commonly used TKIs, including two 4-anilinoquinazolines (lapatinib and vandetanib), one pyrimidine (imatinib) and one indazol (axitinib) to investigate whether these TKIs could inhibit a panel of UGTs activities. To this end, 4-MU was used as a global probe substrate for UGT enzymes (except for UGT1A4) based on the following. The kinetic behavior of 4-MU glucuronidation mediated by different UGT enzymes has been fully characterized 22 . Also, specific probe substrates for most individual UGTs have not been identified yet 23 while 4-MU has been demonstrated to be a robust probe substrate in assessing the activity of most recombinant human UGTs in many studies 12,24,25 . Thus, characterization of UGT inhibition using 4-MU as a probe compound should provide solid preliminary data to estimate UGT inhibitory potentials of perpetrator drugs. Important inhibition is uncommon for compounds with IC 50 values greater than about 75-100 μ M because sufficiently high levels are not clinically achieved 26 . So in the current study, for UGT isoforms whose activity is inhibited by >50% by individual TKIs at 100 μ M, IC 50 and K i values of TKIs were further estimated. Most notable finding in our study was UGT inhibition by imatinib and lapatinib. Our study has revealed that imatinib is a broad inhibitor of several UGTs. Imatinib is a potent inhibitor of   UGT2B17; an intermediate inhibitor of UGT1A1; and a weak inhibitor of UGT1A7, 1A8, 1A9, 1A10, 2B7, and 2B15. The human UGT2B enzymes are involved in the metabolism of steroid hormones as well as bile acids, acidic steroids, and fatty acids 27 . Among UGT2Bs, UGT2B17 exhibits high levels of activity against androgens and xenobiotics. UGT2B17 substrate drugs include carcinogens, coumarins, anthraquinones, flavonoids, nonsteroidal anti-inflammatory drugs, monoterpenoids, phenols, suberoylanilide hydroxamic acid (SAHA), and exemestane [28][29][30] . Therefore, the potent inhibition of UGT2B17 by   imatinib could potentially have a significant clinical effect on the metabolism of UGT2B17 substrates. In fact, our results from quantitative prediction of DDI risk indicate that the co-administration of imatinib at clinical doses can result in over 2-fold increases in AUC of a UGT2B17 substrate drug. Our data offer in vitro evidence that lapatinib is a potent inhibitor of UGT1A1. UGT1A1 is broadly expressed in human organs including liver, intestines, and kidney [31][32][33] ; its expression levels in the intestines and kidney are one third as high as that in liver 11 . About 15% of top 200 prescribed drugs in the United States in 2002 are eliminated mainly via glucuronidation by UGT1A1 6 , and the inhibition of UGT1A1 can have clinically significant impacts on drug therapy with a narrow therapeutic index drug such as irinotecan. Irinotecan is a chemotherapeutic agent commonly used for the treatment of colorectal cancer. Irinotecan requires metabolic activation by carboxylesterase 2 to the active metabolite SN-38. SN-38 is mainly eliminated via glucuronidation by UGT1A1, with minor contribution by UGT1A3, 1A6, and 1A9 19 . Because UGT1A1 inhibition can exhibit substrate-dependency 18 , two UGT1A1 substrates, 4-MU and SN-38, were used in the kinetic study of lapatinib. Interestingly, the kinetic profile of inhibition of SN-38 glucuronidation was different from that of 4-MU glucuronidation. This finding offers new experimental evidence that the reduction of UGT1A1 activity might vary with the substrate, and also for the opinion that UGT1A1 has two or more binding sites for xenobiotics and endobiotics 18 . Our results from the quantitative prediction of DDI risk indicate that the coadministration of lapatinib at clinical doses could result in around 25% increase in AUC of SN-38. Because the extent of UGT inhibition in in vivo systems can be underestimated when the prediction is based solely on in vitro parameters 34 , our prediction of SN-38 and lapatinib interaction appears consistent with the clinical observation that co-administration of lapatinib increased the AUC of SN-38 by an average of 41% 16 .
Different from above two TKIs, vandetanib was found as an intermediate inhibitor of UGT1A9, which is also involved in the glucuronidation of a number of drugs 9 . UGT1A9 is expressed in both human liver and some extrahepatic tissues including the gastrointestinal tract 9 . Significant evidence exists supporting a role for gastrointestinal UGTs as modifiers of pharmacokinetics and biological responses to drugs and xenobiotics 35 . Our data show that vandetanib might influence the first-pass effect and bioavailability of more orally administered drugs. This issue requires additional attentions, especially when modern medicinal chemistry tends to synthesize polar chemicals to avoid the metabolism by CYPs, which makes Phase II enzymes become the main metabolizing enzymes such as UGTs.
It is noteworthy that the extrapolation from in vitro data to in vivo drug interactions should be taken with caution. Various factors may affect the accuracy of prediction, including enzyme resources, incubation systems, protein binding, active transporters, the selection of probe substrates, and the accumulation of products in the reaction medium so on. In addition, UGTs in the gastrointestinal tract may contribute significantly to the first-pass metabolism of orally administered drugs that undergo glucuronidation. The concentrations of substrates in the intestine may be different from the concentrations used here to predict the AUC ratio. Therefore, further systemic studies are needed to clarify the in vivo effects.
In summary, the present findings highlight potential DDIs involving specific TKIs and UGT substrates, namely by inhibition of UGT1A1 and 2B17 by lapatinib and imatinib, respectively. Our study provides a basis for design of clinical studies for investigation of DDIs involving TKIs.  12 . For UGT1A4, TFP was selected as the substrate. TFP (final concentration: 40 μ M) was incubated with one of four TKIs and UGT1A4 (0.1 mg/ml). The following known UGT inhibitors were used as positive controls (at 100 μ M): diclofenac for UGT1A1, 1A6, 1A7, and 1A9; hecogenin for UGT1A4, androsterone for UGT1A3, 2B7, and 2B15; and phenylbutazone for UGT1A8 and 1A10, respectively 12,22,34 . As there are no known inhibitors of UGT2B4 and 2B17, positive controls for these enzymes were not included. All the inhibitors and 4-MU or TFP were dissolved in DMSO. The final concentration of DMSO in the incubation system was 1% (v/v). The UGT reaction mixtures were pre-incubated at 37 °C for 5 min, and reactions were initiated by adding UDPGA (final concentration 5 mM). The following incubation times were used for different UGT isoforms: 120 min for UGT1A1, 1A10, 2B4, 2B7, 2B15, and 2B17; 75 min for UGT1A3; and 30 min for UGT1A4, UGT1A6, 1A7, 1A8, and 1A9. Rate of product formation for each isoform was linear with respect to protein concentration and incubation time. All the reactions were quenched by adding chilled acetonitrile (100 μ l) containing phenytoin (1 μ M) as an internal standard. The mixtures were centrifuged at 16,000 g for 16 min to obtain the supernatant. Samples were then analyzed in LC-MS/MS. The relative rate of 4-MUG or trifluoperazine glucuronide (TFPG) formation for each isoform was linear with respect to protein concentration and incubation time.

Chemicals and Reagents. Axitinib
4-MUG concentrations were determined by using an Agilent 1200 HPLC interfaced with Applied Biosystems Qtrap 3200 equipped with an electrospray ion source. Chromatographic separation was carried out with a Waters XTerra MS C18 column (2.1 × 50 mm, 3.5 μ m; Waters Corporation, Milford, MA). The mobile phase consisted of 0.1% formic acid in water (A) and acetonitrile (B) and was delivered at 250 μ l/min. The gradient was initiated at 85% mobile phase A to 90% mobile phase B over 1 min, held constant for 1.5 min, and then restored to the initial composition until 10 min. The MRM parameters for 4-MUG and phenytoin as the internal standard were 353.05/177.20 and 253.30/182.20 in positive ion mode.
Signals for TFPG were detected by using Applied Biosystems Qtrap 4000 with an electrospray ion source interfaced with a Shimazu Prominence Modular HPLC. The mobile phase consisted of 0.1% formic acid in water (A) and acetonitrile (B). The column was first equilibrated at 25% mobile phase B for 1.5 min at a 200 μ l/min flow rate. The elution was then ramped linearly to 90% mobile phase B over 1.5 min, maintained for 5 min, and followed by a return to initial conditions. TFPG was detected using MRM pair of 584/408 in positive ionization mode 36 . Phenytoin was used as the internal standard.

Inhibition of SN-38 Glucuronidation Assay. SN-38 glucuronidation activity was determined with
a slight modification of a previously published method 13 . SN-38 was incubated in the absence or presence of different concentrations of lapatinib (0-200 μ M). Reactions were performed using recombinant UGT1A1 protein (0.1 mg/ml) or pooled HLMs (0.4 mg/ml). When HLMs were used, 50 μ g/mg protein alamethicin was added into incubation systems for 15 min preincubation on ice. Alamethicin was dissolved in 90/10 incubation buffer/ethanol. The final concentration of ethanol was less than 0.1%. After preincubation for 5 min, the reaction was started by adding UDPGA (final concentration 5 mM). Incubation was performed at 37 °C for 30 min and stopped by adding chilled acetonitrile. The samples were processed as described above for analysis using LC-MS/MS.
The concentration of SN-38G was determined by using an Agilent 1200 HPLC interfaced with Applied Biosystems Qtrap 5500. The MRM setting for SN-38G was 569.2/393.2 in the positive mode. The same internal standard was used for quantification. The mobile phase consisted of 0.1% formic acid in water (A) and acetonitrile (B). The column was first equilibrated at 10% mobile phase B for 1 min at 250 μ l/ min. The elution was then ramped linearly to 90% mobile phase B over 2 min, maintained for 2 min, and followed by a return to initial conditions. Determination of IC 50 . The same incubation condition as for the preliminary inhibition study was used for each UGT isoform, except that the concentration of each inhibitor was ranged from 0 to 200 μ M. The relative formation rate of 4-MUG, TFPG or SN-38G in the presence vs. absence of a TKI was estimated, and the half maximal inhibitory concentration (IC 50 ) value for each TKI was estimated by using GraphPad Prism 5 software (La Jolla, CA).
Determination of K i . 4-MU at four concentrations (ranging from 0.5 × K m to 4 × K m ) was incubated with recombinant UGT1A1, 1A3, 1A7, 1A8, 1A9, 1A10, 2B7, 2B15 or 2B17 at a final protein concentration as described in the UGT glucuronidation inhibition assay in the presence or absence of TKI at four different concentrations (ranging from IC 50 /4 to 4 × IC 50 ). The incubation time for each UGT isoform was the same as that used to determine IC 50 . SN-38 (2, 4, 10 or 20 μ M) was incubated with UGT1A1 (0.1 mg/ml) or HLMs (0.4 mg/ml for) in the presence of lapatinib (0, 0.1, 0.5, 1 or 2 μ M). K i values were calculated using nonlinear regression according to the models for competitive inhibition, noncompetitive inhibition, or mixed inhibition. Global nonlinear regression analysis in the kinetics module of GraphPad Prism 5 software was used to simultaneously fit all data to the equations for competitive inhibition, non-competitive inhibition, and uncompetitive inhibition. The type of inhibition was determined based on Lineweaver-Burk plot and the best fitting of data to the equations. Goodness of fit to kinetic and inhibition models was assessed from the F statistic, r 2 values, parameter standard error estimates and 95% confidence intervals. Kinetic constants are reported as the mean value ± standard error of the parameter estimate. Lineweaver-Burk plot and Dixon plot were also used as graphical representation of the type of inhibition and K i . where AUC i and AUC are the AUC in the presence and absence of inhibitor, respectively; K i is inhibition constant from the in vitro inhibition experiment; f m is the fraction of the UGT substrate metabolized by the inhibited enzyme; and [I] is the TKI concentration at the enzyme active site.

Estimation of in vivo TKI
In view of the general assumption that only unbound drug is available for interaction with the enzyme active site, and the aim of DDI research is to prevent the highest potential risk, the maximum unbound hepatic input concentration ([I] in,u ) was used for [I] in this study. Of note, clinically observed maximum unbound systemic plasma or blood concentration (C max,u ) was not used because the C max,u is known to be much lower than the [I] in,u 37 . Also, for the UGTs expressed in gastrointestinal track (such as UGT1A7), the maximum unbound systemic plasma concentration ([I] max,u ) was used to represent [I]. The prediction of AUC ratio is for any potential UGT substrates with undetermined f m , a range of f m between 0.1 and 1 was included for calculating AUC i /AUC ratio.